USING CITIZEN SCIENCE TO DEVELOP MAST PRODUCTION INDICES IN MICHIGAN  By  Alexander Kyle Killion                      A THESIS  Submitted to  Michigan State University in partial fulfillment of the requirements for the degree of  Fisheries and Wildlife  Master of Science   2016        ABSTRACT USING CITIZEN SCIENCE TO DEVELOP MAST PRODUCTION INDICES IN MICHIGAN By  Alexander Kyle Killion  Variation in mast production influences wildlife populations. Annual mast production is logistically difficult to measure and thus robust yearly estimates generally do not exist. Hence, wildlife managers rarely have access to information on mast occurrence and production that is spatially extensive and temporally replicated. The purpose of my research was to assess the feasibility of using citizen scientists to collect reliable data that can be used by wildlife managers to produce annual mast production estimates throughout Michigan. In Chapter 1, I present the design, development, and recruitment of a citizen-science program called MI-MAST: Wildlife Food Tracker. Following field validation of volunteer submitted data form 2014-2015, I concluded that untrained volunteers were capable of contributing reliable data on mast occurrence and production. In Chapter 2, I use data collected in 2015 to demonstrate how hard- and soft-mast production indices could be generated using the citizen-science sampling protocol. I describe a process to assess the spatial scale at which mast production variability occurs and suggest how data should be collected and combined into ecoregional units to estimate mast production for wildlife planning. I conclude that using the model described herein, citizen science is capable of producing sustainable annual mast production indices that can likely improve population models used to manage wildlife species. I also recommend that increased communication with account holders and recruitment of new participants will be needed to produce robust annual indices throughout Michigan.iii  ACKNOWLEDGEMENTS   Funding for this research was provided by the Michigan Department of Natural Resources - Wildlife Division with funds from the federal Pittman-Robertson Wildlife Restoration Act grant administered by the U.S. Fish and Wildlife Service (Michigan's Statewide Wildlife Research, Surveys, and Monitoring Program (W-155-R)), Safari Club International Michigan Involvement Committee, and Michigan State University College of Agriculture and Natural Resources. This research was reviewed and approved by the Institutional Review Board (IRB x13-1141e; 044942).   I would like to thank my advisor, Gary Roloff, for his endless support, patience, and guidance during this research and other endeavors. I would also like to thank my committee members Henry (Rique) Campa, III, Scott Winterstein, and Sarah Mayhew for their support and involvement during the development of this program and throughout my research. I am grateful for the willingness of D. Youngdyke, A. Mitterling, M. Donovan, and K. Carter to assist in recruitment and marketing efforts and T. Dokes and T. Swem for data collection. I would like to thank A. Ziegler, G. Carballo, and C. Peterson for their patience and dedication in the development of program technologies. Lastly, thank you to all friends and family who supported my efforts during this time, particularly Ashlee Tiernan and Kelli, Lauren, and Calla Killion.          iv TABLE OF CONTENTS  LIST OF TABLES                  vi LIST OF FIGURES                 vii INTRODUCTION                    1 LITERATURE CITED                              3 CHAPTER 1                     7 DESIGNING, IMPLEMENTING, AND VALIDATING A CITIZEN-SCIENCE PROGRAM TO QUANTIFY MAST PRODUCTION IN MICHIGAN                                                          7   Abstract                       7 Key Words                                                                                                          8 1.1. Introduction                   8 1.2. Methods                  11 1.2.1. Collaborative Structure              11 1.2.2. Mast Sampling Protocol               13 1.2.3. Participant Recruitment               16 1.2.4. Streamlining Participant Interactions with MI-MAST                     16 1.2.5. Identifying Plant Species               17 1.2.6. Data Validation                                                           18 1.3. Results                                                                                                                         18 1.4. Discussion                 20 1.5. Acknowledgements                26 APPENDIX                  27 LITERATURE CITED                35  CHAPTER 2                              44 USING CITIZEN-SCIENCE DATA TO DEVELOP ANNUAL MAST PRODUCTION INDICES IN MICHIGAN                 43             Abstract                  43        Kew Words                   44 2.1. Introduction                 44 2.2. Methods                  47  2.2.1. Mast Sampling Protocol              47  2.2.2. Regional Variation                          48  2.2.3. Within-region Variation              49  2.2.4. Using Ecoregions to Index Mast Production                       50  2.2.5. Mast Index Calculation                                    51 2.3. Results                             52 2.4. Discussion                                                  53 2.5. Acknowledgements                56v APPENDIX                  57 LITERATURE CITED                63 CONCLUSIONS                 67                         vi LIST OF TABLES  Table 1.1. Optional demographic questionnaire participants were prompted to complete when creating an account on www.mimast.org.                                                   28  Table 1.2. Hard and soft mast species and genera included in MI-MAST: Wildlife Food Tracker. Mast species were included based on intensity of use by large game animals in Michigan, USA, based on MIDNR biologist recommendations and Martin et al. 1951.                                                                                                  29  Table 2.1. Total mast observation amounts by species and genera in Michigan, USA,  2015.                                            58                                                      Table 2.2.  Average size (ha) of regional clusters classified by the K-means method for 2, 3, and 4 clusters. p- production variability among clusters.                                                  59  Table 2.3. Average size (ha) of within-region clusters and average distance (km) from centroids (when regions are classified into 3 clusters).           59  Table 2.4. Within-region clusters (when regions are classified into 3 clusters) classified by the K-means method for 2, 3, 4, 5 and 6 clusters. p-for mast production variability among clusters.             59             vii  LIST OF FIGURES  Figure 1.1.  MI-MAST: Wildlife Food Tracker logo.                         30  Figure 1.2.  Locations of recruitment events, volunteer data submissions, and webpage views (provided by Google Analytics) for MI-MAST: Wildlife Food Tracker in Michigan, USA, 2014-2015.                                                                                  31  Figure 1.3. Ages of MI-MAST: Wildlife Food Tracker account holders who completed the demographic questionnaire upon registration, 2014-2015.                                    32                                          Figure 1.4. Previous experience identifying plants of MI-MAST: Wildlife Food Tracker account holders who completed the demographic questionnaire upon registration, 2014-2015.                                                                                                             32  Figure 1.5. Age and previous experience identifying plants of MI-MAST: Wildlife Food Tracker account holders who completed the demographic questionnaire upon registration, 2014-2015.                                                                                         33  Figure 1.6. Primary outdoor interests of MI-MAST: Wildlife Food Tracker account holders who completed the demographic questionnaire upon registration, 2014-2015.    34                                Figure 2.1.  Mast production data used in cluster classification (Quercus bicolor, Q. alba, Q. palustris, Q. rubra, and Prunus serotina; n=704) and subsection ecoregional (n=22) in Michigan, USA, 2015.                                                                           60  Figure 2.2. Regional clusters of all oak (Quercus) observations and within-region clusters for regional cluster C, Michigan, USA, 2015.                                                             62      1  INTRODUCTION Hard and soft mast meet important dietary needs for many wildlife species, such as white-tailed deer (Odocoileus virginianus), black bear (Ursus americanus), and wild turkey (Meleagris gallopavo; Feldhamer 2002, Steffen et al. 2002, Ryan et al. 2004, Reynolds-Hogland et al. 2007). Hard mast is defined as a fruit with a hard exterior, and soft mast as a fruit with a fleshy exterior. Individual trees and shrubs produce varying amounts of mast during any given year (Kelly and Sork 2002). This irregular pattern of production is referred to as masting and can directly affect populations of wildlife that rely on these food sources (Martin et al. 1951, Elowe and Dodge 1989, Ostfeld et al. 1996, McShea 2000, Inman and Pelton 2002). Quantitative measures of annual food production can improve the utility of statistical models used to estimate wildlife populations and harvest limits (Malcolm and Van Deelen 2010, Bridges et al. 2011). However, mast is highly variable in space and time, and rigorous monitoring efforts are often arduous and outside the logistical capabilities of wildlife management agencies (Koenig and Knops 2000, Greenberg and Warburton 2007, McDonald-Madden et al. 2010). Studies have attempted to identify the mechanisms responsible for mast patterns (Silvertown 1980, Norton and Kelly 1988, Herrera et al. 1998, Levey and Benkman 1999, Kelly et al. 2001), but consensus has yet to be reached (Kelly and Sork 2002). Due to the complexity of masting, accurately estimating mast production requires that many plants be sampled annually. Citizen science, or the use of untrained volunteers to assist in scientific studies, is a field that has grown immensely in the past decade due largely to the increased accessibility of new technology for data collection and storage (Silvertown 2009). Despite several successful applications, critics remain skeptical about the reliability of data produced by citizen scientists and whether those data should be used to answer scientific questions and manage natural   2  resources (Nichols and Williams 2006, Szabo et al. 2012). Studies are continuing to advance the methods used in citizen science, and educational resources exist that promote best practices in program development. These best practices are intended to instill scientific integrity when using volunteers to collect data (Bonney et al. 2009, Pasrsons et al. 2011). Information derived from citizen scientists that may have otherwise gone unrecorded can assist in answering a new suite of scientific questions (Tulloch et al. 2013).  Citizen scientists are proving they can play a valuable and much needed role in ecological monitoring where data at large scales can be more efficiently collected (Tulloch and Szabo 2012, McKinley et al. 2015), expanding the outcomes from limited conservation budgets (Danielsen et al. 2005, Robertson et al. 2010, Aceves-Buneo et al. 2015).  The importance of mast to a variety of wildlife, and the fact that masting is temporally and spatially variable makes citizen science an appealing approach for collecting data on masting. No consistent, large-scale monitoring efforts on mast production existed in the Great Lakes Region when information on masting was requested by Michigan biologists to improve management of wildlife populations. I tested the feasibility of using citizen scientists to generate data on annual mast-production in Michigan. In Chapter 1, I describe the development of a citizen-science program (called MI-MAST: Wildlife Food Tracker) and test the accuracy of data submitted by volunteers. I also comment on techniques for engaging participants and marketing the program. In Chapter 2, I use data collected via the MI-MAST interface to demonstrate how hard- and soft-mast production indices can be generated in Michigan. The findings of my study can be used to guide the development, marketing, and use of a citizen science program for indexing mast production.        3                         LITERATURE CITED                         4  Aceves-Bueno, E., A. S. Adeleye, D. Bradley, W. T. Brandt, P. Callery, M. Feraud, K. L. Garner, R. Gentry, Y. Huang, I. McCullough, I. Pearlman, S. A. Sutherland, W. Wilkinson, Y. Yang, T. Zink, S. E. Anderson, and C. Tague. 2015. Citizen science as an approach for overcoming insufficient monitoring and inadequate stakeholder buy-in in adaptive management: criteria and evidence. Ecosystems 18:493-506.  Bonney, R., C. B. Cooper, J. Dickinson, S. Kelling, T. Phillips, K. V. Rosenberg, and J. Shirk. 2009. Citizen science: a developing tool for expanding science knowledge and scientific literacy. Bioscience 59:977-984.  Bridges, A. S., M. R. Vaughan, and J. A. Fox. 2011. Reproductive ecology of American black bears in the Alleghany Mountains of Virginia, USA.  Journal of Wildlife Management 75:1137-1144.  Danielsen, F., N. D. Burgess, and A. Balmford. 2005. Monitoring matters: examining the potential of locally-based approaches. Biodiversity and Conservation 14:2507-2542.  Elowe, K. D., and W. E. Dodge. 1989. Factors affecting black bear reproductive success and cub survival. Journal of Wildlife Management 53:962-968.  