This dissertation investigated the practicality and expediency of applying remote sensing data fusion products to the analysis of dryland vegetation phenology. The objective of the first study was to verify the quality of the output products of the spatial and temporal adaptive reflectance fusion method (STARFM) over the dryland Arizona study site. Synthetic 30 m resolution images were generated from Landsat-5 Thematic Mapper (TM) data and a range of 500 m Moderate Resolution Imaging Spectroradiometer (MODIS) surface reflectance datasets and assessed via correlation analysis with temporally coincident Landsat-5 imagery. The accuracy of the results (0.61 < R2 < 0.94) justified subsequent use of STARFM data in this environment, particularly when the imagery were generated from Nadir Bi-directional Reflectance Factor (BRDF)-Adjusted Reflectance (NBAR) MODIS datasets. The primary objective of the second study was to assess whether synthetic Landsat data could contribute meaningful information to the phenological analyses of a range of dryland vegetation classes. Start-of-season (SOS) and date of peak greenness phenology metrics were calculated for each STARFM and MODIS pixel on the basis of enhanced vegetation index (EVI) and normalized difference vegetation index (NDVI) time series over a single growing season. The variability of each metric was calculated for all STARFM pixels within 500 m MODIS extents. Colorado Plateau Pinyon Juniper displayed high amounts of temporal and spatial variability that justified the use of STARFM data, while the benefit to the remaining classes depended on the specific vegetation index and phenology metric. The third study expanded the STARFM time series to five years (2005-2009) to examine the influence of site characteristics and climatic conditions on dryland ponderosa pine (Pinus ponderosa) forest phenological patterns. The results showed that elevation and slope controlled the variability of peak timing across years, with lower elevations and shallower slopes linked to higher levels of variability. During drought conditions, the number of site variables that controlled the timing and variability of vegetation peak increased. / Ph. D.

The need for an efficient dimensionality reduction technique has remained a critical challenge for effective analysis of hyperspectral data for vegetation applications. Discrete wavelet transform (DWT), through multiresolution analysis, offers oppurtunities both to reduce dimension and convey information at multiple spectral scales. In this study, we investigated the utility of the Haar DWT for AVIRIS hyperspectral data analysis in three different applications (1) classification of three pine species (Pinus spp.), (2) estimation of leaf area index (LAI) using an empirically-based model, and (3) estimation of LAI using a physically-based model. For pine species classification, different sets of Haar wavelet features were compared to each other and to calibrated radiance. The Haar coefficients selected by stepwise discriminant analysis provided better classification accuracy (74.2%) than the original radiance (66.7%). For empirically-based LAI estimation, the models using the Haar coefficients explained the most variance in observed LAI for both deciduous plots (cross validation R² (CV-R²) = 0.79 for wavelet features vs. CV-R² = 0.69 for spectral bands) and all plots combined (CV R² = 0.71 for wavelet features vs. CV-R² = 0.50 for spectral bands). For physically-based LAI estimation, a look-up-table (LUT) was constructed by a radiative transfer model, DART, using a three-stage approach developed in this study. The approach involved comparison between preliminary LUT reflectances and image spectra to find the optimal set of parameter combinations and input increments. The LUT-based inversion was performed with three different datasets, the original reflectance bands, the full set of the wavelet extracted features, and the two wavelet subsets containing 99.99% and 99.0% of the cumulative energy of the original signal. The energy subset containing 99.99% of the cumulative signal energy provided better estimates of LAI (RMSE = 0.46, R² = 0.77) than the original spectral bands (RMSE = 0.69, R² = 0.42). This study has demonstrated that the application of the discrete wavelet transform can provide more accurate species discrimination within the same genus than the original hyperspectral bands and can improve the accuracy of LAI estimates from both empirically- and physically-based models. / Ph. D.

