Master of Science (MS), Bowling Green State University, 2005, Geology

The work in this thesis used passive hyperspectral remote sensing analysis with data collected form the high altitude MODIS Airborne Simulator (MAS). This study examined multiple remote sensing band ratios designed to capitalize on methane's 3.314µm absorption feature. Other ratios were also developed to detect atmospheric gas changes associated with possible methane plumes. Much of the analysis utilized datasets covering two California regions known to contain active oil/gas seeps and production. It was determined that no single MAS ratio algorithm could be used to confidently detect a methane gas plume. The presence of other atmospheric gases has the potential to affect the algorithms and produce complications for interpretation. However, by using a concurrence of ratio algorithm results, a suspected plume was thought to be detected in one of the off-shore datasets. The analysis for the land datasets found the high degree of surface material and temperature variations dramatically interfered with the ability to interpret the algorithm results with any significant confidence.

Master of Science (MS), Ohio University, 2017, Geography (Arts and Sciences)

Lake Bonneville was the largest water body that accumulated in the Great Basin during the late Pleistocene. Its latest major lacustral cycle lasted from 30 ka to 12 ka and much evidence of the lake remains are still evident in the landscape today. This thesis investigates the use of Landsat-8 multispectral imagery for mapping the Quaternary deposits in the Tule Valley portion, and Hyperion (EO-1) hyperspectral data for mapping part of the adjacent Snake Valley of Lake Bonneville. Maximum likelihood classification was applied for Landsat 8 data, and the two spectral analysis approaches of linear spectral unmixing and spectral angle mapper (SAM) were applied to the Hyperion dataset. Furthermore, X-ray diffraction (XRD) analysis of a Lake Bonneville marl sediment sample characterized the dominant minerals in that sample. This investigation relied on Sack's (1990) Quaternary geologic map of Tule Valley as the reference for the remote sensing analysis. This study investigates if those sources of information can approach in quality and detail the traditional map that relies on fieldwork and air photo interpretation. Results illustrate that hyperspectral and multispectral data have potential value for Quaternary geological mapping. Maximum likelihood classification yielded overall accuracy of 51% with successful discrimination of Qlf, Qeg, Qes, Qlm, Qac, and bedrock. However, complete separation between several lacustrine and alluvial classes was not achieved. In general, the Hyperion spectral angle mapper (SAM) and spectral unmixing results discriminated relatively well among the three endmembers of calcite, gypsum, and quartz across portions of the Snake Valley study area. The high fraction abundance values on the fractional images reliably represented pixels dominated by calcite, gypsum, or quartz. Some confusion between classifications are attributeded to the local mixing of classes at the pixel scale, overlap in mineralogy, similarities in the nature of surface weathe (open full item for complete abstract)

Doctor of Philosophy, The Ohio State University, 2012, Electrical and Computer Engineering

Range-resolved interferometric phase and Doppler spectra are two subjects of interest with regard to the retrieval of sea surface height profiles from coherent marine radar measurements. The studies of this dissertation attempt to improve understanding of the properties and associated measurement errors of these quantities through the use of numerically simulated low-grazing-angle backscatter data. In the first part of the dissertation, studies of the interferometric phase are described. Backscattered fields computed using the method of moments for one dimensional ocean-like surface profiles are used to examine statistical properties of the single-look interferometric phase estimator, in order to investigate the applicability of standard expectations for height retrieval accuracy in this problem. The results show that shadowing and multipath propagation effects cause errors in interferometric phase estimation beyond those caused by speckle effects alone. In addition, the decorrelation between the fields received at two antennas is found to be impacted by shadowing and multipath propagation effects, making standard models for this quantity less applicable as well. These results show that modeling the expected performance of interferometric sea surface height retrieval approaches at low grazing angles is difficult. The second part of the dissertation involves studies of the range-resolved Doppler spectra at low-grazing-angles. Backscattered fields are computed for a single realization of a one-dimensional ocean-like surface profile as the realization evolves in time. Transformation into the range-Doppler domain enables examination of properties of the resulting Doppler spectra (for both HH and VV polarizations) and their relationship to properties of the surface profile. In general, a strong correspondence between the long wave orbital velocity of the surface and the Doppler centroid frequency is observed for visible portions of the surface, as well as some evidence (open full item for complete abstract)

Master of Science in Computer Engineering, University of Dayton, 2024, Electrical and Computer Engineering

