Search results

Title: Development and application of hierarchical models for monitoring avian soundscapes, populations, and communities
Creator: Doser, Jeffrey W.
Date: 2022
Collection: Electronic Theses & Dissertations
Description: Climate change, land use change, and other anthropogenic pressures are increasing species extinctions, phenology shifts, and drastic population declines. Avian populations and communities are particularly vulnerable to global change given their mobile and migratory life history strategies. Avian abundance has drastically declined throughout North America over several decades, which is compounded by phenological shifts in breeding periods and migratory patterns. Informed management and...
Show moreClimate change, land use change, and other anthropogenic pressures are increasing species extinctions, phenology shifts, and drastic population declines. Avian populations and communities are particularly vulnerable to global change given their mobile and migratory life history strategies. Avian abundance has drastically declined throughout North America over several decades, which is compounded by phenological shifts in breeding periods and migratory patterns. Informed management and conservation of avian populations and communities requires large-scale monitoring programs, as well as associated inferential tools to provide statistically robust inference using multiple data sources. In this dissertation, I develop a suite of hierarchical modeling approaches to understand avian soundscapes, populations, and communities. I leverage a hierarchical Bayesian modeling framework, which is ideally suited for complex wildlife data with numerous types of observation error and dependencies among data points. In Chapter 1, I provide a brief overview of avian monitoring approaches and their associated statistical analysis frameworks. In Chapters 2 and 3, I develop hierarchical models for the analysis of complex avian soundscape data, and apply these approaches to two case studies. In Chapter 2, I apply a two-stage hierarchical beta regression model to quantify the relationship between anthropogenic and biological sounds in avian soundscapes in western New York. In Chapter 3, I use a multivariate linear mixed model to assess disturbance impacts of a shelterwood logging on avian soundscapes in northern Michigan. In Chapter 4, I develop a multi-region, multi-species abundance model to quantify trends of avian species and communities using point count data across a network of National Parks in the northeastern US. In Chapters 5 and 6, I use a model-based data integration approach to yield improved inference on avian population and communities. In Chapter 5, I integrate automated acoustic recording data with point count data to estimate avian abundance, which I apply to a case study on the Eastern Wood Pewee (Contopus virens) in a National Historical Park in Vermont. In Chapter 6, I develop an integrated community occupancy model that combines multiple types of detection-nondetection data for inference on species-specific and community level occurrence dynamics, which I use to assess occurrence dynamics of a foliage-gleaning bird community in New Hampshire. These results exhibit the value of hierarchical models to partition ecological data into distinct observation and ecological components for improved inference on avian population and community dynamics. Future work should continue to leverage complex data sources within hierarchical modeling frameworks to address pressing conservation and management questions on avian populations, communities, and the ecosystem services they provide.
Show less

Title: Computationally efficient hierarchical spatial models for large datasets : a case study for the assessment of forest characteristics across the Lake States
Creator: Zhu, Huirong
Date: 2011
Collection: Electronic Theses & Dissertations
Description: The scientific community is moving into an era where data rich environments provide extraordinary opportunities to understand the spatial complexity of ecological processes. Across scientific fields, researchers face the challenge of coupling these data with imperfect models to better understand variability in their system of interest. In the environmental sciences there is recognized urgent need to develop and disseminate methodology capable of accurately accounting for multiple sources of...
Show moreThe scientific community is moving into an era where data rich environments provide extraordinary opportunities to understand the spatial complexity of ecological processes. Across scientific fields, researchers face the challenge of coupling these data with imperfect models to better understand variability in their system of interest. In the environmental sciences there is recognized urgent need to develop and disseminate methodology capable of accurately accounting for multiple sources of uncertainty. Accordingly, the goal of this thesis was to explore and illustrate the properties of promising new modeling tools that will enable researchers to extract more information from large spatial datasets. In particular, this thesis was motivated by a larger project's need to analyze a large forest inventory dataset with the intent to better understand the potential of managing forests for increased complexity as a climate change mitigation and adaptation strategy. The thesis yields results from the analysis of synthetic and forestry datasets that clearly demonstrate how model misspecification, specifically ignoring spatial dependence among model residuals, can result in incorrect inference about regression parameters of interest. These results have important implications for hypothesis testing and ultimately forest management and policy decisions. The thesis details some modeling tools and useful guidelines that allow practitioners to more fully accommodate model assumptions and draw correct inference for large spatial datasets.
Show less