Automation of urban tree inventory using remote sensing is needed not only to reduce inventory costs but also to support carbon accounting for urban planners and policy-makers. However, urban areas are heterogeneous and complex, and a more sophisticated approach is needed for using remote-sensing technology like lidar for tree inventory in urban areas than is required for forested environments. Based on remote sensing and field data from a suburban residential area in the central United States, this dissertation presents a methodology for utilizing airborne small-footprint lidar data to inventory urban trees. This dissertation proposes approaches that have the potential to automate three main activities of urban tree inventory -- identifying the locations of trees, classifying the trees into taxonomic categories, and estimating biophysical parameters of individual trees -- using airborne lidar data. Mathematical morphological operations followed by a marker-controlled watershed segmentation were found to perform well (r = 0.82 to 0.92) to delineate individual tree crowns in urban areas, especially when the trees occur in relatively isolated conditions. Using various distribution metrics of lidar returns, random forests were used to classify individual trees into different taxonomic classes (broadleaves/conifers, genera, and species). A classification accuracy of 80.5% was obtained when separating trees only into broadleaf and conifer classes, 50.0% for genera, and 51.3% for species. Using spectral metrics from high-resolution satellite imagery in addition to lidar-derived predictors improved the classification accuracies by 10.4% (to 90.9%) for broadleaf and conifer, 8.4% (to 58.4%) for genera and 8.8% (to 60.1%) for species compared to using lidar metrics alone. Prediction models to estimate several biophysical parameters such as height, crown area, diameter at breast height, and biomass were developed using lidar point cloud distributional metrics from individual trees. A high level of accuracy was attained for estimating tree height (R²=0.89, RMSE=1.3m), diameter at breast height (R²=0.82, RMSE=9.1cm), crown diameter (R²=0.90, RMSE=0.7m) and biomass (R²=0.67, RMSE=1.2t). Our results indicate that, while using lidar data alone can achieve the automation of major urban forest inventory tasks to an acceptable level of accuracy, a synergistic use of lidar data with other spectral data such as hyperspectral or orthoimagery, which are usually available at least in the United States for most urban areas, can considerably improve the performance of the lidar-based method. / Ph. D.

The overall objective of this study was to determine if remotely sensed data could be used to model scenic beauty. Terrestrial digital images from within forest stands located in Prince Edward Gallion State Forest near Farmville, Virginia were rated for their scenic beauty by a group of students to obtain scenic beauty estimates (SBEs). Since the inter-rater reliability was low for the SBEs, they were not used in the modeling efforts. Instead, stand parameters (collected on tenth acre plots) that have been used in scenic beauty prediction models, like mean diameter at breast height (dbh), were the dependent variables in regression analyses. A color-infrared aerial photograph from the National Aerial Photography Program (NAPP) was scanned to achieve a pixel ground resolution of one meter. The digital aerial photograph was rectified and used as the remotely sensed data. Since the aerial photograph was taken in April, only conifer stands were used in the analyses. Summary statistics were obtained from a 23 by 23 window around plot locations in three images: the original image, a texture image created with the variance algorithm and a 7x7 window, and the first principal component image. The summary statistics were used as the independent variables in regression analyses. The mean texture digital number for the green band predicted the mean dbh of a plot with an R2 of 0.623. A maximum of 44.3 and 27.4 percent of the variability in trees per acre and basal area per acre, respectively, was explained by the models developed in this study. It seems unlikely that the remotely sensed forest stand variables would perform well as surrogates for field measurements used in scenic quality models. / Master of Science

Remote sensing offers many advantages to previous forest measurements, such as limiting costs and time in the field. Light detection and ranging (lidar) has been shown to enable accurate estimates of forest height. Lidar does produce precise measurements for ground elevation and forest height, where and when it is available. However, it is expensive to collect and does not have wall-to-wall coverage in the United States. In this study, we estimated height using the National Agricultural Imagery Program (NAIP) photogrammetric point clouds to create a predicted height map for managed loblolly pine stands in the southeastern United States. Recent studies have investigated the ability of digital aerial photogrammetry (DAP), and more specifically NAIP, as an alternative to lidar as a means of estimating forest height due to its lower costs, frequency of acquisition, and wall-to-wall coverage across the United States. Field-collected canopy height for 534 plots in Virginia and North Carolina were regressed against the 90th percentile derived from NAIP point clouds. The model for predicted pine height using the 90th percentile of height (P90) is predicted pine height = 1.09(P90) – 0.43. The adjusted R^2 is 0.93, and the RMSE is 1.44 m. This model is being used to produce a 5 m x 5 m canopy height model for all pine stands across Virginia, North Carolina, and Tennessee. NAIP-derived point clouds are thus a viable means of predicting canopy height in southern pines. / M.S. / Collecting accurate measurements of pine plantations is essential to managing them to maximize various ecosystem goods and services. However, it can be difficult and time-consuming to collect these measurements by hand. This study demonstrates that point clouds derived from digital stereo aerial photograms enable calculating forest height to an accuracy sufficient for pine plantation management. We developed a simple linear regression model to predict forest canopy height using the 90th percentile of the photo-derived heights above the ground in a given area. With this model, we created a map of pine plantation canopy heights (consisting of 5 m x 5 m grid cells, each containing a canopy height estimate) for pine forests in Virginia, North Carolina, and Tennessee. Digital aerial photography from the National Agricultural Imagery Program (NAIP) is repeated every three years for a given state, allowing growth to be mapped over time. Photography collected by NAIP and similar programs also has uniform acquisition parameters in a given year applicable over large regions. State- and national photography programs like NAIP are also less expensive than other data sets, like airborne laser scanning data, that enable estimation of tree height.