Feldhamer, G. A. 2002. Acorns and white-tailed deer: interrelationships in forest ecosystems. Pages 215223 in W. J. McShea and W. M. Healy, editors. Oak forest ecosystems: ecology and management for wildlife. The Johns Hopkins Press, Baltimore, Maryland, USA.   Greenberg, C. H., and G. S. Warburton. 2007. A rapid hard-mast index from acorn presence-absence tallies. Journal of Wildlife Management 71:1654-1661.  Herrera, C. M., P. Jordano, J. Guitián, and A. Traveset. 1998. Annual variability in seed production by woody plants and the masting concept: reassessment of principles and relationship to pollination and seed dispersal. The American Naturalist 152:576-594.  Inman, R. M., and M. R. Pelton. 2002. Energetic production by soft and hard mast foods of American black bears in the Smoky Mountains. Ursus 13:57-68.  Kelly, D., D. E. Hart, and R. B. Allen. 2001. Evaluating the wind pollination benefits of mast seeding. Ecology. 82:117-126.  Kelly, D., and V. L. Sork. 2002. Mast seeding in perennial plants: why, how, where? Annual Review of Ecology and Systematics 33:427-447.  Koenig, W. D., and J. M. Knops. 2000. Patterns of annual seed production by northern hemisphere trees: a global perspective. The American Naturalist. 155:59-69.  Levey, D. J., and C. W. Benkman. 1999. Fruitseed disperser interactions: timely insights from a long-term perspective. Trends in Ecology and Evolution. 14:41-43.   5   Malcolm, K. D., and T. R. Van Deelen. 2010. Effects of habitat and hunting framework on American black bear harvest structure in Wisconsin. Ursus 21:14-22.  Martin, A. C., H. S. Zim, and A. L. Nelson. 1951. American wildlife and plants: a guide to wildlife food habits. McGraw-Hill, New York, New York, USA.  McDonald-Madden, E., P. W. J. Baxter, R. A. Fuller, T. G. Martin, E. T. Game, J. Montambault, and H. P. Possingham. 2010. Monitoring does not always count. Trends in Ecology and Evolution 25:547550.  McKinley, D. C., A. J. Miller-Rushing, H. L. Ballard, R. E. Bonney, H. Brown, D. M. Evans, R. A. French, J. K. Parrish, T. B. Phillips, S. F. Ryan, L. A. Shanley, J. L. Shirk, K. F. Stepenuck, J. F. Weltzin, A. Wiggins, O. D. Boyle, R.D. Briggs, S. F. Chapin III, D. A. Hewitt, P. W. Preuss, and M. A. Soukup. 2015. Can investing in citizen science improve natural resource management and environmental protection? Issues in Ecology 19.  McShea, W.J. 2000. The influence of acorn crops on annual variation in rodent and bird populations. Ecology 81:228-238.  Nichols, J. D., and B. K. Williams. 2006. Monitoring for conservation. Trends in Ecology and Evolution 21:66873.  Norton, D. A., and D. Kelly. 1988. Mast seeding over 33 years by Dacrydium cupressinum Lamb. (rimu) (Podocarpaceae) in New Zealand: the importance of economies of scale. Functional Ecology 2:399-408.  Ostfeld, R. S., C. G. Jones, and J. O. Wolff. 1996. Of mice and mast.  BioScience 46:323-330.  Parsons, J., R. Lukyanenko, and Y. Wiersma. 2011. Easier citizen science is better. Nature 471: 37-37.  Reynolds-Hogland, M. J., L. B. Pacifici, and M. S. Mitchell. 2007. Linking resources with demography to understand resource limitation for bears. Journal of Applied Ecology 44:1166-1175.  Robertson, M. P., G. S. Cumming, and B. F. N. Erasmus. 2010. Getting the most out of atlas data. Diversity and Distributions 16:363-375.  Ryan, C. W., J. C. Pack, W.K. Igo, J.C. Rieffenberger, and A. B. Billings.  2004.  Relationship of mast production to big-game harvests in West Virginia. Wildlife Society Bulletin 32:786-794. Silvertown, J. W. 1980. The evolutionary ecology of mast seeding in trees. Biological Journal of the Linnean Society 14:235-250.    6  Silvertown, J. 2009. A new dawn for citizen science. Trends in Ecology and Evolution 24:467-471.  Steffen, D. E., N. W. Lafon, and G. W. Norman. 2002. Turkeys, acorns, and oaks. Pages 241-255 in M. J. McShea and W. M. Healy, editors. Oak forest ecosystems: ecology and management for wildlife. The Johns Hopkins Press, Baltimore, Maryland, USA.  Szabo, J. K., R. A. Fuller, and H. P. Possingham. 2012. A comparison of estimates of relative abundance from a weakly structured mass-participation bird atlas survey and a robustly designed monitoring scheme. Ibis 154:468-479.   Tulloch, A. I. T., H. P. Possingham, L. N. Joseph, J. Szabo, and T. G. Martin. 2013. Realising the full potential of citizen science monitoring programs. Biological Conservation 165:128-138.  Tulloch, A. I. T., and J. K. Szabo. 2012. A behavioural ecology approach to understand volunteer surveying for citizen science datasets. Emu 112.                                7  CHAPTER 1 DESIGNING, IMPLEMENTING, AND VALIDATING A CITIZEN-SCIENCE PROGRAM TO QUANTIFY MAST PRODUCTION IN MICHIGAN  Abstract  Citizen science is increasingly being used as a conservation tool by researchers and managers when extensive monitoring efforts are needed at large spatial or long temporal scales. Estimating annual spatiotemporal variations in hard- and soft-mast production in Michigan could improve wildlife population models currently used to set harvest limits. I investigated whether a citizen-science program relying on volunteers could produce reliable mast production indices. The program was designed with input from the Michigan Department of Natural Resources - Wildlife Division, Michigan State University, and Michigan United Conservation Clubs. I engaged multiple partners to maximize usefulness, effectiveness, and to help increase the likelihood that the program was sustainable. I developed a field sampling protocol for untrained volunteers that could produce accurate mast production indices. I also developed a website equipped with instructional materials, a data submission portal, personal account databases, and mobile phone applications. The focus of technological development was to use readily accessible and easy to use tools that reduced the effort required by MI-MAST participants. In 2014, MI-MAST was introduced to nearly 150 potential participants to test technology and capabilities of volunteers. In 2015, I worked with project partners to market MI-MAST to an additional ~500 potential participants throughout Michigan. In total, 273 people registered for a MI-MAST account and 30 people submitted data. On average, each person that submitted data recorded 4 mast observations. I field validated 60 mast submissions submitted by 15 participants to verify the reliability of data. Of the 60 sites I visited, 100% of the plants were correctly identified and 97% of records describing mast amounts were accurate. 73% of registered users listed hunting as   8  an activity they were interested in and frequent interactions with participants revealed that using the program tools to track mast production on their own property was a primary attraction for enrollment.  My results suggest that with enough participation MI-MAST can produce reliable mast production estimates to inform wildlife population models. The biggest challenge facing natural resource agencies wanting to use MI-MAST is in recruiting and retaining individuals to be able to develop mast production indices that are robust across large spatial and temporal extents. Key Words Acorns, berries, citizen science, fruits, mast production, Michigan, volunteers, wildlife food  1.1. Introduction   Hard and soft mast plays an important dietary role for many wildlife species, such as white-tailed deer (Odocoileus virginianus), black bear (Ursus americanus), and wild turkey (Meleagris gallopavo) (Feldhamer 2002, Steffen et al. 2002, Ryan et al. 2004, Reynolds-Hogland et al. 2007). Wildlife populations are affected by varying types and amounts of mast produced each year (Martin et al. 1951, Elowe and Dodge 1989, Ostfeld et al. 1996, McShea 2000, Inman and Pelton 2002). Human-wildlife conflict has also been linked to fluctuations in mast production (Ryan et al. 2007, Howe et al. 2010). Current models estimating wildlife populations used by managers to set harvest limits of game species that rely on mast could be improved with the addition of mast production data. However, quantifying mast production at scales that correspond to wildlife management units can be difficult because mast production exhibits high spatial and temporal variability (Kelly and Sork 2002). Various hypotheses have been proposed to explain masting patterns but no consensus has been reached (Silvertown 1980, Norton and Kelly 1988, Herrera et al. 1998, Levey and Benkman 1999, Kelly et al. 2001).   9  Extensive sampling is thus required to monitor mast production, but these monitoring requirements often exceed the logistic capabilities of wildlife management agencies. Recently the use of volunteers lacking formal scientific training to conduct ecological monitoring as a means to achieve research and management goals has increased (Bonney et al. 2009, Aceves-Bueno et al. 2015). However citizen science is not a new idea and was first conducted in the 1800s to report birds striking lighthouses, and was followed by the popular Christmas Bird Count in 1900 that still is conducted today (Droege 2007). Programs now exist in a variety of disciplines such as wildlife biology (Evans et al. 2005), water quality (Ohrel and Register 2006), astronomy (Raddick et al. 2010), criminal justice (Vitos et al. 2013), and oncology (Pharoah 2014).  Important large-scale ecological information that would otherwise go unrecorded due to lack of human and monetary resources or data collection restrictions on private lands is now being collected by citizen scientists (Danielsen et al. 2005, Dickinson et al. 2010). For example, using citizen-science data, researchers have investigated long-term abundance and occupancy trends of bird species (Webb at al. 2007), effects of forest fragmentation on wildlife populations (Villard et al. 1999), and the spread of invasive species and infectious disease (Crowl et al. 2008).  Participation in citizen-science programs increases public involvement in natural resource policy, appreciation of the outdoors, and ecological education (Ryan et al. 2001, Newman et al. 2003, Lawrence 2006, Aceves-Buneo et al. 2015). These results are often viewed as secondary to data collection, but often align with state and federal natural resource goals (Stepenuck 2013, McKinley et al. 2015). Citizen science can be an advantageous tool for certain ecological studies; however it has multiple limitations that must be recognized before collecting data to answer new questions. Sampling biases can occur in many forms and standardized sampling protocols may   10  not be appropriate for volunteers (Dickinson et al. 2010). Differences in the amount of time participants dedicate to sampling can result in a bias in detection probability (Link and Sauer 1999), location of sampling can create bias when data is recorded in an area that does not represent the larger landscape (Bart et al. 1995), and differences in the ability of participants to report accurate information can lead to data inconsistencies (Fitzpatrick et al. 2009).  Hence, a citizen science program should be easy to use, intuitive, require minimal economic or time investment by participants, and produce tangible benefits to participants. Developing a successful citizen-science program relies on skills and tools beyond those typically required of a standard scientific study, including broad understandings of institutional collaboration, scientific principles, and public interests (Penrose and Call 1995, Darwall and Dulvy 1996, Ryan et al. 2001, Bonney et al. 2009, Robertson et al. 2010, Rotman et al. 2012). To fulfill the need for mast production data in Michigan, I developed a citizen-science program (MI-MAST: Wildlife Food Tracker) that relies on volunteers to monitor and report annual mast production that can theoretically be used by biologists to improve wildlife population models. Drawing on previous citizen-science successes, my objectives were to 1) create a sustainable and productive collaboration between the Michigan Department of Natural Resources, Michigan State University, Michigan United Conservation Clubs, and other wildlife stakeholder groups, each supplying different expertise, needs, or access to potential volunteers, 2) develop a sampling protocol appropriate for untrained volunteers that produces reliable and usable data on mast production, and 3) understand participant motivations and interests to sustain recruitment and retention of volunteers.      11  1.2. Methods  1.2.1. Collaborative Structure  Citizen-science is capable of producing large quantities of accurate data and influencing policy and management decisions (Stepenuck 2013, Fuccillo et al. 2014). However, to effectively utilize the economic and scientific benefits of using a large network of volunteers, standard protocols must exist to recruit, train, and manage volunteers and to collect and compile their data (Bonney et al. 2009, Sullivan et al. 2009). To alleviate these demands from any one organization, strategic collaboration with diverse partners who each play a specific role can build the capacity of the program to grow and adapt. A common strategy in operating a citizen-science monitoring program is to create a semi-independent organization that may appear as its own entity but is operated by personnel from multiple organizations (Bonney et al. 2009, Silvertown 2009, Shirk et al. 2012).  