Title: MONITORING AND MODELING ECOHYDROLOGICAL PROCESSES IN VEGETATED WATERSHEDS
Creator: Pham, Leo Triet
Date: 2022
Collection: Electronic Theses & Dissertations
Description: Ecohydrology links ecological and hydrological processes and considers interactions between water resources and ecosystems. Modeling tools are not only important for studying the mechanisms of ecological patterns and processes but also for assessing the effects of environmental change on hydrological and ecological processes, providing insights and solutions to issues in water management. This thesis explores various data-driven approaches to monitor and model these processes at 95 watersheds...
Show moreEcohydrology links ecological and hydrological processes and considers interactions between water resources and ecosystems. Modeling tools are not only important for studying the mechanisms of ecological patterns and processes but also for assessing the effects of environmental change on hydrological and ecological processes, providing insights and solutions to issues in water management. This thesis explores various data-driven approaches to monitor and model these processes at 95 watersheds in western USA using a combination of seasonal and annual climate, hydrometric, and remotely sensed vegetation data. In one analysis, we show that a trend in earlier peak in spring vegetation activity may be a linked to reduced runoff availability during drought years compared to non-drought years. We also provide evidence that increase drought severity is consistent with a decrease in runoff ratio in forested catchments through regression analysis, supporting the hypothesis that the relationship among water-balance components may shift during drought events. In another analysis, we show that the type and amount of vegetation coverage, among other catchment characteristics, can affect the accuracy of data-driven runoff models. These results suggest that a better understanding of the ecohydrologic processes and characteristics is vital to development of effective long-term strategies to improve the resilience of watersheds.
Show less

Title: Unlocking the forest inventory and analysis database : applications to nation-wide forest health monitoring
Creator: Stanke, Hunter
Date: 2020
Collection: Electronic Theses & Dissertations
Description: Forest Inventory and Analysis (FIA) is a US Department of Agriculture Forest Service program that aims to monitor changes in forests across the US. FIA hosts one of the largest ecological datasets in the world, though its complexity limits access for many potential users. rFIA is an R package designed to simplify the estimation of forest attributes using data collected by the FIA Program. Specifically, rFIA improves access to the spatio-temporal estimation capacity of the FIA Database via...
Show moreForest Inventory and Analysis (FIA) is a US Department of Agriculture Forest Service program that aims to monitor changes in forests across the US. FIA hosts one of the largest ecological datasets in the world, though its complexity limits access for many potential users. rFIA is an R package designed to simplify the estimation of forest attributes using data collected by the FIA Program. Specifically, rFIA improves access to the spatio-temporal estimation capacity of the FIA Database via space-time indexed summaries of forest variables within user-defined population boundaries. The package implements multiple design-based estimators, and has been validated against official estimates and sampling errors produced by the FIA Program. The package has been made open-source is freely available for download from the Comprehensive R Archive Network.In recent decades, forests of the western US have experienced unprecedented change in climate and forest disturbance regimes, and widespread shifts in forest composition, structure, and function are expected in response. However, large-scale, comprehensive assessments of tree population performance have yet to be conducted in the region. We develop an index of forest population performance based on repeated censuses of field plots, and apply this index to assess the status of the most abundant tree species in the western US. Our study provides empirical evidence to suggest tree species in the western US are exhibiting strong divergence in population performance, with over half (70%) of species experiencing range-wide population decline. We found spatial variation in population performance across the ranges of all species, indicating range shifts are already underway. Our results further indicate that species decline can seldom be attributed to a single forest disturbance agent, highlighting the importance of considering multiple risks factors in broad-scale forest management.
Show less