Northwest Montana has seen a significant increase in its population in the past twenty years. The increase in population, and associated development, is thought to be associated with "amenity migration"; people moving to an area to exploit the recreational opportunities that are unique to that area. Impervious surfaces can serve as a suitable proxy for tracking the spread of various anthropogenic influences on an ecosystem; it impacts groundwater recharge, increases overall surface runoff as well as pollution and sediment load, and fragments landscapes. In this study, an Artificial Neural Network model was developed to update NLCD impervious surface product (2011) in Flathead County, Montana. Four Landsat 8 images from 2015 and 2016 were used to characterize imperviousness. This multi-temporal analytical method was designed to reduce the spectral confusion between impervious surface and soil/agricultural lands. We compared the neural network-predicted impervious surface maps with 2011 NLCD. When all four neural network prediction images agreed with a change of 50% or more from the 2011 NLCD map, the average of those four images replaced that pixel from the 2011 imperviousness map. Compared to the ground truth, the method used showed significant promise, with an R2 of 0.73 and RMSE of 0.123. A comparison of the artificial neural network model results and the 2011 NLCD data showed a continuation of urbanization trends; the urban cores of towns in the study remain static while the majority of impervious surface development takes place along the perimeter of urban areas. / Master of Science / Remotely sensed Landsat data can be used to rapidly detect and estimate changes in impervious surface cover. This study used artificial neural networks in conjunction with the National Landcover Database’s 2011 Percent Developed Imperviousness layer and Landsat 8 data from four dates between the summer of 2015 and fall of 2016 to predict impervious surface cover in 2016, by deriving spectral relationships between Landsat data and impervious surfaces. We found that by requiring agreement between the four dates’ neural networks outputs, we eliminated many of the false positives that arose from exposed soil. Using this method, we achieved an R2 of 0.73 and RMSE of .123, sampling only the areas along a rural-urban gradient, in an area with significant seasonal spectral variability.