For the program I developed, MI-MAST, an intentional collaboration was formed with three organizations, where each played unique roles in the authoritative center to share responsibility, maximize benefits for all parties, and ensure program sustainability. There are many types of organizational models that been developed to best achieve different types of goals, thus it is important to decide what goals the program intends to achieve and to make that decision process transparent. In my case, the Michigan Department of Natural Resources - Wildlife Division (MIDNR) identified a need to monitor mast production throughout Michigan. The MIDNR recognized that they could not produce this information internally at the spatial and temporal scales that would be meaningful to managers hence they invited the support of local people to contribute information. In this case, local people were not intended to take part in the decisions made with their contributed information, but were recruited knowing that their   12  contributions would be used in the best interest of conservation planning. This structure is best described as top-down participation, where an authoritative center of experts is requesting support from the public (Lawrence 2006). The primary goal of asking the public to monitor mast production was to produce an annual mast report that could be used by biologists to better manage the natural resources of Michigan. This differs from a bottom-up approach where local people identify a problem or question and enlist experts to help guide them through the scientific process (Lawrence 2006). Often these types of programs mimic activist behavior and occur on a smaller scale to produce information to guide local decision making.  Michigan State University researchers supplied expertise in developing and testing a scientifically rigorous sampling protocol that could be used by untrained volunteers. In return, Michigan State University was provided with new data that could be used in future research. To efficiently collect and store data from volunteers scattered throughout Michigan, experience with database management and recent technological developments was necessary. I relied on the Applied Spatial Ecology and Technical Services (ASETS) laboratory at Michigan State University to create and maintain technology for MI-MAST. The ASETS laboratory had experience in developing and managing citizen science web pages, mobile applications, and databases and was compensated for the tools they developed, routine system maintenance, and was given the opportunity to contribute to the advancement of technology that scientists could use for future studies.  Efficiently recruiting volunteers at a large scale requires partnerships with diverse organizations that frequently interact with the public (Bell et al. 2008). Michigan United Conservation Clubs (MUCC) is the largest statewide conservation organization in the United States and provides services to member-affiliated clubs with diverse outdoor interests (MUCC   13  2015). Partnering with one large umbrella organization, like MUCC, reduces complexity in the authoritative center to allow greater focus on primary goals of the program. MUCC routinely interacts with the portion of the public most interested in conservation topics and possessed the proper communication mechanisms to promote MI-MAST. Recruiting participants gave MUCC an opportunity to involve stakeholders in actionable science, improve education, and improve Michigan conservation.  1.2.2. Mast Sampling Protocol   One constraint of using citizen scientists to collect data is that participants commonly lack formal scientific training, and data can become unreliable if requested tasks are beyond the ability or interests of the volunteers (Penrose and Call 1995, Newman et al. 2003, Galloway et al. 2006). In some citizen-science programs, in-person or online training modules are required before volunteers can participate if certain technical skills are needed (Bloniarz and Ryan 1996).  Some participants may be interested in the opportunity to learn new things whereas others may not be willing to dedicate the necessary time to go through training (Parsons et al. 2011). Programs requiring little to no technical skills from their volunteers may be in a better position to safely assume that data are reliable; accurate data are capable of being collected regardless of previous training if simple instructions are followed. Data reliability is a concern for all scientific studies even when conducted by professionals and is not limited to citizen-science (Szabo et al. 2012). Citizen-science program designers must find a balance between asking volunteers to perform duties beyond their skill level and requesting over-simplified data that cannot properly answer the scientific questions for which the program was designed (Brandon et al. 2003). This process can result in sampling standards that are more rigorous than traditional studies because extra attention is given to observer error, which is a common concern in almost all ecological   14  studies (Sauer et al. 1994, Kendall et al. 1996).  Significant attention should also be given to specifying the required time commitments of volunteers when developing the sampling protocol. Higher effort requirements can lead to obtaining more data per participant, but the associated time commitments may initially deter contributors who would rather play a smaller role (Parsons et al. 2011, Silvertown et al. 2013).  I developed a sampling protocol for MI-MAST by first identifying the amount and type of data necessary to accurately estimate hard and soft mast production in Michigan. A successful target monitoring design is implemented based on a priori hypotheses (Nichols and Williams 2006), and often is capable of informing more management decisions than undirected monitoring approaches (McNie 2007, Sarewitz and Pielkejr 2007). Focusing on the capabilities of current scientific methods, rather than volunteer interests or capabilities, ensured that rigorous standards were used and that project goals could be met (Bonney et al. 2009).  Several, often labor-intensive, sampling strategies have long existed to estimate hard and soft mast production (Daubenmire 1959, Whitehead 1969). Recent improvements to these methods suggest estimates of hard mast production can be produced using short visual scans that are highly correlated to estimates from the previous labor-intensive methods (Greenberg and Warburton 2007). I designed the sampling protocol for hard mast by adopting methodologies described in Greenberg and Warburton (2007; described below). Less established protocols existed for soft mast production. Soft mast methods suitable for large areas typically involve the establishment of plots and counting of plant stems and fruits (Shaw et al. 2010, Lashley et al. 2011, 2014). I modified an approach used by the West Virginia Division of Natural Resources (Uhlig and Wilson 1952, Richmond et al. 2015), with the modification providing better alignment to the protocol used for hard mast indexing.    15  I asked users to identify mast producing plants in locations they frequently visited, report the amount of mast present once a year, and then return to that location in subsequent years to again report the amount of mast. Reporting on the same plants each year produces greater estimation accuracy because the index values are only relative to previous years (Greenberg and Warburton 2007). Participants recorded the amount of mast present on a plant by selecting one of three possible options. If reporting on hard mast, users described it qualitatively as 1) None <1/3 d o berries are uits The database for a participant that revisits plants is filled with the locations of their plants and a list of production amounts from previous years. The MI-MAST sampling protocol was based on scientifically credible processes (i.e., Uhlig and Wilson 1952, Greenberg and Warburton 2007, Richmond et al. 2015), while theoretically remaining simple enough to maximize the number of participants. I did not establish minimum participation requirements (e.g., minimum number of plants) and ensured that the sampling protocol would not require extensive training. Submitting a mast record the first time only takes a few minutes, and submitting subsequent observations on the same plant requires even less time for observing mast amounts and data entry. I also included an individual database that allowed participants to see their monitoring results. This feature was an important marketing function as some participants (e.g., hunters, landowners) informally expressed interest during our interactions in better understanding wildlife food production at locations they were familiar with. I made the individual online database portal (described below) private so   16  participants were comfortable submitting information that may be valuable to other people interested in outdoor activities like hunting.  1.2.3. Participant Recruitment  To increase attention to the program, allow all partners to be equally represented, and to make the program appear as established, I designed a unique MI-MAST brand. Because volunteers participate in activities that align with their interests (Rotman et al. 2012), creating a brand that clearly represents the program is important for recruitment. The program was named MI-MAST: Wildlife Food Tracker and a logo incorporating a white-tailed deer, black bear, and wild turkey inside of an acorn was created (Figure 1.1). Popular game species that rely on mast were included to attract hunters, who were identified as a potential primary audience for MI-MAST. Hunters often pay attention to masting patterns to better predict where animals will be, and are the audience that directly associate with the goal of MI-MAST, i.e., to improve management information for game species. The nare specific to Michigan conservation and also paralleled the naming format of other state agency wildlife food abundances to increase viewing or hunting successes. To inform future recruitment strategies, MI-MAST participants were given the option to specify their age, sex, previous experience identifying plants, and primary and secondary interests in outdoor activities when registering for an account (Table 1.1). Using Google Analytics, additional information was collected regarding how visitors interacted with the website (described below). 1.2.4. Streamlining Participant Interactions with MI-MAST  To efficiently coordinate volunteers and collect information throughout the state, all instructions and tools necessary to participate in MI-MAST were provided on a website   17  (http://www.mimast.org/). Here participants could find details regarding the purpose of the program, partners, instructional materials, and field identification guides (described below). Within the website users created an account and completed an optional demographic questionnaire (Table 1.1), watched a video tutorial explaining how to use the online features, and accessed a private portal to enter records and manage their personal database. Smartphone applications were developed for iOS 5.1 or later and Android 3.2 or later devices that also included instructional materials, field guides, and a personal database for participants to record, view, and edit their records. The mobile applications were an alternative to submitting data through the website and reduced sampling effort by utilizing the GPS capabilities of smartphones to record locational information. Users were also given the option to use the camera in their phone to include pictures with their data. All records were stored locally on the phone until mobile data or wireless internet connectivity was established, allowing the user to synchronize their records with those of the main database and their website account. This feature enabled data collection in remote areas lacking cellular reception, which is common throughout Michigan.  1.2.5. Identifying Plant Species  As mast production can vary among species or genus, it is important to accurately identify plants that are recorded (Kelly and Sork 2002, Lusk et al. 2007). Many mast-producing species found in Michigan can be difficult to accurately identify to species even with formal training. The ability of untrained volunteers to correctly identify plants using structured protocols has been demonstrated, but accuracy decreases when volunteers need to differentiate among similar species (Bloniarz and Ryan 1996, Brandon et al. 2003, Crall et al. 2011, Jordan et al. 2012, Fuccillo et al. 2014). I compiled a list of the most-common mast producing species found in Michigan that were used by game animals (Table 1.2). To help reduce identification errors, I   18  restricted participants to only submitting data on plants in Table 1.2, but I also provided the opportunity for participants to request that new species be added to the database. Field identification guides including pictures of several parts of the plant and non-technical descriptions were provided to assist in identification and included on both the website and smartphone applications.  1.2.6. Data Validation  A subset of mast data submitted by 15 participants in 2014 and 2015 (n=60/82 sites) were field validated for accuracy within 2 weeks of submission. I visited data sites and tested the data submissions for 1) accuracy of GPS location tools by comparing submitted GPS locations with those associated with the plant of interest using a separate GPS handheld device, 2) correct species identification, and 3) correct mast amount classification compared to what I would have recorded based on the data collection protocol.  1.3. Results  In 2014, the first year of the program, 10 conservation groups were visited in Michigan and given a presentation introducing MI-MAST and inviting them to participate (~ 150 individuals). A small audience was purposely targeted to test the programs instructional materials, technology, and sampling protocol before marketing it to a larger audience. In 2015, additional outreach to conservations groups occurred (~ 200 individuals), and hyperlinks directing people to the MI-MAST website were featured on MIDNR webpages, MUCC hosted multiple blog posts featuring MI-MAST, and MI-MAST was featured on MIDNR and MUCC social media outlets. MI-MAST was also featured using a booth at two large outdoor expositions (~ 150 individuals interacted with the booth). MIDNR staff was also incentivized to participate in the program by being entered into a drawing to win wildlife memorabilia. Throughout these   19  informal interactions common themes of potential participants included 1) interests in keeping a private database to track mast production on their own property, 2) interests in what sort of products would be publicly available, and 3) how the data would be used by the MIDNR. During this time little interest was expressed in helping the MIDNR or being involved in scientific inquiry.  