The lack of Hyperspectral imagery (HSI) is an issue for many researchers and fields that wish to utilize the sheer amount of data a HSI cube contains. Given this along with the cost and the effort associated with gathering HSI, a way to generate them using existing would be very useful. Other works have generated synthetic images, images that contain the characteristics of a HSI cube, but that do not actually map to any real world location. This work attempts to show that it is possible to generate those cubes with easier to gather datasets and less data. This is done by using a paired image generation deep learning model, a Generative Adversarial Network. The HSI cubes were gathered from USGS's Earth Explorer and the sensor used was Earth Observing-1's Hyperion. The network was trained on four different input types in four regions and tested on three different regions. The four input types were 5 bands, 10 bands, 10 bands with no bands from the middle 100 bands, and 20 bands. The results and accuracy of the model were based on various metrics and a separate model was trained on each input until those metrics plateaued. A comparison of input vs generated spectra as well as the various metrics were then used to verify the accuracy of the test dataset. It was found the models each generalized well and that even individual bands of the greater HSI cube generated quite well to the target.

MS, University of Cincinnati, 2024, Engineering and Applied Science: Environmental Engineering

Phosphorus pollution is a major water quality issue impacting the environment and human health. Traditional methods limit the frequency and extent of total phosphorus (TP) measurements across many rivers. However, remote sensing can accurately estimate riverine TP; nevertheless, no large scale assessment of riverine TP using remote sensing exists. Large-scale models using remote sensing can provide a fast and consistent method for TP measurement, important for data generalization and accessing extensive spatial-temporal change in TP. Our study uses remote sensing and machine learning to estimate the TP in rivers in the contiguous United States (CONUS). Initially, we developed a national scale matchup dataset for Landsat detectable rivers (river width >30m) using in situ TP and surface reflectance. We used in-situ data from the Water Quality Portal (WQP), alongside water surface reflectance data from Landsat 5, 7, and 8 spanning from 1984 to 2021. Then, we used this dataset to develop a machine learning (ML) model using different preprocessing methods and algorithms. We found that using high-level vegetation in the clustering approach and over-sampling or under-sampling our training data in the sampling approach improved our model estimation accuracy. We compared XGBLinear, XGBTree, Regularized Random Forest (RRF), and K-Nearest Neighbors (KNN) ML algorithms and selected XGBLinear as the best model with an R2 of 0.604, RMSE of 0.103 mg/L, MAE of 0.83, and NSE of 0.602. Finally, we identified human footprint, elevation, river area, and soil erosion as the main attributes influencing the accuracy of estimated TP from the ML model.

Master of Science (M.S.), University of Dayton, 2024, Electrical Engineering

RGB imagery provides detail which is usually sufficient to perform computer vision tasks. However, images taken in low-light appear vastly different from well-lit imagery due to the diversity in light intensity. Polarimetric data provides additional detail which focuses on the orientation of the light rather than intensity. Scaling our classic RGB images using polarimetric data can maintain the RGB image type, while also enhancing image contrast. This allows transfer learning using pre-trained RGB models to appear more feasible. Our work focuses on developing a large dataset of paired polarimetric RGB images in a highly controlled laboratory environment. Then, we perform transfer learning on a pre-trained image segmentation model with each of our image product types. Finally, we compare these results in both well-lit and low-light scenarios to see how our polarimetrically enhanced RGB images stack up against regular RGB images.

Master of Science, The Ohio State University, 2005, Graduate School

Master of Science, The Ohio State University, 2006, Graduate School

MS, Kent State University, 2024, College of Arts and Sciences / Department of Geography

Wildfire and drought are key drivers of shrubland expansion in southwestern US landscapes. Stand-replacing fires in dry conifer forests induce shrub-dominated stages, and changing climatic patterns may cause a long-term shift from coniferous forests to deciduous shrublands. This study assessed recent changes in deciduous fractional shrub cover (DFSC) in the eastern Jemez Mountains from 2019-2023 using topographic and Sentinel-2 satellite data in a random forest model. Sentinel-2 provides multispectral bands at 10 and 20 meters, including three 20 meter red edge bands, which are highly sensitive to variation in vegetation. There is no consensus in the literature on whether upscaling imagery to 20 meters or downscaling to 10 meters is more advantageous. Therefore, an additional goal of this study was to evaluate the impact of spatial scale on DFSC model performance. Two random forest models were built, a 10 and 20 meter model. The 20 meter model outperformed the 10 meter model, achieving an R-squared value of 0.82 and an RMSE of 7.85, compared to the 10 meter model (0.76 and 9.99, respectively). The 20 meter model, built from 2020 satellite imagery, was projected to the other years of the study, by replacing the spectral variables with satellite imagery from the respective year, resulting in yearly predictions of DFSC from 2019-2023. DFSC decreased from 2019-2022, coinciding with severe drought and a 2022 fire, followed by a significant increase in 2023, particularly within the 2022 fire footprint. Overall trends showed a general increase in DFSC despite high interannual variability, with elevation being a key topographic variable influencing these trends. This study revealed yearly vegetation dynamics in a semi-arid system and provided a close look at post-fire regeneration patterns in deciduous resprouting shrubs. Understanding this complex system is crucial for informing management strategies as the landscape continues to shift from conifer forest to shrubland du (open full item for complete abstract)