Title: Environmental controls on phenoregions across an East African megatransect
Creator: Desanker, Gloria
Date: 2019
Collection: Electronic Theses & Dissertations
Description: Semi-arid and savanna-type (SAST) systems in East Africa have unique plant species compositions and characteristics that make quantifying this area's seasonality and inter-annual variability difficult. Phenoregion classification offers a way to use seasonality of vegetation growth dynamics to help understand the phenology of complex landscapes. Here, we used Normalized Difference Vegetation Index (NDVI) time series from the Landsat 8 imagery to map phenoregions in scenes centered around...
Show moreSemi-arid and savanna-type (SAST) systems in East Africa have unique plant species compositions and characteristics that make quantifying this area's seasonality and inter-annual variability difficult. Phenoregion classification offers a way to use seasonality of vegetation growth dynamics to help understand the phenology of complex landscapes. Here, we used Normalized Difference Vegetation Index (NDVI) time series from the Landsat 8 imagery to map phenoregions in scenes centered around national parks from Mt. Kenya National Park (Kenya) to Limpopo National Park (Mozambique) to assess whether landscape-scale controls on phenology are consistent across the region or if they differ on a latitudinal gradient. We used MODIS Land Cover to assess land cover composition in each phenoregion, and discriminant analysis to determine the role that elevation, slope and aspect play in driving phenological differences. There was no clear latitudinal pattern seen in land cover or geologic composition. Most of the site's phenoregions showed no unique composition of either of the variables, meaning that land cover or geology type did not help in differentiating phenoregions. The discriminant analysis showed that topography was a strong predictor of many of the phenoregions, however, these also did not reveal any clear latitudinal pattern. Using seasonality of the NDVI time series to generate phenoregions provides different and even in some cases more ecologically relevant information, compared to past studies that use only land cover to generate ecoregions. With a significant population of humans and animals that live in and depend on SAST ecosystems, it is important to better understand vegetation processes and the factors that affect them as climate change becomes an increasingly pertinent issue in dry systems.
Show less

Title: Hierarchical Bayesian models for small area estimation of biophysical and social forestry variables
Creator: Ver Planck, Neil Ryan
Date: 2018
Collection: Electronic Theses & Dissertations
Description: Forest inventories and surveys, accounting for time and cost constraints, are typically designed to yield accurate and precise estimates of population means and totals over large spatial domains. In many instances, these inventories and surveys also offer reliable inference for smaller subpopulations with sufficient sample observations; however, there is growing demand for valid and precise estimates at levels that have smaller sample sizes based on the original sample design. One solution to...
Show moreForest inventories and surveys, accounting for time and cost constraints, are typically designed to yield accurate and precise estimates of population means and totals over large spatial domains. In many instances, these inventories and surveys also offer reliable inference for smaller subpopulations with sufficient sample observations; however, there is growing demand for valid and precise estimates at levels that have smaller sample sizes based on the original sample design. One solution to this problem is application of small area estimationmethods. Small area estimation (SAE) is a model-based approach that couples a direct estimate and possible covariates to improve the estimate precision and, in some cases, accuracy. Unlike a standard linear regression, the SAE framework is comprised of two components: a sampling model and a linking model. Estimation of the SAE parameter of interest accounts for and balances between the sampling (i.e., direct estimator) and linking model errors. The linking model is a linear model with random effects that relate the small areas of interest with some error. Additional spatial structure might still remain in the linking model after accounting for possible covariates. Such residual structure can be further modeled using spatial random effects.This dissertation presents SAE methods within a hierarchical Bayesian (HB) framework. This framework is applied to common biophysical forest inventory outcomes of interest (i.e., aboveground biomass, basal area, volume, and tree density) at the stand level, and to thesocial forestry survey outcomes of private forest landowner populations. Furthermore, an in depth examination of the direst estimator, in the presence of nonresponse, is assessed for private forest landowner population size. The primary objectives of this dissertationare: i) to apply a HB framework to increase the precision of estimates for biophysical forest variables at the stand level by borrowing strength across all stands through the use of LiDAR covariates; ii) to apply a conditional autoregressive structure to the stand-levelrandom effects to assess gains in precision of biophysical forest variables; iii) to evaluate the current National Woodland Owner Survey estimators of private forest area and private forest landowner population size for a known population at the state level; iv) to presentan alternative estimator of private forest landowner population size that explicitly accounts for various nonresponse scenarios; v) to evaluate the impacts of nonresponse biases on each of these estimators; vi) to produce county-level private forest ownership datasets for two complete states; vi) to define and assess SAE models to improve county-level inference of the number of private forest ownerships, and; vii) to develop open source software to fit proposed SAE models.
Show less