Based on results from Google Analytics, during 2014-2015, 7,725 people visited the website and 608 returned to the website at least once (8%). Most web-site visitors (55%) connected to MI-MAST via a link from an MIDNR web page, with most (44%) of the other visitors engaged by direct communication or email. Most individuals on the MI-MAST web page were located in the southern portion of the Lower Peninsula of Michigan, even though recruitment efforts specifically focused on northern Michigan (Fig. 1.2). The MI-MAST smartphone application had 239 iOS and 199 Android downloads, pointing to the importance of providing both mobile platforms. I had 273 users register for an account and 222 of those users completed a demographic questionnaire (Table 1.1). Of those users who completed the questionnaire, 75% were male and the majority (35%) of respondents were 45-54 years old (Fig. 1.3). About one-third of the respondents received formal plant identification training, 43% had identified plants before, and 25% had never identified plants before (Fig. 1.4). Most (93%) of the participants in the youngest age group (14-24 years old) had some experience identifying plants (Fig. 1.5). The majority (61%) of respondents listed hunting as their primary outdoor interest (Fig. 1.6).  After arriving at the website homepage, 52% of users next visited the mast species fact sheets, 22% the program description page, and 13% the registration page. During these visits   20  users spent an average of 31 seconds on the homepage, 32 seconds on the fact sheets, 46 seconds on the program description, and 3 minutes and 44 seconds on the registration page.  Overall 30 users submitted at least one record. Those 30 users submitted 82 plant records in total and recorded 70 mast amounts (12 plants were identified earlier in the year but were not revisited to report mast amounts). I visited 60 of the 82 locations (73%) that were submitted by 15 participants.  All 60 of the plants were correctly identified to 19 different genera or species (depending on the plant) by MI-MAST participants. All 23 hard mast amounts were correctly recorded and 29 out of 32 (91%) soft mast amounts were correctly recorded, based on the amount I would have recorded following the sampling protocol instructions. The most common soft mast error was over-reporting mast amounts on dwarf red raspberry (Rubus pubescens). 1.4. Discussion  In developing MI-MAST my first objective was to create a sustainable and productive collaboration among the founding organizations and between wildlife stakeholder groups. During program creation I found the collaboration between the MIDNR, MUCC, and MSU to be well-balanced in regard to devotion of time spent on program objectives. Relying on the ASETS laboratory to develop and maintain the technology behind the program proved to be most beneficial. Although tools exist for novices to create their own websites, mobile applications, and databases (Hartung et. al 2010); the program would have taken much longer to develop if a team was not entirely devoted to technical development and maintenance. The ASETS laboratory also had previous experience building tools for conservation programs, and thus I assumed they were able to better handle budgetary constraints while maintaining basic functionality than were potentially offered by private technical development companies.   21  During initial recruitment I attempted to present MI-MAST to other conservation organizations not affiliated with MUCC, and was seldom successful at recruiting new participants. Common responses included shortage of staff time to organize a presentation meeting, or no upcoming planned member activities that were relevant to the topic of MI-MAST. Building new relationships with volunteer organizations often already overwhelmed with duties takes more time than is often available when trying to beta test a new program. Relying on the trust MUCC has built with their organizations proved to be beneficial when needing to arrange presentations and gain confidence from large new audiences. All MUCC affiliated organizations that I contacted were willing and enthused to have MI-MAST introduced to their members, and allowed me to successfully beta test the program in 2014. Similar initial interests were observed with MIDNR biologists and during several of those meetings the biologists provided comments on how the beta program could be improved and insights on how the public may react to certain features. My second objective was to develop a sampling protocol appropriate for untrained volunteers that produced reliable and usable data on mast production. I found no indications that the design of MI-MAST impeded the effective collection of data that could be used to calculate mast production indices in Michigan. My field validation showed that MI-MAST participants correctly identified mast-producing plants and accurately portrayed hard (100% correct) and soft (91% correct) mast amounts. My results were consistent with other citizen science programs that have shown volunteers can correctly identify plants when given the appropriate resources and simple instructions (Brandon et al. 2003, Crall et al. 2011, Fuccillo et al. 2014). All locations that were field validated proved to have recorded accurate coordinates based on a separate GPS   22  reading on the plant of interest and suggested users could properly use the interactive online map and mobile applications when submitting data.  I caution that my validation results are only based on data submitted by 15 different volunteers who came from a population of predominantly hunters, who generally had experience in identifying plants. If a broader audience is engaged to participate in the program the accuracy of data may be different than what I found in this study. I recommend data validation be conducted in future years when more users participate to ensure data are reliable for producing accurate mast production estimates.    A limitation in the sampling protocol was that participants were not given a specific time to record their data to account for temporal variations in mast ripening throughout Michigan, and instead were instructed to report mast amounts at a time when mast was ripe. If participants collect data on a plant too early, mast may have not developed yet and it would be underreported. However, if the plant is visited too late, soft mast may have been eaten. McCarty et al. (2002) found that ripe fruit survived predation ranging anywhere from 3-165 days. Providing additional information in the field guides suggesting ripe periods for each species could reduce this temporal bias without having to set a firm data collection time requirement for all soft or hard species, potentially reducing participation interest.      Continued use of this sampling protocol may also require additional investigation of participant sampling behavior that may influence production estimates. I found that participants could correctly identify species and mast amounts, but I did not place any restrictions on where participants should record data or how many plants to sample in an area. This was done to make submitting data to the program easier for participants (Parsons et al. 2011). Using opportunistic data instead of data collected on fixed transects may result in collection bias in the first few   23  years.  Participants may be more likely to begin monitoring a plant that was producing mast in the initial year because it caught their attention as a mast-producing plant (Kendall et al. 1996, Fitzpatrick et al. 2009).  However, asking participants to revisit the same plants in subsequent years should eliminate this bias in the future. Although data submitted by MI-MAST participants did not reveal this pattern, year one observations should still be used with caution when comparing pr-been documented in other citizen-science programs where data collection behavior of the participants changes once they become more comfortable with the methods (Kendall et al. 1996, Sauer et al. 1994). In such instances coordinators often choose not to use those data. My third objective was to understand participant motivations and interests to sustain recruitment and retention of volunteers. Even after extensive marketing and recruiting, only a small percentage (11%) of account holders submitted data. I predicted that creating an account would be the largest impediment to getting volunteers to participate in MI-MAST. I assumed that once participants created an account they would likely submit mast observations because of the little effort needed to participate. Silvertown et al. (2013) also found low participation rates in a citizen science program, but their program required greater sampling effort. In the Silverton et al. (2013) program, few participants submitted data and an even smaller group produced the majority of data. In my study, notifications were sent to account holders twice via email reminding them it was an appropriate time to collect mast data. My results suggest that other and more frequent communication methods may be needed to ensure more account holders are submitting data.   Bell et al. (2008) found that participants highly valued social experiences and communication amongst other participants, and suggested more frequent communication may be   24  necessary. This communication strategy may be particularly effective for engaging a younger cohort of MI-MAST participants as these individuals are known to rely on social connectivity (Lenhart et al. 2010). Another strategy for improving participation may be to appeal to the competitive tendencies of individuals.  Sullivan et al. (2009) tripled the number of individuals who submitted data to a citizen science program immediately after implementing features that allowed participants to view their own reports and compare them with the reports of other participants. Creating a community culture among participants throughout Michigan that communicate and share experiences may increase participant interest and hold users more accountable. Creating an online message board or forum may serve this purpose. The demographics of MI-MAST account holders aligned with our marketing audience. Generally, hunters in Michigan are middle-to-late aged males (Frawley 2004). I recognize that many conservation organizations without hunting interests consist of members who frequently volunteer to support wildlife conservation. Members from these organizations often are already participating in citizen-science programs (e.g., stream monitoring, bird surveys, invasive species reporting). Ryan et al. (2001) found that participants of this demographic were most motivated by helping the environment and learning. The interests of conservation or stewardship members may make them more willing to participate if they know the data are being used by managers for conservation purposes, whereas the interest among hunters may be more directed towards individual benefits that will ultimately increase their own hunting success or preserving hunting values. This was observed when recruiting participants, where hunters were more interested in the way they could apply our tools for their own hunting purposes.  I noticed that at initial presentations of MI-MAST, many attendees appeared to be >60 years old. It is commonly accepted that this generation lacks technological skills or access to   25  smartphones or tablets (Roupa et al. 2010). However, many attendees had smartphones and proved capable of operating both the website portal and mobile applications. Considering that roughly one-third of volunteers in America are 55 years or older, they should not be overlooked as an important demographic in technology-based citizen science programs (Bureau of Labor Statistics 2015). The majority of conservation organizations that were introduced to MI-MAST with an in-person presentation were located in the northern Lower Peninsula, and coincided with the majority of volunteer data locations (Figure 1.2). The majority of webpage views occurred in large urban areas in the southern Lower Peninsula and is likely related to the larger resident population of these areas (Figure 1.2). The two outdoor expositions where MI-MAST was marketed were located in the southern portions of Michigan and may have contributed to more webpage views, but did not correlate to the amount of data submitted near these locations. These findings showed that members of the conservation organizations visited in the northern Lower Peninsula were less likely to view the webpages than audiences in the southern portion of the state, but may have been more likely to submit data. This could suggest that two different audience types exist in Michigan; one that is more likely to visit the webpage and learn about the program but less likely to get involved, and another that participates more readily but may be less concerned about new online tools or features.   I found no indications that features of the sampling protocol or collaboration among organizations would inhibit a sustainable and successful citizen-science program. MI-MAST was only marketed in 2015 and continued recruitment will be necessary in future years to increase participation and data collection to produce mast production indices throughout Michigan. I   26  suggest the additional recruitment strategies and new insights into audience behavior as found in this study be used to guide program development and participation in future years.  