Master of Science (MS), Bowling Green State University, 2024, Applied Geospatial Science

Cities around the globe are undergoing significant transformations due to rapid urbanization, fueled by factors such as population growth, economic development, and the migration of people from rural to urban areas. By 2050, an estimated two-thirds of the global population will reside in urban areas, posing significant challenges for sustainable development. Remote sensing data, combined with machine learning modeling approaches play a crucial role in monitoring and analyzing urban sprawl. This study investigates the potential of a machine learning (ML) classification algorithm coupled with data fusion remote sensing techniques to improve land-use and land-cover (LULC) change detection in Kiambu County, Kenya, for the period 2000-2022. It utilizes Landsat data from 2000 to 2022, augmented by Harmonized Landsat Sentinel-2 (HLS) and Sentinel-1 SAR data from 2013 to 2022, for urban land use/land cover (LULC) change detection. Google Earth Engine (GEE) facilitated preprocessing and analysis, refining Synthetic Aperture Radar (SAR) imagery and employing Random Forest (RF) for classification. Integrating Landsat 8/HLS and SAR data enhanced classification accuracy, supported by feature selection, hyperparameter tuning, and spectral band ratios to mitigate data errors. Key indices like NDBI, NBR2, BSI, NDWI, NDVI, and SAVI were crucial for classifying land cover types. From 2000 to 2022, Landsat-based analysis shows significant urbanization. Urban areas grew from 17.8% in 2000 to 22.4% by 2005, 25.7% in 2010, 29.6% in 2015, and 31.9% by 2022. Specifically, for 2015, using Landsat 8 alone, urban areas covered 23.4% (594.0 km²), while fusing Landsat 8 with SAR data raised this to 28.7% (729.4 km²) with improved testing accuracy of 91.7% and validation accuracy of 87.5%. Integrating optical data (HLS and Landsat 8) with SAR and applying ML techniques on GEE, the classification accuracy improved by 5.7% compared to optical data alone. Overall, urbanization in Kiambu Co (open full item for complete abstract)

Master of Science, The Ohio State University, 2006, Graduate School

Master of Science, The Ohio State University, 2005, Graduate School

Master of Science, The Ohio State University, 1960, Graduate School

Doctor of Philosophy, The Ohio State University, 2024, Earth Sciences

Seasonal snow plays a large role in the water cycle and local ecosystem dynamics in snow dominated regions. However, two characteristics of the snowpack, the snow water equivalent (SWE) and density, are challenging to measure at scale. Modeling and remote sensing allow for the estimation of these characteristics at wide spatial scales, but practical limitations remain on our ability to estimate at a fine spatial fidelity, wide spatial extent, and daily temporal resolution. Regional Climate Models (RCMs) have been shown to successfully estimate SWE at basin-wide scales but remain too computationally expensive to run at sub-kilometer resolutions over large domains. In this thesis, I present two alternative methods to estimate daily SWE at a high spatial and temporal resolution on a basin-wide scale. The first, Blender, presented in Chapter 2, merges 9 km RCM estimates of SWE, precipitation, and top of the snowpack energy balance from the Weather and Research Forecasting (WRF) model with remotely sensed snow cover fraction (SCF) measurements to produce 500 m estimates of SWE timeseries. Blender re-solves the mass and energy balance of the snowpack with a constrained non-linear optimization, forced by the timing of the snow on and off dates from the SCF data. Compared against 50 m LiDAR estimates of SWE from 18 Airborne Space Observatory (ASO) flights, Blender has an average spatial RMSE of 11.5% of maximum SWE, while the prior from WRF has an average spatial RMSE of 17% of maximum SWE. The mean absolute bias of the total basin snow water storage (SWS) for the Blender estimates is 7.3% in the winter, and 31.6% for the WRF prior. This method, Blender, requires ~ 20% extra computing time on top of the original WRF runs, and improves both the spatial RMSE and basin SWS absolute bias, all while better matching the melt timing to the remotely sensed SCF. In Chapter 3 we present the second method, Linear Blender, is a linearized version of Blender, Chapter 2. This meth (open full item for complete abstract)