Primary productivity is a fundamental ecosystem process driven by vascular plants in most terrestrial ecosystems and by microbes in more extreme ecosystems. In dense associations, microbial organisms can form visually conspicuous layers on sediment, soil, and rock surfaces, called microbial mats and biological soil crusts (i.e., biocrusts). Both microbial mats and biocrusts consist of cyanobacteria, moss, diatoms, and green algae, and also support diverse heterotrophic taxa. These communities exist in harsh environments worldwide such as hypersaline environments, tundra ecosystems, and hot and cold deserts where they are foundational taxa, providing most of the primary production and nitrogen fixation, as well as promoting cohesion and stability to soil surfaces. In the McMurdo Dry Valleys of Antarctica, microbial mats are the main source of fixed carbon in lentic and lotic environments, but their contribution to soil carbon and nitrogen cycling has not been systematically examined. In my dissertation, I investigated the relationships between microbial mats and the soil environments in which they occur. Using a combination of field surveys, soil analyses, and remote sensing, my objectives were to examine the influence of microbial mats and biocrusts on underlying soils and model the main drivers of their distribution and abundance. In Chapter 2, I investigated the relationships between underlying soil chemistry and microbial mat distribution, composition, and function in the Taylor Valley, finding that microbial mats enrich underlying soils, contributing to soil organic carbon and nitrogen. In Chapter 3, I assessed the spectral detectability of patchy biocrusts using multispectral satellite imagery to examine the environments in which biocrusts occur, finding that spectral unmixing of satellite imagery can successfully detect the presence of biocrust and its association with seasonal snow patches. As a direct continuation, in Chapter 4, I created a habitat suitability model using machine learning algorithms to determine the distribution and abundance of biocrusts in the Lake Fryxell basin. I found that biocrusts contribute a significant amount of carbon to the surface soil in the Lake Fryxell basin, with biocrust presence primarily driven by snow frequency, moisture content, and salinity. This dissertation contributes to ongoing questions about the sources of energy fueling soil food webs and regional carbon balance in the Taylor Valley, and how we can use remote sensing techniques for researching these critical soil communities in the dynamic Antarctic landscape. / Doctor of Philosophy / Photosynthesis is the process where plants and other organisms use sunlight to transform carbon dioxide into chemical energy. This is crucial because it provides the energy and nutrients that support all other life forms. In this dissertation, I focused on colonial microorganisms, which are the main primary producers in extreme environments, like deserts. I used a combination of field surveys and satellite imaging to study these organisms in the McMurdo Dry Valleys, Antarctica, which is a harsh polar desert environment that lacks vascular plants. Microbes colonize the surface of soil and form mm-cm thick microbial mats and biological soil crusts (called biocrusts). These organisms are found within the glacial-melt streams that flow on and off for only a few weeks each year, and they also occur on the stream margins and other periodically wet areas like near snow patches. This dissertation investigates the ecological importance of microbial mats and biocrusts, the ability to measure where they are using satellite imagery, and how much organic material they contribute to the broader landscape. Field work in the McMurdo Dry Valleys and laboratory analyses were required for each of these chapters. In Chapter 2, I investigated the relationships between microbial mats and the soils below them, and I found that microbial mats increase the organic matter and nutrient content in the soils. In Chapter 3, I assessed whether satellite imagery could be used to study the presence of sparse biocrusts and examined the environments in which biocrusts occur. I discovered that satellite imagery can successfully detect the presence of biocrust and that biocrusts occurred near melting snow patches. Lastly, in Chapter 4, I created models to determine where biocrusts occur in the Lake Fryxell basin and why biocrusts occur in those areas. I found that biocrusts occur over a significant area of the Lake Fryxell basin, containing a lot of organic material, and that biocrusts thrive in wet areas near snow patches where the soils are less salty. This dissertation contributes to ongoing questions about the sources of nutrients fueling soil food webs and contributing to the amount of organic material in the McMurdo Dry Valleys, and how we can use satellite imagery for monitoring these important soil communities in the changing Antarctic landscape.

In regions characterized by semi-arid climates where households’ welfare primarily relies on rainfed agricultural activities, extreme weather events such as droughts can present existential challenges to their livelihoods. To mitigate these risks, numerous social protection programs have been established to assist vulnerable households affected by weather events. Despite efforts to monitor environmental changes through remotely sensed technology, estimating the impact of weather variability on livelihoods remains challenging. This is compounded by the need to select appropriate statistical distribution for weather anomaly measures and household characteristics. We address these challenges by analyzing household consumption data from the Living Standards Measurement Study survey in Niger and systematically evaluating how each input factor affects vulnerability estimates. Our findings show that the choice of statistical distribution can significantly alter outcomes. For instance, using alternative statistical distribution for vegetation index readings could lead to differences of up to 0.7%, which means around 150,000 more households might be misclassified as not vulnerable. Similarly, variations in household characteristics could result in differences of up to 10 percentage points, equivalent to approximately 2 million households. Understanding these sensitivities helps policymakers refine targeting and intervention strategies effectively. By tailoring assistance programs more precisely to the needs of vulnerable households, policymakers ensure that resources are directed where they can make the most impact in lessening the adverse effects of extreme weather events. This enhances the resilience of communities in semi-arid regions. / Master of Science / In drought-prone regions where many families rely on rainfed farming, extreme weather can devastate livelihoods. Governments have created aid programs to assist the most vulnerable households during these climate crises, but identifying who needs help is extremely challenging. Part of this difficulty lies in selecting the right statistical methods for analyzing weather data and household information. In this paper, we focus on Niger, a country that experiences frequent droughts and where over 80% of the population depends on rainfed agriculture. By evaluating household consumption data, we aim to assist in identifying the households who has high probability of becoming poor as a result of unfavorable weather events and thus needs support from social protection programs. In our analysis, we systematically evaluate how each input factor (including household characteristics and statistical distributions) affects households likelihood of becoming poor in the event of weather crises. We find that compared to alternative statistical distributions, using a conventional normal distribution could lead to misclassifying around 150,000 households as non-vulnerable, leaving them without vital assistance. Similarly, using different sets of household characteristics can result in up to 10 percentage points which equivalents to 2 million households that would miss out on much-needed support. Understanding these sensitivities is crucial for policymakers in refining how aid programs identify the vulnerable populations and include them into the protection programs. The improved targeting approach will enhance the resilience of communities in semi-arid regions facing increasing weather variability.