1.5. Acknowledgements  I would like to thank D. Youngdyke, A. Mitterling, M. Donovan, and K. Carter for their efforts in recruitment, A. Tiernan for branding design, and A. Ziegler, G. Carballo, and C. Peterson for their work in developing online tools.                                         27                          APPENDIX                        28  Table 1.1. Optional demographic questionnaire participants were prompted to complete when creating an account on www.mimast.org.         Age Select Age:  o 14-24 o 25-34 o 35-44 o 45-54 o 55-64 o 65+ Gender  Select Gender:  o Male o Female Plant Identification  Select ID Experience:  o None o I have identified plants before o I have received formal training in identifying plants Primary Interests  Select your primary outdoor interest:  o Hunting o Bird Watching o Hiking/Walking o Natural Areas o Outdoor Sports o Other  Secondary Interests  Select your secondary outdoor interest:  o Hunting o Bird Watching o Hiking/Walking o Natural Areas o Outdoor Sports o Other   29  Table 1.2. Hard and soft mast species and genera included in MI-MAST: Wildlife Food Tracker. Mast species were included based on intensity of use by large game animals in Michigan, USA, based on MIDNR biologist recommendations and Martin et al. 1951.           Species   Common Name         Hard Mast     Quercus bicolor   Swamp white oak  Quercus alba      White oak  Carya cordiformis  Bitternut hickory  Corylus cornuta  Beaked hazelnut  Carya ovata  Shagbark hickory  Fagus grandifolia  American beech  Juglans cinerea  Butternut  Juglans nigra  Black walnut  Quercus macrocarpa  Bur oak  Quercus palustris  Pin oak  Quercus rubra  Northern red oak  Quercus velutina  Black oak Soft Mast     Amelanchier arborea  Serviceberry  Cornus spp.  Dogwood  Crataegus spp.  Hawthorn  Elaeagnus umbellata  Autumn Olive  Gaylussacia baccata  Huckleberry  Malus coronaria  Crabapple  Malus pumila  Apple  Phytolacca americana  Pokeweed  Prunus serotina  Black cherry  Prunus virginiana  Chokecherry  Rubus allegheniensis  Blackberry  Rubus pubescens  Dwarf red raspberry  Sambucus canadensis  Elderberry  Vaccinium myrtilloides  Blueberry  Viburnum acerifolium  Mapleleaf viburnum            30  Figure 1.1.  MI-MAST: Wildlife Food Tracker logo.                            31  Figure 1.2. Locations of recruitment events, volunteer data submissions, and webpage views (provided by Google Analytics) for MI-MAST: Wildlife Food Tracker in Michigan, USA, 2014-2015.           32  Figure 1.3. Ages of MI-MAST: Wildlife Food Tracker account holders who completed the demographic questionnaire upon registration, 2014-2015.     Figure 1.4. Previous experience identifying plants of MI-MAST: Wildlife Food Tracker account holders who completed the demographic questionnaire upon registration, 2014-2015.     0102030405060708014-2425-3435-4445-5455-6465+Account Holders (N)Age (Years)Account Holder AgeFemaleMale0102030405060708090100Formal PlantIdentification TrainingIdentified Plants BeforeNoneAccount Holders (N)Previous Experience Identifying PlantsAccount Holder ExperienceFemaleMale  33  Figure 1.5. Age and previous experience identifying plants of MI-MAST: Wildlife Food Tracker account holders who completed the demographic questionnaire upon registration, 2014-2015.                            010203040506070809010014-2425-3435-4445-5455-6465+Account Holders (%)Age (Years)Account Holder Experience and AgeNo ExperienceIdentified Plants BeforeFormal Plant ID Training  34  Figure 1.6. Primary outdoor interests of MI-MAST: Wildlife Food Tracker account holders who completed the demographic questionnaire upon registration, 2014-2015.                           020406080100120140160BirdWatchingHikingHuntingNaturalAreasOutdoorSportsOtherAccount Holders (N)Primary InterestAccount Holder InterestsFemaleMale  35                          LITERATURE CITED                        36  Aceves-Bueno, E., A. S. Adeleye, D. Bradley, W. T. Brandt, P. Callery, M. Feraud, K. L. Garner, R. Gentry, Y. Huang, I. McCullough, I. Pearlman, S. A. Sutherland, W. Wilkinson, Y. Yang, T. Zink, S. E. Anderson, and C. Tague. 2015. Citizen science as an approach for overcoming insufficient monitoring and inadequate stakeholder buy-in in adaptive management: criteria and evidence. Ecosystems 18:493-506.  Bart, J., M. Hofschen, B. G. Peterjohn. 1995. Reliability of the breeding bird survey: effects of restricting surveys to roads. Auk 112:758-61.  Bell, S., M. Marzano, J. Cent, H. Kobierska, D. Podjed, D. Vandzinskaite, H. Reinert, A. Armaitiene, M. -Jurczak, and R2008. What counts? Volunteers and their organisations in the recording and monitoring of biodiversity. Biodiversity and Conservation 17:3443-3454.  Bloniarz, D. V., and H. D. P. Ryan. 1996. The use of volunteer initiatives in conducting urban forest resource inventories. Journal of Arboriculture 22:75-82.  Bonney, R., C. B. Cooper, J. Dickinson, S. Kelling, T. Phillips, K. V. Rosenberg, and J. Shirk. 2009. Citizen science: a developing tool for expanding science knowledge and scientific literacy. Bioscience 59:977-984.  Brandon, A., G. Spyreas, B. Molano-Flores, C. Carroll, and J. Ellis. 2003. Can volunteers provide reliable data for forest vegetation surveys? Natural Areas Journal 23:254-261.  Bureau of Labor Statistics. 2015. Volunteering in the United States News Release. United States Department of Labor, Washington, D.C., USA.  Crall, A. W., G. J. Newman, T. J. Stohlgren, K. A. Holfelder, J. Graham, and D. M. Waller. 2011. Assessing citizen science data quality: an invasive species case study. Conservation Letters 4:433-442.  Crowl, T. A., T. O. Crist, R. R. Parmenter, G. Belovsky, and A. E. Lugo. 2008. The spread of invasive species and infectious disease as drivers of ecosystem change. Frontiers in Ecology and the Environment 6:238-46.  Danielsen, F., N. D. Burgess, and A. Balmford. 2005. Monitoring matters: examining the potential of locally-based approaches. Biodiversity and Conservation 14:2507-2542.  Darwall, W. R. T., and N. K. Dulvy. 1996. An evaluation of the suitability of non-specialist volunteer researchers for coral reef fish surveys. Mafia Island, Tanzania-a case study. Biological Conservation, 78:223-231.   Daubenmire, R. 1959. A canopy-coverage method of vegetational analysis. Northwest Science 33:43-64.    37  Dickinson, J. L., B. Zuckerberg, and D. N. Bonter. 2010. Citizen science as an ecological research tool: challenges and benefits. Annual Review of Ecology, Evolution, and Systematics 41:149-172.  26 in C. McEver, R. Bonney, J. Dickinson, S. Kelling, K. Rosenberg, and J. Shirk, editors. Citizen Science Toolkit Conference. Cornell Laboratory of Ornithology.  Elowe, K. D., and W. E. Dodge. 1989. Factors affecting black bear reproductive success and cub survival. Journal of Wildlife Management 53:962-968.  Evans, C., E. Abrams, R. Reitsma, K. Roux, L. Salmonsen, and P. P. Marra. 2005. The Neighborhood Nestwatch Program: participant outcomes of a citizen-science ecological research project. Conservation Biology 19:589-594.   Feldhamer, G. A. 2002. Acorns and white-tailed deer: interrelationships in forest ecosystems. Pages 215223 in W. J. McShea and W. M. Healy, editors. Oak forest ecosystems: ecology and management for wildlife. The Johns Hopkins Press, Baltimore, Maryland, USA.   Fitzpatrick, M., E. Preisser, A. Ellison, J. Elkinton. 2009. Observer bias and the detection of low-density populations. Ecological Applications 19:1673-79.  Frawley, B. J. 2004. Demographics, recruitment, and retention of Michigan hunters. Michigan Department of Natural Resources, Wildlife Division Report 3426, Lansing, USA.  Fuccillo, K. K., T. M. Crimmins, C. E. de Rivera, and T. S. Elder. 2014. Assessing accuracy in citizen science-based plant phenology monitoring. International Journal of Biometeorology 1-10.  Galloway, A., M. Tudor, and W. Vander Haegen. 2006. The reliability of citizen science: a case study of Oregon White Oak stand surveys. Wildlife Society Bulletin 34:1425-29.  Greenberg, C. H., and G. S. Warburton. 2007. A rapid hard-mast index from acorn presence-absence tallies. Journal of Wildlife Management 71:1654-1661.  Hartung, C., A. Lerer, Y. Anokwa, C. Tseng, W. Brunette, and G. Borriello. 2010. Open data kit: tools to build information services for developing regions. In Proceedings of the 4th ACM/IEEE International Conference on Information and Communication Technologies and Development.  Herrera, C. M., P. Jordano, J. Guitián, and A. Traveset. 1998. Annual variability in seed production by woody plants and the masting concept: reassessment of principles and relationship to pollination and seed dispersal. The American Naturalist 152:576-594.    38  Howe, E. J., M. E. Obbard, R. Black, and L. L. Wall. 2010. Do public complaints reflect trends in human-bear conflict? Ursus 21:131-142.  Inman, R. M., and M. R. Pelton. 2002. Energetic production by soft and hard mast foods of American black bears in the Smoky Mountains. Ursus 13:57-68.  Jordan, R. C., W. R. Brooks, D. V. Howe, and J. G. Ehrenfeld. 2012. Evaluating the performance of volunteers in mapping invasive plants in public conservation lands. Environmental Management 49:425-434.  Kelly, D., D. E. Hart, and R. B. Allen. 2001. Evaluating the wind pollination benefits of mast seeding. Ecology. 82:117-126.  Kelly, D., and V. L. Sork. 2002. Mast seeding in perennial plants: why, how, where? Annual Review of Ecology and Systematics 33:427-447.  Kendall, W. L., B. G. Peterjohn, and J. R. Sauer. 1996. First-time observer effects in the North American Breeding Bird Survey. Auk 113:823-29.  Lashley, M. A., C. A. Harper, G. E. Bates, and P. Keyser. 2011. Deer forage available following silvicultural treatments in upland hardwood forests and warm-season plantings. Journal of Wildlife Management 75:1467-1476.  Lashley, M. A., J. R. Thompson, M. C. Chitwood, C. S. DePerno, and C. E. Moorman. 2014. Evaluation of methods to estimate understory fruit biomass. PloS ONE 9:e96898. doi: 10.1371/journal.pone.0096898  Lawrence, A., 2006. No personal motive? Volunteers, biodiversity, and the false dichotomies of participation. Ethics, Place Environment 9:279-298.  Lenhart, A., K. Purcell, A. Smith, and K. Zickuhr. 2010. Social media and Mobile internet use among teens and young adults. Pew Internet & American Life Project, Washington, D.C., USA.  Levey, D. J., and C. W. Benkman. 1999. Fruitseed disperser interactions: timely insights from a long-term perspective. Trends in Ecology & Evolution. 14:41-43.  Link, W. A., and J. R. Sauer. 1999. Controlling for varying effort in count surveys: An analysis of Christmas Bird Count data. Journal of Agricultural, Biological, and Environmental Statistics 116-125.  Lusk, J. J., R. K. Swihart, and J. R. Goheen. 2007. Correlates of interspecific synchrony and interannual variation in seed production by deciduous trees. Forest Ecology and Management 242:656-670.    39  Martin, A. C., H. S. Zim, and A. L. Nelson. 1951. American wildlife and plants: a guide to wildlife food habits. McGraw-Hill, New York, New York, USA.  McCarty, J. P., D. J. Levey, C. H. Greenberg, and S. Sargent. 2002. Spatial and temporal variation in fruit use by wildlife in a forested landscape. Forest Ecology and Management 164:277-291.  McKinley, D. C., A. J. Miller-Rushing, H. L. Ballard, R. E. Bonney, H. Brown, D. M. Evans, R. A. French, J. K. Parrish, T. B. Phillips, S. F. Ryan, L. A. Shanley, J. L. Shirk, K. F. Stepenuck, J. F. Weltzin, A. Wiggins, O. D. Boyle, R.D. Briggs, S. F. Chapin III, D. A. Hewitt, P. W. Preuss, and M. A. Soukup. 2015. Can investing in citizen science improve natural resource management and environmental protection? Issues in Ecology 19.  McNie, E., 2007. Reconciling the supply of scientific information with user demands: an analysis of the problem and review of the literature. Environmental Science and Policy 10:17-38.  McShea, W.J. 2000. The influence of acorn crops on annual variation in rodent and bird populations. Ecology 81:228-238.  Michigan United Conservation Clubs [MUCC]. 2015. MUCC about page. http://www.mucc.org/about. Accessed 12 Dec 2015.  Newman, C., C. Buesching, and D. MacDonald. 2003. Validating mammal monitoring methods and assessing the performance of volunteers in wildlife conservation-stodiety  Conservation 113:189-97.  Nichols, J. D., and B. K. Williams. 2006. Monitoring for conservation. Trends in Ecology and Evolution 21:668-73.  Norton, D. A., and D. Kelly. 1988. Mast seeding over 33 years by Dacrydium cupressinum Lamb. (rimu) (Podocarpaceae) in New Zealand: the importance of economies of scale. Functional Ecology 2:399-408.  Ohrel, R. L., and K. M. Register. 2006. Volunteer estuary monitoring: a methods manual. EPA, Washington, D.C., USA.  Ostfeld, R. S., C. G. Jones, and J. O. Wolff. 1996. Of mice and mast.  BioScience 46:323-330.  Parsons, J., R. Lukyanenko, and Y. Wiersma. 2011. Easier citizen science is better. Nature 471:37-37.  Penrose, D., and S. Call. 1995. Volunteer monitoring of benthicmacroinvertebrates: regulatory  14:203-209.    40  Pharoah, P. D., 2014. Cell Slider: Using crowd sourcing for the scoring of molecular pathology. Cancer Research 74:303-303.  Raddick, M. J., G. Bracey, P. L. Gay, C. J. Lintott, P. Murray, K. Schawinski, A. S. Szalay, and J. Vandenberg. 2010. Galaxy Zoo: exploring the motivations of citizen science volunteers. Astronomy Education Review 9:18.  Reynolds-Hogland, M. J., L. B. Pacifici, and M. S. Mitchell. 2007. Linking resources with demography to understand resource limitation for bears. Journal of Applied Ecology 44:1166-1175.  Richmond, E.S., C. W. Ryan, C.P. Carpenter, and M. L. Peters. 2015. West Virginia mast survey and hunting outlook. West Virginia Division of Natural Resources Section. Wildlife Resources Bulletin Number 15-3.   Robertson, M. P., G. S. Cumming, and B. F. N. Erasmus. 