DEFORESTATION DEGRADES RAIN FOREST STREAM HABITAT AND BIODIVERSITY OVER TIME IN THE RAMA - KRIOL INDIGENOUS TERRITORY, SOUTHEAST NICARAGUA By Joel T homas Betts A THESIS Submitted to Michigan State University i n partial fulfillment of the requirements f or the degree of Fisheries and Wildlife Master of Science 2019 ABSTRACT DEFORESTATION DEGRADES RAIN FOREST STREAM HABITAT AND BIODIVERSITY OVER TIME IN THE RAMA - KRIOL INDIGENOUS TERRITORY, SOUTHEAST NICARAGUA By Joel Thomas Bet ts In southeast Nicaragua, recent waves of illegal deforestation for cattle pasture are damaging the Indio - Maíz Biological Reserve (IMBR) and Rama - Kriol territory (RKT) , with negative consequences to aquatic ecosystems and the people who rely on services t hey provide. Deforestation and subsequent land use are causing shifts in stream community structure that are mediated by changes in stream habitat. This study integrated temporally explicit land use informati on with stream habitat, macroinvertebrate, fresh water shrimp, and fish community data studied protected rainforests. The new calculation, deforestation history index (DFI) , a product of deforestation amount and ti me since deforestation for the catchment draining to each stream reach, was the best linear predictor of most taxa responses better than other habitat metrics and raw forest cover at multiple scales. Stream r eaches that were deforested for a longer time an d to a larger extent thus having higher values for the DFI had less large wood, organic debris, macroalgae, and macrophytes; more stream bank erosion and sedimentation; degraded riparia; lower diversity and a bundance of m acroinvertebrates, shrimp, and fish ; higher invertebrate evenness; and distinct changes in invertebrate community composition. All deforested reaches also had smaller sized game fish. New registers of fish species and insect genera were record ed for Nicaragua. As this is the first aquatic s tudy in these watersheds of the IMBR and RKT, this region should be a high priority for further research and conservation investment before it is lost. R ESUMEN LA DEFORESTACIÓN DEGRADA EL HABITAT Y LA BIODIVE RSIDAD DE LOS RÍOS A LO LARGO DEL TIEMPO EN EL T ERRITORIO INDÍGENA RAMA - KRIOL, SURESTE DE NICARAGUA Por Joel Thomas Betts En el sureste de Nicaragua, l as recientes actividades de deforestación ilegal para ganadería, está impactando la Reserva Biológica Indio - Maíz (RBIM) y el territorio Rama - Kriol (TRK) , con consecuencias negativas para los ecosistemas acuáticos y las personas que dependen de los servicios que estos brindan . La deforestación y el uso de suelo aledaño provocan cambios en la estru ctura de la comunidad ecológica de los ríos, debido a cambio s en su hábitat acuático. Este estudio integró la información de uso de suelo a través del tiempo de las cuencas con datos de hábitat, macroinvertebrados (MI) , crustáceos y peces para evaluar los impactos de la deforestación en 15 ríos de cabecera en una z ona poc a estudiad a del sureste de Nicaragua . Un nuevo índice , el índice de historial de deforestación ( DFI ), un produc to que integra la cantidad y el tiempo de deforestación , calculado a nivel de la microcuenca , fue el mejor predictor lineal para la mayorí a de las respuestas biológicas cuantificadas , y fue mejor que otras métricas de hábitat y cobertura forestal . Los ríos que fueron deforestados durante más tiempo y en mayor medida , tuvieron menos madera, material orgánico, macroalgas, y vegetación acuática ; más erosión y sedimentación; bosque ribereño m ás degradado; menor diversidad y abundancia de MI , camarones y peces ; mayor uniformidad de MI ; y cambios distintos en la composición de la comunidad. Todos los sitios deforestados también tenían peces de meno r tamaño. Se encontra ron nuevos registros de peces e insectos acuáticos para Nicaragua . E sta región debería tener una alta prioridad para futuras investigaciones y esfuerzos de conservación antes de que se pierda . iii ACKNOWLEDGEMENTS This work was funded by a United States Student Fulbright Award for Nicaragua and Costa Rica from the Institute of International Education, and by the following awards at Michigan State University: The Robert C. Ball and Betty A. Ball Fisheries and Wildlif e Fellowsh ip, the Rose Graduate Fellowship Fund in Water Research Graduate Student Award , the College of Agriculture and Natural Resources Critical Needs Summer Fellowship, the Center for Latin American and Caribbean Studies Graduate Student Research Grant , and four semesters of a Graduate Teaching Assistantship with Lyman Briggs College. This research would not have been possible without the support of countless people, in Michigan, Nicaragua, and Costa Rica. First, I would like to thank my advisor Dr. Ger ald Urquha supervision, for all his support editing grant proposals and earlier versions of this thesis, for connecting me with people in Nicaragua, and for meeting with me for in numerable hours in support of my thesis and other work throughout the last three years at Michigan State. I would like to thank Dr. Chris Jordan for all his support in research idea development and for introducing me to a whole network of scienti sts, stude nts, and community leaders in Nicaragua. Thanks to Chris, as well as Dr. Kendra Cheruvelil and Dr. Eric Benbow, for serving on my committee. Thanks to these professors as well as Dr. Dana Infante and Dr. Pablo Gutiérrez - Fonseca for their advice a nd comment s on my analysis and earlier versions of the thesis. I would like to thank my labmates and associates Lauren Phillips, Matt Cleary, and Armando Dans for their ideas and support throughout this program. I extend thanks to the Department of Fisher ies and Wi ldlife at Michigan State for providing an inspiring, applied, and iv productive context for learning and research, and to the Michigan State Center for Statistical Training and Consulting (CSTAT), specifically Hope Akaeze and Andrew Denhardt, who of fered inva luable statistical advice. I would like to thank the numerous people who assisted with the field work: Jossly Flores Mc.rea, Nestor Joel Gonzalez Aleman, and Keren Matus from Bluefield is an Indian and Alicia, and Vale rio from t he community of Sumukat; and many others who helped to prepare food, offered places to stay, and gave advice during field work. I am very grateful to the Rama and Kriol community leaders and forest rangers who from the beginning were supportive o f my work and offered their consent and advice to conduct the work in their communities and territory. Thanks to the Consejo Regional Autónomo Costa Caribe Sur and the professors and deans at BICU - FARENA for supporting me with the appropriate permits and p aperwork r equired for the study, and to the Centro de Investigaciones Acuáticas de la BICU (CIAB) for their collaboration and lab space in Bluefields, Nicaragua. Many thanks to Professor Monika Springer, who generously provided space and equipment for this project i n her new aquatic entomology lab at the University of Costa Rica in San José. Thanks to Monika Springer, Dr. Pablo Gutiérrez - Fonseca, Dr. Wills Flowers, and Alejandra Jiménez Fretes for lending their expertise in macroinvertebrate identification. Special t hanks to Jareth Román - Heracleo, Marycruz Velasquez , Paola Campos Arce, Alvaro Cerdas Cedeño, and Darha Solano - Ulate, who lent their expertise and spent months working with me to sort and identify macroinvertebrate samples. Thanks also to the Univ ersity of Costa Rica v Zoological Museum for a space to keep my preserved specimens in perpetuity, and for help with fish identification specifically from Jorge San Jil, Arturo Angulo Sibaja, and Carlos Garita - Alvarado . Thanks to the staff at the Institute f or Interna tional Education and the U.S. embassies in Nicaragua and Costa Rica for working to make it possible for me to complete the second half of the program with Fulbright in Costa Rica in light of the Nicaraguan political crisis of spring 2019. I am es pecially i ndebted to my partner Laura, my parents and family, and my friends and housemates in Lansing, Bluefields, and at Casa Adobe in Santa Rosa , who were my source of joy vi TABLE OF CONTENTS LIST OF TABLES ................................ ................................ ................................ ................................ .. viii LIST OF FIGURES ................................ ................................ ................................ ................................ .. ix INTRODUCTION ................................ ................................ ................................ ................................ ...... 1 C ontext of t ropical d eforestation ................................ ................................ ................................ 1 Deforestation and s tream e cosystems ................................ ................................ ......................... 1 Tropical deforestation and stream habita t ................................ ................................ ................ 2 Tropical d eforestation and s tream m acroinvertebrat es ................................ ........................... 4 Tropical d eforestation and s tream f ish ................................ ................................ ...................... 6 The i mportance of the h istory of d eforest ation and l and u se ................................ .................. 7 The c ontext of d eforestation in S outheast Nicaragua ................................ ............................... 8 O bjectives , research question, and h ypotheses ................................ ................................ ....... 1 1 M ETHODS ................................ ................................ ................................ ................................ ............... 12 Site selection ................................ ................................ ................................ ................................ 12 Site set - up ................................ ................................ ................................ ................................ ..... 16 Stream habitat sampling ................................ ................................ ................................ ............. 17 Riparian condition sampling ................................ ................................ ................................ ..... 20 Macroinvertebrate sampling and identification ................................ ................................ ..... 21 Shrimp sampling and identifica tion ................................ ................................ .......................... 22 Fish sampling and identification ................................ ................................ .............................. 22 Spatial data p rocessing ................................ ................................ ................................ .............. 23 Catchment - scale forest cover parameters ................................ ................................ ................ 24 Deforestation history index ................................ ................................ ................................ ........ 27 Habitat and landscape v ariable s election ................................ ................................ ................ 28 Habitat n onparametric c omparisons ................................ ................................ ........................ 29 Analysis of m acroi nvertebra te , shrimp, and f ish community response metrics .................. 29 Multivariate analysis of the macroinvertebrate comm unity ................................ .................. 31 Linear r egression analyses ................................ ................................ ................................ ........ 32 RESULTS ................................ ................................ ................................ ................................ .................. 34 Differences in habitat condition ................................ ................................ ................................ 34 Deforestation history as a predictor of stream habitat ................................ .......................... 3 8 Macroinvertebrate , fi sh , and shrimp summary ................................ ................................ ....... 38 Differences in m acroinvertebrate , shrimp, and fish community response metrics ............ 39 Differences in fish length ................................ ................................ ................................ ............ 44 Changes in m acroinvertebrate community structure ................................ ............................. 48 Taxa specific responses ................................ ................................ ................................ .............. 51 Deforestation history and habitat as predictors of the stream community ......................... 54 vii DISCUSSION ................................ ................................ ................................ ................................ ........... 63 Instream habitat response mediated by deforestation eff ects ................................ ................ 63 Changes in riparia caused by hurricane and deforestation effects ................................ ...... 65 Consistent reductions of m acroinvertebrates , shrimp, and fish ................................ ............ 66 Taxa response and stream h abita t ................................ ................................ ............................ 69 History of d eforesta tion as the best predictor of t axa r esponse s ................................ .......... 72 Thresholds of habitat and biotic disturbance ................................ ................................ .......... 73 The BMWP index appropriate for assessing deforestation impacts to streams? ............... 73 Study l imitations ................................ ................................ ................................ .......................... 74 Novel f indings and future r esearch p riorities ................................ ................................ .......... 76 A n ew i ndex ................................ ................................ ................................ ................................ .. 78 Relevance to c onservation ................................ ................................ ................................ .......... 79 Conservation r ecommendations ................................ ................................ ................................ 79 Conclusion ................................ ................................ ................................ ................................ .... 8 2 APPENDICES ................................ ................................ ................................ ................................ .......... 83 APPENDIX A : Raw Data ................................ ................................ ................................ .......... 84 APPENDIX B : Additional statistics and graphs ................................ ................................ .... 92 REFERENCES ................................ ................................ ................................ ................................ ....... 100 viii LIST OF TABLES Table 1: Nonparametric test results by habitat variable ................................ .............................. 36 Table 2: Single regression comparisons of the deforestation history index at the watershed scale (X) as predictors of habitat responses (Y) ................................ ................................ ............. 38 Table 3 : T - test and ANOVA and Tukey post - hoc pairwise comparisons for macroinvertebrate community summary statistics ................................ ................................ ....... 41 Table 4: ANOVA results for fish standard lengths by species ................................ .................... 45 Table 5: Indicator analysis and SIMPER results ................................ ................................ ........... 52 Table 6: Single regression comparisons of landscape and habitat variables (X) as predictors of macroinvertebrate taxa responses (Y) ................................ ................................ ......................... 56 Table 7: Single regression comparisons of landscape and habita t variables (X) as predictors of fish and shrimp responses (Y) ................................ ................................ ................................ ........ 58 Table A.1: Reach details ................................ ................................ ................................ ....................... 84 Table A.2: Full list of macroinvertebrate taxa abundances, by reach ................................ ....... 85 Table A.3: Full list of fish taxa, by reach ................................ ................................ .......................... 90 Table B.1: Correlations of taxa response metrics ................................ ................................ ........... 92 ix LIST OF FIGURES Figure 1: Forest cover and deforestation in the protected areas of Southeast Nicaragua .... 1 3 Figure 2. Forest cover and d eforestation in the catchment above (draining to) each study reach (study sites ................................ ................................ ................................ ................................ .... 14 Figure 3 : Transect set - up for instream habitat, riparian, and macroinvertebrate metrics . 17 Figure 4 : Forest cover in the catchment and buffer over time ................................ .................... 25 Figure 5: Macroinvertebrate community summary statistics for two forested watersheds and a recently and less recen tly deforested wa tershed ................................ ................................ .. 42 Figure 6: Fish and shrimp community summary statistics for two forested watersheds and a recently and less recently deforested watershed ................................ ................................ .......... 43 Figure 7: Fish standard lengths by species for two fore sted, and a recen tly deforested and less recently deforested watershed ................................ ................................ ................................ ..... 47 Figure 8: Stress plot for non - metric multidimensional scaling analysis ................................ ... 49 Figure 9: Non - metric multidimensional scaling ordination plots of macroinvertebrate communit y matrix ................................ ................................ ................................ ................................ .. 49 Figure 10: Single linear regression comparisons of the deforestation history index at the catchment scale (X) as a predictor of habitat responses (Y) ................................ ....................... 59 Figure 11: Single regression comparison s of the deforestation history index at the catchment scale (X) as a predictor of macroinvertebrate, fish, and shrimp responses (Y) . 61 F igure 12: Examples of streams in each watershed ................................ ................................ ........ 70 Figure B.1: Non - metri c multidimensional scaling o rdination plots of macroinvertebrate community matrix ................................ ................................ ................................ ................................ .. 93 Figure B.2: Habitat metrics for two forested watersheds a recently and less recently deforested watershed ................................ ................................ ................................ ............................. 95 1 INTRODUCTION Context of t ropical d eforestation The rate of deforestation in primary rain forests is high throughout much of Latin America (Wright, 2005; Hansen et al., 2013). This is a global problem, as neotropical rainforests y and are critically important for global climate change mitigation (Bonan, 2008). Deforestation and subsequent conversion to pasture, agricult ure, urban area, or other anthropogenic land use s threaten all components of the forest ecosystem, including aeri al , canopy, terrestrial, subterranean, and aquatic organisms and ecosystems processes, as well as the people who rely on the services they provide (Foley et al. , 2007 ). Deforestation and s tream e cosystems Freshwater organisms and their habitat can be seve rely affected by deforestation and land use change. In 2003, Bens tead, Douglas, & Pringle conservatively estimated that globally, each year in the humid tropics >5 x 10 5 km of stream channel are impacted by deforestation. Habitat degradation from land use change i s one of the most significant threats to freshwater biodiversity and ecosystem function (Dudgeon et al., 2006; Reid et al., 2018). Inland fisheries are an important ecosystem service provided by freshwater biodiversity and are increasingly threaten ed by human - induced environmental change (Phang e t al ., 2019). Therefore, it is critical to consider the impacts of landscape changes when studying, managing, or conserving stream ecosystems ( Fausch, Torgersen, Baxter, & Li , 2002 ; Allan 2004). Many studies have shown that impacts to stream community structure from deforestation are caused by i ts effects on water quality and instream habitat ( Harding, Benfield, Bolstad, Helfman, & Jones , 1998 ; Gergel, Turner, Miller, Melack, & Stanley, 2002 ; Iwa ta, Nakano, & 2 Inoue , 2003; Leitão et al. , 2017 ; ) . Allan (2004) presented impacts of land use to stream habitat in six main categories as they effect stream biota: sedimentation, nutrient enrichment, contaminan t pollution, hydrologic alteration, riparian clearing/canopy opening, and loss of large woody deb ris. The magnitude and form of these impacts to habitat depends not only on the land use type, history, and the proximity of the disturbance to the stream chan nel, but also on natural hydrogeological and climatic conditions. These interrelated influences of deforestation determine the specific mechanisms of impact to stream biota, and their response s differ based on the requirements and tolerance of each specie s. For example, cattle ranching following deforestation in temperate regions tends to result in decreased shade and increased stream temperature, eroded banks and siltation , simplification of stream bottom habitat, and eutrophication from nutrient overlo ad related to excrement, which together cause a loss of sensitive macroinvertebrate taxa and domi nance by burrowing taxa (see Figure 1 from Strand & Merrit t , 1999, 14). In general, as anthropogenic influence s increase stream conditions move beyond threshol ds of tolerance , and most organisms adapted to natural conditions ultimately decrease in abundanc e (Allan, 2004). Many, but not all lessons learned from temperate systems apply to tropical systems (Dodds, Gido, Whiles, Daniels, & Grudzinski, 2014). In the past 15 years there has been a proliferation of studies showing how the dynamics of land use cha nge, particularly deforestation, a ffect tropical stream habitat and biota . But a comprehensive review does not yet exist. Tropical deforestation and stream ha bitat The habitat impact categories from Allan (2004) (bolded below) are also relevant to tropical streams. Deforestation and associated land use causes sedimentation ( Heartsill - Scalley & Aide , 2003 ; Iwata et al., 2003 ) , which can lead to decreased bed sta bility ( Iwata et al., 2003, 3 Leitão et al., 2017; Macedo, Hughes, Kaufmann, & Callisto, 2018), lo ss of interstitial spaces (higher embeddedness) and subsequent declines in fish and invertebrate taxa richness and periphyt on mass (Iwata et al., 2003 ). In temp erate streams, Schwendel, Death, Fuller, & Joy ( 2010 ) observed declines in taxa richness and periphyton mass and increases in evenness related to decreased bed stability. Sedimentation can also result in higher rates of macroinvertebrate drift ( , Jocqué, & Kelly - Quinn, 2015). Nutrient enrichment has been related to pasture and agricultural land use ( Mori, de Paula, de Barros Ferraz, Camargo, & Martinelli, 2015) and was linked to higher macroinvertebrate drift ( , and lower dissolved oxygen ( Teresa, Casatti, & Cianciaruso, 2015 ; Tanaka, de Souza, Moschini, & de Oliveira, 2016 ). Hydrologic alterations such as bank erosion (Iwata et al., 2003, Chaves et al., 2008; Wantzen & Mol, 2013; Leitão et al., 2017 ), variation in depth ( Leal et al., 2016), increases of bankfull width/depth ratio ( Leitão et al., 2017), decreases in stream depth (Montag et al., 2019) , decreases in discharge ( Coe, Costa, & Soares - Filho, 2009), increases in wet - season surface flows ( Chaves et al., 2008), and increased flashiness and flooding (Bradshaw, Sodhi, P eh , & Brook, 2007; Chaves et al., 2008; Recha et al, 2012; Arancibia, Bruijnzeel, Mulligan, & van Dijk, 2019) can also result from deforestation and were related in many of these cases to shifts in the b iot ic community . Riparian clearing/canopy opening can cause decreases in mid - channel shade (Leal et al., 2016), higher periphyton biomass ( Bojsen & Jacobsen, 2003; Lobón - cerviá , Mazzoni, & Rezende , 2016; Feijó - Lima et al., 2018 ) , higher water temperature ( Benstead et al., 2003 ; Fugère, Kasangaki, & Chapman, 2016 ; Leal et al., 2016), lower levels of benthic organic matter or leaf litter ( Bojsen & Ba rriga, 2002; Bojsen & Jacobsen, 2003; Benstead et al., 2003; Brejão et al., 2018; Montag et al., 2019 ) , increas ed aquatic vegetation ( Leitão et al., 2017), and declines in 4 terrestrial insect inputs ( Chan, Zhang, & Dudgeon , 2008; da Silva Gonçalves, de Souza Braga, & Casatti , 2018; as found in Nakano, Miyasaka, & Kuhara, 1999). L oss of instream large woody debris re sults from deforestation ( Heartsill - Scalley & Aide , 2003 ; De Paula, Gerhard, Wenger, Ferreira, Vettorazzi, Ferraz, 201 1 ; Leal et al., 2016 ), and has been related to shifts in the biotic community (Wright & Flecker, 2004; Valente - Neto, Koroiva, Fonseca - Gess ner, & de Oliveira Roque, 2015 ; Leitão et al., 2017; Brejão et al., 2018; Montag et al., 2019 ). Most of these studies related these shifts in habitat to shifts in diversity, community composition, and other invertebrate and fish indicator and species respo nses. But since many of these habitat changes co - occur, it is challenging to connect specific changes in habitat to specific biotic responses ( Gergel et al., 2002) . Because of this, c atchment - scale deforestation and subsequent anthropogenic land use can be used as an integrator, and therefore a strong predictor of changes to instream habitat and biota (Leal et al., 2016; Molina, Roa - Fuentes, Zeni, & Casatti, 2017). Tropical d eforestation and s tream m acroinvertebrates Forested streams have consistently highe r macroinvertebrate taxa richness than deforested streams in many tropical studies ( Paaby, Ramirez, & Pringle, 1998; Iwata, Nakano, & Inoue , 2003; Lorion & Kennedy, 2009a; Iñiguez Armijos, Leiva, Frede, Hampel, & Breuer, 2014; Fugère et al. , 2016; Tanaka e t al., 2016 ; Montag et al., 2019 ). Higher richness is often due to the maintenance of especially sensitive or specialized taxa, such as those in the orders Ephemeroptera, Plecoptera, Trichoptera, and Odonata ( Siegloch, Schmitt, Spies, Petrucio, & H ernández , 2017; Brito et al., 2018). Taxa evenness/dominance has also been used to assess deforestation impact, but has shown variable results, with some studies showing no difference 5 ( Iwata et al., 2003 ; Iñiguez Armijos et al., 2014), and others showing forested sites having higher evenness (Fugère et al., 2016). Indices of biotic integrity (IBIs) are often used to summarize the response s of sensitive taxa to disturbances, and many of these have recently been develop ed for tropical streams. Some e fforts from Latin America include the Biological Monitoring Working Party ( BMWP ) Index for Col o mbia (Zamora, 2007), Costa Rica ( Springer, Ramírez, & Hanson , 2010), Panama ( Cornejo, 2010) and Cuba (Naranjo et al., 2005) ; the Índice Biológico a Nivel de Familia de Invertebra dos Acuaticos for El Salvador ( IBF - SV - 2010 : Sermeño et al., 2010) ; and a variety of other multi - me t ric indices ( Helson & Williams, 2013; Chen et al., 2017 ). These IBIs are commonly used by neighboring countries with similar ecology , such as the BMWP index for Costa Rica in Nicaragua ( González, Mateo, & Valdivia, 201 3 ; Salvatierra, 2014) . C atchment - scale deforestation and subsequent anthropogenic land use s can be strong predictor s of changes in IBI indices ( Ligeiro et al., 2013; Iñiguez Armijos et al., 2014) . In addition to changes in these metrics, differences in community composition have been commonly reported. In multiple studies, the macroinvertebrate community was found to be significantly different between forested and deforested stream reaches accordi ng to multivariate techniques based on community similarity (Benstead et al., 2003 ; Lorion & Kennedy, 2009a ; Fugère et al., 2016 ; I ñiguez Armijos et al., 2014) . In all these studies, forested sites were also more similar to each other than deforested sites , which were more variable in community composition. These studies show that deforestation changes the macroinvertebrate community, but not necessarily in a consistent way among sites. These changes in community composition are driven by individual tax on r esponses to disturbance and subsequent trophic effects that are highly context dependent. Specific taxa 6 responses are too contingent on study conditions and region to provide constructive background for comparative studies . But e ffects on important ecosyst em functions can be caused by the declines in abundance of even single taxa due to deforestation, as was the case in the mountains of Ecuador where the decreases in abundance of an important leaf litter shredding genus of caddisfly at deforested sites sign ificantly decreased large organic matter processing rates compared to forested sites (Encalada, Calles, Ferreira, Canhoto, & Graca, 2010). Tropical d eforestation and s tream f ish Neotropical freshwater fishes face a variety of threats, many of which are lin ked to deforestation and land use change and associa ted habitat degradation (Pelicice et al., 2017). Multiple studies have shown that stream fish abundance, community assemblage, and trophic structure shift in response to impacts from deforestation. Defore station in the catchments draining to streams has been linked to decreases in beta - diversity of fish , resulting in deforested sites with homogenized community structure, dominated by species adapted to deforested conditions (Bojsen & Barriga, 2002; Lorion & Kennedy, 2009b; Teresa et al. , 2015 ; et al., 2016 ). Studies have shown community changes are related to shifts in instream habitat. Wright and Flecker (2004) found higher abundance of most species, and especially rare species, in streams where woody debris was not removed (loss of woody debris is coincident with deforestation). Bojsen and Barriga (2002) correlated shifts in the fish communi ty to increased sunlight and lower instream leaf abundance from loss of canopy cover. Teresa et al. (2015) showed increases in hypoxia tolerant i ndividuals after deforestation , which implies that shifts in water chemistry associated with deforestation also drive changes in the fish community. In Brazil, et al. (2016) found that following conversion to agriculture and subsequent substrate siltation and sunlight exposu re, macrophytes abounded and 7 there was a shift from benthic and lithophilic fish towards nektonic, macrophyte - associated fish. In the same system, Leitão et al. (2017) found that deforestation was linked to declines in the functional evenness of assemblage s as mediated by increases in macrophytes , and that riverscape fragmentation from road crossings from logging and agricultural roa ds was linked to reductions of functional diversity and evenness in streams. Shifts in the fish communit ies can also result fr om shifts in diet related to impacts from deforestation . For example, multiple tropical studies have emphasized the importan ce of terrestrial arthropod s from forested riparian zones in fish diets ( Chan et al. , 2008 ; da Silva Gonçalves et al. , 2018 ) . The lack of this could result in shifts in the fish community . Lobón - cerviá e t al. (2016) and Bojsen and Barriga (2002) showed increased dominance of periphyton in fish diet at deforested sites. They found that periphyton - feeding loricariids made up more than 50% of fish at deforested sites and were less abundant or absent at fores ted sites. The i mportance of the h istory of d eforestation and l and u se Impacts to streams from deforestation change over time and are related to the unravelling of processes that only begin with initial deforestation and depend on the type of subsequent l and use s . However, few studies have analyzed the effect of time since deforestation on stream responses. Time since deforestation and land use history could be just as or more important than the extent of forest loss for describing changes to stream habita t (Leal et al., 2016; Molina et al., 2017 ) and biotic response s ( Brejão et al., 2018; Casatti, 2019 ) . Brejão et al. (2018) found both time since deforestation and current extent of catchment deforestation to be i mportant predictors of changes to the fish community in Brazil. Zeni et. al. (2019) found that fish functional diversity was reduced in streams with a longer 8 history of deforestation. Even when reforested, there can be remnant effects from deforestation on habitat and biota tha t last decades ( Harding et al. , 1998 ; Iwata et al., 2003). The c ontext of d eforestation in S outheast Nicaragua Nicaragua is losing 1330 square kilometers of forest each year, mostly in protected reserves ( Alvarez, 2016 ) . Much of this is occurring near the Atlantic coast, as the agricultural frontier expands eastward (Jordan, 2015; Phillips, 2017). Nearly the entirety of the southern Atlantic region of Nicaragua is included in the massive Rio San Juan UNESCO Biosphere Reser ve, much of which is no longer forested. Its core area, the Indio - Maíz Biological Reserve, is one of the last and largest intact regions of primary forest left in Nicaragua and hosts pristine river systems and very high biodiversity of plants, fish, and wildlife ( Dans , Luna, & Jordan , 2015). Each of the limited number of published studies from Southeast Nicaragua calls for more research in these understudied and threatened systems ( Fenoglio , Badino, & Bona, 2002; Organiz ación de los Estados Americanos, 2005; Jordan, St evens, Urquhart, Kramer, & Roe, 2010; Dans et al ., 2015 ; Jordan, Schank, Urquhart, & Dans, 2016 ; , & Meyer, 2017 ). Indio - Maíz makes up the southern half of the Rama - Kriol indigenous territory, which was protected to provide space to sustain subsistence agriculture, fishing, hunting, and gathering by indigenous Rama and afro - desce ndant Kriol communities, on their t radit ional lands. The northern half of the territory is composed of the Cerro Silva and Punta Gorda National Reserves. Illegal deforestation by mestizo migrants from western Nicaragua over the last three decades has conve rted much of these northern reserves from primary rainforest to cattle pasture, and it is rapidly encroaching on the intact Indio - Maíz to the south (see Figure 1). H unting and fishing by mestizo migrants ha ve also taken their toll near the deforested areas ( Jordan, Galeano, & Alonzo, 9 2014). In most cases of deforestation, the forest is being slashed , burned , and converted to pasture for beef and dairy cattle production. In some cases crops are also grown. Long - term residents of the Rama - Kriol territory are alarmed by the changes in t heir landscape being driven by mestizo colonists invading the territory . Rama - Kriol leaders and forest rangers, along with local conservation organizations, are working to document impacts of the agricultural frontier on their r esources, in order to use this information in advocacy and management. As the Rama communities rely heavily on river fish and shrimp in their diets, knowledge of the effects of illegal deforestation and fishing on river fish and shrimp populations and the river ecosystem that supports them is of utmost interest to community leaders . This is a part of the Rama - - Maíz ( Gobierno Territorial Rama y Kriol , 201 8 ). In many cases, streamside areas are the first to be d eforeste d in these landscapes. Rivers and streams are entry points into the landscape, and focal points for starting new cattle ranches ( unpublished data, Gobierno Territorial Rama y Krio l). This has been shown to be true with the invasion and destruction of prima ry forests in other parts of the world as well (Ferraro, 1994). Unlike in many areas throughout the tropics, other disturbances to these rivers from infrastructure have been minimal there are no dams and very few road crossings in these watersheds, and in some streams in the Indio - Maíz reserve there are no human disturbances. This context not only allows for relevant conservation application of research on the impacts of deforestation on streams, but also provides a unique opportunity to document the ecolog ical effects of deforestation for cattle ranching within a gradient ranging from pristine primary rainforest streams to streams recently defo rested to those deforested much earlier . 10 Stream macroinvertebrates and freshwater fish are being increasingly studi ed in Latin America, although many gaps in research still exist ( Smith & Bermingham, 2005; Ramirez & Gutiérrez - Fonseca, 2014; Pelicice et al. , 2017). In Nicaragua, aquatic surveys have been scarcer , in particular along the Caribbean Coast. Species, genus, and even family presence are still being described ( Maes & Salvatierra - Suarez , 2014 ). The most recent comprehensive list of fish species in Nicaragua was in 1982 ( Villa, 1982), which was depauperate of registers from the southern Caribbean coast. S ince the n many updates to taxonomy and species lists in Costa Rica ( Angulo Sibaja, Bussing, Ga rita - Alvarado, & López, 2013) and Central America (Rican, Pialek, Dragova, & Novak, 2016) have been reported , many of these taxa exist in Nicaragua . Although Indio - Maíz i s more than 3,150 square kilometers in size, and one of the best protected primary rainforests in Central America, only one formally published macroinvertebrate study (Fenoglio et al., 2002) and no formally published fish stu dies (not including the San Jua n River, bordering the reserve to the south) exist from the rivers of the I ndio - Maíz Biological Reserve. Fenoglio et al. (2002) is very limited in scope and geographic distribution. It is likely that many undescribed species exist in these rivers. In addit ion, the ecology and range of many aquatic species in the region are poorly described ( Maes & Salvatierra - Suarez , 2014 ; Härer et al., 2017) . There have been no studies to date that assessed the relationships between deforestation/land use, stream habitat, and stream biota in Nicaragua, and very few studies of this nature in all of Central America (Lorion & Kennedy, 2009a,b; ). Given the impending threats from deforestation and cattle ranching to this data poor region, it is a high pr iority area for research. 11 O bjectives , research question, and h ypotheses The objectives of the study were (1) t o describe and assess the complex impacts of deforestation to stream habitat and com m unities in the Rama - Kriol territory; (2) to fill knowledge gaps on distributions and ecology of aquatic species in SE Nicaragua; and (3) to provide new information and resources to scientists, conservationists, and i ndigenous leadership further scientific research. T his study intend ed to answer the question: what are the effects of deforestation and subsequent cattle ranching on stream macro invertebrate (including shrimp) and fish communities and their habitat in the protected areas of so utheastern Nicaragua? It was hypothesized that c hanges in s tream and riparian habitat due to impacts from deforestation and cattle ranching over time within each catchment and its reach buffer would predict shifts in the stream biota . Specifically, that i n stream and riparian disturbances would be more evident in deforested watersheds, especially those with a longer deforest ation history including increased sedimentation, decreased stream bed and channel stability, damaged riparian condition, increased algae cover, increased temperature, decreased large wood, and decreased leaf litter, among other impacts and that these impac ts would lead to lower m acroinvertebrate richness, BMWP score, diversity, evenness, and density ; lower fish taxa richness, abundance, an d average lengths ; and differen ces in community composition in deforested catchments, especially those with a longer deforestation history . 12 METHODS Site selection Sampling was carried out in the Rama - Kriol t erritory and the national reserves of Southeas t Nicaragua, including fifteen headwater stream reaches each with a distinct catchment (Figure 1). E ight stream reaches were in primary forested watersheds (Indian River , N=5 and Corn River , N=3) and seven reaches were in deforested watersheds (Pijibaye Ri ver , N=3 and Kukra River , N=4) (Figure 2 ). Each reach was on different stream s draining to these larger rivers. Thus, reaches were considered independent, as no reach had another reach downstream of it (no catchment overlap). Each watershed represented a u nique disturbance class: The I ndian River watershed is primary rainforest but with some hurricane damage , the Corn River watershed is primary rainforest and without hurricane damage, the Pijibaye River watershed is recently becoming deforested, and the Kuk ra River watershed has been in the process of deforestation throughout the last 3 decades . An additional nine sites were planned (4 forested and 5 Petriello & Joslin , 2019) cut the field season short by two months. Data collection occurred during the dry season ( February to April ) of 20 18, with six field trips : Kukra River (5 - 11 Feb.), Indian River (16 - 28 Feb.), Indian River (10 - 14 Mar.), Kukra River (21 - 27 Mar.), Corn River (8 - 16 Apr.), and Pij ibaye River (17 - 25 Apr.). Due to the remoteness of sites, sometimes multiple days were spent t raveling by boat, dugout canoe, horse, or foot before reaching the headwaters of the rivers . Selecting and sampling each stream reach took 1.5 to 2.5 days. 13 Figur e 1: Forest cover and deforestation in the protected areas of Southeast Nicaragua . Study reaches are visualized by stars. Study reaches occur in headwater streams of Kukra, Pijibaye, Corn, and Indian River watersheds , which occur from north to south, respe ctively. Forest l oss year data grouped in 4 - 5 - year intervals for visualization. Hu rricane damaged forest is treated as forest land cover in all analyses. No Forest Pre - 2001 represents pixels without forest in 2000. See methodology for more detailed descrip tion. The heavily invaded Cerro Silva and Punta Gorda Reserves and the largely int act Indio - Maíz Reserve are from north to south, respectively, each of which overlaps with the Rama - Kriol territory . 14 Figure 2 . Forest cover and deforestation in the catchment above (draining to) each study reach (study sites). A) Five forested catchments with varying hurricane damage above sample reaches in the Indian River watershed. B) Three forested catchments above sample rea ches in the Corn River watershed. Loss year data grouped in 4 - 5 - year intervals for visualization, as in Figure 1 . Hurricane damaged forest treated as forest land cover in all analyses. Catchment data generated in ArcMap from ASTER DEM (90M) and forest cove r data from Hansen/UMD/Google/USGS /NASA (Hansen et al ., 2013). A) B) 15 Figure 2 C) Pijibaye River watershed. D) above sample reaches in the Kukra River watershed . C) D) 16 Site set - up Site set up and data collection used an adapted protocol based on the US EPA Ecosystem Monitoring and Assessment Program (EMAP), accordin g to Hughes and Peck (2006) and Kaufman, Levine, Robison, Seeliger, & Peck (1999). This methodology has also been applied to tropical streams in Brazil (Leal et al ., 2016 ; Terra, Hughes, & Araújo, 2016 ) . In each watershed streams were selected between thr ee and fifteen meters mean wetted width and in plane - bed or pool - riffle gradient class (Montgomery & Buffington, 1997; Lorio n & Kennedy, 2009a). Streams were at remote locations, so a topographic map and local guides were consulted to estimate gradient and stream size, which were then verified upon arrival. Starting points for site selection were at least 500 meters upstream of the confluence with the larger river. Site length was 40X mean wetted width, or 150 meters (m) for streams less than 3.75 m wide. M ean wetted width used in reach set - up was established via 10 measurements upon arrival to a proposed reach , at least 15 m ap art, within the proposed sample reach. Habitat assessment and macroinvertebrate sampling occurred at eleven transects per reach, at intervals of 4X mean wetted width (Figure 3 ) or 15 m for streams less than 3.75 m wide. Base transects were set at the upstr eam edge of a riffle habitat in each stream reach. Ma croinvertebrate, habitat, and riparian parameters were sample d at or between ea ch transect, starting at the downstream - most transect A and moving upstream to transect K. Fish sampling was throughout the whole reach. Since total sampling effort for a stream reach took 10 - 18 hours, typically 3 - 5 transects were left to finish in the a ft ernoon of the second day, after fish sampling. A GPS point was taken at each furthest downstream transect, and a track created for the sample reach distance by walking the whole stream channel from transect K to A, once all sampling was complete. All sam pl ing was carried out under the appropriate regional and local permits . 17 Figure 3 : Transect set - up for instream habitat, riparian, and macroinvertebrate metrics. Macroinvertebrate sample locations oscillated between river right, center, and left, with one per transect. Fishing conducted throughout the reach. Example longitudinal section of 15 m shown, for an example stream of mean wetted width 3.75 m , reach length 150 m . Figure from Leal et al. (2016, Supplementary Material, 9), Figure S1. Stream habitat s ampling Water temperature, conductivity, and pH were measured in the morning, mid - day, and late afternoon at the furthest upstream transect sampled to that point in time using Hannah 1301). At these (measured in NTU) (model 77096). The tube was filled to the top with water and allowed to drain until the disk at the bottom becomes visible, the n the height of water was recorded. If the disk at the bottom of the tube was clearly visible when the tube was filled with water to the top, NTU was record ed as <5 NTU (Myre & Shaw, 2006). 