Title: Bayesian hierarchical spatial models to improve forest variable prediction and mapping with light detection and ranging data sets
Creator: Babcock, Chad
Date: 2014
Collection: Electronic Theses & Dissertations
Description: Light Detection and Ranging (LiDAR) data has shown great potential to estimate spatially explicit forest variables, including above-ground biomass, stem density, tree height, and more. Due to its ability to garner information about the vertical and horizontal structure of forest canopies effectively and efficiently, LiDAR sensors have played a key role in the development of operational air and space-borne instruments capable of gathering information about forest structure at regional,...
Show moreLight Detection and Ranging (LiDAR) data has shown great potential to estimate spatially explicit forest variables, including above-ground biomass, stem density, tree height, and more. Due to its ability to garner information about the vertical and horizontal structure of forest canopies effectively and efficiently, LiDAR sensors have played a key role in the development of operational air and space-borne instruments capable of gathering information about forest structure at regional, continental, and global scales. Combining LiDAR datasets with field-based validation measurements to build predictive models is becoming an attractive solution to the problem of quantifying and mapping forest structure for private forest land owners and local, state, and federal government entities alike. As with any statistical model using spatially indexed data, the potential to violate modeling assumptions resulting from spatial correlation is high. This thesis explores several different modeling frameworks that aim to accommodate correlation structures within model residuals. The development is motivated using LiDAR and forest inventory datasets. Special attention is paid to estimation and propagation of parameter and model uncertainty through to prediction units. Inference follows a Bayesian statistical paradigm. Results suggest the proposed frameworks help ensure model assumptions are met and prediction performance can be improved by pursuing spatially enabled models.
Show less