2010. Getting the most out of atlas data. Diversity and Distributions 16:363-375.  Rotman, D., J. Preece, J. Hammock, K. Procita, D. Hansen, C. Parr, D. Lewis, and D. Jacobs. 2012. Dynamic changes in motivation in collaborative citizen-science projects. In Proceedings of the ACM Conference on Computer Supported Cooperative Work, New York, USA.  Roupa, Z., M. Nikas, E. Gerasimou, V. Zafeiri, L. Giasyrani, E. Kazitori, and P. Sotiropoulou. 2010. The use of technology by the elderly. Health Science Journal 4-2.  Ryan, R. L., R. Kaplan, and R. E. Grese. 2001. Predicting volunteer commitment in environmental stewardship programmes. Journal of Environmental Planning and Management, 44:629-648.  Ryan, C. W., J. C. Pack, W.K. Igo, J.C. Rieffenberger, and A. B. Billings. 2004.  Relationship of mast production to big-game harvests in West Virginia. Wildlife Society Bulletin 32:786-794.  Ryan, C. W., J. C. Pack, W. K. Igo, and A. Billings. 2007. Influence of mast production on black bear non-hunting mortalities in West Virginia. Ursus 18:46-53.  Sarewitz, D., and R. Pielkejr. 2007. The neglected heart of science policy: reconciling supply of and demand for science. Environmental Science and Policy 10:5-16.  Sauer J. R., B. G. Peterjohn, and W. A. Link. 1994. Observer differences in the North-American breeding bird survey. Auk 111:50-62.  Shaw C. E., C. A. Harper, M. V. Black, and A. E. Houston. 2010. Initial effects of prescribed burning and understory fertilization on browse production in closed canopy hardwood stands. Journal of Fish and Wildlife Management 1:64-71.   41   Shirk, J. L., H. L. Ballard, C. C. Wilderman, T. Phillips, A. Wiggins, R. Jordan, E. McCallie, M. Minarchek, B. Lewensteun, M. Krasny, and R. Bonney. 2012. Public participation in scientific research: a framework for deliberate design. Ecology and Society 17:29.  Silvertown, J. W. 1980. The evolutionary ecology of mast seeding in trees. Biological Journal of the Linnean Society 14:235-250.  Silvertown, J. 2009. A new dawn for citizen science. Trends in Ecology and Evolution 24:467-471.  Silvertown, J., C. D. Buesching, S. K. Jacobson, and T. Rebelo. 2013. Citizen science and nature conservation. Key topics in conservation biology 2:124-142.  Steffen, D. E., N. W. Lafon, and G. W. Norman. 2002. Turkeys, acorns, and oaks. Pages 241-255 in M. J. McShea and W. M. Healy, editors. Oak forest ecosystems: ecology and management for wildlife. The Johns Hopkins Press, Baltimore, Maryland, USA.  Stepenuck, K. F. 2013. Improving understanding of outcomes and credibility of volunteer environmental monitoring programs. Dissertation. University of Wisconsin, Madison, USA.   Sullivan B. L., C. L. Wood, M. I. Iliff, R. E. Bonney, D. Fink, S. Kelling. 2009. eBird: A citizen-based bird observation network in the biological sciences. Biological Conservation 142:2282-2292.  Szabo, J. K., R. A. Fuller, and H. P. Possingham. 2012. A comparison of estimates of relative abundance from a weakly structured mass-participation bird atlas survey and a robustly designed monitoring scheme. Ibis 154:468-479.   Uhlig, H. G., and H.L. Wilson. 1952. A method of evaluating an annual mast index. Journal of Wildlife Management 16:338-343.  Webb, T. J., D. Noble, and R. P. Freckleton. 2007. Abundance-occupancy dynamics in a human dominated environment: linking interspecific and intraspecific trends in British farmland and woodland birds. Journal of Animal Ecology 76:123-34.   Whitehead, C. J. 1969. Oak mast yields on wildlife management areas in Tennessee. Tennessee Game and Fish Commission, Nashville, USA.  Villard, M. A., M. K. Trzcinski, and G. Merriam. 1999. Fragmentation effects on forest birds: relative influence of woodland cover and configuration on landscape occupancy. Conservation Biology 13:774-83.    42  Vitos, M., J. Lewis, M. Stevens, and M. Haklay. 2013. Making local knowledge matter: supporting non-literate people to monitor poaching in Congo. In Proceedings of the 3rd ACM Symposium on Computing for Development, Bangalore, India.                                              43  CHAPTER 2 USING CITIZEN-SCIENCE DATA TO DEVELOP ANNUAL MAST PRODUCTION INDICES IN MICHIGAN  Abstract  Some animals rely on mast production for a significant portion of their diet. Patterns of mast are often variable, producing a large crop some years and scarce amounts in others. This large fluctuation in food availability can affect wildlife populations. Including quantitative information about annual mast production into wildlife population models may produce more accurate estimates of population demographics. However, annual statewide monitoring efforts would not be feasible for most wildlife management agencies. I show how point-based data collected by citizen scientists can be used to generate hard- and soft-mast production indices. These indices, if based on data that are spatially and temporally extensive, could be used to compare relative crop sizes among regions and years. In 2015, I collected 673 records of hard mast production and 218 records of soft mast production in several locations throughout Michigan. I introduce a process to identify spatial scales of production variability that can inform how data should be distributed and combined into similar ecoregional units. Using these results, I calculated a hard and soft mast index based on previously developed empirical equations for 1 ecoregion for 3 combined species of oak (Quercus alba, Q. palustris, and Q. rubra) and black cherry (Prunus serotina). Indices can be used to compare differences in production amounts among years. Based on previous long term studies on mast production indices and associated subjective crop rankings, the indices I calculated for both oak (0.54) and cherry (0.71) suggested near crop failure in the ecoregion I selected for this analysis. The accuracy of calculated mast indices is directly related to the number of records and the spatial and temporal extents of those records within an assessment area. I caution that my results are based on data from relatively few   44  species and concentrated in only one portion of the state; however the process I describe can be replicated when more data are collected. Given sufficiently detailed and properly distributed data, I show that a simple data collection protocol compatible with citizen science programs can produce relative estimates of hard and soft mast production that are spatially explicit and potentially useful for managing wildlife.  Key Words Acorns, berries, fruits, hard mast, soft mast, citizen science, Michigan, wildlife foods  2.1. Introduction Some plants exhibit a behavior of masting, or the bimodal pattern of seed production, where many seeds are produced some years and few in others (Kelly 1994, Koenig and Knops 2000, Kelly and Sork 2002). This behavior is more commonly associated with hard-mast producing trees like oak (Quercus spp.) and beech (Fagus grandifolia) (Gysel 1971, Greenberg and Parresol 2002). Trees closer in proximity often show synchrony and produce large mast crops in unison, yet a consistent spatial scale at which synchrony occurs has not been observed (Kelly and Sork 2002). Several hypotheses have been proposed to explain why this pattern in seed production exists, yet researchers have struggled to provide overwhelming evidence in support of a single hypothesis (Kelly and Sork 2002). One hypothesis, the resource-matching hypothesis, suggests that temperature and precipitation determine mast production patterns. The hypothesis suggests that it is less costly for trees to reproduce in years with abundant water and favorable temperatures than in years with unfavorable weather, and as a result natural selection has favored producing mast during favorable conditions (Norton and Kelly 1988, Sork 1993, Kelly 1994, Houle 1999, Pearse et al. 2014). Since variation in weather alone does not reflect the bimodal pattern observed in mast production, other factors are likely influencing production (Koenig and   45  Knops 2000). Another hypothesis, the predator satiation hypothesis, suggests that crop production is determined by predators, where producing more seeds at once and synchronizing production with neighbors overwhelms seed predators and thereby creates a higher probability of seedling survival (Janzen 1971, Silvertown 1980). If this were the case, synchrony of mast production should occur at a scale comparable to the geographic extents of seed predator movements (Janzen 1971), with synchrony being less defined among trees that attract multiple types of predators with varying ranges (Kelly and Sork 2002). Plant communities lacking tree species diversity exhibit stronger masting behavior (Janzen 1971, Boucher 1981), which may be a product of single or similar types of predators living in those communities. No single explanation has yet accounted for the spatiotemporal variation in masting patterns, leading to the conclusion that multiple interacting factors influence the behavior (Herrera et al. 1998, Kelly and Sork 2002).  Masting is less common in small trees or shrubs that produce fleshy fruits (Silvertown 1980). These soft-mast species are more likely to reproduce in regular intervals and yield similar amounts of fruits (Kelly and Sork 2002). However spatiotemporal variability can still occur among soft mast species and appears most correlated with large temperature and precipitation fluctuations during the summer when fruits are ripening and during autumn when flower buds are forming (Selas 2000). A variety of growing patterns exist among soft mast species, suggesting that weather effects are most likely species specific (Martin et al. 1951).  Hard and soft mast is an important source of food for many wildlife species (Martin et al. 1951, Wolff 1996). Masting can cause changes in body condition, reproduction, and ultimately population size and distribution in game species such as the white-tailed deer (Odocoileus virginianus), black bear (Ursus americanus), and wild turkey (Meleagris gallopavo) (Feldhamer   46  2002, Steffen et al. 2002, Ryan et al. 2004, Reynolds-Hogland et al. 2007). A proper understanding of wildlife population dynamics is needed by managers to set seasonal harvest limits that are sustainable. Population models of wildlife species that rely on mast could be improved with the addition of quantitative information describing the annual amount of mast available on the landscape relative to previous years.  To produce a dataset that portrays mast production at scales corresponding to harvest management of wildlife would require extensive sampling over space and time to collect enough data to understand the spatiotemporal variability of mast production (Greenberg and Warburton 2007). Methods for estimating the size and extent of acorn production exist, including direct measurement with seed traps (Greenberg and Parresol 2002), qualitative rankings based on visual surveys (Sharp 1958, Graves 1980), and laborious quantitative surveys counting acorns or tree stems (Whitehead 1969, Koenig et al. 1994). Seed traps can best describe actual crop sizes but they are impractical to implement across large scales. Labor intensive quantitative visual surveys as described by Whitehead (1969) produce hard-mast indices that can be used to compare crop production to previous years and are highly correlated with density estimates calculated by seed traps (Perry and Thill 1999, Greenberg and Warburton 2007). By using a greater quantity of qualitative visual surveys, Greenberg and Warburton (2007) found that high production years were correlated with the amount of trees producing acorns.  Methods exist to estimate soft-mast yield at small spatial scales, but these also require individual fruit or stem counts making it difficult to apply to large areas (Daubenmire 1959, Lashley et al. 2014). Since soft mast production exhibits less variability than does hard mast, qualitative measures of production can describe patterns at large scales. These types of   47  qualitative assessments are performed by several state wildlife agencies (Uhlig and Wilson 1952, Richmond et al. 2015).  To demonstrate how spatially explicit mast production indices could be calculated from relatively simple data like those collected by citizen science, my objectives were to 1) identify the spatial scale where regional differences in mast production occur, 2) delineate ecoregional mast production units in accordance with the regional spatial scale, 3) identify the spatial scale where mast production varies within regions to estimate how data should be dispersed within units, and 4) calculate hard- and soft-mast production indices for select areas in Michigan based on field data collection.  2.2. Methods 2.2.1.  Mast Sampling Protocol The hard-mast sampling protocol I used was based on Greenberg and Warburton (2007), who showed that reliable hard-mast indices could be produced by sampling the proportion of trees bearing acorns in a specified area. By using a relatively fast, visual survey of the tree canopy, Greenberg and Warburton (2007) suggested that a tree with acorns on >33.5% of the canopy be designated as high production. Although the study was limited to oak, I assumed that this observation could apply to other hard mast species (e.g., beech, walnut (Juglans spp.), hickory (Carya spp.). In my study, I asked volunteers to opportunistically select potential mast producing trees and describe  I used a similar three-choice rating system for soft-mast data collection. Considering that soft-mast species exhibit a greater variety of life forms (i.e., individual trees, shrubs, patches of plants) than do hard-mast species, I developed a rating system similar to the hard mast scheme,   48  but applicable to all life forms. I asked volunteers to opportunistically select individual plants or patches of plants and rate present and they may be sFruits are abundant and in good  2.2.2. Regional Variation Mast production can vary at both coarse and fine spatial scales (Kelly and Sork 2002). I first assessed regional variation in mast production, as constrained by a sample limited to disparate areas of the northern Lower Peninsula of Michigan (Figure 2.1). Here, I define a region as an area with clustered samples where mast production significantly differs from other clusters of samples. I first used a -function test to determine the scale(s) at which my point data exhibited clustering (Dixon 2002).  