18 At each of the eleven transects, stream wetted width was measured. Stream substrate was estimated using a standard pebble count as defined by Kaufmann, Faustini, Larsen, & Shirazi (2008). Five equidistant samples were taken at each of the 11 transects in each reach (55 samples per reach), starting one seventh of the way across the transect. Substrates were divided into eight - gauge (1984) to measure small substrates and a collapsib le 2 meter - stick for larger substrates (Organic detritus; fines: <0.06 mm; sand: 0.06 2.0 mm; small grav el: 2.0 16 mm; Coarse gravel: >16 64; Cobble: 64 - 250 mm; Small Boulder: >250 1000 mm; Large boulder: >1000 4000 mm; and bedrock: >4000 mm), using the sh ortest substrate axis to determine substrate category. These data were summarized across each reach as g eometric mean diameter (Dgm) 1 and percent of each substrate class (Kaufmann et al ., 2008; Terra et al . , 2016). Depth (x.x cm) was m easured and percent e mbeddedness ( Watzin, & Hession, 2004) estimated at each substrate point. Between each transect, using a collapsible two - meter stick, maximum depth (thalweg) was measured at ten consecutive points, following the deepest channel. The dista nce between each point was 1/100 th of the calculated reach length. Each subsequent transect was set up a t the 10 th thalweg measurement. Thus there were two depth summary measurements, average depth (based on the 55 points, 5 evenly spread across each channel). According to EM AP protocols (Kaufmann , Levine, Robison, Seeliger, & Peck , 1999), percent cover of different habitat features in the wetted stream channel was visu ally estimated 1 Dgm will be determined as described in Kaufmann et al. (2008, 153 - Dgm was calculated by nominally assigning to each particle the geo metric mean diameter of the upper and lower bounds of its size class (e.g., 5.66 mm for fine gravel) and then calculatin g the geometric mean as the antilog of the arithmetic mean of the logarithms of those frequency - weighted class midpoint values. Bedrock and fines, respectively, were assigned class midpoint bute towards the Dgm. 19 within an area five meters upstream and five meters downstream of each of the eleven transects (Figure 3 ). All visual estimates were done by JT Betts, to maintain consistency. These features included periphytic macroalgae, macrophytes, large woody debris (>0.3 m diameter), small woody and leafy debris (<0.3 m diameter), live trees and roots, overha nging vegetation (within 1 meter of the water surface), boulders, and artificial structures. Proportion of stream bank actively being eroded within five meters of the transect was also estimated (see Figure 12 for example). For each reach and habitat chara cteristic, the percentages of the eleven transects were averaged to create a value representing the reach. Large wood number of pieces and total v olume within the bankfull channel for each reach was calculated using EMAP protocol from Kaufmann et al. (199 9). Large wood was defined as w oody material with diameter of at least 10 cm and length of at least 1.5 m. Wood was classified into four diameter c lasses (0.1 m to < 0.3 m, 0.3 m to < 0.6 m, 0.6 m to < 0.8 m, and > 0.8 m) and three length classes (1.5m to < 5.0 m, 5 m to < 15 m, and > 15 m), only counting the portion of the log that has diameter > 0.1m. Diameter and length class were visually estimat ed by JT Betts. The number of logs in each length - diameter category between each transect was tallied. Large wood abundance and volume was summarized into multiple reach - scale values ( Kaufmann et al., 1999). For analysis, large wood volume per 100 m of stream was used. To calculate this value for each diameter and length category, a representative value was assig ned ([Upper limit - lower limit] * [ 1/3 ] + lower limit) and volume of a cylinder calculated (length * (Diameter/2) 2 For example, for a l og in the smallest category, (length class 1.5 to <5 m and diameter class 0.1 to <0.3 m), volume was is calculated as {[ (5 - 1.5 ) * ( 1/3 ) + 1.5 ] * ([ (0.3 - 0.1 ) * ( 1/3 ) + 0.1 ] / 2) 2 = { * [( /2)] 2 } = 0.0582 m 3 ( Kaufmann et al. , 1999; Robinson , 1998). Total volume per 100 m was calculated as the sum of volumes of all the large 20 wood i n the transect/transect length ( m )*100 . At two reaches in the Indian River watershed, values from a transect with a major log jam was replaced with average values from the rest of the transects a t the reach, because these transects heavily biased the measu rements at the log jam reaches. Large wood measures serve as useful indicators of instream habitat and cover, as well as the extent of impact from Hurricane Otto (November 2016). Riparian conditi on sampling A densiometer was used at left edge, right edge, and in each direction from the center of each transect to estimate percent shade, according to Kaufmann et al. (1999). Densiometer readings were summarized into average percent shade per reach. Riparian condition was visually estimated in 10 x 10 m plots a t each transect following Kaufmann et al . (2008) with parameters as defined in Kaufmann et al. (1999) (Figure 3 ). Percent cover of large (>0.3 m diameter) and small (<0.3 m diameter) trees in the upper canopy (>5 m tall), and percent cover of woody and non - woody vegetation in the mid - canopy (between 0.5 m and 5 m tall) and ground layers (<0.5 m) were estimated. Val ues of each summary measure for all eleven transects were averaged. Presence of riparian human disturbances was recorded in 12 categories, weight ed by proximity to the stream edge (presence of roads, dams, trails, pasture, crops, pipes, etc.) (Kaufmann et al., 1999). A sum was calculated that represented the riparian human disturbance index (W1_HALL). Using this information and the following formu la, the riparian condition index (RCOND) was calculated (Kaufmann et al. , 1999, Kaufman & Hughes, 2006, Kaufman n et al., 2008). The RCOND index is determined by riparian % cover of large trees, woody vegetation at all three canopy layers, and proximity of different human disturbances to the stream bank. The equation for the riparian condition index is as follows, f rom Kaufmann et al. (2008): 21 RCOND = {( Mean u pper c anopy l arge t rees % c over) * otal woody veg. % cover in all three canopy layers) * [1 / (1 + W1_HALL)]} 1/3 Macroinvertebrate sampling and identification Macroinvertebrates were sampled using a Surber Sampler (Wildco 243 µm Nitex net, sample area 0.0929 m 2 =1 ft 2 ), at 11 locat ions for each reach (1 per transect), changing between river right, center, and left at each consecutive transect . Rocks were scraped clean and the sample area agitated with gloved hands until the substrate was loose within the sample area to a depth of 5 cm (2 - 4 minutes). The Surber sample area was not disturbed before sampling. For consistency, all macroinvertebrate samples were done by JT Betts. All samples were preserved i n the field using 95% ethanol in 250 ml containers labelled inside and out. When t he sample was more than three - quarters full of debris, contents were split, and an additional container was used. Samples were transported to the University of Costa Rica [ex ported under Law °N 28, Decree °N 3584, resolution °N 1076 - 22 - 08 - 2018 (SERENA) and dictate °N 31 - 2108 - 2018 (Consejo Regional), Nicaragua; and imported under Law °N 7317, Ordinary Session °N 088 - SETENA, Costa Rica] (Appendix C). Lab work was carried out und er supervision of Monika Springer in the Aquatic Entomology lab in the School of B iology at the University of Costa Rica. Springer et al. (2010), Domínguez and Fernández (2009), Roldán (1988), and Merritt and Cummins (1996) were consulted for identificatio n. JT Betts, J Román - Heracleo, P Campos, D Solano - Ulate carried out identification s, and consulted M Springer (Trichoptera, others), P Gutiérrez - Fonseca (Plecoptera, others), and W Flowers (Ephemeroptera) with unknowns. Invertebrates were identified to best taxonomic resolution possible, typically genus ( see Appendix A for a full list ). Identified specimens are cataloged at the University of Costa Rica Zoological Museum in the School of Biology (Contact: M Springer). Specimens were sorted i n 22 small glass vials (capped with permeable cotton) containing each distinct taxa of invertebrate fo und at a given reach. These vials were submerged in 90% ethanol in sealed jars for each reach (N=15) for future reference. Shrimp sampling and identification In pools throughout each reach four mesh pyramidal and five metal cylindrical traps were set overn ight with dog food as bait. The holes of the metal traps were adjusted to 4 inches in diameter to accommodate large Macrobrachium shrimp (Covich, Crowl, & He artsill - Scalley, 2006). The main purpose of the traps was to catch these freshwater shrimp, but fis h captures were also recorded. T raps were set out at the end of the first day of sampling at a reach and removed the following morning . Shrimp were identified to genus and morphospecies ( Atya [2 - 3 morphospecies] or Macrobrachium [3 morphospecies]) by JT Be tts, with help from N Gonzalez - Aleman . Fish sampling and identification The fish community was sampled at each reach using hook and line (Montaña & Winemiller, 2010) and cast - net methods. See Bojsen and Barriga (2002) for somewhat similar mixed methods. El ectrofishing was not possible due to remoteness of sites (some being >50 km from electricity or infrastructure) and the difficulty and risk of bringing expensive research equipment into Nicaragua. Consistent effort of each technique was applied, in attempt to have similar Catch p er u nit e ffort (CPUE) at each reach. Fish were sampled f irst thing in the morning on the second day, starting at the base transect and moving upstream. Three individuals fished with hook and line ( Gamakatsu C12U size 14, 10 , and 8 ) using worms and raw fish ( Astyana x sp.) caught on site. Hook and line effort consisted of thorough coverage of all pools and glides in the reach (1.75 - 2.5 hours depending on river size). Using a cast net, one person followed behind 23 the three hook and line anglers, attempting to cast every surface area of the stream. Due to limited sampling techniques and time, some species present at each reach could have been not captured, and inferences about the fish and shrimp communit ies are made with caution ( Hetrick & Bromaghin, 2006) . F o r all fish caught, species was recorded, s tandard length was measured in cm, and photos taken for those that could not easily be identified in the field. Fish and shrimp were kept with a bubbler in a bucket when caught and released af ter measurement at the end of sampling. A few specimens were kept in 95% ethanol for identification. These are cataloged at the Zoological Museum at the University of Costa Rica. JT Betts identified fish species in the field, and A Angulo - Sibaja, J San Gil , CA Garita - Alvarado, and N Gonzalez - Aleman helped with photo and specimen ID, with reference to Bussing (1998). These efforts are the first ever recorded for these streams and can serve as an initial species list for more in - depth future investigation. F i sh sampling methodolog y was approved via the Institutional Animal Care and Use Committee (IACUC) office at Michigan State University, AUF# 12/17 - 220 - 00 . Spatial data p rocessing All spatial data processing was done in ArcGIS 10.5.1 . The base transect point was used along with the NASA S huttle Radar Topography Mission (SRTM GL3) 90 - meter global digital elevation model ( van Zyl , 2001 ; Rodriguez , Morris, & Belz, 2006 ) to calculate the catchment area above the base transect of each reach ( Brenden et al., 2006 ; Leal et al. , 2016 ) . The reach track was used to calculate 100 - meter riparian buffers around each study reach using the Buffer tool ( ArcGIS 10.5.1 ) . These shapefiles were used as areas for forest cover analyses (see below). To calculate catchments draining to the base transect at each reach, the Hydrology toolset in Spatial Analyst in ArcToolbox was used ( ArcGIS 10.5.1 ) . The DEM was used to create fill , 24 flow direction , and flow accumulation raster layers, and Snap Pour Point was used to find cells of high ac cumulated flow nearest to each base transect. These pour points were used to create catchments for the whole stream network upstream of the base transect. Catchments were checked against topographic maps for accuracy ( Dirección Gen eral de Cartografía de Ni caragua, 1988 ). Catchment - s cale f orest c over p arameters Percent forest cover and forest loss for each catchment (land area contributing to each sample reach) and the 100 m buffer around each reach was calculated, using the Global Forest Change dataset ( Han sen et al. , 2013) (Figure 4) . This is a well - known raster dataset based on NASA Landsat satellite imagery (30 m pixel resolution). R aster files loss event ( lossyear and 0 ( treecover2000 were do wnloaded using the 20N, 90W extent, which includes Nicaragua. The lossyear raster was divided into 18 separate files representing forest loss by year (2001 - 201 8). Using the treecover2000 raster , which has values representing % can opy closure for each pixel on a scale of 0 (full closure) to 100 (no closure), a binary forest cover 2000 layer was created by extracting and combining all 3 0 3 0 to represent forest, with the idea that if a pixel is determined as more than 3 0% de forest ed , it is considered deforested. Both above functions were done using the Extract by Attributes f unction from Spatial Analyst Tools in ArcToolbox ( ArcGIS 10.5.1 ) . To calculate forest cover in the year of sampling (2018), the sum of lossyear pixel values (2001 - 2018) was subtracted from the number of forested pixels in 2000. Percent forest cover in each catchment and reach buffer in the year of sampling, and each yea r prior until 2000 was calculated . This was done by u sing the catchme nt and buffer shapefiles to clip the forest cover raster files, using the clip function in raster processing in ArcToolbox 25 Figure 4 : Forest cover in the catchment and buffer over time. A) Percent forest in catchment above (draining to) each study reach . Catchment data generated in ArcMap from ASTER DEM ( 90 m ) and forest cover data from Hansen et al . ( 2013). Corresponds directly to areas visualized in Figure 2. Loss year data grouped in 4 - 5 - year intervals for visualization. Hurricane damaged forest treated as forest land cover in all analyses. Loss Pre - 200 0 represents pixels without forest in 2000. See methodo logy for more detailed description. Note that Kukra River watersheds tend to be deforested much earlier than Pijibaye River watersheds. 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Mountain Cow Guinea Banana Vieja She Tiger Long Falls Boca Tapadas Moga La Combinación El Coco La Perra El Salto Papa Abrahán El Limón Chacalín Limonero Indian River Corn River Pijibaye River Kukra River Forest Cover Loss 2014-2018 Loss 2010-2013 Loss 2006-2009 Loss 2001-2005 Loss Pre-2000 Hurricane Damage (2017) 26 B) Percent forest in 100 - meter buffer area ar ound each study reach. Buffer data generated from study . Note that deforestation is higher in the buffer than the watershed for most sites. 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Mountain Cow Guinea Banana Vieja She Tiger Long Falls Boca Tapadas Moga La Combinación El Coco La Perra El Salto Papa Abrahán El Limón Chacalín Limonero Indian River Corn River Pijibaye River Kukra River Forest Cover Loss 2014-2018 Loss 2010-2013 Loss 2006-2009 Loss 2001-2005 Loss Pre-2000 Hurricane Damage (2017) 27 ( ArcGIS 10.5.1 ). These data are visualize d by 4 - 5 year intervals of forest loss for each catchment and buffer in Figure 4 and mapped with the same intervals and color code in Figure 2. Deforestation history index A time - weighted index of deforestation history was developed based on the forest co ver data described above from Hansen et al. ( 2013 ). To calculate the index, the area of interest (catchment or buffer) was used to clip the lossyear and treecover2000 rasters, and data from the attribute table extracted into excel. For each year in the los syear raster (2001 - 2018) for the area of interest , the number of pixels classified as deforested in a particular lossyear was multiplied by the number of years before present, for example 100 pixels lost in 2006 (13 years before present) would be 100 * 13 = 1300. The calculation was repeated for each yea r 2001 - 2018 and these values were summed for all years. To include forest loss before 2001, number of pixels 3 0 (at least 30% deforested) from the treecover2000 raster for the area of interest w as summed, and this value multiplied by 20 (~20 y ears before present) and added to the sum of multiplied values for lossyear 2001 - 2018. This total was divided by the number of pixels in the area of interest (catchment of buffer) in order to standardize com parisons. This is represented by the equation bel ow: Deforestation History Index = lossyear2001 * (2019 - 2 001), lossyear2002 * (2019 - lossyear2018 * (2019 - 2018)]+[ treecover2000 ( ) * 20]} / [# of pixels in area of interest] This process created an index typically on a scale of 0 - 10 that portrays a time - weighted deforest ation value for use in analysis. Its application is for situations in which the impacts of deforestation are accumulative over time , and recent deforestation is not the sam e in impact as deforestation years ago. In other words, where current percent defore station in a study area does 28 not fully capture its impact on a study system. This has been shown to be true for stream habitat and biota in relation to land - use history ( Ha rding et al., 1998). Habitat and landscape v ariable s election Habitat v ariables were consolidated into 64 summary variables at the stream reach level. They were organized into categories of stream size, hydrology and substrate, bank disturbance, water quality, wood and debris, in - channel algae and plants, and riparian forest condition. A on matrix was calculated and variables within each category with a correlation coefficient (R) of 0.6 or larger were thinned to one per category prioritizing variables based on best judgment of ecological importance of the variable and i ts suitability as a representative summary measure of the category (for example, geometric mean substrate diameter was chosen over %boulder or %cobble, even though all were correlated) ( Ferreira et al., 2014) . Ordination techniques like PCA or PCoA were no t appropriate for v ariable selection due to relatively low ratio of number of samples to number of habitat variables ( McGarigal, Cushman, & Stafford, 2013) . A few variables were kept because of their ecological importance and distinctness, even though they significantly corr elated with another variable within the same category. These include small woody and leafy debris % cover and large wood volume per 100 m , which can respond differently to disturbance and interact with each other in the stream ( Bilby & L ikens , 1980) ; and m id - c anopy p lant % c over and the riparian condition index , which show distinct aspects of riparian habitat quality (Kaufman et al., 1999). Twenty variables were selected (see Table 1). Four landscape predictors were selected for use in an alyses. They i ncluded the deforestation history index and percent forest cover for both the catchment draining to each base transect and the 100 m buffer around the study reach. 29 Habitat n onparametric c omparisons Mann - Whitney U and Kruskal - Wallis tests were used to assess differences of all twenty habitat metrics between forested and deforested stream reaches, and by watershed, Pairwise comparisons were carried out using Mann - Whitney U tests. These nonparametric al ternatives to T - tests and ANOVAs were select ed because assumptions of normality and equal variance were not met for many comparisons, according to Shapiro - Wilk and Exact p - values were used unless there were tied values where a normal approx imation was used . Tests were done using func tions wilcox.test and kruskal.test from the package stats for R version 5.3.1 (R Core Team , 2016). Analysis of m acroinvertebrate , shrimp, and f ish community response metrics Macroinvertebrate taxa lists were organized in systematic order, as in Domínguez & Fernández (2009). Abundances were recorded for each distinct taxon based on the sum of all 11 samples for each reach. Taxa richness, and Shannon - Weiner diversity (H) and eve nness (EH) indices (Jost, 2006) were calculated according using the number of dist inct taxa in the reach, typically at genus level, but sometimes at family (Hydroscaphidae, for example) or higher level (Oligochaeta, for example). If a specimen was found bu t not identified past a coarser taxonomic level, while other taxa in the same coar se taxonomic level were identified to a finer taxonomic level, the coarse taxonomic level individual was not included as a unique taxa unless it was clearly not the same taxa (for example, trichopteran pupae that could not be identified past the order leve l were never included as unique taxa because other trichopterans were identified to family or genus level at the same reach). The BMWP index for the Costa Rican Caribbean ( Sp ringer et al. , 2010; Salvatierra, 2014 ) was calculated to generate values represen 30 still in the process of verification and has not yet been formally adapted to Nicaragua (Pers . comm . , M Springer & T Salvatierra). Density was calculated as the total number of individual invertebrates in a ll Surber samples, divided by the total area of the Surber samples (11 square feet= 1.02193 square meters). Long Falls, a reach in Indian River, had o nly 10 transects, since one invertebrate sample was lost in transport, thus density calculations were adju sted accordingly. 2006): } chness: Shannon Evenness (EH) } / S Where p is the proportion of the total number of individuals comprised by taxa i and s is the taxa richness, or total number of unique taxa identified for a given reach. Fish taxa lis ts were organized taxonomically, as in Bussing (1998). Fish and Shrimp metrics were based on consistent effort of fishing the whole reach with cast net, hook and line, overnight traps, and Surber sampler (some shrimp) . Cast - netting at one reach in Kukra Ri ver (Papa Abrahán Creek) was not performed due to one team member not being present with the net. This may have biased fish abundances and richness estimates to be lower for that reach, although they were in the range of other sites. Fish in the family Cic hlidae (hereafter overall fish abundance for a variety of reasons. Sampling effort for cichlids was more comprehensive than for most other taxa, because hook and line and cast net surveys during the day were quit e effective for al l five species of cichlids commonly captured. Most other common taxa, such as Rhamdia , Eleotris , and Awaous , were more elusive. Small c haracidae/ Astyanax was 31 always present, but abundances were not recorded due to the sheer number capture d in most reaches. These species of Cichlids are also important for the local fishery, thus important to this study. Cichlids also represent a range of niches, so they are vulnerable to changes in habitat ( Rican et al., 2016 ). Fish and shrimp metrics were not standa rdized by stream size because these comparisons were part of subsequent regression analyses. T - tests and ANOVAs were carried out to assess differences in means of all seven invertebrate, fish, and shrimp metrics between forested and deforested st ream reach es, and by watershed . For all comparisons, assumptions of normality and equal variance were tested using Shapiro - Wilk and Invertebrate density and Cichlid abundance were ln() transformed in order to meet assumptions. All models we re fitted using functions t.test, anova, and tukey.test from the package stats for R version 5.3.1 (R Core Team , 2016). Plots were made with package ggplot2 for R version 3.1.1 ( Wickham, 2016) . Invertebrate d iversity (H) was not included in reporting becau se it is weighted by evenness and richness, and evenness and taxa richness showed clear opposite trends in our data, thus diversity was non - significant and interpretation confusing for nearly all comparisons. ANOVAs ( with subsamples nested by reach ID ) wer e used to compare standard length of fish captured in forested and deforested reaches and by watershed. These were run for each species that had at least 5 individuals captured in both forested and deforested reaches. Multivariate analysis of the m acroinve rtebrate c ommunity Non - metric multidimensional scaling ( Faith, Minchin, & Belbin, 1987) of the macroinvertebrate community was carried out using the metaMDS function and plots were made with function ordiplot from the package vegan for R version 2.5. 