Master of Science in Computer Engineering, University of Dayton, 2024, Electrical and Computer Engineering

Event-based sensors (EBS) report pixel-asynchronous changes in scene intensity called events. Hence, its sparse event stream is well-suited for many computer vision tasks— particularly object tracking. However, relative motion between the sensor and scene will generate extraneous events caused by the translation of static scene features across the sensor. We present OFE-EBS, an optical flow equation-inspired event-based sensor for low-Earth orbit (LEO) ground moving target indication (GMTI). Owing to the predictable velocity of a satellite platform in LEO, we augment the EBS pixel with additional cross-row subtraction hardware to remove static background features. Pixel adaptivity is modified to ensure dynamic foreground features generate fewer events, further reducing event rate. Finally, using our analytical sensor model, we show that OFE-EBS outperforms conventional EBS in spatial resolution and event rate, considering the effects of pixel nonuniformity.

Master of Arts, Miami University, 2024, Geography

Rapid urbanization and the Urban Heat Island (UHI) effect are key global challenges, particularly in the Global South. Dhaka City, Bangladesh, one of the fastest growing and most densely populated cities in this region, facing increasing heat trends which threaten its public health, social and environmental conditions. Mitigating future UHI requires a geospatial analysis of urbanization and temperature trends at the neighborhood-scale. However, proprietary geospatial data and applications are often prohibitively costly for planners. This study combines cost-effective, publicly-accessible geospatial applications and time series satellite data from 2003 - 2023 to analyze land use and land cover (LULC) and land surface temperature (LST) changes in Dhaka City at regional and neighborhood-scales. For monitoring LULC, Landsat images were analyzed through supervised classification at a regional scale, and spectral mixture analysis (SMA) helped understand complex urban development patterns at the neighborhood-scale. This study analyzed MODIS daily LST images to understand diurnal temperature trends and found a strong positive correlation between urban development intensity and increased day and nighttime temperatures, contributing to neighborhood-specific UHI impacts. The study emphasizes the importance of developing publicly-accessible and inexpensive geospatial methods to support UHI mitigation planning that can benefit other similar cities.

Master of Science, The Ohio State University, 2024, Food, Agricultural and Biological Engineering

Waste coal in abandoned mine lands poses significant environmental challenges, affecting nearby communities, rivers, and streams. Effective management of these piles is essential due to concerns such as acid mine drainage, soil and water contamination, coal fires, and methane emissions. Various strategies have been proposed for managing waste coal, including potential utilization for rare earth element recovery, soil amendment, construction aggregates, and energy generation. However, the implementation of these strategies remains uncertain due to the lack of precise location and volume data on waste coal piles. Traditional methods for gathering these data rely on field visits and Global Navigation Satellite System surveying, which are costly and labor-intensive. Advances in satellite technologies and the availability of digital elevation models (DEMs) offer an opportunity to estimate waste coal volume on a regional scale in a timely and cost-effective manner. Thus, the objective of this thesis was to develop a robust data analytical framework to locate and estimate the volume of waste coal piles on a regional scale, using the Muskingum River Basin (MRB) in Ohio as the study area. Initially, a prototype was developed to determine the most effective machine learning (ML) model to map waste coal piles in a historical coal mine site within the MRB. While all four ML models effectively identified dominant classes such as Grassland and Forest, the Random Forest (RF) model demonstrated superior performance in classifying the more complex waste coal class, with a precision of 86.15% and recall of 76.71%. Subsequently, the greatest disturbance and reclamation mapping of these waste coal piles were conducted using the LandTrendr algorithm to distinguish waste coal piles in abandoned mine lands from those in active mining areas. Moreover, this study utilized publicly available elevation models to estimate waste coal volume in the MRB. However, since historical terrain mo (open full item for complete abstract)

MS, Kent State University, 2024, College of Arts and Sciences / Department of Geography