Given that the point data I was using exhibited clustering, I subsequently tested for variability in mast production by classifying all of my point data into different sized clusters using the K-means classification algorithm (Jain 2010). This algorithm partitions data points into a user-specified number of clusters by randomly placing centroids throughout the study area, and subsequently assigning points to the nearest centroid that reduces the squared error (distances between points and their assigned centroids). This process is iterated until the algorithm converges to a local minimum (Jain 2010). I evaluated 3 different regional scales by inputting 2, 3, and 4 different clusters into the K-means algorithm. For each cluster classification, I used 1,000 centroid initializations and 100,000 iterations to ensure convergence occurred, in program R 3.1.2 (R Core Team 2014). The proportion of trees bearing nuts in >33% of the canopy and the proportion of soft mast plants producing abundant fruits in each cluster were calculated and tested for differences among other clusters at the same scale using a Fishers exact test (McDonald 2009).  est was used in place of a   49  Chi-square test because it is more accurate when expected values are < 5, which occurred in this analysis because data representing abundant mast observations were relatively rare (McDonald 2009) (Table 2.1). Regions were identified and mapped as the smallest scale (most clusters) that showed significant variation in mast production among the point clusters. I performed the clustering and comparison process for each genus, subgenus, and species with >30 observations in a localized area, with those localized areas separated by >50 km. Only enough data from the oak genus, red oak subgenus (Q. rubra and Q. palustris), and black cherry (Prunus serotina) were available to implement the regionalization procedures.  2.2.3. Within-region Variation Regional clusters can account for larger scale variation in mast production, but variability can also exist at smaller scales (Kelly and Sork 2002). To identify the scale that mast production varied within regions, I used methods similar to the regional classification process. Data points within a region were classified into smaller clusters again using the K-means method (Figure 2.2). Each region was partitioned into 2, 3, 4, 5, and 6 clusters. Subsequently, the proportion of trees bearing nuts on > 33% of the canopy and the proportion of soft mast plants producing abundant fruits in each cluster were calculated. I again tested for differences among clusters within a region using xact test. 5 different scales, and the largest scale (fewest clusters) showing significant differences in mast production among clusters was deemed the optimal spatial configuration within a region. I averaged the distances separating the centroids of the optimal configuration to estimate the spacing of the masting pattern as a guide to the minimum distance that data collection should be distributed within a region.     50  2.2.4.  Using Ecoregions to Index Mast Production Variability in mast production among species is partially determined by weather patterns (Norton and Kelly 1988, Sork 1993, Kelly 1994, Houle 1999, Pearse et al. 2014). For example, Pearse et al. (2014) found temperatures during the previous two spring flowering periods to be correlated with production amounts. Thus, mast production indices should be calculated for regions with similar biotic and abiotic characteristics (Kelly and Sork 2002). Michigan ecoregions represent areas with similar climate, physiography, soil, and biotic communities (Albert 1995). The hierarchical classification of Albert (1995) includes 4 levels: Section - primarily driven by long-term climatic records (n = 5 in Michigan, mean = ~42,500 km2), Subsection - climate and physiography (n = 21, mean = ~6,800 km2; excluding Lac Veaux Desert Outwash Plain), and Sub-subsection - physiography and soil (n = 38, mean = 3,400 km2; Albert 1995). Based on results from the regional and within-region spatial variation of mast production, I used the subsection level of the hierarchy to demonstrate calculation of a mast production index using a land classification scheme that might help account for among region variability. The size of an ecoregion for calculating mast production indices should be small enough that variation among major vegetation cover types (e.g., mast producing species richness) is minimized but large enough to represent broad-scale phenomenon known to affect mast (e.g., precipitation, temperature). Ecoregions that are too large may yield average mast indices that fail to accurately portray the entire area, whereas ecoregions that are too small may result in over-delineation of similar mast producing areas and thus require more total data (i.e., not an efficient sampling strategy). Given the results of my regional analysis, I selected the Subsection level of the Albert (1995) hierarchy because 1) it was comprised of units that had an area most similar to   51  the regional scale of mast production variability in my study, and 2) was comprised of units that were large enough to accommodate the minimum separation distance among sample points that was identified during my within region analysis. 2.2.5. Mast Index Calculation  Using data collected by volunteers and researchers, I calculated one hard-mast index and one soft-mast index in one ecoregional unit. To estimate a hard-mast index using data collected by quick visual surveys, I used the predictive equation shown to calculate an index comparable to that calculated using a more labor-intensive sampling scheme (e.g., Whitehead 1969) developed in a 20-year study of 5 oak species (Quercus alba, Q. prinus, Q. coccinea, Q. rubra, and Q. velutina) in western North Carolina (Greenberg and Warburton 2007):  = b0 + b1  Where  is the predicted hard-mast index, b0 is the model intercept resulting from a reduced major axis regression model in their study (0.403), b1 is the coefficient estimated from that model (0.069), and   is the percentage of trees bearing nuts in a given year. I used the Whitehead (1969) equation because a time series of mast production data (15-20 years; Greenberg and Warburton 2007) did not exist for Michigan. Greenberg and Warburton (2007) recommend that >165 observations per area (~7,500 km2 in their study) and per species or genus are needed to produce hard mast production indices that are within the 80% confidence level.  To calculate a soft-mast index, I adapted a technique from the West Virginia Division of Natural Resources (Richmond et al. 2015). To my knowledge, no studies have evaluated the sample sizes required to accurately describe soft mast production. I generated the soft mast index on a 0 - 10 scale (where 0 represents poor and 10 represents high production) as:   52  Soft-Mast Index (SMI) = (0.5 ÷ total observations) X 10) Lastly, to estimate the minimum amount of data needed throughout the state to accurately produce mast indices, I tested the maximum percentage of data needed from the original dataset to calculate an index value that did not significantly differ from the original value. I randomly removed data at 10% intervals (10,000 iterations), calculated a new index, and tested these values to the original index value using a one-sample t-test (McDonald 2009).  2.3. Results In 2015, volunteers submitted 40 mast production observations and I supplemented these data with 843 observations in the northern Lower Peninsula concentrated in the Gladwin State Forest, Manistee National Forest, and Benzie County (Figure 2.1).  I used 617 of these points to assess regional and within-region variation for the oak genus, 433 points for red oak subgenus, and 89 points for black cherry (Table 2.1). The average area of regional clusters ranged from 70,518-313,211 ha for hard mast and 6,518-51,877 ha for black cherry (Table 2.2). Significant differences (p < 0.05) among clusters were only observed for all oaks combined (Table 2.2). All oaks combined partitioned into 3 clusters, with an average area of 127,298 ha as the smallest scale where significant variability in mast production occurred (Table 2.2). Four clusters for all oak (p = 0.052) and 2 clusters for red oak (p = 0.058) also warrant evaluation once more data are collected (Table 2.2).  The average area of within-region clusters ranged from 1,187-18,140 ha for hard mast and 441-3,858 ha for black cherry (Table 2.3). Significant differences (p < 0.05) of mast production within a regional cluster were only detected for the oak genus and red   53  oak subgenus groups, but not in every region (Table 2.4). The largest scale where I detected significant within region variability occurred with 4 clusters and an average area of 4,034 ha for all oaks combined (Tables 2.3; 2.4). The average distance between the 4-cluster centroids was 31.87 km and this was the shortest average distance of all oak genus cluster scales (Table 2.3).  The smallest regional scale that showed significant variability (3 clusters, mean=127,298 ha) most closely corresponded to the Albert (1995) sub-subsection scale (mean=337,363 ha). However, within-region variability in mast production from my sample occurred at distances (31.87 km) larger than the linear dimensions of several sub-subsection boundaries, suggesting that the hierarchical level of subsection may provide more appropriate mast production boundaries. The subsection was subsequently used to calculate mast index values. Only one subsection (Newaygo Outwash Plain) contained enough hard- or soft-mast observations that were dispersed at distances of at least 31.87 km throughout the majority of the area. This area contained 21 pin oak, 149 northern red oak, 128 white oak, and 70 black cherry observations. Resultant index values were 0.54 for oak and 0.71 for black cherry production. No significant difference (One Sample T-test, p < 0.05) was observed in either index value when compared to values calculated with lesser amounts of data.  2.4. Discussion To demonstrate how mast production indices could be calculated from relatively simple data like those collected by citizen science, I quantified the spatial scales where differences in mast production could be detected to help guide how these data should be summarized to support wildlife management. Recognizing that these results may vary by year, I found that mast production significantly differed for all oaks among areas that averaged 127,299 ha. I failed to detect a significant region effect for the red oak subgenus or black cherry. Other studies have   54  failed to find a consistent distance at which mast synchrony occurs, with distances ranging between 135 km (Koenig et al. 1994) and 1,000 km (Koenig and Knops 1998). Within these areas, I found that oak (all oaks combined and the red oak subgenus) production varied on average every 3,500  4,000 ha. Additionally, the average distance between clusters of points that were producing different amounts of hard mast was ~32 km (oak genus) and 33 km (red oak subgenus), suggesting that mast samples can be collected from relatively dispersed locations within regions. These results may prove useful in allocating sampling effort.   Mast production varies with factors like soils, temperature, and precipitation which are components of ecological land classification schemes (Kelly and Sork 2002). Given that I detected significant patterns in large- and mid-scale hard mast production in Michigan, I used my results to help identify a scale of the Albert ecoregional classification scheme (Albert 1995) that might prove useful for mast data collection. Although my region-level analysis suggested that the sub-subsection of the Albert classification was most appropriate, the linear distances over which differences in hard mast production were detected suggested that the subsection level was more appropriate for sampling efficiency. Sub-subsections are smaller than subsections and many sub-subsections are not large enough to sample at distances beyond 33 km. Choosing the larger subsection of the ecoregional hierarchy would not necessarily result in sampling less total area, but would allow more optimal placement of sample locations.  The process of calculating mast production indices demonstrated here was only based on 1,137 data points collected in 2015 and not evenly dispersed throughout the state. Only 5% of the total records indicated that mast production was abundant, suggesting that 2015 was a low producing year. The resultant HMI of 0.54 in my study indicated a crop failure based on the interpretation of values provided by Greenberg and Warburton (2007). However, to best interpret   55  the results from using their equations, long term mast production data (15-20 years) is needed to know what the true maximum proportion of trees bearing acorns is in the study area. It is unrealistic to assume it is possible for 100% of trees to bear acorns in a given year, meaning an index range of 0-10 may not be accurate. By knowing this maximum potential percentage witnessed over a long period of time, results can be standardized by dividing the percentage of trees bearing acorns in a given year, by the maximum percentage witnessed. By doing so, a subjective rating of mast production (Failure, Poor, Average, Good, Bumper) can be assigned to this standardized value as well as the historically used HMI value (Warburton and Greene 2007). Until the maximum potential percentage is identified, using the percentage of trees bearing acorns as the hard-mast index will be the value most easily interpreted and accurate in comparing production amounts amongst years.    