4 (Oks anen et al., 2013). For NMDS, PERMANOVA, SIMPER, and Indicator Analysis, rare 32 whole study were excluded from analysis (Lorion & Kennedy, 2009a) . Taxa were maintained at the lowest taxonomic level possible, and unknown genera were excluded unless they could only be identified to the same coarser taxono mic level at all sites (for example, Oligochaeta). Densities were W isconsin - standardized and square - root transformed according to the default algorithm in metaMDS . The community matrix was generated from invertebrate densities at each reach by taxa, using Bray - Curtis dissimilarity distances. I terations using 2, 3, and 4 - ax e s were attempted, and the 3 - axis solution selected becaus e it yielded a sufficiently low stress 3 - axis solution (stress <0.1, R 2 <0.9) (Figure 8 and 9 ) . PERMANOVA ( Anderson, 2014) and SIM PER ( Warton, Wright, & Wang, 2012) analyses were carried out in P AST Statistical software Version 3.20 ( Hammer, Harper & Ryan, 2001). These tests used the Bray - Curtis dissimilarity distances of square root transformed densities. To test the significance of differences in community composition between stream reach groups, P ERMANOVA tests were run with reaches grouped as forested and deforested and grouped by watershed. SIMPER was used to calculate the percent contribution of each taxa to the differences in c ommunity composition between these groups . Indicator analysis was carried out using the procedure ( Dufrêne & Legendre, 1997; De Cáceres & Legendre, 2009) on square root transformed invertebrate densities, to determine which taxa w ere significantly associated with forested and d eforested reaches, as well as with each watershed . Analysis was carried out with the multipatt function with IndVal.g in package indicspecies R version 1.7.6 ( De Cáceres & Legendre, 2009 ). Linear r egression a nalyses Individual linear regressions were run with each of the twenty habitat variables and four landscape variables as predictors for the seven selected taxa response variables . Linear 33 regressions were also run with the Deforestation History Index for th e catchment as a predictor with each of the twenty habitat variables as a response . Models were fit using the Regression : Linear and Correlate : Bivariate functions in SPSS Statistics version 26.0 (IBM SPSS , 2019). For each individual model , care was taken to see whether residuals met assumptions of normality, linearity, and homogeneity of variance. SPSS Linear Regression function was used to generate plots of standardized predicted versus standardized residuals to visually test for linearity and homogeneity of variance, and SPSS Descriptive Statistics: Explore function was used to conduct Shapiro - Wilk tests for normality of residuals (IBM SPSS , 2019). Combinations that did not meet these assumptions were not included. In one case (Periphytic M acroalgae), it was apparent that the relationship was more logarithmic than linear, so a log 10 regression was calculated in addition to the linear regression . 34 RESULTS Differences in h abitat condition Considering the 20 habitat variables retained for analysis, variables that represent bank disturbance, water quality, instream habitat, and riparian condition metrics tended to differ significantly between forested and deforested reaches, while variables f or stream size, hydrology, and substrate metrics did not differ, acco rding to non - parametric Mann - Whitney U and Kruskal - Wallis testing (Table 1 and Appendix B, Figure B. 2 ). Notably , conductivity was significantly higher at deforested sites (Diff.=22, p=0.0 08), while multiple instream habitat and riparian condition metrics w ere significant lower, including macrophytes % cover (Diff.= - 1.36, p=0.0361), instream live trees and roots % cover (Diff.= - 1.360, p=0.0229), large wood volume (Diff.= - 76.165, p=0.009) , mid - canopy plant % cover (Diff.= - 48.29, p=0.000), and the riparian condition index (Diff.= - 7.514, p=0.000). Other riparian and instream metrics followed a similar pattern. For example, p roportion of stream bank eroded was higher, and small woody and le afy debris % cover was lower at deforested reaches (p=0.0726 and 0.05 58, respectively). M ore recently deforested streams had periphytic macroalgae concentrations up to 27 % cover, whereas forested streams ranged from 5 to 15 % cover, hurricane impacted streams ranged from 4 - 20 % cover ( though one naturally erosional foreste d reach had <1 %), and longer deforested streams only ranged from 0 to 4 % cover (though these patterns were not significant). Turbidity, temperature, and pH appeared highly dependent on recent weather conditions and were excluded from analysis. Large wood volume and small woody and leafy debris were highest at reaches in the Indian watershed, where there was the gre atest impact from Hurricane Otto in 2017 (Figure 1 and 3). Riparian impact from the hurricane was also evidenced through upper canopy large tre e 35 % cover (above 15M, >0.3M diameter) and % shade (densiometer) metrics, as both were significantly lower at reac hes in Indian watershed than reaches in Corn watershed (P=0.036 for both metrics). All reaches in both watersheds were >99% forested, but Corn watershed was not as impacted by the hurricane. One deforested site in the Pijibaye watershed (La Perra Creek) ha d abnormally high large wood and leafy/woody debris levels because it was being actively deforested and cut riparian trees were left in the riv er. 36 Table 1: Nonparametric test results by habitat variable for comparison of forested (N=8) and deforested (N=7) reaches (Mann - Whitney U) and by watershed (Kruskal - Wallis) Forested (Indian, N=5 and Corn, N=3), Recently Deforested (Pijibaye, N=3), and Long er Deforested (Kukra, N=4). A positive difference implies a higher value at deforested reaches. Exact p - values were used unless signified by red text, which implies tied values where a normal approximation was used . Pairwise comparisons (Mann - Whitney U) ar e represented by the first letter of the watershed name. Le tters A and B represent significance groupings between Indian (I), Corn (C), Pijibaye (P), and Kukra (K) River watersheds, respectively. P - or lettering. Significant p - ee methods for calculations of variables (Most values based on a reach level mean of 11 transects and associated subsamples). Mann - Whitney U (Forested vs. Deforested) Kruskal - Wallis (By Watershed) Habitat Variable Difference p - Value p - Value Pairwise Sign ificance (Mann - Whitney U) Stream Size Stream Size (Reach Volume) - 235.095 0.4634 0.8757 A Hydrology and Substrate % Pool 0.02 6 0.8168 0.49 7 A % Fines 0 1 0.37 9 A % Sand - 0.010 0.9536 0.684 A Geometric Mean Substrate Size - 9.228 0.7789 0.9 86 A Embeddedness in Riffles and Rapids 5 0.3519 0.202 A (IC=0.072) Standard Deviation Embeddedness - 0.335 0.9551 0.13 2 A (CK=0.05714) Bank Disturbance Proportion of Stream Bank Eroded 0.177 0.0726 0.080 A (IC=0.072) Water Quality Conduc tivity (µS) 22 0.0321 0.008 A,B,B,B (IC= 0.036 ,IP= 0.036 ,IK= 0.016 ,CP=0.077) Instream Habitat Periphytic Macroalgae % Cover - 3.190 0.1642 0.16 6 A (IK=0.085,CK=0.057) Macrophytes % Cover - 1.36 0.0361 0.023 3 A,AB,AB,B (IC=0.099,IK= 0.015 ) Instream Liv e Trees and Roots % Cover - 1.360 0.0229 0.108 AB,A,AB,B (CK= 0.050 ) Overhanging Vegetation % Cover - 2.471 0.1015 0.119 A,AB,AB,B (IK= 0.027 ) Small Woody and Leafy Debris % Cover - 8.72 6 0.0558 0.078 A,AB,AB,B (IK= 0.032 ) 37 Large Wood Volume per 100 m - 76.16 5 0.009 0.05 1 A,AB,AB,B (IK= 0.032 , CK=0.057) Habitat Complexity - 11.365 0.281 0.697 A Riparian Condition Upper Canopy Large Trees % Cover 0.855 1 0.0378 A,B,AB,AB (IC= 0.036 ,CP=0.10) Mid - Canopy Plant % Cover - 48.29 0.0 00 0.0197 A,AB,B,B (IP= 0.036 ,IK= 0.032 ,CK=0.05714,CP=0.1) % Shade (Densiometer) - 0.024 0.8617 0.0587 A,B,AB,AB (IC= 0.036 ,CP=0.077) Riparian Condition Index - 7.514 0.000 0.009 A,AB,B,B (IP= 0.036 ,IK= 0.016 ,CP=0.1,CK=0.057) 38 Deforestation history a s a predictor of stream habitat The deforestation history index for the catchment draining to each reach was a significant predictor of bank disturbance, instream habitat, and riparian condition metrics. Proportion of stream bank eroded increased as time - weighted % of catchment deforeste d increased ( R²= 0.392, p =0.012) (Table 2, Figure 10) . As the index value increased (more deforestation for a longer time), instream s mall woody and leafy debris % cover decreased ( R² =0.436, p =0.007), large wood volume decr eased (R² =0.425, p =0.008), and m acrophytes % cover decreased ( R² = 0.416, p =0.009) . O verhanging vegetation % cover, i nstream l ive t rees and r oots % c over, and p eriphytic m acroalgae % c over also decreased ( R² 0.238, p =0.065 ; R² 0.222, p =0.076 ; & R² = 0.210, p =0.086, respectively). Table 2: Single regression comparisons of the deforestation history index at the watershed scale (X) as predictors of habitat responses (Y). 2 , and p are listed. Bolded are listed in order of significance. Although regressions were run on all 20 habitat predictors, only habitat predictors with a p - included. Italicized items did not meet assumptions of normality of residuals (Shapiro - Wilk p<0.05). Deforest ation History Index ( Catchment ) Predictor Habitat Response Transformation R R² p Mid - Canopy Riparian Plant % Cover ArcSIN - 0.678 0.460 0.005 Small Woody and Leafy Debris % Cover ArcSIN - 0.660 0.436 0.007 Large Wood Volume per 100 m ln - 0.652 0.425 0.00 8 Macrophytes % Cover ArcSIN - 0.645 0.416 0.009 Proportion of Stream Bank Eroded ArcSIN 0.626 0.392 0.012 Riparian Condition Index ln - 0.591 0.349 0.020 Overhanging Vegetation % Cover ArcSIN - 0.488 0.238 0.065 Instream Live Trees and Roots % Cover Arc SIN - 0.471 0.222 0.076 Periphytic Macroalgae % Cover ArcSIN - 0.458 0.210 0.086 Macroinvertebrate s, f ish , and shrimp summary Among all sites, 107 distinct aquatic insect taxa and 15 other distinct invertebrate taxa were captured and identified. Of the in sect taxa, 92 were identified to the genus and 15 to the 39 subfamily or family. Other invertebrates varied more in taxonomic resolution (see Appendix A, Table A.1 for a full list with reach coordinates ). Site level taxa richness ran ged from 37 at El Limón in the Kukra watershed, to 71 at El Coco in the Pijibaye watershed. Three species of Macrobrachium shrimp and at least two species of Atya shrimp were captured, as well as at least one species of freshwater crab ( Pseudothelphusidae ). Twenty distinct fish tax a were captured in our study reaches, and 11 more in other surveys. All but Characidae/ Astyanax spp. and Rhamdia spp. were identified to species ( s ee Appendix A , Table A.2 for a full list with reach coordinates ) . All insect, other invertebrate, and crustac ean specimens are preserved in the University of Costa Rica Zoological Museum in the School of Biology (Contact: M Springer), and photos of fish are with author JT Betts. Differences in m acroinvertebrate , shrimp , and fish community response metrics In comp aring mean taxa richness, densit y macroinvertebrate community, only evenness differed significantly between forested and deforested stream reaches, being lower at forested reaches ( df =10.32, T=2.25, p=0.047). But there were significant differences for each metric when compared by watershed ( Table 3 , Figure tended to have low er taxa richness, density, and BMWP score, and a h igher evenness than forested sites (Indian watershed, N=5 and Corn watershed, N=3) and more recently deforested sites (Pijibaye watershed, N=3) ( Table 3 , Figure 5). Reaches in Corn watershed (forested) had by far the lowest evenness, as there were particul arly high densities (dominance) of two subfamilies of Chironomidae, and relatively low abundances of rarer taxa. Taxa richness, BMWP score, and density ( LN ) were negatively correlated with evenness (R 2 =0.17 4, p=0.122; R 2 =0.347, p=0.021; 40 and R 2 =0.281, p=0.0 42, respectively) ( see Appendix B, Table B.1 for a list of correlations and p values between taxa response variables ). Considering mean fish taxa richness, Cichlid fish abundance , and shrimp abundance, only shrimp abundance differed significantly between f orested and deforested sites, being higher at forested sites ( df =12.00, T= - 2.64, p=0.022). Cichlid abundance differed significantly whe n compared by watershed in separate ANOVA tests by metric ( df =3, F=5.22, p=0.017). Tukey pairwise comparisons showed that longer deforested sites (Kukra, N=4) tended to have lower fish taxa richness, cichlid abundance, and shrimp abundance than forested (I ndian, N=5 and Corn, N=3) and more recently deforested sites (Pijibaye, N=3) ( s ee Table 3 and Figure 6), although these c omparisons were only significant for cichlid abundance between Indian and Kukra (p=0.046) and Pijibaye and Kukra (p=0.020) watersheds. 41 Table 3 : T - test and ANOVA and Tukey post - hoc pairwise comparisons for macroinvertebrate community summary statistics comparing means of deforested (N=7) and fore sted reaches (N=8) , and means for all four watersheds f orested (Indian, N=5 and Corn, N=3), r ecentl y d eforested (Pijibaye, N=3), and l onger d ef orested (Kukra, N=4). Positive T - value implies higher value at deforested reaches. Significant p - Macroinvertebrate community statistics based on sum of eleven Surber samples (0.092903 M 2 ). Tukey post - hoc tests were run on pairwise comparisons of summary statistics for each watershed pair. L etters represent significance groupings between Indian, Corn, Pijibaye, and Kukra River watersheds, respectively. P - listed but d name. T - Tests Forested vs. Deforested ANOVA b y Watershed df T p df F p Pairwise Significance (Tukey) Macroinvertebrate Community Invert. Taxa Richness 10.03 - 1.37 0.20 3 4.27 0.031 AB,AB,A,B (IK=0.065,CK=0.099,PK=0.043) Invert. Density ( ln ) 7.82 - 1 .17 0.28 3 6.03 0.011 AB,A,A,B (IC=0.086,CK=0.024,PK=0.016) BMWP Score 12.32 - 1.89 0.083 3 4.17 0.034 AB,A,AB,B (CK=0.023) Invert. Evenness (EH) 10.32 2.25 0.047 3 7.10 0.006 A,B,AB,A (IC=0.037,CP=0.076,CK=0.0038) Fish and Shrimp Community Fish Taxa Ri chness 12.15 - 1.03 0.321 3 1.09 0.39 A Cichlid Abundance ( ln ) 9.06 - 1.24 0.248 3 5.22 0.017 A,AB,A,B (IK=0.046,CK=0.080,PK=0.020) Shrimp Abundance 12.00 - 2.64 0.022 3 3.06 0.074 A (IK=0.097,CK=0.098) 42 Figure 5: Macroinvertebrate community summary statistics for two forested watersheds and a recently and less recently deforested watershed. A) Macroinvertebrate Taxa Richness (mostly genus level). B) Macroinvertebrates per m 2 . Statistics run on ln of d ensity, to meet normality assumptions. Raw density is displayed. C) Biological Monitoring Working Party water quality score adapted for Costa Rica. Represents taxa richness, weighted by family sensitivity to pollution. D) Shannon Reach values repres ented by points. Community statistics based on sum of eleven Surber samples (0.092903 m 2 ) subsamples. T - tests were carried out lumping forested and deforested reaches . ANOVA and Tukey Pairwise tests were run between each watershed letters represent signifi letters implies no significance. A B AB A B B A B A A B B A B A AB B B A B AB A B A) B) D ) C ) 43 Figure 6: Fish and s hrimp community summary statistics for two forested watersheds and a recently and less recently deforested watershed. A) Number of fish ta xa caught in each reach . B) Total number of fishes in the family Cichlidae caught in each reach. Statistics run on ln of abundance, but r aw abundance is displayed. C) Total number of shrimp (genera Atya an d Macrobrachium ) caught in each reach. When lumped by forested and deforested, forested sites ha d significantly more shrimp than deforested sites (T= - 2.64, p=0.022). Reach valu es represented by points. Community statistics based on consistent effort of fishing the whole reach with cast net, hook and line, overnight traps, and Surber sampler. T - tests were carried out comparing forested to deforested reaches . ANOVA and Tukey Pairw ise tests were run between each watershed letters represent significance groupings o implies no significance. A) B) C ) A AB A B B 44 Differences in f ish l ength The four common fish species that are important for local subsistence tended to be significantly larger at forested sites than deforested sites. These species are rela tively larger and can easily be caught by hook or spear. ANOVAs with subsamples by reach showed significantly higher standard length at forested stream reaches than deforested reaches for Brycon guatemalensis ( df =1, F=2.096, p =0.025), Cribroheros alfari ( d f =1, F=5.923, p =0.016), Parachromis dovii ( df =1, F=63.029, p =0.000), and Tomocichla tuba ( df =1, F=19.364, p =0.000) (Table 4 , Figure 7). Other locally important subsistence species such as Gobiomorus dormitor and Rhamdia sp. were not present in high enough abundances at any reach to detect trends in size. These trends were not present for other species, most of which were only caught on ver y small hooks (Size 14) or by cast net in our surveys (these fishing techniques are not utilized in most communities). R oeboides bouchelli and Amatitlania nigrofasciata were notably more common at deforested reaches, while Amatitlania septemfasciata was ab undant at forested reaches and absent in deforested reaches. Uncommon species Bramocharax bransfordii , Neetrop l us nemato pus , Eleotris p isonis , Sicydium altum , Phallichthys amates , and Priapichthys annectens were only found in forested reaches. No species were completely unique to deforested reaches (Table 4 , Figure 7) . 45 Table 4 : ANOVA results for fish standard l engths by species comparing two forested, a recently deforested, and longer deforested watershed . Community statistics based on consistent effort of fishing the whole reach with cast net, hook and line, and overnight traps. ANOVA on fish length for each sp ecies (with individual fish lengths as subsamples for each reach). Comparisons between forested and deforested reaches and by watershed. Tukey Pairwise tests were run between each watershed are visualized by letters in Fi gure 7. Only fish that had at least 5 individuals in both forested and deforested sites were included in statistical analysis. Mean standard length and sum of captured individuals from all reaches within each watershed is included for all species. Astyanax spp. w ere highly abun dant in all b ut one reach, so counting and measuring was limited to 20 individuals per reach, thus it is not included here. I=Indian, C=Corn, P=Pijibaye, and K=Kukra. n.d. signifies that the fish were not measured, although individuals were recorded in the watershed. P Family Taxa Mean Length # Captured ANOVA (Length) Forested vs. Deforested ANOVA ( Length) b y Watershed I C P K I C P K df F p df F p Characidae Astyanax spp. NA (Multiple Species) NA - - - - - - Bramocharax bransfordii - 6.2 - - 0 1 0 0 - - - - - - Brycon guatemalensis 14.4 10.4 9.9 8.5 43 2 10 1 1 2.096 0.025 3 2.096 0.113 Roeboides bouchelli 6.7 - 6.5 6.2 4 0 41 12 - - - - - - Heptapteridae Rhamdia sp. 20.6 16.6 - 12.2 2 2 0 10 - - - - - - Poeciliidae Alfaro cultratus n.d. 5.6 4.1 - P 1 1 0 - - - - - - Phallichthys amates n.d. - - - P 0 0 0 - - - - - - Poecilia gillii 5.8 6.5 5.9 7.2 P 29 54 6 1 2.169 0.145 3 2.361 0.077 Priapichthys annectens 4.4 - - - P 0 0 0 - - - - - - Mugi lidae Agonostomus monticola 7.2 8.7 8.3 8.5 25 6 21 10 1 3.389 0.072 3 2.028 0.122 Cichlidae Amatitlania nigrofasciata 5.0 5.9 5.5 5.0 8 3 66 18 1 0.249 0.619 3 2.031 0.116 Amatitlania septemfasciata 5.4 7.4 - - 40 32 0 0 - - - - - - Cribroheros alfar i 6.1 9.7 6.4 7.3 86 44 53 4 1 5.923 0.016 3 31.874 < 0.00 1 Neetrop l us nematopus 6.3 - - - 11 0 0 0 - - - - - - Parachromis dovii 14.6 10.9 6.7 8.6 31 18 25 6 1 63.029 < 0.001 3 25.930 < 0.001 Tomocichla tuba 11.2 - 7.2 7.3 17 0 7 17 1 19.364 < 0.001 3 9 .687 < 0.001 Gobiidae Awaous banana 8.5 n.d. 10.6 18.5 1 2 2 1 - - - - - - Sicydium altum 9.3 - - - 2 0 0 0 - - - - - - 46 Table 4 Eleotridae Eleotris pisonis - 12.7 - - 0 1 0 0 - - - - - - Gobiomorus dormitor 16 .0 14.2 17.0 11.2 3 6 5 2 1 0.098 0.765 3 1.444 0.320 47 Figure 7: Fish standard lengths by species for two forested, and a recently deforested and less recently deforested watershed. The four species that are important to the local fishe ry and relatively high abundance in our study are highlighted: A) Brycon guatemalensis B) Cribroheros alfari C) Parachromis dovii D) Tomocichla tuba . For all four species, deforested reaches tend toward s lower average lengths. When Brycon guatemalensis was lumped by forested and deforested, forested sites were significantly higher than deforested sites (F=2.096, p=0.025). Individual fish lengths represented by points. Community statistics based on consis tent effort of fishing the whole reach with cast net, hook and line, and overnight traps (see methods). ANOVAs (with reaches nested) were carried out, categorized by forested and deforested and by watershed, with Tu key pairwise tests run between each watershed letters represent significance groupings of water 4 ) . A AB A B B A B B B B A B C BC C ) A ) B ) D ) 48 Changes in m acroinvertebrate c ommunity structure Non - metric multidimensional scaling of the macroinvertebrate community matrix (Wisc onsin standardized and square root transformed densities) using a Bray - Curtis dissimilarity dis tance yielded a low stress 3 axis solution (stress <0.0945, R 2 =0.929) (Figure 8). Ordinations of axis 1 and 2, 1 and 3, and 2 and 3 all showed a similar pattern of sites displayed, so axes 1 and 2 were visualized (Figure 9) (see Appendix B, Figure B. 1 for other ax e s displays ). PERMANOVA analysis of the community matr ix showed that stream reaches cluster significantly by forested (Indian, N=5 and Corn, N=3) and deforested ( Pijibaye, N=3 and Kukra, N=4 ) groupings (F= 1.88, p=0.0317). R eaches clustered with muc h higher significance when analyzed by watershed (F=2.445, p=0 .0001) , which suggests that deforestation history is an important contributor to structuring the macroinvertebrate community. Forested reaches cluster together on both axes, with the exception that Guinea Creek (GU), a relatively erosional site in Indian watershed, is an outlier on axis 1. Recently deforested reaches (Pijibaye) cluster with forested reaches (Indian and Corn) on axis 1, but with longer deforested reaches (Kukra) on axis 2. When t he deforestation index was fit as a gradient on the NMDS plot, it appears that both axis 1 and 2 capture the differences in the macroinvertebrate community that could be attributed to impa cts from deforestation history (See Figure 9.C.) The placement of ta xa on these plots aligns well with indicator analysis results described below and could be useful to help determine associations of taxa with streams degraded by deforestation. 49 Figure 8 : Stress plot for non - metric multidimensional scaling analysis of macroinvertebrate community matrix (taxa densities by reach). 3 axis solution used; stress level is 0.0945. With stress <0.1 and linear fit R 2 >0.9, the model is an excellent fit. Figure 9: Non - metric multidimensional scaling ordination plots of ma croinvertebrate community matrix (taxa densities by reach). A) Ordination plot with reaches visualized. Axis 1 and 2 represented (plots with axis three group similarly). Polygons show watershed groupings. According to PERMANOVA, reaches group significantly as forested (Indian, N=5 and Corn, N=3) and deforested (Pijibaye, N=3 and Kukra, N=4) (F=1.88, p=0.0317). Reaches group with higher significance by watershed (F=2.445, p =0.0001 ) . A ) 50 B) Ordinat ion plot with taxa visualized. C) Ordination plots A and B with the Deforestation History Index gradient visualized. Plot C could be used to infer taxa that are sensitive to the impacts of deforestation over time, aligning results to indicator analysis or SIMPER. 51 Taxa - s pecific responses From the indicator analysis, four m ayfly genera (Order Ephemeroptera) in two families were significant eight distinct 5 ). When broken down by watershed, only Moribaetis remained as a significant indicator of longer deforested co nditions (Kukra), while recently deforested (Pijibaye), and each forested (Indian and Corn) watershed had a diverse assemblage of indicator taxa. The lack of indicators for Kukra can be attributed to depressed abundances of all taxa in these stream reaches . The NMDS plot (Figure 9.B.) displays strong trends . Taxa that indicate deforested reaches tend to be located positive on axis 1 and negative on axis 2 (towards the bottom right), and taxa that indicate forested reaches tend to be located negative on axi s 1 and positive on axis 2 (towards the top left). SIMPER showed some similar results to indicator analysis but was more heavily weighted by density. Ther efore SIMPER was less clear to interpret than indicator analysis. For example, Chironominae had the hi ghest percent contribution to the differences between watersheds (6.425 %) and does show up as a significant indicator for Corn watershed. T he next three highest contributions according to SIMPER Microcylloepus , Orthocladiinae , and Smicridea account for 4. 239 %, 4.125 %, and 3.116 % of the difference between watersheds, but do not show up as significant indicators of any watershed in the indicator ana lysis . 52 Table 5 : Indicator analysis and SIMPER results for key taxa (mostly genus level), showing taxa th at were identified by Indicator Analysis as indicators (p<0.1) for forested and deforested reaches, as well as by watershed. Significant p - these taxa, SIMPER percent contribution to the Bray - Curtis dissimilarity between catego ri es is also recorded. Percent contributions are heavily weighted by dominant taxa. All analyses were based on the square - root of reach level densities of m acroinvertebrates, based on sum of eleven Surber samples (0.092903 m 2 ). Mean densities (individuals pe r m 2 ) are also recorded for each category. Low evel assignments from Ramírez and Gutiérrez - Fonseca (2014). Indication Category Tax a BMWP Score Functional Feeding Group Indicator Value Indicator Sig . ( p ) SIMPER % Contrib. Mean Invertebrate Density (per m 2 ) For. Def. Forested Cryphocricos 4 Pr 0.846 0.017 1.678 15.9 4.47 - - Oligochaeta 1 NA 0.844 0.019 3.410 92.2 17 .6 - - Palaemnema 7 Pr 0.780 0.023 1.547 30.6 13.6 - - Leucotrichini Gen. undet. 