Heterogeneous geospatial big data, from multi-modal Earth Observation (EO) data to geo-social media data, has become more and more accessible in recent years. This provides a potential data source for automatically extracting and mapping key geographical characteristics, hence mitigating the global mapping problem using current data mining methods. These automated geographic feature mapping techniques, especially for man-made infrastructure, are crucial to a lot of our socio-economic existence. Machine learning techniques, among many other data mining methodologies, have demonstrated better performance across a wide range of academic domains, most notably natural language processing and computer vision. In recent times, there has been a growing interest in research on ML-based Geospatial Artificial Intelligence (GeoAI), particularly in its ability to support autonomous mapping with heterogeneous geographical data. Though the potential is high and obvious, it remains a major challenge to handle inherent heterogeneity and empower data synergy when building robust and scalable GeoAI models for large-scale automated mapping purposes. For geospatial analysis, citizen science initiative is seen to be the most effective. This is a result of the fast development of Web 2.0 and crowdsourcing/Volunteered Geographic Information (VGI) technologies. These technologies enable even regular users or volunteer mappers to develop, gather, and distribute geospatial data using a variety of digital devices (such as desktop computers, mobile tablets, and smartphones). The technological obstacles to digital mapping have been significantly reduced by ongoing crowdsourcing and VGI efforts. In the real world, though, problems with global mapping have persisted for a considerable amount of time even in higher-income nations. Intelligent automated mapping techniques for geospatial analysis are desperately needed in this situation since they may effectively and efficiently close significant data gaps across nations. The research effort reported in this dissertation explores the possibilities of using citizen science or VGI to conduct geospatial analysis of different man-made infrastructures using ML from diverse geospatial data sources (e.g., multi-modal EO data, OSM, and GIS data). Three main research questions (RQs), derived from data-driven, method-driven, and application-driven research perspectives, are established to better address the issue of geospatial analysis with remote sensing and citizen science. The thesis especially goes in this direction by i) investigating the data-driven issue that combines ML for segmentation tasks; ii) creating strategies to deal with VGI data noises; and iii) using the created strategies in various mapping tasks. This creates even more intriguing possibilities for related works in the future.

Light reflectance from plants can be used as a non-invasive predictor of health and yield for many cropping systems, and has been investigated to a lesser extent with managed turfgrass systems. The frequent agronomic inputs associated with maintaining golf course grasses allow for exceptional stand quality under harsh growing conditions, but often expend resources inefficiently, leading to either stand loss or unnecessary inputs in localized areas. Turfgrass researchers have adopted some basic principles of light reflectance formerly developed for cropping systems, but field radiometric-derived narrow-band algorithms for turfgrass-specific protocols are lacking. Research was conducted to expand the feasibility of using radiometry to detect various turfgrass stressors and improve speed and geographic specificity of turfgrass management. Methods were developed to detect applied turfgrass stress from herbicide five days before visible symptoms developed under normal field growing conditions. Soil volumetric water content was successfully estimated using a water band index of creeping bentgrass canopy reflectance. The spectral reflectance of turfgrass treated with conventional synthetic pigments was characterized and found to erroneously influence plant health interpretation of common vegetation indices because of near infrared interference by such pigments. Finally, reflectance data were used to estimate root zone temperatures and root depth of creeping bentgrass systems using a gradient of wind velocities created with turf fans. Collectively, these studies provide a fundamental understanding of several turfgrass-specific reflectance algorithms and support unique opportunities to detect stresses and more efficiently allocate resources to golf course turf. / Ph. D.