In 2018, the largest wildfire to ever occur in the Santa Monica Mountains National Recreation Area burned 88% of National Park Service (NPS) land. Located near Los Angeles, this park is the largest urban national park in the U.S. and home to mountain lions (Puma concolor) that are severely threatened. High levels of urbanization force them to live in overlapping and too small of home ranges, leading to intraspecific conflicts and inbreeding. The frequent wildfires add another threat, killing pumas directly or damaging their habitat. Current research conflicts as to how pumas select habitat post-fire, and most do not incorporate remote sensing metrics or consider how movements change with time since fire. In this study, I used global positioning system (GPS) collar data supplied by the NPS to analyze post-fire puma habitat selection. I conducted integrated step selection functions (iSSFs) at individual and population levels, for every 6-month seasonal period following the 2018 fire through 2023. I analyzed nine static variables to account for abiotic landscape variability and three variables derived from multi-temporal remote sensing to capture the dynamic, biotic environment, mainly focused on burn severity and vegetation condition and structure metrics. Habitat selection and variable importance were compared within each time period, as well as throughout the study period. I focused results on the population level analyses only. Results indicated that pumas consistently selected for increased vegetation vigor and selected for higher landscape heterogeneity and structure for the majority of time periods. Vegetation vigor also appeared as one of the most important variables to movement, along with terrain ruggedness and slope. Seasonal trends emerged for some variables post-fire. This study suggests that pumas are considerate of vegetation condition and fire impacts when selecting habitat, highlighting key habitat characteristics that pumas prefer post-fire. The influ (open full item for complete abstract)

Master of Science (MS), Bowling Green State University, 2024, Geology

Precise estimation of canopy chlorophyll content (CCC) and canopy nitrogen content (CNC) is essential for effective monitoring of crop growth conditions. The field of satellite imaging spectroscopy, known as hyperspectral imaging, offers a non-destructive, large-area, and real-time approach to monitoring of CCC and CNC over large areas. With the introduction of recent science-driven mission such as the Environmental Mapping and Analysis Program (EnMAP), imaging spectroscopy has emerged as a crucial component of empirical and physical modelling. This study introduces two approaches for retrieving CCC and CNC using EnMAP hyperspectral imagery acquired over Kellogg Biological Station, Michigan in the summer 2023: i) Machine learning regression algorithms (MLRAs), and ii) a hybrid model that combines a radiative transfer model with machine learning. We assessed the performances of six regression techniques including kernel ridge regression (KRR), least squares linear regression (LSLR), partial least squares regression (PLSR), Gaussian process regression (GPR), neural network (NN), and random forest (RF) for retrieving CCC and CNC. In the hybrid model workflow, each of the six techniques were integrated with PROSAIL, a widely known radiative transfer model. For both the approaches, the final CCC and CNC models were validated against field data. For CCC retrieval, KRR demonstrated the best performance in the MLRA approach (RMSE = 10.01, NRMSE = 9.81%, R2 = 0.93), while GPR exhibited the best performance in the hybrid approach (RMSE = 9.62, NRMSE = 9.43%, R2 = 0.93). Regarding CNC retrieval, KRR outperformed other models in both the MLRA (RMSE = 10.10, NRMSE = 8.13%, R2 = 0.94) and hybrid approach (RMSE = 16.98, NRMSE = 13.67%, R2 = 0.83). While the hybrid model outperformed the standalone MLRA approach for CCC retrieval, the MLRA approach surpassed the hybrid approach in CNC retrieval. In CCC retrieval, the most significant bands of EnMAP were identified in the visible to (open full item for complete abstract)

Master of Science in Electrical Engineering, University of Dayton, 2024, Electrical Engineering

Remote sensing techniques are capable of capturing large scenes of data over several sensing domains. Hyperspectral imagery (HSI), often accompanied with lIDAR and orthoimagery sensors during collection, can provide deeper contextual information for a wide range of applications in many different fields. Complex characteristics across spectral bands in addition to high-dimensionality of HSI data present challenges to accurate classification. Generally, dimensionality reduction of the input hyperspectral data cube is performed through multi-phase analytical algorithms as a pre-processing step before further analysis to include machine learning networks. These networks commonly rely on labeled training data for segmentation. Annotating ground truth aerial data can prove to be a cumbersome endeavor that may require specific expertise for accurate assessment. This inspires exploratory research for useful unsupervised feature-learning approaches that can withdraw essential information from HSI data to map scenes without labeled data thereby providing a start-to-finish scene segmentation process.