6 Generally Pc - Hb, Sc, CG 0.791 0.029 1.289 12.6 0 - - Lutrochus 7 Sh - Dt, Hb 0.738 0.053 0.576 1.5 0.14 - - Nematoda 1 NA 0.768 0.073 1.981 33.5 0.28 - - Helico psyche 5 Sc 0.707 0.077 0.555 1.76 0 - - Deforested M oribaetis 5 CG 0.845 0.008 0.782 0 2.8 - - Fallceon 5 CG 0.805 0.051 0.857 2.35 6.15 - - Vacupernius 5 Generally CG, few Ft 0.776 0.079 2.190 1.83 30.1 - - Leptohyphes 5 Generally CG, few Ft 0.815 0.093 2.803 22.5 82.2 - - I C P K Indian Leucotrichini Gen undet. 6 Generally Pc - Hb, Sc, CG 0.830 0.034 1.157 19.8 0.652 0 0 Camelobaetidius 5 CG 0.645 0.041 0.895 12.3 1.96 3.26 3.42 Hexacylloepus 5 Generally CG, Sc, Sh - Hb 0.707 0.068 0.57 8 1.98 0 1.63 0 Metrichia 6 Generally Pc - Hb, Sc, CG 0.676 0.086 1.477 18.1 3.26 4.89 6.12 Corn Atya NA Ft 0.942 0.008 0.550 0 4.24 0 0.245 Chironominae 2 Generally CG, Ft 0.643 0.013 6.425 141 730 353 54.6 Macronema 5 Generally Ft. Some Pr & Sc 0.7 55 0.034 1.96 10.4 3.26 0.245 1.96 Palaemnema 7 Pr 0.586 0.048 1.367 26.6 37.2 10.8 15.7 Ceratopogoninae NA Generally Pr, few CG 0.684 0.070 1.553 2.07 23.8 14 1.22 Collembola NA NA 0.672 0.089 0.338 0.196 1.3 0 0.245 53 Pijibaye Tricorythodes 5 Generally CG, few Ft 0.681 0.001 3.013 44.5 39.1 211 16.4 Argia 4 Pr 0.693 0.004 1.254 4.54 8.48 46 4.65 Limnocoris 4 Pr 0.789 0.006 1.385 1.96 0.326 25.1 4.16 Ancylidae NA Sc 0.859 0.007 1.035 1.17 0.326 14 0.979 Vacupernius 5 Generally CG, few Ft 0.851 0.014 2.204 2.94 0 63.3 5.14 Oecetis 8 Pr, Facultative Sh - Hb. 0.693 0.022 0.852 3.13 0.652 8.48 1.71 Tanypodinae 2 Pr 0.650 0.036 1.164 5.42 10.4 22.5 0.979 Epigomphus 7 Pr 0.709 0.057 0.587 0.822 0.979 3.59 0.489 Psephenus 7 Sc 0.594 0.063 2.555 53.3 78.3 116 15.7 Petrophila 5 Sc, Facultative Sh - Hb 0.640 0.075 2.041 35.8 8.81 52.8 6.36 Leptohyphes 5 Generally CG, few Ft 0.666 0.080 2.755 33.1 4.89 133 44 Kukra Moribaetis 5 CG 0.768 0.046 0.670 0 0 3.26 2.45 54 Deforestation h istory and h abitat as p redictors of the s tream c ommunity The deforestation history index for the catchment draining to each reach was the best predictor of all invertebrate taxa responses except evenness, when compared to 3 other lan dscape - scale and 20 habitat - level predictors in a series of linear regressions (See Table 6 & Figure 11 for R, R 2 , and p values). As the index value increased, invert ebrate taxa richness, BMWP score, and invertebrate density ( ln transformed) all decreased signif icantly ( R²= 0.484, p =0.004; R²= 0.445, p =0.007; & R²= 0.393, p =0.012, respectively). In the case of macroinvertebrate community evenness the relationship was the opposite as the index increased, evenness increased. The deforestation index for the 100 m buffer around the reach was the strongest landscape predictor (+) (R²= 0 .323, p =0.027), followed by % Forest Cover in the catchment ( - ) (R²= 0 .272, p =0.046), and the deforestation history index for the catchment (+) (R²= 0.258, p = 0.053). Invertebrate taxa r ichness was also significantly predicted by % fines (+) ( R²= 0.335 , p = 0.024) , small woody and leafy debris % cover (+) ( R²= 0.303 , p = 0.034) , and large wood volume (+) ( R²= 0.291 , p = 0.038) (Table 6). BMWP score was significantly predicted by embeddedness in ri ffles and rapids ( - ) ( R²= 0.428 , p = 0.008) , the riparian condition index (+) ( R² =0.365 , p = 0.017) , proportion of stream bank eroded ( - ) ( R²= 0.321 , p = 0.027) , periphytic macroalgae % cover (+) ( R²= 0.278 , p = 0.043 ) , and habitat complexity (+) ( R²= 0.265 , p = 0.049 ) (Table 6). Density had no significant habitat predictor s. Evenness was significantly predicted by large wood volume ( - ) ( R²= 0.265 , p = 0.050 ) . The residuals of linear regressions with periphytic macroalgae % cover as a predictor were clearly more logarithmic than linear, and this variable became a strong positive predictor of invertebrate taxa richness, BMWP score, and invertebrate 55 density ( ln transformed) when a logarithmic fit to the regression was employed ( R² >0.4, p <0.01). Other predictors (p<0.1) are li sted in Table 6. The deforestation history index for the catchment draining to each reach and for the 100 m buffer around each reach were significant negative predictors of cichlid abundance ( ln transformed) (Catchment: R²= 0.430 , p = 0.008; Buffer: R²= 0.523 , p = 0.002) and shrimp abundance (Catchment: R²= 0.342 , p = 0.022; Buffer: R²= 0.303 , p = 0.034) . The deforestation history index for the catchment was also a negative predictor for fish taxa richness , though not significant ( R²= 0.199 , p = 0.095 ). These landscape sc ale metrics better predicted fish and shrimp response metrics than nearly all 20 habitat p redictors, when compared in a series of linear regressions (Table 7 & Figure 11). Fish taxa richness was predicted by the standard deviation of embeddedness ( - ) ( R²= 0 .472 , p = 0.005 ) . Cichlid abundance had no significant habitat predictors. Shrimp abundance was significantly predicted by embeddedness in riffles and rapids ( - ) ( R²= 0.483 , p = 0.004 ) , standard deviation of embeddedness ( - ) ( R²= 0.280 , p = 0.042 ) , and mid - canopy riparian plant % cover (+) ( R²= 0.264 , p = 0.050 ) . Other predictors (p<0.1) are listed in Table 7. 56 Table 6: Single regression comparisons of landscape and habitat variables (X) as predictors of macroinvertebrate taxa responses (Y). 2 , an d p are listed. Bolded items are Landscape parameters are shown above and habitat parameters below, in order of significance. Although regressions were run on all 20 habitat predictors, only habitat predictors with a p - included. All regressions were linear, except for peri phytic macroalgae, which was a better fit under a logarithmic relationship. Italicized items did not meet assumptions of normality of residuals or were influenced by an outlier (Shapiro - Wilk p<0.05). T he deforestation history index for each study catchment was the best predictor at the landscape scale (except for with evenness, where the index at the buffer scale was better). It also predicted better than all but one of the 20 other habitat variables fo r each macroinvertebrate taxa summary response. Invert . Taxa Richness BMWP Score Density ( ln ) Evenness (EH) Landscape Predictors R Deforestation History Index ( Catchment ) - 0.696 Deforestation History Index ( Catchment ) - 0.667 Deforestation History Index ( Catchment ) - 0.627 Deforestation History Index (100 m Buffer) 0.568 R² 0.484 0.445 0.393 0.323 P 0.004 0.007 0.012 0.027 R % Forest Cover Catchment 0.544 % Forest Cover Catchment 0.56 % Forest Cover Catchment 0.438 % Forest Cover Catchment - 0.52 2 R² 0.296 0.314 0.191 0.272 P 0.036 0.030 0.103 0.046 R Deforestation History Index (100 m Buffer) - 0.517 Deforestation History Index (100 m Buffer) - 0.478 Deforestation History Index (100 m Buffer) - 0.422 Deforestation History Index ( Catchment ) 0.508 R² 0.267 0.229 0.178 0.258 P 0.048 0.071 0.118 0.053 R % Forest Cover 100 m Buffer 0.267 % Forest Cover 100 m Buffer 0.243 % Forest Cover 100 m Buffer 0.128 % Forest Cover 100 m Buffer - 0.394 R² 0.071 0.059 0.016 0.155 P 0.337 0. 383 0.649 0.146 Habitat Predictors R % Fines 0.579 Embeddedness in Riffles and Rapids - 0.654 Periphytic Macroalgae % Cover 0.510 Large Wood Per Transect Per 100 m - 0.515 R² 0.335 0.428 0.260 0.265 P 0.024 0.008 0.052 0.050 R Small Woody & Le afy Debris % Cover 0.55 Riparian Condition Index 0.604 Embeddedness in Riffles and Rapids - 0.486 Small Woody & Leafy Debris % Cover - 0.481 R² 0.303 0.365 0.236 0.231 P 0.034 0.017 0.066 0.070 R Large Wood Volume Per 100 m 0.539 Proportion of Stre am Bank Eroded - 0.567 % Sand - 0.458 R² 0.291 0.321 0.210 P 0.038 0.027 0.086 R Instream Live Trees and Roots % Cover 0.486 Periphytic Macroalgae % Cover 0.527 R² 0.236 0.278 P 0.066 0.043 R Proportion of Stream Bank Eroded - 0.470 Habitat Complexity 0.515 R² 0.221 0.265 P 0.077 0.049 57 R Small Woody & Leafy Debris % Cover 0.506 R² 0.256 P 0.054 R Large Wood Volume Per 100 m 0.460 R² 0.212 P 0.084 R Periphytic Macroalgae % Cover (Logarithmic) 0.659 Periphytic Macroalgae % Cover (Logarithmic) 0.750 Periphytic Macroalgae % Cover (Logarithmic) 0.697 Periphytic Macroalgae % Cover (Logarithmic) - 0.497 R² 0.434 0.563 0.4 86 0.247 P 0.008 0.001 0.004 0.059 58 Table 7: Single regression comparisons of landscape and habitat variables (X) as predictors of fish and shrimp responses (Y). 2 , and p are listed. Bolded items are significant 0.05). Landscape parameters are shown above and habitat paramet ers below, in order of significance. Although regressions were run on all 20 habitat predictors, only habitat predictors with a p - or periphytic macroalgae for shrimp abundance, which was a bett er fit under a logarithmic relationship. Italicized items did not meet assumptions of normality of residuals or were influenced by an outlier (Shapiro - Wilk p<0.05). The deforestation history in dex for each study catchment was the best predictor at the land scape scale (except for with evenness, where the index at the buffer scale was better). It also predicted better than all but one of the 20 other habitat variables for each macroinvertebrate ta xa summary response. Fish Taxa Richness Cichlid Abundance ( ln ) Shrimp Abundance Landscape Predictors R Deforestation History Index ( Catchment ) - 0.446 Deforestation History Index (Buffer) - 0.723 Deforestation History Index ( Catchment ) - 0.585 R² 0.199 0.523 0.342 P 0.095 0.002 0.022 R % Forest Cover Catchment 0.416 Deforestation History Index ( Catchment ) - 0.656 Deforestation History Index (Buffer) - 0.55 R² 0.173 0.430 0.303 P 0.123 0.008 0.034 R Deforestation History Index (Buffer) - 0.3 71 % Forest Cover Catchment 0.567 % Forest Cover Catchment 0.502 R² 0.138 0.321 0.252 P 0.173 0.027 0.057 R % Forest Cover Buffer 0.187 % Forest Cover Buffer 0.284 % Forest Cover Buffer 0.316 R² 0.035 0.080 0.100 P 0.504 0.305 0.252 Ha bitat Predictors R Standard Deviation of Embeddedness - 0.687 Periphytic Macroalgae % Cover (Linear) 0.497 Embeddedness in Riffles and Rapids - 0.695 R² 0.472 0.247 0.483 P 0.005 0.060 0.004 R Stream Size (Reach Volume) 0.497 % Shade (Densiometer) - 0.491 Standard Deviation of Embeddedness - 0.529 R² 0.247 0.241 0.280 P 0.060 0.063 0.042 R Mid - Canopy Riparian Plant % Cover 0.514 R² 0.264 P 0.050 R Riparian Condition Index 0.496 R² 0.246 P 0.060 R Prop ortion of Stream Bank Eroded - 0.478 R² 0.228 P 0.072 59 0 0.5 1 1.5 2 2.5 3 0 2 4 6 8 10 Mid - Canopy Riparian Plant % Cover (ArcSIN) Deforestation History Index (Catchment) Forested Recently Deforested Longer Deforested R 2 =0.460, p=0.005 A) 0 0.2 0.4 0.6 0.8 1 1.2 0 2 4 6 8 10 Small Woody and Leafy Debris % Cover (ArcSIN) Deforestation History Index ( Catchment ) Forested Recently Deforested Longer Deforested R 2 =0.436, p=0.007 B) 0 1 2 3 4 5 6 7 0 2 4 6 8 10 Large Wood Volume per 100 M (ln) Deforestation History Index ( Catchment ) Forested Recently Deforested Longer Deforested R 2 =0.425, p=0.008 C) 0 0.1 0.2 0.3 0.4 0.5 0.6 0 2 4 6 8 10 Macrophytes % Cover (ArcSIN) Deforestation History Index (C atchment ) Forested Recently Deforested Longer Deforested R 2 =0.416, p=0.009 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 0 2 4 6 8 10 Proportion of Stream Bank Eroded (ArcSIN) Deforestation History Index ( Catchment) Forested Recently Deforested Longer Deforested R 2 =0.392, p=0.012 0 0.5 1 1.5 2 2.5 3 3.5 0 2 4 6 8 10 Riparian Condition Index (ln) Deforestation History Index (Ctchment) Forested Recently Deforested Longer Deforested F) R 2 =0.349, p=0.020 Figure 10: Single linear regression comparisons of the deforestation history index at the catchment scale (X) as a predictor of habitat responses (Y). A) ArcSIN transformed mid canopy riparian plant % cover, B) ArcSIN transformed small woody and leafy debr is % cover, C) Natural log transformed large wood volume per 100 M D) ArcSIN transformed macrophytes % cover, E) ArcSIN transformed pro portion of stream bank eroded, F) ln ri parian condition index. R 2 and associated p - values listed. The deforestation history index for the catchment significantly predicted six habitat variables ( Table 2 ). The other 14 were non - significant (only p <0.1 visualized here) . D) E) 60 0 0.2 0.4 0.6 0.8 1 1.2 0 2 4 6 8 10 Periphytic Macroalgae % Cover (ArcSIN) Deforestation History Index ( Catchment ) Forested Recently Deforested Longer Deforested I) R 2 =0.210, p=0.086 0 0.2 0.4 0.6 0.8 1 1.2 0 2 4 6 8 10 Overhanging Vegetation % Cover (ArcSIN) Deforestation History Index ( Catchment ) Forested Recently Deforested Longer Deforested G) R 2 =0.238, p=0.065 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0 2 4 6 8 10 Instream Live Trees and Roots % Cover (ArcSIN) Deforestation History Index ( Catchment ) Forested Recently Deforested Longer Deforested H) R 2 =0.222, p=0.076 G) ArcSIN transformed overhanging vegetation % cover, H) ArcSIN transformed instream live trees and roots % cover, and I) ArcSIN transformed periphytic macroalgae % cover. 61 Figure 11: Sin gle regression comparisons of the deforestation history index at the catchment scale (X) as a predictor of macroinvertebrate, fish, and shrimp responses (Y). A) Invertebrate taxa richness, B) ln invertebrate density (per m 2 ). C) BMWP score D) Invertebrate R 2 listed and associated p - values listed. The deforestation history index for the catchment was the best predictor at the landscape scale (except for with evenness and cichlid abundance, where the index at the buffer scale w as better) and predicted best or second best when compared to all 20 other habitat predictors (Tables 6 & 7). 0.4 0.5 0.6 0.7 0.8 0.9 1 0 2 4 6 8 10 Shannon's Evenness (EH) Deforestation History Index ( Catchment ) Forested Recently Deforested Longer Deforested R 2 =0.258, p=0.053 D) 5 5.5 6 6.5 7 7.5 8 8.5 9 0 2 4 6 8 10 Invertebrate Density (ln) Deforestation History Index ( Catchment ) Forested Recently Deforested Longer Deforested R 2 =0.393, p=0.012 B) 50 70 90 110 130 150 170 190 0 2 4 6 8 10 BMWP Score Deforestation History Index ( Catchment ) Forested Recently Deforested Longer Deforested R 2 =0.445, p=0.007 C) 30 35 40 45 50 55 60 65 70 75 0 2 4 6 8 10 Invertebrate Taxa Richness Deforestation History Index ( Catchment ) Forested Recently Deforested Longer Deforested R 2 =0.484, p=0.004 A) 62 E) Fish taxa richness F) Natural log transformed cichlid abundance, G) Shrimp abundance. 0 2 4 6 8 10 12 14 16 18 0 2 4 6 8 10 Shrimp Abundance Deforestation History Index ( Catchment ) Forested Recently Deforested Longer Deforested R 2 =0.342, p=0.022 G) 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 2 4 6 8 10 Cichlid Abundance (ln) Deforestation History Index ( Catchment ) Forested Recently Deforested Longer Deforested R 2 =0.430, p=0.008 F) 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 Fish Taxa Richness Deforestation History Index ( Catchment ) Forested Recently Deforested Longer Deforested R 2 =0.199, p=0.095 E) 63 DISCUSSION Deforestation and subsequent land use change can cause a variety of different impacts to stream biota and habitat, depending on its extent, timing, type of land use, and natural conditions (Allan, 2004; Leitão et al ., 2017). In the Rama - Kriol te rritory and reserves of Southeast Nicaragua the impacts of deforestation and subsequent conversion to pasture on streams are distinct yet, as hypothesized, align closely with what has been found in many other tropical stream studies : d eforestation and catt le ranching had pronounced impacts on stream biota , mediated by changes in stream habitat , which were strongly predicted by level of deforestation and time since history i deforestation, it appears that it plays a substantial role in structuring str eam habitat and biotic responses when cattle ranching is the main subsequent land use. Instream habitat response mediated by deforestation effects There were strong impacts of deforestation and cattle ranching on instream habitat, which were generally more severe for streams whose catchments were deforested longer and to a larger extent. Metrics of bank disturbance and instream habitat evidence this trend. Bank erosion varied from 0 - 12 % in forested conditions to 25 - 48% in deforested conditions and was sign ificantly explained by the deforestation history index. One reach in Pijibaye and one in Kukra had lower bank erosion (7 % and 5 %, respectively), but both also had more intact riparian forest and were mostly fenced off to cattle . Cattle access to streams appeared to be a major contributor to instream and bank destabilization, as shown in many studi es ( Strand & Merrit t , 1999; Wantzen & Mol, 2013). 64 Both autochthonous and allochthonous sources of plant material were lowest in longer deforested streams (Table 1) and were significantly explained by the deforestation history index (Table 2). Lower amounts of large wood and small woody and leafy debris in streams is a common impact of deforestation in a catchment ( Bojsen & Barr iga, 2002 ; Benstead et al., 2003; Bojsen & Jacobsen, 2003; Wright & Flecker, 2004; De Paula et al., 201 1 ; Leal et al., 2016; Leitão et al., 2017 ; Brejão et al., 2018 ; Montag et al., 2019 ), and is related to reduced riparian vegetation. It could also be related to decreased flow consistency, as deforestation could be resulting in increasingly flashy streams ( Chaves et al., 2008; Recha e t al., 2012; Peña - Arancibia, Bruijnzeel, Mulligan, & v an Dijk , 2019) with less stabilizing large wood structure which tend s to flush out leaf litter and smaller debris (Bilby & Likens , 1980). Notably, a quatic vegetation (macrophyte % cover) was highest in forested streams, and at 0 % cover in all longer deforested, one recently deforested, and 1 more erosional forested stream , and was significantly predicted by the deforestation history index . Periphytic macroalgae had the highest percentage in more recent ly deforested streams, but lowest in longer deforested streams. In other systems, increases in both metrics is often associated with more available sunlight to the stream channel because of riparian deforestation (Bojsen & Jacobsen, 2003; Lobón - cerviá et a l., 2016; Leitão et al., 2017; Feijó - Lima et al., 2018). This sunlight - instream productivity subsidy effect (Allan, 2004) may be occurring with macroalgae recently deforested streams and in streams affected by the hurricane , but the opposite was true for l onger deforested streams. This indicates that there may be a threshold of stream stability above which/below which macroalgae can flourish in deforested stream s. Sediment from eroded banks could be smothering aquatic vegetation and macroalgae and scouring the stream b otto m of suitable stable 65 substrates for establishment ( ; Schwendel et al., 2010 ) , despite positive sunlight conditions for growth. Although sedimentation and decreased bed stability were not effectively recorded in this stud y, they were apparent in longer deforested streams (Figure 12). Percent embeddedness in riffles and rapids, % fines, and % sand were intended to capture these dyna mics (Allan, 2004; ; Kaufman et al., 2008; Lorion & Kennedy, 2009a), and results showed higher embeddedness in longer deforested streams, but results were nonsignificant. Bed stability can be measured by a variety of different technique s ( Schwendel et al., 2010) , all of which involved more time and labor - intensive measurements than possible in this study. Yet given that these streams were relatively high gradient, it is possible that sediments move through the system rapidly, and therefo re bed instability would not be captured by measurements of instream fine substrates and embe ddedness ( ). The relatively high gradient nature of these streams appeared to determine substrate parameters more than any feature associated w ith disturbance in this system. When considering the high rates of bank erosion and relativel y high gradient of streams, in addition to the declines in aquatic vegetation, macroalgae, small woody and leafy debris, and macroinvertebrate density in the most impacted streams, it is likely that deforestation is causing decreased bed stability. Increas ing flashiness and flooding from deforestation (Bradshaw et al., 2007; Chaves et al., 2008; Recha et al., 2012; Peña - Arancibia et al., 2019) could exacerbating these issues, but seasonal patterns in discharge rates were not measured for these streams. Chan ges in r iparia caused by hurricane and deforestation effects Riparian habitat was clearly degraded around all deforested strea ms, according to the riparian condition index. Understory riparian plant metrics ( particularly mid - canopy plant % 66 cover ) also capt ured this, as they were significantly higher in forested reaches , and significantly explained by deforestation history index (Table 2). The impacts of Hurricane Otto are shown by the finding that upper canopy large trees % cover and % shade were significan tly lower and large wood and small woody and leafy debris were s ignificantly higher at reaches impacted by the hurricane. T he hurricane downed many of the large riparian trees , which ended up in the streams, opening up the canopy. The e ffects of this on st ream habitat and biota could be substantial while the forest reg enerates , including increases in temperature or algal growth. Because of these hurricane dynamics, upper canopy large tree % cover did not differ significantly between forested reaches and rec ently deforested or longer deforested reaches. But streams in th e Corn River watershed had consistently high large tree cover (>50 - 57%) whereas recently deforested and longer deforested streams had variable large tree cover (10 - 48% and 19 - 53%, respectively ). These results together show that cattle ranchers in the study area often maintain large riparian trees but remove the woody understory for pasture, often right up to the stream bank. In fact, at most deforested sites there was evidence of active cattle grazing right up to the stream bank. This could also affect futu re riparian tree recruitment, as cattle trampling and grazing could limit seedling recruitment ( Griscom, Griscom, & Ashton, 2009; D e Paula et al., 2011 ) . Consistent reductions of m acroinverteb rates , shrimp, and fish Comparisons of macroinvertebrate community metrics showed significantly lower taxa richness and density only in streams where deforestation in the catchment has been occurring for a longer time and to a larger extent (Kukra watersh ed), whereas recently deforested streams were more like forested streams. BMWP score showed the same pattern (Figure 5). This aligns with studies that have found decreases in taxa richness ( Paaby et al., 1998 ; Iwata et al., 2003; Lorion & Kennedy, 2009a ; I ñiguez Armijos et al., 2014; Fugère et al., 2016; Tanaka et al., 2016 ; 67 Montag et al., 2019 ) and density ( Paaby et al., 1998 ; Iwata et al., 2003 ) in streams with deforested catchments. NMDS and PERMANOVA showed the same trend, where forested streams were si gnificantly gr ouped apart from deforested streams, but where forested streams and recently deforested streams group more closely together than longer deforested streams (Figure 9). Given both of these results together, it follows that the macroinvertebrate communities i n streams in the Pijibaye watershed, which have been deforested less time and to a lesser extent, have not yet been impacted the same amount as the communities in streams in the Kukra watershed, which have been deforested longer, and to a lar ger extent. The importance of deforestation not just in extent but also over time is also captured by the deforestation history index and its explanatory power (Table 6, Figure 11; see discussion below). Forested streams were also more similar to each oth er (cluster more tightly) than deforested streams, which has been seen in other tropical studies (Figure 9) ( Benstead et al., 2003; Lorion & Kennedy, 2009 a ; I ñiguez Armijos et al., 2014; Fugère et al., 2016) . This suggests that d eforestation changes the ma croinvertebrate community in an in consistent way over time between streams . This explanation is supported