Title: The development and application of spatio-temporal methods to understand and predict broad-scale patterns of forest change
Creator: Itter, Malcolm S.
Date: 2018
Collection: Electronic Theses & Dissertations
Description: "The function, composition, and health of regional forest systems are driven by factors operating at a range of spatio-temporal scales. Climate shapes regional species composition at centennial-to-millennial timescales, but may also contribute to more rapid forest change through the occurrence of climate extremes. Disturbance events operate at scales ranging from individual trees to landscape-level metacommunities impacting forest dynamics and resetting forest succession and development over...
Show more"The function, composition, and health of regional forest systems are driven by factors operating at a range of spatio-temporal scales. Climate shapes regional species composition at centennial-to-millennial timescales, but may also contribute to more rapid forest change through the occurrence of climate extremes. Disturbance events operate at scales ranging from individual trees to landscape-level metacommunities impacting forest dynamics and resetting forest succession and development over decadal-to-centennial time frames. At the local-scale, forest function, composition, and health at a given time are determined by forest demographic processes including growth, mortality, and regeneration. Understanding and predicting broad-scale patterns of forest change requires methods to integrate these different factors synthesizing information across spatio-temporal scales. The research presented here focuses on the development and application of spatio-temporal, Bayesian hierarchical methods to advance understanding of the processes and factors driving large-scale forest change. The methods seek to make inference about latent forest processes of interest based on noisy observations of forest demographics, climate, and disturbance events. The impacts of novel climatic conditions forecast to occur over the next century on forest ecosystem function are difficult to predict given potential interactions between climate, disturbance events, and forest characteristics such as species composition, density, and tree size/age distribution. The first three chapters of the following dissertation focus on the development and application of methods to advance understanding of such interactions. First, a dynamic Bayesian hierarchical model is presented allowing forest growth responses to climate variables to vary over time in relation to past climate extremes, disturbance events, and forest dynamics. The model was applied to tree-ring data from a range of sites within northeastern Minnesota. Results revealed significant growth responses to soil water availability triggered by large climatic water deficits across multiple seasons and years, forest tent caterpillar defoliation events, and high forest density following large regeneration events. Building on these results, the interactive effects of past water deficit and insect defoliation stress on forest growth were further explored using broad-scale tree-ring and defoliation data from two regions of the Canadian boreal forest with contrasting species compositions, primary insect defoliators, and regional climates. A series of novel methods were developed to quantify the ecological memory of boreal trees to antecedent water and insect defoliation stress. Results highlighted the temporal persistence of drought and defoliation stress on boreal tree growth dynamics and provided an empirical estimate of their interactive effects. Finally, a Bayesian state space framework for the assimilation of tree-ring and forest inventory data with a forest growth and yield model (Forest Vegetation Simulator) was developed to reconstruct forest dynamics with explicit uncertainty. The framework allows for the use of tree-ring data to inform growth-climate relationships and inventory data to inform estimates of past forest composition, density, and tree size/age distribution. The unique inference afforded by the framework is demonstrated through its application to red pine plantation data from northern Minnesota. The final chapter of the dissertation presents a Bayesian point process model for the reconstruction of past fire regimes using sediment charcoal data. The framework was applied to a network of boreal forest lakes in interior Alaska demonstrating a significant reduction in the uncertainty of past fire identification compared to existing methodologies. Further, results highlighted shifts in the regional fire regime coincident with changes in regional species composition over the past 223C10,000 years. The methods developed herein and their application to a range of forest data types provide increased understanding of the multi-scale factors contributing to changes in forest growth and mortality over time and space. Still missing, however, is a process-based framework that integrates the various spatio-temporal methods presented to gain mechanistic understanding of forest responses to extreme climate and disturbance events. Future work is needed to develop such a framework and apply it to extensive regional forest data sets to advance mechanistic understanding and predict forest responses to the novel environmental conditions of the 21st century."--Pages ii-iii.
Show less

Title: Bayesian hierarchical models for environmental datasets
Creator: Matney, Jason Andrew
Date: 2014
Collection: Electronic Theses & Dissertations
Description: This thesis explores the applicability of Bayesian spatial models for predicting the occurrence of permafrost across Alaska, USA. Exploratory analysis of a large Alaska soil carbon database suggests the impact of some important environmental covariates on permafrost occurrence is non-linear. Also, exploratory analysis using non-spatial regression models shows that substantial spatial autocorrelation among residuals exists even after accounting for available covariates. Spatial regression...
Show moreThis thesis explores the applicability of Bayesian spatial models for predicting the occurrence of permafrost across Alaska, USA. Exploratory analysis of a large Alaska soil carbon database suggests the impact of some important environmental covariates on permafrost occurrence is non-linear. Also, exploratory analysis using non-spatial regression models shows that substantial spatial autocorrelation among residuals exists even after accounting for available covariates. Spatial regression models specifically designed to accommodate non-linearity between covariates and probability of permafrost are developed and tested. Results show the proposed models provide improved fit and predictive ability over conventional modeling techniques. Considerations for applying the proposed models to large spatial domains for creating high-resolution map permafrost products are also discussed.
Show less

MSU Libraries

Digital Repository

Enabled Filters

MSU Libraries

Digital Repository

You are here

Search results