In years when production is better, variation among different spatial scales is likely to be greater. When comparing within region variability in mast production, I found that masting differences occurred at smaller scales depending on the region. This relatively fine-scale variability in masting is common (Kelly and Sork 2002), and is why more data over longer periods of time are needed to more precisely determine the proper scale at which data should be sampled. I suggest that more data be collected at the scales I have identified in this study, and that this analysis be replicated each year until consistent results are found among years.  A sensitivity analysis using my data found that the hard and soft mast production indices were invariant to changes in sample sizes. I suspect that this result most likely relates to the preponderance of plants in my data that did not produce mast in 2015. Greenberg and Warburton (2007) showed that less data are required to accurately estimate hard mast indices in boom or bust years due to the synchrony observed across large scales, meaning more data are required   56  during moderate production years. I recommend that mast amounts be recorded on multiple nearby plants at each location where data is collected to produce a fine-scale estimate of mast production relative to that local area.  I present a sampling protocol that can be used by untrained volunteers to produce data to develop relative hard- and soft-mast production indices in Michigan. Using citizen scientists to collect data can make monitoring statewide wildlife food sources attainable with limited resources. This protocol can easily be adapted to surrounding states to expand data collection to further the understanding of masting patterns in the Midwest. 2.5. Acknowledgements  I would like to thank T. Dokes, T. Swem, and A. Tiernan for their assistance in data collection, and A. Ziegler, G. Carballo, and C. Peterson for their efforts in structuring the online database.                                57                     APPENDIX                             58  Table 2.1. Total mast observation amounts by species and genera in Michigan, USA, 2015.       Amount  Hard Mast   None <33% >33% Total   American beech 44 1 0 45  Beaked hazelnut 2 3 0 5  Bitternut hickory 1 0 0 1  Black walnut 0 2 1 3  Bur oak  0 0 1 1  Northern red oak 284 103 24 411  Pin oak  9 11 2 22  Shagbark hickory 0 0 1 1  Swamp white oak 5 3 0 8  White oak  151 24 1 176     Grand Total 673           Amount  Soft Mast   None Few Many Total   Apple  0 3 1 4  Autumn olive 6 12 5 23  Black cherry 66 15 8 89  Blackberry  7 7 5 19  Blueberry  18 6 1 25  Chokecherry 8 8 0 16  Crabapple  0 0 1 1  Dogwood  1 2 10 13  Dwarf red raspberry 2 5 0 7  Elderberry  0 1 0 1  Hawthorn  7 2 1 10  Highbush cranberry 0 1 2 3  Pokeweed  0 0 1 1  Serviceberry 0 1 1 2  Wild grape 2 2 0 4     Grand Total 218           59  Table 2.2.  Average size (ha) of regional clusters classified by the K-means method for 2, 3, and 4 clusters. p-est for mast production variability among clusters.   Regional Oaks Red Oaks Black Cherry Clusters (n) Area  p-value Area  p-value Area  p-value  (ha) (ha) (ha) 2    313,211  0.011    312,614  0.058    51,877  0.185 3    127,298  0.034    126,787  0.180    13,533  0.275 4      71,068  0.052      70,518  0.161      6,518  0.195  Table 2.3. Average size (ha) of within-region clusters and average distance (km) from centroids (when regions are classified into 3 clusters).  Oaks Red Oaks Black Cherry Within-region Area  Distance Area  Distance Area  Distance Clusters(n) (ha) (km) (ha) (km) (ha) (km) 2      18,140  48.65      17,889  48.90         3,858  18.07 3         6,878  39.05         6,692  39.02         1,780  15.34 4         4,034  31.87         3,446  33.00            950  12.81 5         2,351  32.72         2,229  32.68            679  72.94 6         1,437  89.45         1,187  30.05            441  80.47  Table 2.4 Within-region clusters (when regions are classified into 3 clusters) classified by the K-means method for 2, 3, 4, 5 and 6 clusters. p-variability among clusters.     p-value   Within-region Clusters (n)  Region 2 3 4 5 6 All Oaks A 1.000 1.000 0.707 0.707 0.787  B 0.381 0.080 0.039 0.001 0.004  C 0.224 0.388 < 0.001 < 0.001 < 0.001        Red Oaks A 0.603 0.733 < 0.001 < 0.001 < 0.001  B 1.000 1.000 0.839 0.839 0.859  C 0.366 0.204 0.211 0.004 0.018        Black Cherry A 1.000 1.000 1.000 1.000 1.000  B 0.187 0.259 0.389 0.478 0.589  C 1.000 1.000 1.000 1.000 1.000     60  Figure 2.1.  Mast production data used in cluster classification (Quercus bicolor, Q. alba, Q. palustris, Q. rubra, and Prunus serotina; n=704) and subsection ecoregional (n=22) in Michigan, USA, 2015.         61  Figure 2.1.  Mast production data used in cluster classification (Quercus bicolor, Q. alba, Q. palustris, Q. rubra, and Prunus serotina; n=704) and subsection ecoregional units (n=22) in Michigan, USA, 2015.                     62  Figure 2.2. Regional clusters of all oak (Quercus) observations and within-region clusters for regional cluster C, Michigan, USA, 2015.       63                         LITERATURE CITED                         64  Albert, D. A. 1995. Regional landscape ecosystems of Michigan, Minnesota and Wisconsin: a working map and classification. General technical report NC-178. U.S. Forest Service North Central Forest Experiment Station. St. Paul, Minnesota.  Boucher, D. H. 1981. Seed predation by mammals and forest dominance by Quercus oleoides, a tropical lowland oak. Oecologia 49:409-419.  Clark, J. D., F. T. van Manen, and M. R. Pelton. 2005. Bait stations, hard mast, and black bear population growth in Great Smoky Mountains National Park. Journal of Wildlife Management 69:1633-1640.  Daubenmire, R. 1959. A canopy-coverage method of vegetational analysis. Northwest Science 33:43-64.  Dixon, P. M. 2002. Ripley's K function. Encyclopedia of Environmetrics 3:1796-1803.  Elowe, K. D., and W. E. Dodge. 1989. Factors affecting black bear reproductive success and cub survival. Journal of Wildlife Management 53:962-968.  Feldhamer, G. A. 2002. Acorns and white-tailed deer: interrelationships in forest ecosystems. Pages 215223 in W. J. McShea and W. M. Healy, editors. Oak forest ecosystems: ecology and management for wildlife. The Johns Hopkins Press, Baltimore, Maryland, USA.   Graves, W. C. 1980. Annual oak mast yields from visual estimates. Pages 270-274 in T. R. Plumb, technical coordinator. Proceedings of the Symposium on the Ecology, Management, and Utilization of California Oaks. Pacific Southwest Forest and Range Experiment Station General Technical Report PSW-44, Berkeley, California, USA.   Greenberg, C. H., and B. R. Parresol. 2002. Dynamics of acorn production by five species of Southern Appalachian oaks. Pages 149-172 in W. J. McShea and W. M. Healy, editors. Oak forest ecosystems: ecology and management for wildlife. The Johns Hopkins Press, Baltimore, Maryland, USA.   Greenberg, C. H., and G. S. Warburton. 2007. A rapid hard-mast index from acorn presence-absence tallies. Journal of Wildlife Management 71:1654-1661.  Gysel, L. W. 1971. A 10-year analysis of beechnut production and use in Michigan. Journal of Wildlife Management 35:516-519.  Herrera, C. M., P. Jordano, J. Guitián, and A. Traveset. 1998. Annual variability in seed production by woody plants and the masting concept: reassessment of principles and relationship to pollination and seed dispersal. The American Naturalist 152:576-594.    65  Houle, G. 1999. Mast seeding in Abies balsamea, Acer saccharum and Betula alleghaniensis in an old growth, cold temperate forest of northeastern North America. Journal of Ecology 87:413-422.  Jain, A. K. 2010. Data clustering: 50 years beyond K-means. Pattern Recognition Letters. 31:651-666.  Janzen, D. H. 1971. Seed predation by animals. Annual review of ecology and systematics 2:465-492.  Kelly, D. 1994. The evolutionary ecology of mast seeding. Trends in Ecology and Evolution 9:465-470.  Kelly, D., and V. L. Sork. 2002. Mast seeding in perennial plants: why, how, where? Annual Review of Ecology and Systematics 33:427-447.  Koenig, W. D., and J. M. Knops. 1998. Scale of mast-seeding and tree-ring growth. Nature 396:225-226.  Koenig, W. D., and J. M. Knops. 2000. Patterns of annual seed production by northern hemisphere trees: a global perspective. The American Naturalist 155:59-69.  Koenig, W. D., J. M. H. Knops, W. J. Carmen, M. T. Stanback, and R. L. Mumme. 1994. Estimating acorn crops using visual surveys. Canadian Journal of Forest Research 24:2105-2112.  Lashley, M. A., J. R. Thompson, M. C. Chitwood, C. S. DePerno, and C. E. Moorman. 2014. Evaluation of methods to estimate understory fruit biomass. PloS ONE 9:e96898. doi: 10.1371/journal.pone.0096898  Martin, A. C., H. S. Zim, and A. L. Nelson. 1951. American wildlife and plants: a guide to wildlife food habits. McGraw-Hill, New York, New York, USA.  McDonald, J. H. 2009. Handbook of biological statistics. Sparky House Publishing, Baltimore, Maryland, USA.  Norton, D. A., and D. Kelly. 1988. Mast seeding over 33 years by Dacrydium cupressinum Lamb. (rimu) (Podocarpaceae) in New Zealand: the importance of economies of scale. Functional Ecology 2:399-408.  Pearse, I. S., W. D. Koenig, and J. M. H. Knops. 2014. Cues versus proximate drivers: testing the mechanism behind masting behavior. Oikos 123:179-184.  Perry, R. W., and R. E. Thill. 1999. Estimating mast production: an evaluation of visual surveys and comparison with seed traps using white oaks. Southern Journal of Applied Forestry 23:164-169.   66  R Core Team. 2014. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria.  Reynolds-Hogland, M. J., L. B. Pacifici, and M. S. Mitchell. 2007. Linking resources with demography to understand resource limitation for bears. Journal of Applied Ecology 44:1166-1175.  Richmond, E.S., C. W. Ryan, C.P. Carpenter, and M. L. Peters. 2015. West Virginia mast survey and hunting outlook. West Virginia Division of Natural Resources Section. Wildlife Resources Bulletin Number 15-3.   Ryan, C. W., J. C. Pack, W.K. Igo, J.C. Rieffenberger, and A. B. Billings.  2004.  Relationship of mast production to big-game harvests in West Virginia. Wildlife Society Bulletin 32:786-794.  Selas, V. 2000. Seed production of a masting dwarf shrub, Vaccinium myrtillus, in relation to previous reproduction and weather. Canadian Journal of Botany 78:423-429.  Sharp, W. M. 1958. Evaluating mast yields in the oaks. Pennsylvania State University Agriculture Experiment Station Bulletin 635, University Park, USA.  Silvertown, J. W. 1980. The evolutionary ecology of mast seeding in trees. Biological Journal of the Linnean Society 14:235-250.  Sork, V. L. 1993. Evolutionary ecology of mast-seeding in temperate and tropical oaks (Quercus spp.). Vegetatio 107:133-147.  Steffen, D. E., N. W. Lafon, and G. W. Norman. 2002. Turkeys, acorns, and oaks. Pages 241-255 in M. J. McShea and W. M. Healy, editors. Oak forest ecosystems: ecology and management for wildlife. The Johns Hopkins Press, Baltimore, Maryland, USA.  Uhlig, H. G., and H.L. Wilson. 1952. A method of evaluating an annual mast index. Journal of Wildlife Management 16:338-343.  Whitehead, C. J. 1969. Oak mast yields on wildlife management areas in Tennessee. Tennessee Game and Fish Commission, Nashville, USA.  Wolff, J. O. 1996. Population fluctuations of mast-eating rodents are correlated with production of acorns. Journal of Mammalogy 77:850-856.        67  CONCLUSIONS The purpose of this thesis was to describe the feasibility of using citizen scientists to develop annual mast-production indices in Michigan, and to describe a process for analyzing citizen science data to identify scales of variability in mast production. My work can be used to assess the effectiveness of the methods deployed and to inform decisions as to how to best move forward with the coordination of the program. I developed a sampling protocol based on previous scientific findings and adopted the established techniques to ensure accurate data reporting from untrained volunteers. Results from field-validating an initial subset of participant data suggests that data produced by the program can be assumed to be reliable and can be used to accurately estimate mast production indices. Subsequently, I used data collected with this protocol to introduce a process capable of identifying the spatial scales at which hard and soft mast is variable, and how to use this information to estimate production indices at a statewide scale. In the initial years of mast production analyses, the percentage of trees bearing acorns may be the most easily interpreted value of hard mast production by region until a maximum potential production amount has been identified.  Limitations of this work that should be considered are 1) that field validated data were submitted by only 15 volunteers and may not be an accurate representation of data collection capabilities of all participants and 2) mast production indices created in Chapter 2 were created for the purpose of demonstrating the process one would use in the future once adequate data have been obtained. Additional efforts may need to be taken to ensure data is collected for species other than oaks to account for inter-genus production variability. Considering the complimentary working relationships of the partners, the effectiveness of database tools, and accuracy of data   68  submitted by participants thus far, I recommend the continuation of this citizen-science program with an emphasis on recruitment and retention of volunteers.