by the indicator analysis, which shows four indicator taxa in the more recently deforested reaches (P ij ibaye) and only one ( Moribaetis ) in the longer deforested reaches of the Kukra watershed . This implies that few er taxa consistently thrive in the most impacted conditions (Table 5 ). In this study, higher evenness was not an appropriate measure of macroinvertebrate community health. Evenness was signifi cantly lower in undisturbed forested streams compared to deforeste d streams. This is contrary to other tropical studies, which found either no difference ( Iwata et al., 2003 ; Iñiguez Armijos et al., 2014) or higher evenness in forests (Fugère et al., 2016) . E venness was negatively correlated with BMWP score, and positively correlated with the 68 deforestation index. In this study, the trend in evenness was driven by the dominance of Chironomidae at forested reaches. This has also been reported in the literatur e for forested tropical streams ( Sug a & Tanaka 2013; Gutiérrez - Fonseca, Ramírez, & Pringle, 2018 ). Functional fe eding group assignments can provide useful information to assess ecological impacts of disturbance ( Ramírez & Gutiérrez - Fonseca, 2014). In this study, all four indicator taxa for deforested streams were relatively mobile (swimming and clinging) genera of collector - gatherer mayflies (in two families; Table 3), as was the only indicator for longer deforested streams ( Moribaetis ). Forested a nd recen tly deforested sites had a variety of other feeding groups present as indicators. Baetid and Leptohyphid mayflies may be more resilient to disturbances, as they fill a flexible niche and their mobility allows them to actively seek cover to escape f lashes o f high flow and elevated sediment load. Fish and shrimp were also impacted in deforested streams. Comparisons of fish community metrics followed the same patterns as macroinvertebrate metrics, but differences were less pronounced. Shrimp abundance and Cich lid abundance were significantly affected by deforestation, especially where it has been occurring for a longer time and to a larger extent (Kukra). Fish taxa richness was also higher at forested reaches, but differences were not significant. But t he findi ng that all 20 species found in the whole study were found in forested streams, while 7 of these were not found in deforested streams , points to some impact on diversity and abundance occurring in deforested streams. Limited sampling effort and low sample size limit power and interpretation of fish results. O nly the four commonly fished species were significantly smaller at deforested reaches, while the rest of the smaller, less desirable species did not differ between forested and deforested streams. Larg er individuals of Brycon guatemalensis , Parachromis dov i i , and Tomocichla tuba 69 were nearly absent in both recently deforested and longer deforested streams compared to forested reaches (Figure 7). These findings could be attributed to either fishing pressu re or the habitat effects of deforestation and cattle pasture . Though fishing p ressure was not measured, it is known that fishing is common in these cattle ranching communities, and the data likely speak to its effects given that only the length of larger , more heavily fished species was significantly reduced. Th ese reductions in size and abundance could pose a threat to the local fishery, especially if these species are not allowed to reach prime reproductive age or size . In Nicaragua, there are no size r estrictions or limits on freshwater fish take in streams outsid e of protected areas. Even fishing bans for illegal occupants of Indio - Maíz are ignored (GTRK, unpublished data). Regulation in these remote areas is nil, and the life histories of many of the species in these rivers are poorly studied, so the foundation f or establishing effective size limits and take restrictions is weak. Given these results, regulators should consider further study in these systems, and establishment of size and take limits, b efore fisheries are further reduced. Taxa r esponse and s tream h abitat Associating specific changes in the habitat with specific changes in the biotic community is a consistent problem in stream studies ( Gergel et al., 2002). In this study it was clear that deforestation caused many cooccurring impacts to stream habitat, all of which interact with biota in unique but interrelated ways. Linear regression analyses helped to elucidate these patterns. The dominating influence on macroinvertebrates in this study a ppears to be stream channel and bed instability as it influences instream habitat and food availability. Since periphytic macroalgae (especially when considering a logarithmic relationship), embeddedness in riffles and rapids, stream bank erosion, small wo ody and leafy debris, and large wood are all related to 70 F igure 12: Examples of streams in each watershed , featuring typical levels of disturbed banks and riverbed. (A) Indian River tributary. Note high levels of understory growth and large wood and debris from Hurricane Otto. (B) Corn River tributary. Note the intact primary forest canopy and instream debris. (C) Pijibaye River tributary. Note the pasture up to the strea m bank, as well as some maintained larger riparian trees but no forest understory . (D) Kukra River tributary. Note the sluffing eroded banks, pasture up to the stream bank, and bank sediment covering the stream bottom. This was a particularly a ffected reac h. 71 stream channel and bed instability , this was likely the main mediator of mac roinvertebrate response to deforestation. Wood as a direct provider of habitat (Valente - Neto et al. , 2015), and macroalgae and debris as sources of food could also be important mediators . Nutrient enrichment and contaminant pollution were not measured and could also be contributing to patterns in biotic response (Allan, 2004). Given that habitat conditions in longer deforested streams were especially not stable, and it was in these streams where the strongest taxa response was observed, deforestation over t ime as it impacts instream and bank stability was the most likely ca use of invertebrate declines. Multiple studies, both temperate and tropical, have found stream chan nel and bed instability to be some of the most important factors in determining patterns in the macroinvertebrate community (Townsend, Scarsbrook, & Dolédec, 1997; Schwendel et al., 2010; Ferreira et al, 2014 ). Shrimp abundance was also predicted significantly by embeddedness and riparian condition metrics, and followed similar patterns to mac roinvertebrate density, which lends to the conclusion that shrimp were also sensitive to decreases in stream channel and bed stability in deforested streams ( Table 7). Observed evenness patterns could also be described by stream channel and bed instability . Miyake, Hiura, & Nakano ( 2005 ) found that frequent bed disturbance in a Japanese stream raised evenness in stream patches because Chironomidae were not able to effectively colonize, whereas they were the dominant taxa at undisturbed patches. These dynami cs could be driving evenness patterns in this study, since the sites with the lowest numbers of Chironomidae were the longer and more deforested sites, which were also the most eroded and least stable. Schwendel et al. ( 2010) also observed increased evenne ss with increased bed stability. Many studies have also associated the impacts observed in this study with changes in the fish community, for example: decreased bed stability (Leitão et al., 2017), eroded banks and 72 sedimentation (Iwata et al., 2003), algae abundance ( Bojsen & Barriga , 2002 ; Lobón - cerviá et al., 2016 ), decreased large wood ( Wright & Flecker , 2004 ), and lower woody and leafy debris (Bojsen & Barriga, 2002). This could be impacting the fish community in this study, but changes in the fish comm unity observed in this study were less pronounced, and associations with habitat metrics were weaker (Table 7). The lack of information on the feeding habits and reproductive needs of many species in this study also make further interpretation difficult (B ussing, 1998). History of d eforestation as the best predictor of t axa r esponse s Notwithstanding all of these relationships between habitat and biotic response, the deforestation history index for the catchment draining to each reach was the best linear pre dictor of all invertebrate taxa responses better than all other habitat metrics, and better than % forest cover at the catchment or 100 m buffer scale and the deforestation history index at the 100 m buffer scale (Table 6). Only for evenness did the index for the buffer and forest cover for the catchment better predict than the index for the catchment. The deforestation history index for the catchment was also a top - three predictor for all fish and shrimp community metrics, in every case predicting better t han % forest cover at catchment and buffer scales (Table 7). This supports the i dea that land use change at the catchment scale is an integrator of habitat impacts and can serve as an important predictor for impacts to stream ecosystems ( Gergel et. al., 20 02; Heartsill - Scalley & Aide , 2003 ; Leal et al., 2016; Molina et al., 2017 ; Brejão et al., 2018; Zeni et al., 2019 ). The fact that the deforestation index, which integrates both forest cover extent, and time since deforestation, better predicted biotic res ponse than raw forest cover extent, emphasizes the importance of the tempora l component of impacts of land use (Harding et al., 1998; Iwata et al., 2003; Brejão et al., 2018; Zeni et al., 2019 ). In this context, stream processes set into motion by 73 deforest ation, and then exacerbated by cattle ranching, increasingly a ffect stream habitat and biota over time. Thresholds of h abitat and b iotic d isturbance The presence of thresholds of disturbance for aquatic habitat and biota is a pillar of stream disturbance ecology (Allan, 2004) . Although detailed threshold analysis was not carried out, presence of a threshold response is anecdotally supported by this study. It appears that some threshold of impact after which macroinvertebrate, shrimp, and many habitat metri cs declined has not yet been achieved in the more recently and less deforested streams (Pijibaye), whereas it has in longer and more deforested streams (Kukra) (see Figures 10 & 11) . Stream channel and bed stability have been shown to de cline after a thres hold of anthropogenic land use change occurs (Kaufmann, Larsen, & Faustini, 2009), as have metrics of stream biota in response to thresholds of bed stability ( Schwendel et al., 2010 ). This could serve as an explanation for the patterns s een in this study. The BMWP index appropriate for assessing deforestation impacts to streams? Taxa associated with forested and deforested streams were often contrary to what would be expected by the BMWP index (where taxa with low scores are usually associated with disturbe d conditions). For example, the four mayfly genera that indicated deforested habitat all have a BMWP score of 5. If these genera indeed indicate habitat degraded by deforestation, a BMWP value of 5 could be incorrect in the context, and these genera should be assigned a lower score. This would have to be verified by additional studies. Two of the seven indicators for forested streams had a BMWP score of 1, and Chironominae, with a BMWP score of 2, was an indicator at forested streams in Corn River watershed could be expected at pristine locations, but not their dominance, according to index theory, 74 which typically also considers relative abundance (Hilsenhoff, 1988). This puts into questi on the utility of the index for evaluating impact to streams from deforestation and conversion to cattle al - cause it functions only at the f amily level, while actual tolerances vary at the species level (Hilsenhoff, 1988). But it has been applied to assess impacts of a variety of types of disturbance in Costa Rica and Nicaragua ( Kumar, Colton, Springer, & Trama, 2013; Salvatierra, 2014). Defor estation a also benthic habitat, so the index may not be fully appropriate. Much clearer trends can be found by looking at community measures of diversity and density, and functional feeding groups or life history traits of each t axa as it relates to changes in habitat, and thus analyzing these factors may be a more useful approach than considering the ind e x for evaluating deforestation impacts. Study l imitations Sampling methods used and variables measured limit this study in man y regards. Fish sampling techniques selected for species that are active in the daytime (both methods), eat worms and can bite a t least a size 14 hook (hook and line) or are found in the open channel (cast net). Sampling effort was surely not enough to cap ture all the species present in each stream or even the reach. Terra et al. (201 6 ) , who used the same reach length as this study in streams in Brazil, described that even electrofishing was not sufficient to estimate species richness because of the presenc e of so many relatively rare taxa. But other metrics of assemblage condition can s till be useful for environmental assessment ( Terra et al., 201 6 ). Fishing with pesticides (mainly cypermethrin) occurs in many of these communities, and likely also has play ed a role in structuring fish and invertebrate communities. Forthcoming research w ill further assess the extent of this problem (JT Betts, unpublished data). 75 In summing and averaging of habitat and invertebrate metrics from all 11 transects to attain reac h level values, specific information was lost. Although reach - scale analysis is va luable for landscape - level studies, potentially important finer scale impacts of disturbance were not assessed. Further analysis which separates habitat and invertebrate metr ics by transect will be useful, especially in evaluating taxa - specific habitat pre ferences and responses to disturbance. important stream habitat processes were impossib le to capture. Metrics that vary widely depending on weather and seasonal conditio ns such as temperature, pH, dissolved oxygen, nutrient concentrations, turbidity, discharge, bed stability, and sediment transport, were not possible to accurately measure si nce sampling at each reach occurred in 2 - 3 days. For this same reason, abundances of certain taxa which move in the stream system seasonally may have been over or underestimated. T rue algal biomass was not reco r ded, given that it was a visual assessment of larger (clearly visible) filamentous algae, not even all periphytic algae . This limits interpretations about subsidy effects from increased sunlight and nutrients. These limitations were anticipated, but future study could benefit by considering longer te rm monitoring at sites and quantification of algal biomass. The informative power of this study was limited for a variety of reasons. A sample size of 15 is too low to significantly capture many relationships a nd differences that may be present in both th e habitat and biotic community. For example, low sample size relative to number of habitat variables (20) made it inappropriate to apply multivariate techniques like DCA, PCA, and Random Forest analysis, which h ave been very informative in similar studies with higher sample size (Terra et al., 2016; Leal et al., 2016). More reaches were planned, but when the political crisis arose in April of 2018, the field study was immediately terminated by the funding source 76 (US government). This speaks to the challenge s of carrying out research in politically unstable contexts. Although reaches were independent of each other (no catchment overlap), some spatial autocorrelation was likely in this stud y, since sampling was carr ied out in just 4 larger watersheds. The rele vance of this study is not its large sample size, but the paucity of any prior undescribed stream fauna. Difficulty of acce ss, lack of infrastructure and communication, and political instability in southeast Nicaragua are likely reasons why studies from this area are so rare, compared to other ecologically similar regions like northeast Costa Rica. Novel f indings and f uture r esearch p riorities Given that this study is th e first formal aquatic study in any of these streams, and one of the only for the Indio - Maíz Biological Reserve , there were many novel findings and future research opportunities. Many of the reported genera are new reports for Nicaragua (JM Maes, pers. com m.) , and it is likely that some are undescribed taxa, since this study is the first published macroinvertebrate study from any of these watersheds. For example, one caddisfly genus (Hydropti lidae: Leucotrichini , genus undet.) has never been associated with its adult (M Springer, pers . comm.), and description is underway. Notably, the mayfly family Euthyplociidae ( Euthyplocia ) had not yet been recorded for the country (Maes & Salvatierra - Suarez , 2014) . Thiaridae, an invasive family of aquatic snails, was fou nd only in deforested reaches. It is likely that it is Melanoides tuberculat a , as this is the only species of Thiaridae reported for the country ( Pérez & López de la Fuente , 1993) . Introduced in the aquatic plant trade, it poses risk to native fauna and is a potential disease vector and has been shown to dominate in disturbed stream conditions ( Gutiérrez - Gregoric & Vogler, 2010 ). 77 Sic y dium altum was a new fish sp ecies register for Nicaragua, not known to extend north of Costa Rica (Bussing, 1998). Hypostomus sp. (possibly H. niceforoi ) , an invasive Loricariid catfish, was found in the upper reaches of Indian River. This genera has been recorded in Nicaragua, only in the San Juan River drainage ( Corea, Hernández, Solís, & Aguilar, 2014 ; Härer et al., 2017), an d evidence from interviews f rom a study by the authors (publication forthcoming) with fishermen indicates that it is a recent arrival and could be present in other watersheds as well . In prior publications it was recorded as H. panamensis , a Central Americ an cogener native to south of Nicaragua, but recent genetic evidence from samples from the San Juan drainage suggests it is likely H. niceforo i , a species from Columbia (N Lujan, pers. comm.). Invasive loricariid catfish have been shown to dominate in dist urbed stre am conditions ( Bojsen & Barriga, 2002 ; Leitão et al., 2017 ). Loricariid catfish are of little use to fisheries ( Capps & Flecker, 2015) and have limited predators in invaded systems ( Nico, 2010). They have been shown to reach high densities in som e streams in which the invasion has progressed significantly (up to ~2 per m 2 at some sites) , where they can be a significant threat to fisheries and river ecosystem function by degrading the amount and quality of primary resources ( Capps & Flecker, 2015). E ffort will be made to incorporate these species and genus registers into country and regional taxa lists, and future effort to further identify aquatic insect and shrimp specimens to species could yield important range expansions and even new species. Th e number of important findings even given the limited extent of this study (only 15 headwater streams ) justifies further intensive surveying of the region. There are many reasons it could be important for regional conservation. For example, species such as the Bobo (or hognose) mullet ( Joturus pichardi ) are of conservation concern in Costa Rica due to overharvest and dams ( Anderson, Pringle, & Rojas, 2006), and appear to be common in at least Indian River . 78 But the potential of these rivers as populat ion s trongholds and threats to these populations has not been studied. Another species of concern, the American Eel ( Anguil l a rostrata ), is likely present in these rivers ( unpublished data, JT Betts ), yet no data exists on its presence anywhere from n orthern Co sta Rica to n orthern Nicaragua ( Benchetrit & McCleave, 2015) . A n ew i ndex The predictive power of the novel Deforestation History Index for describing landscape driven processes is another important finding of this research. Since land use data has beco me increasingly accessible, there have been many efforts to find the most useful landscape predictors of stream conditions (Hawkins et al . , 2000 ; Macedo et al., 2018; Sandric et al., 2019). Brejão et al. (2017) used time since deforestation and deforestati on as separate indicators of impacts to streams . O ther studies have considered land use history in addition to extent of land use change i n how it impacts streams (Harding et al., 1998; Iwata et al., 2003; Zeni et. al. , 2019 ). But this study appears to be the first to integrate the time and extent components of land use change into one index . Given that the Deforestation History Index was easily generated with widely accessible software (GIS, Excel) using publicly available data (Hansen et al., 2013), this index could be a useful as a landscape predictor in a variety of contexts. It is useful when the impacts of deforestation are accumulative over the years where current percent deforestation in a study area does not fully capture its impact on a study syst e m. As shown in this study, this is true for conversion of rainforest to pasture as it impacts streams. If integrated with other land use datasets, this index could also be adapted to more complicated land use history situations, where multiple land uses ( n ot just binary forest cover) are considered. It could also easily consider forest gain, as the Hanson et al. (2013) dataset also includes a forest gain layer. In this analysis 79 the gain layer was not included because it currently only includes data up to 2 0 12. Other land uses were not considered because deforestation almost always meant conversion to cattle ranch in SE Nicaragua. Relevance to c onservation The findings of this study show that the agricultural frontier has had serious detrimental effects on s tream ecosystems and associated fisheries in the Rama - Kriol territory. Cattle ranchers are increasingly invading the Indio - Maíz Reserve to remove the for est and create pasture. With these changes in the headwaters of the Indian, Corn, Pijibaye, and Bartola Rivers which drain the reserve, degradation of stream habitat and subsequent degradation of stream macroinvertebrate, fish, and shrimp populations shoul d be expected. In the case of fish and Macrobrachium shrimp declines, this could also threaten the subs istence of Rama and Kriol communities who rely on these animals for food. In more invaded watersheds, older Rama community leaders have complained about the loss of reliance on rivers for food compared to before the invasion ( unpublished data, JT Betts). I ndeed, finding hard evidence for these concerns was an important goal of this study. Conservation r ecommendations The action plan for Indio - Maíz was finalized in 2017 by the Rama - Kriol territorial government, and includes ators for the condition of Indio Maíz ( Gobierno Territorial Rama y Kriol , 201 8 ) . This project serves as one such mon itoring effort. As the action plan for Indio - Maíz is implemented, these current and projected negative impacts of the invasion to streams and their fisheries need to 80 be considered . The following damage mitigation and prevention efforts, not only for Indio - Maíz, but for the whole territory are recommended: 1) Pr omotion of forest conservation and restoration . M aintaining the most primary forest possible in catchments and restoring forest where it has been lost are the best way s to promote aquatic ecosystem healt h and a healthy fishery. Given these results, any deforestation within a catchment will likely have some effect on stream habitat and biota over time. 2) Encouragement of m riparian buffers along streams. Riparian buffers along streams of every order, inc luding ephemeral streams, can mitigate the impact s of land use change on streams such as like temperature increase, nutrient increase, bank erosion, and decreased large wood and leaf litter inputs, among others ( Luke et al., 2019 ). Many studies have shown the e ffectiveness of buffers in maintaining stream macroinvertebrate and fish communities (Lorion and Kennedy, 2009a ,b ; Chellaiah & Yule, 2018 ). Based on extensive review of the literature, Sweeney & Newbold ( 2014) recommended buffers of m . 3) Removal of illegal ranches in Indio - Maíz a nd the Rama - Kriol Territory. The na tional and territorial laws of the reserves and indigenous territories of Nicaragua prohibit the establishment of new cattle ranches and farms within Indio - Maíz and the Rama - Kriol Territory unless in accordance with indigenous communal property laws under law 445 ( Saenz, unpublished report, 2019). The Rama - Kriol authorities have filed numerous legal complaints against new illegal ranches (Gobierno Terr i torial Rama y Krio l, 201 8). The Nicaraguan authorities should respond to these legal complaints and support removal of illegal ranches. 81 4) Alternatives to cattle ranching like sustainable agriculture and agroforestry . Cattle ranching is particularly damaging to streams compare d to many other forms of agriculture (Strand & Merritt, 1999) . Where agriculture is necessary, agroforestry techniques and traditional crop ping systems such as banana, coconut, and fruit tree systems, or corn and bean intercropping and root vegetabl es crop s like malanga, as grown typically by the Rama and Kriol inhabitants of the region are much less degrading alternatives to cattle ranching . 5) Fencing and restrict ed stream access for livestock . Where cattle or other livestock are present, r estricting the acc ess to the stream channel using fences can reduce stream bank erosion and trampling of the riparian zone. Having alternative water sources, or only allowing a few access points to a stream within a pasture can greatly reduce impacts llaghan et al., 20 18). 6) Fish size restrictions and take limits. Size and abundance of important fish species was severely a ffected in cattle ranching communities. Establishing and enforcing fishing regulations could help overfished species like Brycon guate malensis or Parach romis dovii to recover and maintain enough large, reproductive age individuals for a sustainable river fishery, even in areas of higher fishing pressure. 7) Removal of Loricariid catfishes. The Loricariid catfish invasion could become a significant threat to the fishery (Capps & Flecker, 2015). When caught, individuals should not be thrown back. Targeted harvest and removal of Loricariids could also be beneficial to help limit the effects of this nuisance species. 8) Ed ucational activities about the impacts of de forestation and cattle ranching on streams . Although there is general concern, there is little awareness about the 82 mechanisms of impacts of deforestation and cattle ranching to streams in the communities where the research was conducted. T he results of thi s study could be used to create educationa l materials to disseminate into the communities, to increase awareness and help motivate the adoption of some of these recommendations. Conclusion The case of deforestation in Rama - Kriol territory in Southeast Nic aragua is an important example of how rainforest loss impacts aquatic organisms and ecosystems processes, as well as the people who rely on the services they provide (Foley et al. , 2007). Al though they are an important component of biodiversity, stream org anisms are often neglected in conservation initiatives, compared to more charismatic fauna. As the agricultural frontier continues to threaten the Indio - Maíz Biological reserve and the rest of the Rama - Kriol territory, the distinct threats it poses to stre am biodiversity and ecosystem function needs to be considered, if these rivers and the life they support are to be conserved. 83 APPENDICES 84 APPENDIX A Raw Data Table A.1: Reach details, including local stream name s, reach codes, date of first day sampling at the reach, and base transect coordinates. Creeks with asterisks were names that the team created upon arrival, if there was not a known l ocal name. Caño Boca Tapadas and Caño Moga are considered the same creek on topographic maps but are different streams with confluences near each other. Watershed Indian River Watershed Corn River Watershed Stream Name Mountain Cow Creek* Caño Guinea Caño Banana Vieja * She Tiger Creek Long Falls Creek Caño Boca Tapadas Caño M oga Caño L a Combinación Study Reach Code IR18MC IR18GU IR18BV IR18ST IR18LF CR18BT CR18MG CR18CO Date Sampled 2/19/2018 2/22/2018 2/24/2018 2/26/2018 3/12/2018 4/10/2018 4/12/2018 4/14/2018 Coordinates (N) 11.13290 11.11850 11.12733 11.13787 11.12330 11 .28167 11.28114 11.26783 Coordinates (W) 84.04524 84.09462 84.08104 84.06307 84.05836 84.00817 84.00561 83.99248 Watershed Pijibaye River Watershed Kukra River Watershed Stream Name Caño El Coco Caño La Perra Caño El Salto Caño Papa Abrahán Caño El Lim ón Caño Chacalín Caño Limonero Study Reach Code RP18EC RP18LA RP18SA KR18PA KR18EL KR18CH KR18LM Date Sampled 4/19/2018 4/21//2018 4/23/2018 2/7/2018 2/9/2018 3/22/2018 3/25/2018 Coordinates (N) 11.45131 11.43747 11.43631 11.76249 11.73801 11.79634 1 1.80070 Coordinates (W) 83.95861 83.96941 83.93757 84.08453 84.10820 84.11238 84.10738 85 Table A.2: Full list of macroinvertebrate taxa abundances, by reach. Based on sum of eleven Surber samples (0.092903 m 2 ). Domínguez & Fernández (2009). Hydroptilidae: Leucotrichini, genus undet. is a unique t axon, likely undescribed (M Springer, pers. comm.). Grayed out columns are unknowns that were not considered unique taxa unles s there were no other reports from the respective family at a site. These taxa were not included in determination of richness or d iversity measures. These taxa, and taxa with <5 individuals in the study total were not included for NMDS, PERMANOVA, SIMPER, UCR museum (Crambidae undet., Planiplax, Tholymis, and Phoridae). Family Genus (or Subfamily ) IR18MC IR18GU IR18BV IR18ST IR18LF CR18BT CR18MG CR18CO RP18E C RP18LA RP18SA KR18PA KR18EL KR18CH KR18LM Study Total Arthropoda: Collembola: undet. undet. undet. 1 0 0 0 0 0 2 2 0 0 0 1 0 0 0 6 Arthropoda: Insecta: Ephemeroptera Baetidae Americabaetis 1 0 0 4 7 3 0 0 7 0 3 0 1 6 0 32 Apobaetis 0 0 0 0 1 5 0 2 2 0 1 0 0 0 0 11 Baetodes 1 17 2 5 7 1 3 2 8 0 0 0 0 3 44 93 Camelobaetidius 3 13 14 4 26 1 3 2 8 0 2 1 5 4 4 90 Cloeodes 0 0 0 0 0 0 2 0 1 0 1 0 1 2 0 7 Fallceon 1 2 0 4 2 7 3 0 2 5 10 1 7 13 6 63 Guajirolus 0 0 0 0 0 0 0 0 17 0 0 0 0 2 0 19 M ayobaetis 0 1 0 0 1 0 0 0 1 0 0 0 0 0 0 3 Moribaetis 0 0 0 0 0 0 0 0 10 0 0 2 1 2 5 20 Paracloeodes 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 2 undet. 3 2 0 0 10 7 1 3 4 0 1 1 6 1 0 39 Caenidae Caenis 1 0 0 0 0 26 5 7 5 7 29 0 0 1 0 81 Euthyplociidae Euthyplo cia 2 specimens from He Tiger Creek and 2 from Guinea Creek , Indian River, May 2 3 - 5 , 2017 , verified by L. Jacobus Heptageniidae Maccaffertium 0 0 2 0 1 2 0 0 0 0 0 0 0 0 0 5 Leptohyphidae Asioplax 13 0 3 9 10 2 19 2 11 5 0 1 1 12 7 95 Cabecar 0 0 0 0 3 3 2 0 1 0 2 0 0 1 0 12 Epifrades 0 0 1 2 0 1 1 9 2 1 28 0 0 2 0 47 Leptohyphes 6 86 28 48 1 8 2 5 311 71 26 11 6 49 114 772 Tricorythodes 72 37 18 41 54 35 41 44 344 190 112 4 14 45 4 1055 Vacupernius 2 12 0 1 0 0 0 0 6 51 137 0 4 17 0 230 unde t. 0 0 2 0 0 0 0 0 0 0 0 4 0 0 1 7 86 Leptophlebiidae Farrodes 40 5 5 9 41 52 19 49 114 37 47 2 45 14 7 486 Hagenulopsis 1 1 0 2 4 3 0 6 2 1 1 2 0 0 1 24 Hydrosmilodon 0 2 0 3 0 0 0 0 0 0 0 0 0 0 0 5 Thraulodes 10 7 7 0 32 4 70 27 85 34 56 14 18 90 8 92 617 Traverella 1 2 1 1 0 0 0 0 115 10 0 0 0 2 11 143 Arthropoda: Insecta: Odonata Calopterygidae Hetaerina 0 0 0 0 0 0 0 0 2 0 0 0 0 0 1 3 Coenagrionidae Argia 6 8 1 6 2 10 10 6 91 18 32 1 9 4 5 209 Gomphidae Agr iogomphus 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 1 Desmogomphus 3 0 0 0 0 0 1 1 0 2 3 0 0 0 0 10 Epigomphus 1 0 0 1 2 3 0 0 1 3 7 0 1 1 0 20 Perigomphus 1 5 6 4 1 2 0 2 3 0 2 4 0 0 0 30 Phyllocycla 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 1 Libellulidae Brechmorhoga 0 0 6 3 8 0 0 1 14 1 6 0 4 0 9 52 Libellula 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 Planiplax ** 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 1 Tholymis ** 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 1 Libellulinae undet. 3 1 0 0 0 0 0 0 4 1 0 6 2 0 2 19 Megapodagrionidae Heteragrion 0 0 0 0 0 0 1 1 0 2 0 0 0 0 0 4 Platystictidae Palaemnema 6 34 46 28 20 50 29 35 12 19 2 7 2 29 26 345 Arthropoda: Insecta: Plecoptera Perlidae Anacroneuria 2 21 1 4 2 16 1 15 109 3 0 13 11 4 47 249 Arthropoda: Insecta: Hemiptera Gerridae undet. 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 2 Mesoveliidae Mesoveloidea 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 Naucoridae Cryphocricos 9 11 39 3 26 22 8 9 12 14 0 5 0 1 0 159 Naucoridae Limnocoris 4 1 1 4 0 0 1 0 36 9 32 12 5 0 0 105 Veliidae Rhagovelia 0 0 0 4 0 1 1 3 0 1 0 0 0 0 0 10 Arthropoda: Insecta: Megaloptera Corydalidae Chloronia 0 0 0 0 0 0 2 0 0 0 1 0 2 0 0 5 Corydalidae Corydalus 0 6 0 3 0 1 4 1 74 8 0 0 0 0 3 100 Arthropoda: Insecta: Trichoptera Calamoceratidae Phylloicus 0 0 0 0 0 2 0 1 0 0 0 0 0 0 0 3 Gloss osomatidae Culoptila 0 0 0 0 0 1 12 3 0 0 11 0 0 1 0 28 Mortoniella 0 0 3 0 0 2 1 1 8 1 0 0 0 0 0 16 87 Protoptila 0 0 0 0 1 0 0 0 3 0 8 0 0 0 0 12 undet. 0 4 0 0 0 0 0 0 0 0 48 0 0 8 5 65 Helicopsychidae Helicopsy che 0 0 1 5 4 4 0 0 0 0 0 0 0 0 0 14 Hydrobiosidae Atopsyche 0 0 0 0 1 0 0 0 0 0 0 0 0 1 0 2 Hydropsychidae Centromacronema 0 0 1 0 8 0 0 0 0 0 0 0 0 0 1 10 Leptonema 5 34 44 20 38 98 36 53 105 21 6 9 9 33 66 577 Macronema 9 0 0 1 0 11 11 10 0 8 2 0 0 1 0 53 Macrostemum 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 2 Smicridea 20 142 214 61 54 92 131 74 459 83 18 29 3 56 49 1485 Hydroptilidae Alisotrichia 0 0 0 0 3 0 0 0 0 0 0 0 0 0 1 4 Hydroptila 6 0 76 0 3 0 38 0 120 0 7 0 0 0 0 250 Leucotrichia 0 1 1 0 0 3 0 0 0 0 0 0 0 0 0 5 Mayatrichia 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 1 Metrichia 30 1 16 8 34 2 0 8 15 0 0 1 0 5 19 139 Neotrichia 4 0 1 2 1 0 0 0 7 0 0 0 0 1 2 18 Ochrotrichia 0 0 3 13 2 0 0 0 0 8 6 1 0 42 10 85 Oxyethira 3 0 1 2 12 1 1 0 1 1 5 0 0 0 0 27 Zumatrichia 2 2 0 0 0 0 0 1 0 0 0 0 0 0 1 6 Leucotrichini undet. 92 0 3 5 1 0 2 0 0 0 0 0 0 0 0 103 undet. 26 5 19 8 4 0 0 0 2 0 0 0 1 0 25 90 Leptoceridae Nectopsyche 0 0 1 2 3 4 0 0 0 2 5 0 0 0 0 17 Oecetis 5 0 9 2 0 1 0 1 10 4 12 1 0 3 3 51 Triaenodes 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 1 Philopotamidae Chimarra 2 19 7 10 9 19 10 11 189 8 2 1 1 0 7 295 Polycentropodidae Cernotina 13 0 0 4 0 0 3 0 0 2 0 0 1 0 0 23 Polycentropus 0 0 2 0 0 0 0 0 0 2 0 0 0 4 0 8 Polyplectropus 4 9 1 9 0 0 0 0 0 46 1 0 0 4 0 74 Xiphocentronidae undet. 1 0 0 3 0 7 0 1 0 0 0 0 0 3 0 15 undet. undet. 25 1 3 3 0 8 41 0 0 2 28 0 0 2 4 117 Arthropoda: Insecta: Lepidoptera Crambidae Petrophila 27 8 111 29 7 4 13 10 79 29 54 4 0 6 16 397 undet.* * 6 0 0 1 19 0 0 0 0 1 0 1 0 0 0 28 Arthropoda: Insecta: Diptera Ceratopogonidae Ceratopogoninae 2 0 0 2 6 17 40 16 23 15 5 0 1 4 0 131 Dasyheleinae 0 0 0 0 1 0 0 1 2 0 0 0 0 0 0 4 Forcyponinae 0 0 0 1 0 2 0 0 3 1 0 0 0 0 0 7 88 C hironomidae Chironominae 185 66 144 158 152 501 1090 646 478 323 281 65 37 65 56 4247 Orthocladiinae 112 122 109 39 619 250 159 246 389 64 42 39 8 36 124 2358 Tanypodinae 10 1 2 7 7 5 8 19 40 15 14 1 1 1 1 132 Empididae Hemerodromia 3 0 6 0 3 0 1 1 15 0 2 3 0 0 15 49 undet. 5 1 0 2 0 4 0 0 0 0 0 0 0 4 0 16 Phoridae ** undet. 0 0 0 0 1 0 0 0 1 0 0 0 0 0 0 2 Psychodidae Maruina 0 0 2 2 1 5 1 1 10 4 1 1 0 1 3 32 undet. 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0 3 Stratiomyidae undet. 0 8 20 15 14 34 14 30 31 1 2 0 36 8 83 296 Simuliidae Simulium 5 0 0 0 0 0 0 0 0 0 0 41 0 0 1 47 Tipulidae Hexatoma 1 7 0 1 0 3 1 0 0 1 0 5 0 2 2 23 undet. 2 1 4 0 0 5 0 0 0 0 0 0 1 0 2 15 undet. undet. 0 0 0 20 0 1 0 0 0 0 2 0 0 0 1 24 Arthropoda: Insecta: Coleoptera Dryop idae Dryops 0 0 0 0 0 0 0 0 22 0 0 0 0 1 0 23 Elmidae Austrolimnius 7 2 5 1 19 1 5 4 16 1 4 3 3 3 0 74 Cylloepus 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 1 Heterelmis 1 3 4 5 6 4 0 10 77 1 1 4 0 0 0 116 Hexacylloepus 3 0 2 4 1 0 0 0 0 0 5 0 0 0 0 15 Hexancho rus 0 0 0 0 11 0 0 0 1 0 0 0 0 0 0 12 Macrelmis 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0 3 Microcylloepus 116 48 179 67 235 80 30 66 752 31 23 18 3 0 19 1667 Neocylloepus 0 0 1 0 1 0 0 1 0 0 0 0 0 0 0 3 Neoelmis 7 4 2 3 10 10 1 10 60 3 4 2 4 3 6 129 Phanoce rus 0 3 5 10 11 11 0 9 82 5 0 6 2 0 0 144 Stenhelmoides 1 1 2 2 9 0 3 0 2 1 0 1 0 0 0 22 Xenelmis 0 0 0 0 0 0 1 0 1 1 0 0 0 0 0 3 undet. 5 1 0 1 0 2 0 4 126 3 0 7 2 0 1 152 Gyrinidae undet. 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 Hydroscaphidae undet. 5 0 0 0 88 8 11 5 4 6 2 0 0 1 0 130 Lutrochidae Lutrochus 3 0 0 2 3 2 2 0 0 0 1 0 0 0 0 13 Psephenidae Psephenus 28 70 28 90 51 99 42 99 106 173 77 11 5 13 35 927 Ptilodactylidae Anchytarsus 2 0 0 2 4 8 1 3 3 1 0 1 0 0 0 25 Staphylinidae undet. 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 Arthropoda: Malacostraca: Decapoda Atyidae Atya 0 0 0 0 0 7 1 5 0 0 0 1 0 0 0 14 Palaemonidae Macrobrachium 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 89 Pseudothelphusidae undet. 0 0 2 0 0 0 0 0 0 0 0 0 0 0 1 3 undet. undet. 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 1 Arthropoda: Malacostraca: Ostracoda undet. undet. 0 0 0 0 1 0 3 0 0 0 2 0 0 0 0 6 Arthropoda: Arachnida: Hydrachnidia undet. undet. 5 0 2 3 5 2 3 3 9 1 1 2 1 5 3 45 Annelida: undet. undet. undet. 0 0 0 0 0 3 0 0 0 0 0 0 0 0 0 3 Annelida: Oligochaeta: Clitellata: undet. undet. undet. 36 16 110 35 400 36 38 43 52 0 11 1 3 49 10 840 Mollusca: Bivalvia: Veneroida Sphaeriidae undet. 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 1 Mollusca: Gastropoda: Basommatophor a Ancylidae undet. 6 0 0 0 0 0 1 0 21 9 13 0 0 4 0 54 Planorbidae undet. 0 0 0 0 0 0 0 0 0 1 0 0 0 4 0 5 Mollusca: Gastropoda: Neotaenioglossa Thiaridae undet. 0 0 0 0 0 0 0 0 0 0 2 0 0 5 0 7 Hydrobiidae undet. 4 0 0 5 0 4 2 2 0 5 11 2 0 5 1 41 Nema toda: undet. undet. undet. 2 0 1 0 79 0 181 3 2 0 0 0 0 0 0 268 Platyhelminthes: Trepaxonemata: undet. undet. undet. 1 0 2 1 6 6 0 2 5 1 4 0 0 1 0 29 Macroinvertebrate Community Metrics Taxa Richness 63 44 57 66 69 63 60 58 71 59 60 44 37 55 44 - B MWP Score 149 118 144 149 158 172 149 164 149 154 137 123 105 146 123 - Abundance 1029 925 1328 893 2189 1701 2129 1696 4696 1400 1225 357 340 608 962 - Density 1007 905 1299 874 2356 1664 2083 1660 4595 1370 1199 349 333 595 941 - Shannon's Diversit y (H) 3.00 2.91 2.83 3.21 2.65 2.76 2.10 2.49 3.07 2.84 2.96 2.97 2.64 3.22 2.97 - Shannon's Even n ess (EH) 0.72 0.77 0.70 0.77 0.63 0.67 0.51 0.61 0.72 0.70 0.72 0.78 0.73 0.80 0.79 - 90 Table A.3: Full list of fish taxa, by reach. Codes correspond to t Bussing (1998). Numbers are raw abundances from the study. P indicates present, but not caught. * indicates possibly caught but not identified, in just one case. For individuals not caught as part of the s tudy, # caught, location, and date of capture are listed. Family Scientific Name Rama - Kriol Name IR18MC IR18GU IR18BV IR18ST IR18LF CR18BT CR18MG CR18CO RP18EC RP18LA RP18SA KR18PA KR18EL KR18CH KR18LM Osteriophysi: Characiformes Characidae Characidae sp p. ( likely Astyanax spp.) Bilam 20+ 20+ 3 20+ 20+ 20+ 20+ 20+ 20+ 20+ 20+ 20+ 20+ 20+ 20+ Bramocharax bransfordii Bilam 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 Brycon guatemalensis Machaca 7 17 0 19 0 1 1 0 0 9 1 0 0 1 0 Roeboides bouchelli Bilam 1 0 0 3 0 0 0 0 12 26 3 3 3 6 0 Osteriophysi: Siluriformes Heptapteridae Rhamdia nicaraguensis Mulung 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 Rhamdia sp. Mulung 0 0 1 0 0 1 0 1 0 0 0 P 8 0 1 Loricariidae Hypostomus sp. Devil Fish 1 spec . from Indian R. near Guinea Creek, M ay 23, 2017. Species likely H. niceforoi (N Lujan, pers. comm.) Acanthopterygii: Cyprinodontiformes Poeciliidae Alfaro cultratus Tush - Tush 0 0 P 0 0 0 1 0 0 0 1 0 0 0 0 Phallichthys amates Tush - Tush 0 0 P 0 0 0 0 0 0 0 0 0 0 0 0 Poecilia gillii Tush - Tush 0 2 * 0 1 19 1 9 8 35 11 P 0 5 0 Priapichthys annectens Tush - Tush 0 0 P 0 3 0 0 0 0 0 0 0 0 0 0 Acanthopterygii: Atheriniformes Atherinopsidae Atherinella hubbs i NA 1 specimen from Indian River near She Tiger Creek, May 23, 2017, verified by A An gulo Sibaja, UCR Acanthopterygii: Mugiliformes Mugilidae Agonostomus monticola Salin 2 8 4 10 1 1 1 4 4 11 6 0 1 8 3 Joturus pichardi Salin/Bobo 1 specimen from Indian River near She Tiger Creek, May 26, 2017, verified by J Betts Acanthopterygii: Per ciformes Caranjidae Caranx sp. Jackfish 1 specimen from Indian River near Long Falls Creek, Mar. 13, 2018, verified by A Angulo Sibaja, UCR Haemulidae Pomadasys sp. Droma 1 specimen from Indian River near Guinea Creek, Feb. 21, 2018, verified by A Angul o Sibaja, UCR Cichlidae Amatitlania nig rofasciata Contrayat 0 6 0 2 0 1 1 1 17 25 24 6 3 6 3 Amatitlania septemfasciata Contrayat 1 2 31 1 5 12 6 14 0 0 0 0 0 0 0 Amphilophus citronellus NA 1 specimen from Indian River near Guinea Creek, Feb. 25, 201 8, verified by A Angulo Sibaja, UCR Cribroheros alfari Shine - Thru 9 32 32 6 7 16 18 10 9 33 11 2 0 2 0 Cribroheros rostratus Shine - Thru 1 specimen from Indian River near Long Falls Creek, Feb. 19, 2018, verified by J San Gil, UCR 91 Hypsophrys nicaraguensis NA 1 specimen from Indian River near Long Falls Creek, Feb. 19, 2018, verified by A Angulo Sibaja, UCR Neetroplus nematopus Contrayat 0 9 0 2 0 0 0 0 0 0 0 0 0 0 0 Parachromis dovii Sasin 13 9 0 9 0 2 8 8 4 13 8 4 0 1 1 P arachromis loisellei Sasin 1 specimen from Indian River near Guinea Creek, Feb. 21, 2018, verified by J San Gil, UCR Tomocichla tuba Moga 0 8 0 9 0 0 0 0 0 7 0 0 10 6 P Vieja maculicauda Tuba 1 specimen from Corn River near Chirripo Creek, Apr. 8, 2018 , verified by J Betts in field Gobiidae Awaous banana NA 0 1 0 0 0 0 P P 0 1 1 0 1 0 0 Sicydium altum NA 0 1 0 1 0 0 0 0 0 0 0 0 0 0 0 Eleotridae Eleotris pisonis Elik 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 Gobiomorus dormitor Elik 0 1 0 2 0 3 P 2 1 3 1 0 0 1 1 Fish and Shrimp Community Metrics Fish Taxa Richness 7 14 8 12 6 10 13 10 8 11 11 7 7 10 Cichlid Abundance 23 64 63 29 12 31 33 33 30 78 43 12 13 15 Shrimp Abundance 8 16 8 8 4 9 10 10 10 9 0 3 1 4 92 APPENDIX B Additional statistics an d graphs. Table B.1: Correlations of taxa response metrics. Invert. Taxa Richness BMWP Score Invert. Density (LN) Invert. Diversity (H) Invert. Evenness (EH) Fish Taxa Richness Cichli d Abundance (LN) Shrimp Abundance Invert. Taxa Richness R - 0.839 0.787 0.034 - 0.417 0.060 0.324 0.250 P 0.000 0.000 0.905 0.122 0.831 0.238 0.368 BMWP Score R 0.839 - 0.674 - 0.156 - 0.53 0.117 0.325 0.326 P 0.000 0.006 0.578 0.042 0.678 0.238 0.235 Invert. Density (LN) R 0.787 0.674 - - 0.266 - 0.589 0.117 0.333 0.439 P 0.000 0.006 0.338 0.021 0.678 0.225 0.101 Invert. Diversity (H) R 0.034 - 0.156 - 0.266 - 0.893 - 0.141 - 0.122 - 0.173 P 0.905 0.578 0.338 0.000 0.616 0.665 0.538 Invert. Evenness (EH) R - 0.417 - 0.53 - 0.589 0.893 - - 0.178 - 0.279 - 0.278 P 0.122 0.042 0.021 0.000 0.526 0.314 0.315 Fish Taxa Richness R 0.060 0.117 0.117 - 0.141 - 0.178 - 0.632 0.575 P 0.831 0.678 0.678 0.616 0.526 0.011 0.025 Cichlid Abundance (LN) R 0.324 0.325 0.333 - 0.122 - 0.279 0.632 - 0.635 P 0.238 0.238 0.225 0.665 0.314 0.011 0.011 Shrimp Abundance R 0.250 0.326 0.439 - 0.173 - 0.278 0.575 0.635 - P 0.368 0.235 0.101 0.538 0.315 0.025 0.011 93 Figure B.1: Non - metric multidimensional scaling ordin ation plots of macroinvertebrate community matrix (taxa densities by reach). Axis combinations not feature d in the main text are visualized. Polygons show watershed groupings. According to PERMANOVA, reaches group significantly as forested (Indian, N=5 and Corn, N=3) and deforested (Pijibaye, N=3 and Kukra, N=4) (F=1.88, p=0.0317). Reaches group with higher sig nificance by watershed (F=2.445, p=0.0001). A) Ordination plot of Axis 1 and 3 with reaches visualized. B) Ordination plot of Axis 1 and 3 with taxa visualized. A ) B ) 94 C) Ordination plot of Axis 1 and 3 with reaches visualized. D) Ordination plot of Axis 1 and 3 with taxa visualized. C ) D ) 95 Figure B.2: Habitat metrics for two forested watersheds a recently and less recently deforested watershed. A) Stream size (reach volume in M 3 ). B) % Pool. C) % Fines. D) % Sand. Reach values represented by points. See methods for calculations of values (Most values based on mean of 11 transects and associated subsamples). Mann - Whitney U tests were carried out lumping forested and deforested reaches . Kruskal - Wallis and pairwise Mann - Whitney U nonparametric tests were run between each watershed (see Table 4). Letters represent No letters implies no significance. A) B) C ) D ) 96 E) Geometric mean substrate size. F) Embeddedness in riffles and rapids G) Standard deviation of embeddedness. H) Proportion of stream bank eroded. E ) G ) H ) F ) 97 Figure B.2 ( I) Conductivity (µS) J) Periphytic macroalgae % cover. K) Macrophytes % cover . L) Instream live trees and roots % cover A B B 0 0 0 0 0 0 0 0 0 0 B I ) J ) K ) L ) A A B A B B A B A A B B 98 M) Overhanging Vege tation % Cover N) Small Woody and Leafy Debris % Cover. O) Large Wood Volume per 100 M . P) Habitat Complexity. B AB AB A M ) N ) O ) P ) A A B A B B B AB AB A 99 Q) Upper canopy large trees % cover . R) Mid - canopy plant % cover . S) % Shade (densiometer). T) Riparian condition index. Q ) R ) S ) T ) B AB AB A B 1 1 1 1 1 1 1 1 1 1 1 1 AB B A B AB AB A B 1 1 1 1 1 1 1 1 1 1 1 1 AB B A 100 REFERENCES 101 REFERENCES Allan, J. D. (2004). Landscapes and riverscapes: the influence of land use on stream ecosystems. Annual Review of Ecology, Evoluti on, and Systematics , 35, 257 - 284. Alvarez, R.M. (2016). Cada año hay menos bosques en Nicaragua. Managua, Nicaragua: La Prensa. Anderson, M. J. (2014). Permutational multivariate analysis of variance (PERMANOVA). Wiley statsref: statistics reference onli ne , 1 - 15. Anderson, E. P., Pringle, C. M., & Rojas, M. (2006). Transforming tropical rivers: an environmental perspective on hydropower development in Co sta Rica. Aquatic Conservation: Marine and Freshwater Ecosystems , 16(7), 679 - 693. Angulo Sibaja, A., Bussing, W., Garita - Alvarado, C., & López, M. (2013). Annotated checklist of the freshwater fishes of continental and insular Costa Rica: additions and nomenclatural revisions. Check List , 9, 987. Benchetrit, J., & McCleave, J. D. (2015). Current and hist orical distribution of the American eel Anguilla rostrata in the countries and territories of the Wider Caribbean. ICES Journal of Marine Science , 73 (1), 122 - 134. Benstead, J. P., Douglas, M. M., & Pringle, C. M. (2003). Relationships of stream invertebra te communities to deforestation in eastern Madagascar. Ec ological Applications , 13( 5), 1473 - 1490. Bilby, R. E., & Likens, G. E. (1980). Importance of organic debris dams in the structure and function of stream ecosystems. Ecology , 61( 5), 1107 - 1113. Bojse n, B. H., & Barriga , R . ( 2002 ) . Effects of deforestation on fish community structure in Ecuadorian Amazon streams. Freshwater Biology 47 , 2246 2260. Bojsen, B. H., & Jacobsen, D. (2003). Effects of deforestation on macroinvertebrate diversity and assembla ge structure in Ecuadorian Amazon streams. Archiv für Hydrobiologie , 158(3) , 317 - 342. Bonan, G. B. (2008). Forests and climate change: forcings, feedbacks, and the climate benefits of forests. S cience , 320(588 2), 1444 - 1449. Bradshaw, C. J., Sodhi, N. S., P eh , K.S. H., & Brook, B. W. (2007). Global evidence that deforestation amplifies flood risk and severity in the developing world. Global Change Biology , 13(11 ), 2379 - 2395. 102 L. (2018). Threshold responses of Amazonian stream fishes to timing and extent of deforestation. Conservation B iology , 32( 4), 860 - 871. Brito, J. G., Martins, R. T., Oliveira, V. C., Hamada, N., Nessimian, J. L., Hughes, R. M., Ferraz, S. F. , & de Paula, F.R. ( 2018 ) . Biological indicators of diversity in tropical streams: Congruence in the similarity of invertebrate assemblages. Ecological Indicators , 85, 85 - 92. Capps, K. A., & Flecker, A. S. (2015). High impact of low - trophic - position invaders: nonnative grazers alter the quality and quantity of basal food resources. Freshwater Science , 34(2 ), 784 - 796. into tropical streams: the importance of intact riparian forest and influences on fish diet s. Marine and Freshwater Research, 59(8), 653 - 660. Chaves, J., Neill, C., Germer, S., Neto, S. G., Krusche, A., & Elsenbeer, H. (2008). Land management impacts on runoff sources in small Amazon watersheds. Hydrological Processes: An International Jour nal , 22(12), 1766 - 1775. Chellaiah, D., & Yule, C. M. (2018). Riparian buffers mitigate impacts of oil palm plantations on aquatic macroinvertebrate community structure in tropical streams of Borneo. Ecological indicators , 95, 53 - 62. Chen, K., Hughes, R. M., Brito, J. G., Leal, C. G., Leitão, R. P., de Oliveira - Júnior, J. M., de Oliveira, V. C., Dias - Silva, K., Ferraz, S. F., Ferreira, J. , & Hamada, N. ( 2017 ) . A multi - assemblage, multi - metric biological condition index for eastern Amazonia streams. Ecological indicators , 78, 48 - 61. Coe, M. T., Costa, M. H., & Soares - Filho, B. S. (2009). The influence of historical and potential future deforestation on the stream flow of the Amazon River Land surface processes and atmospheric feedbacks. Journal of Hydrology , 3 69( 1 - 2), 165 - 174. Corea, J. T., Hernández, G. M., Solís, V., & Aguilar, A. J. (2014). Distribución y abundancia de peces de la familia Loricariidae (Pleco ) y su relación con los peces de interés comercial en los alrededores de la Isla de Ometepe. Encuentr o , 98, 44 - 59. Cornejo, R.A. ( 2010 ) . Macroinvertebrados acuáticos bioindicadores de la calidad del agua en Panamá: Propuesta de Índice BMWP/PAN . Villahermosa, Tabasco, México : Reunión especial , Macroinvertebrados dulceacuícolas en Mesoamérica (MADMESO) . C ovich, A. P., Crowl, T. A., & Heartsill - Scalley, T. (2006). Effects of drought and hurricane disturbances on headwater distributions of palaemoni d river shrimp ( Macrobrachium spp.) in the Luquillo Mountains, Puerto Rico. Journal of the North American Benth ological Society , 25( 1), 99 - 107. 103 d a Silva Gonçalves, C., de Souza Braga, F. M., & Casatti, L. (2018). Trophic structure of coastal freshwater stream fishes from an Atlantic rainforest: evidence of the importance of protected and forest - covered areas to fi sh diet. Environmental Biology of Fishes , 101 (6), 933 - 948. , X., Olden, J. D., Becker, F. G., Guimarães, T. D. F., & Melo, A. S. in South B razilian grasslands. Freshwater Biology , 61(11), 1921 - 1934. Dans, A . , G. Luna, & Jordan , C. A . (2015). Estado poblacional del Almendro como indicador de la disponibilidad de hábitat de Lapa Verde en el Sureste de Nicaragua. Ciencia e Interculturalidad , 15 (2), 114 - 124. De Cáceres, M ., & Legendre, P. ( 2009 ) . Associations between species and groups of sites: indices and statistical inference. Ecology , 90(12): 3566 - 3574. De Paula, F. R., de Barros Ferraz, S. F., Gerhard, P., Vettorazzi, C. A., & Ferreira, A. (2011). Large woody debris input and its influence on channel structure in agricultural lands of Southeast Brazil. Environmental management , 48( 4), 750. Dodds, W. K., Gido, K., Whiles, M. R., Daniels, M. D., & Grudzinski, B. P. (2014). The stream biome g radient concept: factors controlling lotic systems across broad biogeographic scales. Freshwater Science , 34(1), 1 - 19. Domínguez, E., & Fernández, H. R. (2009). Macroinvertebrados bentónicos sudamericanos: sistemática y biología . Tucumán , Argentina : Funda ción Miguel Lillo. Dudgeon, D., Arthington, A. H., Gessner, M. O., Kawabata, Z. I., Knowler, D. J., Lévêque, C., Naiman, R. J., Prieur - Richard, A. H., Soto, D., Stiassny, M. L. , & Sullivan, C. A. ( 2006 ) . Freshwater biodiversity: importance, threats, statu s and conservation challenges. Biological R eviews , 81(2), 163 - 182 . Dufrêne , M., & Legendre, P. (1997). Species assemblages and indicator species: the need for a flexible asymmetrical approach. Ecological M onographs, 67(3), 345 - 366. Encalada, A. C., Calle s, J., Ferreira, V., Canhoto, C. M., & Graca, M. A. (2010). Riparian land u se and the relationship between the benthos and litter decomposition in tropical montane streams. Freshwater Biology , 55(8), 1719 - 1733. Faith, D. P., Minchin, P. R., & Belbin, L. ( 1987). Compositional dissimilarity as a robust measure of ecological distance. Vegetatio , 69(1 - 3), 57 - 68. Fausch, K. D., Torgersen, C. E., Baxter, C. V., & Li, H. W. (2002). Landscapes to riverscapes: bridging t he gap between research and conservation of stream fishes: A continuous view of the river is needed to understand how processes interacting among scales set the context for stream fishes and their habitat. BioScience , 52( 6), 483 - 498. 104 Feijó - Lima, R., Mcleay, S. M., Silva - Junior, E. F., Tromboni, F., Moulton, T. P., Zandonà, E., & Thomas, S. A. (2018). Quantitatively describing the downstream effects of an abrupt land cover transition: buffering effects of a forest remnant on a stream impacte d by cattle grazing. Inland Waters , 8(3), 294 - 311. Fenoglio , S., Badino , G. , & Bona , F . (2002). Benthic macroinvertebrate communities as indicators of river environment quality: an experience in Nicaragua. Revista de Biología Tropical, 50(3 4), 1125 1131. Ferraro, P. J. (1994). Natural resource use in the southe astern rain forests of Madagascar and the local impacts of establishing the Ranomafana National Park. , North Carolina. Ferreira, W. R., Ligeiro , R., Macedo, D. R., Hughes, R. M., Kaufmann, P. R., Oliveira, L. G., & Calli sto, M. (2014). Importance of environmental factors for the richness and distribution of benthic macroinvertebrates in tropical headwater streams. Freshwater Science , 33( 3), 860 - 8 71. Foley, J. A., Asner, G. P., Costa, M. H., Coe, M. T., DeFries, R., Gibbs , H. K., Howard, E. A., Olson, S., Patz, J., Ramankutty, N. & Snyder, P. ( 2007 ) . Amazonia revealed: forest degradation and loss of ecosystem goods and services in the Amazon Basin. Frontiers in Ecology and the Environment , 5(1 ), 25 - 32. Fugère, V., Kasanga ki, A., & Chapman, L. J. (2016). Land use changes in an afrotropical biodiversity hotspot affect stream alpha and beta diversity. Ecosphere , 7(6 ), e01355. Gergel, S. E., Turner, M. G., Miller, J. R., Melack, J. M., & Stanley, E. H. (2002). Landscape indic ators of human impacts to riverine systems. Aquatic S ciences , 64( 2), 118 - 128. Gobierno Territorial Rama y Kriol . (2018). Action Plan: Protection and sustainable use of the Indio - Maíz Biological Reserve within the Rama and Kriol Territory and adjacent are as. Bluefields, Nicaragua : Gobierno Territorial Rama y Kriol. González, N., Mateo, S. S., & Valdivia, Á. M. (2013) . Macroinvertebrados bentónicos como bioindicadores de calidad de agua del trópico húmedo en las microcuencas de los alrededores de Bluefield s, RAAS. Wani , 68, 53 - 63. Griscom, H. P., Griscom , B. W., & Ashton, M. S. (2009). Forest regeneration from pasture in the dry tropics of Panama: effects of cattle, exotic grass, and forested riparia. Restoration Ecology , 17(1 ), 117 - 126. Gutiérrez - Fonseca , P. E., Ramírez, A., & Pringle, C. M. (2018). Large - scale climatic phenomena drive fluctuations in macroinvertebrate assemblages in lowland tropical streams, Costa Rica: The importance of ENSO events in determining long - term (15y) pattern s. P LO S ONE , 13( 2 ), e0191781. 105 Gutiérrez - Gregoric, D. E., & Vogler, R. E. (2010). Riesgo de establecimiento del gasterópodo dulceacuícola invasor Melanoides tuberculatus (Thiaridae) en el Río de la Plata (Argentina - Uruguay ). Revista M exicana de B iodiversidad , 81(2), 573 - 57 7. Hammer, O. , Harper, D. A. T., & Ryan, P. D. ( 2001 ) . PAST: Palaeontological Statistics software package for education and data analysis. Palaeontologia Electronica , 4(1) , 9 Hansen, M. C., Potapov , P. V. , Moore , R. , Hancher , M. , Turubanova , S. A. , Tyuk avina , A. , & Kommareddy , A. (2013). High - resolution global maps of 21st - century forest cover change. Science , 342(6160), 850 - 853. Harding, J. S., Benfield, E. F., Bolstad, P. V., Helfman, G. S., & Jones, E. B. D. (1998). Stream biodiversity: the ghost of land use past. Proceedings of the N ational A cademy of S ciences , 95(25 ), 14843 - 14847. Canal zone. Conservation Biology , 31(1 ), 86 - 95. Hawkes, H. A. (1998). Origin and development of the biological monitoring working party score system. Water Resear ch , 32( 3), 964 - 968. Hawkins, C. P., Norris, R. H., Gerritsen, J., Hughes, R. M., Jackson, S. K., Johnson, R. K., & Stevenson, R. J. (2000). Evaluation of the use of landscape classifications for the prediction of freshwater biota: synthesis and recommenda tions. Journal of the North American Benthological Society , 19(3), 541 - 556. Heartsill - Scalley, T., & Aide, T. M. (2003). Riparian vegetation and stream condition in a tropical agriculture secondary forest mosaic. Ecological Applications , 13( 1), 225 - 234. Helson , J. E., & Williams, D. D. (2013). Development of a macroinvertebrate multimetric index for the assessment of low - land streams in the neotropics. Ecological Indicators , 29, 167 - 178. Hetrick, N. J., & Bromaghin , J. F. (2006). Sampling b ias of h l ine g ear u sed to c apture r ainbow t rout in Gertrude Creek, Alaska. North American Jo urnal of F isheries M anagement , 26(1 ), 13 - 23. Hilsenhoff, W. L. (1988). Rapid field assessment of organic pollution with a family - level biotic index. Journal of the North American Benthological Society , 7(1 ), 65 - 68. Hughes, R. M., & Peck, D. V. (2008). Acquiring data for large aquatic resource surveys: the art of compromise among science, logistics, and reality. Journal of the North American Benthological Society , 27( 4), 837 - 859. IBM SPS S . 2019. IBM SPSS Statistics for W indows, v ersion 26.0. IBM Corp., Armonk, NY . 106 Iñiguez Armijos, C., Leiva, A., Frede, H. G., Hampel , H., & Breuer, L. (2014). Deforestation and benthic indicators: how much vegetation cover is needed to sus tain healthy Andean streams? PLOS ONE , 9( 8), e105869. Iwata, T., Nakano , S. , & Inoue , M. ( 2003 ) . Impacts of past Riparian Deforestation on Stream Communities in a Tropical Rain Forest in Borneo. Ecological Applications 13(2): 461 473. Jordan, C. A. (2015 ). The dynamics of wildlife and environmental knowledge in a bioculturally diverse coupled natural and human system in the Caribbean region of Nicaragua (PhD thesis). Michigan State University , Michigan. Jordan, C. A., Galeano, M. R., & Alonzo, A. S. (201 4). La cacería histórica de tapires Centroamericanos (Tapirus bairdii) en la RAAS, Nicaragua. Estud ios Ambient ales , 1(1), 73 - 87. Jordan, C. A., Schank, C. J., Urquhart, G. R., & Dans, A. J. (2016). Terrestrial m ammal o ccupancy in the c ontext of w idespread f orest l oss and a p roposed i nteroceanic c anal in Nicaragua's d ecreasingly r emote s outh Caribbean r egion. PLOS ONE, 11(3), e0151372. Jordan, C. A., Stevens, K. J., Urquhart, G. R., Kramer, D . B., & Roe, K. (2010). A new record of Ta pirus bai rdii in Nicaragua and potential implications. Tapir Conservation Newsletter , 19( 1), 11 - 15. Jost, L. (2006). Entropy and diversity. Oikos , 113( 2), 363 - 375. Kaufmann, P. R., Faustini, J. M., Larsen, D. P., & Shirazi, M. A. (2008). A roughness - corrected ind ex of relative bed stability for regional stream surveys. Geomorphology , 99( 1), 150 - 170. Kaufmann, P. R., Larsen, D. P., & Faustini, J. M. (2009). Bed stability and sedimentation associated with human disturbances in Pacific Northwest streams . Journal of the American Water Resources Association , 45(2), 434 - 459. Kaufmann, P.R., Levine, P., Robison, E.G., Seeliger, C., Peck, D.V. ( 1999 ) . Quantifying Physical Habitat in Wadeable Streams. Washington, DC : US Environmental Protection Agency . Kumar, A., Colton, M. B. M., Springer, M., & Trama, F. A. (2013). Macroinvertebrate communities as bioindicators of water quality in conventional and organic irrigated rice fields in Guanacaste, Costa Rica. Ecological Indi cators , 29, 68 - 78. Leal, C. G., Pompeu, P. S., Gardner, T. A., Leitão, R. P., Hughes, R. M., Kaufmann, P. R., Zuanon, J., de Paula, F. R., Ferraz, S. F., Thomson, J. R. , & Mac Nally, R. ( 2016 ) . Multi - scale assessment of human - induced changes to Amazonian instream habitats. Landscape Ecology , 31(8), 1725 - 1745. Leitão, R. P., Zuanon, J., Mouillot, D., Leal, C. G., Hughes, R. M., Kaufmann, P. R., Villéger, S., Pompeu, P. S., Kasper, D., de Paula, F. R. , & Ferraz, S. F. ( 2017 ) . Disentangling the pathways of 107 land use impacts on the functional structure of fish assemblages in Amazon streams. Ecography . 41(1), 219 - 232. Ligeiro, R., Hughes, R. M., Kaufmann, P. R., Macedo, D. R., Firmiano, K. R., Ferreira, W. R., Oliveira, D., Melo, A. S. , & Callisto, M. ( 2013 ) . Defining quantitative stream disturbance gradi ents and the additive role of habitat variation to explain macroinvertebrate taxa richness. Ecological I ndicators , 25 , 45 - 57. Lobón - cerviá, J., Mazzoni , R. , & Rezende , C.F . ( 2016 ) . Effects of riparian forest removal on the trophic dynamics of a Neotropica l stream fish assemblage. Journal of Fish Biology , 89(1): 50 - 64. Lorion, C.M. , & Kennedy, B. P . (2009). Relationships between deforestation, riparian forest buffers and benthic macroinvertebrates in neotropical headwater streams. Freshwater B iology , 54.1: 1 65 - 180. Lorion, C. M., & Kennedy, B. P. (2009). Riparian forest buffers mitigate the effects of deforestation on fish assemblages in tropical headwater streams. Ecological Applications , 19(2), 468 - 479. Luke, S. H., Slade, E. M., Gray, C. L., Annammala, K . V., Drewer, J., Williamson, J., Agama, A. L., Ationg, M., Mitchell , S. L., Vairappan, C. S. , & Struebig, M. J. ( 2019 ) . Riparian buffers in tropical agriculture: Scientific support, effectiveness and directions for policy. Journal of Applied Ecology , 56(1 ), 85 - 92. Macedo, D. R., Hughes, R. M., Kaufmann, P. R., & Callisto, M. (2018). Development and validation of an environmental fragility index (EFI) for the neotropical savannah biome. Science of The Total Environment , 635, 1267 - 1279 . Maes J. M., & Salvati erra - Suare z, T. (2014). Nicaragua. Alonso - EguíaLis, P., Mora, J. M., Campbell, B., & Springer, M. (Eds.). Diversidad, conservación y uso de los macroinvertebrados dulceacuícolas de México, Centroamérica, Colombia, Cuba y Puerto Rico . Querétaro , Mexico: Ins tituto Mexicano de Tecnología del Agua, Universidad Autónoma de Querétaro geomorphic state in relation to ecological integrity: evidence using habitat assessments an d macroinvertebrates. Environmental Management , 34(5 ), 669 - 683. McCullough, W. F. (1984). Sand - gauge. McGarigal, K., Cushman, S. A., & Stafford, S. (2013). Multivariate statistics for wildlife and ecology research . Berlin, Germany: Springer Science & Business Media. Merritt, R. W. , & Cummins, K. W. (1996). An introduction to the a quatic insects of North America . Dubuque, Iowa: Kendall Hunt. 108 Miyake, Y., Hiura, T., & Nakano, S. (2005). Effects of frequent streambed disturbance on the diversity of stream invertebrates. Archiv für Hydrobiologie , 162( 4), 465 - 480. Molina, M. C., Roa - Fuen tes, C. A., Zeni, J. O., & Casatti, L. (2017). The effects of land use at different spatial scales on instream features in agricultural streams. Limnologica , 65, 14 - 21. Montag, L. F., Leão, H., Benone, N. L., Monteiro - Júnior, C. S., Faria, A. P. J., Nicaci o, G., Ferreira, C. P., Garcia, D. H., Santos, C. R., Pompeu, P. S. , & Winemiller, K. O. ( 2019 ) . Contrasting associations between habitat conditions and stream aquatic biodiversity in a forest reserve and its surrounding area in the Eastern Amazon. Hydrobi ologia , 826(1 ), 263 - 277. diversity of species assemblages of cichlid fishes in a tropical floodplain river. Ecology of Freshwater Fish , 19( 2), 216 - 227. Montgomery, D . R., & Buffington, J. M. (1997). Channel - reach morphology in mountain drainage basins. Geological Society of America Bulletin , 109(5), 596 - 611. Mori, G. B., de Paula, F. R., de Barros Ferraz, S. F., Camargo, A. F. M., & Martinelli, L. A. (2015). Influenc e of landscape properties on stream water quality in agricultural catchments in Southeastern Brazil. Annales de Limnologie - International Journal of Limnology , 51 (1), 11 - 21 . Myre, E., & Shaw, R. (2006). The turbidity tube: simple and accurate measurement o f turbidity in the field. Houghton, Michigan: Michigan Technological University . Nakano, S., Miyasaka, H., & Kuhara , N. (1999). Terrestrial aquatic linkages: riparian arthropod inputs alter trophic cascades in a stream food web. Ecology , 80( 7), 2435 - 2441. Naranjo, C., Garcés, G., González, D., Brandimarte, A., Muñoz, S. y Musle, Y. ( 2005 ) . Una metodología rápida y de fácil aplicación para la evaluación de la calidad del agua utilizando el índice BMWPCub para ríos cubanos. Tecnura , 17 , 65 76. Nico, L. G. (2010). Nocturnal and diurnal activity of armored suckermouth catfish (Loricariidae: Pterygoplichthys) associated with wintering Florida manatees ( Trichechus manatus latirostris ). Neotropical Ichthyology , 8 (4), 893 - 898. , M., & Kell y - Quinn, M. (2015). Nutrient - and sediment - induced macroinvertebrate drift in Honduran cloud forest streams. Hydrobiologia , 758( 1), 75 - 86. - hUallacháin, D . ( 2018 ) . Impact of Cattle Access to Watercourses: Literature Review on Behalf of the COSAINT Project. Washington, DC: US Environmental Protection Agency. Oksanen, J., Blanchet, F. G., Kindt, R., Legendre, P., Minchin, P. B., Simpson, G. L., Solymos, P., Stevens, M. H. H., Wagner, H. , & Oksanen, M. J. ( 2013 ) . Community ecology package, version 2(9). 109 Organización de los Estados Americanos. (2005). Estudio de diagnóstico de la cuenca del río San Juan y lineamientos del plan de acción. Washington, DC: Organization of American States. Paaby, P., Ramirez, A., & Pringle, C. M. (1998). The benthic macro invertebrate community in Caribbean Costa Rican streams and the effect of two sampling methods. Revista de B iologia T ropical , 46, 185 - 199. Pelicice, F. M., Vitule, J. R., Orsi, M. L., Lima Junior, D. P., Magalhães, A. L., Pompeu, P. S., Petrere Jr, M., & Agostinho, A. A. (2017). Neotropical freshwater fishes imperilled by unsustainable policies. Fish and fisheries , 18(6), 1119 - 1133 . Peña - Arancibia, J. L., Bruijnzeel, L. A., Mulligan, M., & van Dijk, A. I. (2019). Forests as - season flows across the tropics. Journal of Hydrology , 574, 946 - 963. Pérez, A. M., & López de la Fuente, A. (1993). Estado actual del conocimiento de la malacofauna continental de Nicaragua. Encuentro, 40, 23 - 38. Petriello, M. A., & Joslin, A. J. (2019). The e mbers of r adical e cology and r evolutionary i deology in Nicaragua's p rotests. Journal o f Latin American Geography , 18( 1), 203 - 209. Phang, S. C., Cooperman, M., Lynch, A. J., Steel, E. A., Elliott, V., Murchie, K. J., Cooke, S. J., Dowd, S. , &, Cowx, I. G. ( 2019 ) . Fishing for conservation of freshwater tropical fishes in the Anthropocene. Aq uatic Conservation: Marine and Freshwater Ecosystems , DOI: 10.1002/aqc.3080 . Phillips, L. T. (2017). Modeling the effects of subsistence livelihoods on mammalian occupancy and understanding the importance of values held by M iskito forest users for carnivo re conservation in working forests . Michigan State University , Michigan. R Core Team . 2016 . R: A language and environment for statistical computing. R Foundation for Statistical Computing , Vienna , Austria . Ramírez, A., & Gutiérrez - Fonsec a, P. E. (2014). Functional feeding groups of aquatic insect families in Latin America: a critical analysis and review of existing literature. Revista de Biología Tropical , 62, 155 - 167. Recha, J. W., Lehmann, J., Walter, M. T., Pell, A., Verchot, L., & Jo hnson, M. (2012). Stream discharge in tropical headwater catchments as a result of forest clearing and soil degradation. Earth Interactions , 16( 13), 1 - 18. Reid, A. J., Carlson, A. K., Creed, I. F., Eliason, E. J., Gell, P. A., Johnson, P. T., Kidd, K. A., MacCormack, T. J., Olden, J. D., Ormerod, S. J. , & Smol, J. P. ( 2018 ) . Emerging th reats and persistent conservation challenges for freshwater biodiversity. Biological Reviews , 94(3), 849 - 873. 110 Rican, O., Pialek, L., Dragova, K., & Novak, J. (2016). Divers ity and evolution of the Middle American cichlid fishes (Teleostei: Cichlidae) with revised classification. Vertebrate Zoology , 66( 1), 3 - 102. Rodriguez, E., Morris, C. S., & Belz, J. E. (2006). A global assessment of the SRTM performance. Photogrammetric Engineering & Remote Sensing , 72(3), 249 - 260. Roldán, G. (1988). Guía para el estudio de los macroinvertebrados acuáticos del Departamento de Antioquia . Medellín , Columbia: Universidad de Antioquia, Fondo FEN . Salvatierra, T. (2014). Macroinvertebrados Ac uáticos como indicadores de la calidad de las aguas en tres microcuencas en Tola, Rivas, Nicaragua. Revista Científica Agua y Conocimiento , 1( 1), 1 - 12. P., David, D. C., Paciogl u, O., Weiperth, A., Gál, B. , & Pîrvu, M. ( 2019 ) . Integrating catchment land cover data to remotely assess freshwater quality: a step forward in heterogeneity analysis of river networks. Aquatic Sciences , 81 (2), 26. Schwendel , A. C., Death, R. G., Fuller, I. C., & Joy, M. K. (2010). Linking disturbance and stream invertebrate communities: how best to measure bed stability. Journal of the North American Benthological Society , 30(1), 11 - 24. Sermeño, J.M., Serrano, L., Springer, M., Paniagua, M, R., Pérez, D ., Rivas, A . W., Menjivar, R, A., Bonilla, B, L., Carranza, F, A., Flores, J . M., Gonzáles, C., Gutiérrez, P,E., Monterrosa, A, J,. y Arias, A, Y. (2010). Determinación de la calidad ambiental de las aguas de los ríos de El Salvador, utilizando invertebrado s acuáticos: índice biológico a nivel de familias de invertebrados acuáticos en El Salvador (IBF - SV - 2010). San Salvador, El Salvador : Editorial Universitaria , Universidad de El Salvador . Siegloch , A. E., Schmitt, R., Spies, M., Petrucio, M., & Hernández, M. I. M. (2017). Effects of small changes in riparian forest complexity on aquatic insect bioindicators in Brazilian subtropical streams. Marine and Freshwater Research , 68(3 ), 519 - 527. Smith, S. A., & Bermingham, E. (2005). The biogeography of lower Meso american freshwater fishes. Journal of Biogeography , 32( 10), 1835 - 1854. Springer, M., Ramirez, A., & Hanson, P. (2010). Fresh water macroinvertebrates of Costa Rica I. Revista de B iologia T ropical , 58, V - XIII. Strand, M., & Merritt, R. W. (1999). Impacts of livestock grazing activities on stream insect communities and the riverine environment. American Entomologist , 45( 1), 13 - 29. Suga, C. M., & Tanaka, M. O. (2013). Influence of a forest remnant on macroinvertebrate communities in a degraded tropical str eam. Hydrobiologia , 703 (1), 203 - 213. 111 Tanaka, M. O., de Souza, A. L. T., Moschini , L. E., & de Oliveira, A. K. (2016). Influence of watershed land use and riparian characteristics on biological indicators of stream water quality in southeastern Brazil. Agr iculture, Ecosystems & Environment , 216, 333 - 339. Teresa, F. B., L. Casatti, & M.V. Cianciaruso. (2015). Functional differentiation between fish assemblages from forested and deforested streams. Neotropical Ichthyology , 13(2): 361 370. Terra, B. D. F., H ughes, R. M., & Araújo, F. G. (2016). Fish assembla ges in Atlantic Forest streams: the relative influence of local and catchment environments on taxonomic and functional species. Ecology of Freshwater Fish , 25( 4), 527 - 544. Townsend, C. R., Scarsbrook, M. R., & Dolédec, S. (1997). The intermediate disturbance hypothesis, refugia, and biodiversity in streams. Limnology and O ceanography , 42(5 ), 938 - 949. van Zyl, J., ( 2001 ) . The shuttle radar topography mission (SRTM): A breakthrough in remote sensing of topo graphy, Acta z , 48(5 12) , 559 565 Villa, J. (1982). Peces nicaragüenses de agua dulce . Managua , Nicaragua : Fondo de Promoción Cultural del Banco de América. Wantzen, K., & Mol, J. (2013). Soil erosion from agriculture and mining: a threat to tropical str eam ecosystems. Agriculture , 3( 4), 660 - 683. Warton, D. I., Wright, S. T., & Wang, Y. (2012). location and dispersion effects. Methods in Ecology and Evolution, 3(1), 89 - 101. Wickham, H. ( 2016 ) . ggplot2: Elegan t Graphics for Data Analysis. New York, New York.: Springer - Verlag . Wright, J. P. , & Flecker , A.S. ( 2004 ) . Deforesting the riverscape: the effects of wood on fish diversity in a Venezuelan piedmont stream. Biological Conservation , 120 , 439 - 447. Wright, S. J. (2005). Tropical forests in a changing environment. Trends in E cology & E volution , 20(10), 553 - 560 Zamora, H. ( 2007 ) . El Índice BMWP y la evaluación biológica de la calidad del agua en los ecosistemas acuáticos epicontinentales naturales de Colombia. Revista de la Asociación Colombiana de Ciencias Biológicas , 19 , 73 - 81. ( 2019). How deforestation drives stream habitat changes and the functional structure of fish assembl ages in different tropical regions. Aquatic Conservation: Marine and Freshwater Ecosystems . DOI: 10.1002/aqc.3128 .