{"id":{"repo_id":"vt","oai_identifier":"oai:vtechworks.lib.vt.edu:10919/140545"},"canonical_url":"https://search.dev.ndltd.org/etd/vt/oai:vtechworks.lib.vt.edu:10919/140545","repository":{"repo_id":"vt","name":"Virginia Tech","base_url":"https://vtechworks.lib.vt.edu/oai/request"},"display":{"title":"Enhanced Soil Moisture and Streamflow Estimation in Ungauged or Data-Scarce Watersheds","abstract":"Watershed models are important tools for quantifying hydrologic processes and evaluating water management practices. These models simplify complex systems through parametrization and assumptions, particularly in semi-distributed frameworks like Soil and Water Assessment Tool (SWAT). Clear conceptualization is critical to realize adequate representation of internal hydrologic processes, particularly in saturated excess runoff dominated watersheds where the spatio-temporal dynamics of saturated areas influence both the distributed (soil moisture state, runoff generation, or pollutant export) and integrated (streamflow) watershed responses. While streamflow is commonly used to constrain model parameters due to its availability and integrative nature, the issue of equifinality, where multiple parameter sets yield similar streamflow, but divergent internal process estimates, can lead to significant uncertainty. This dissertation addresses these challenges by introducing new and improved techniques for representing field scale soil moisture patterns and using satellite soil moisture data for model calibration in ungauged and data-scarce watersheds, especially for watersheds where variable source area runoff mechanism dominate. Chapter 2 proposes the topographic index (TI) as a tool to represent the spatial soil moisture patterns. Using unsupervised machine learning on in-situ soil moisture data from a 4.2 ha field, three distinct soil moisture groups were identified. TI values were classified into three groups using various approaches (equal-interval, equal-area, k-means, Fisher) and digital elevation model (DEM) sources (United States Geological Survey (USGS) and drone-based LiDAR) generating topographic index classes (TIC). Performance was evaluated using Spearman's correlation and misclassification rates. Results showed that low-resolution LiDAR and USGS DEMs outperformed high-resolution LiDAR DEMs, with equal-interval classification yielding the best performance. The result showed resampling to a lower resolution improved the performance of LiDAR DEMs, while TICs derived from publicly available USGS DEMs outperformed those from LiDAR DEMs. Among classification approaches equal-interval provided the highest performance. Overall, the result showed a three classes TIC can represent field scale soil moisture pattern and the effect of DEM type, resolution, and classification approach can be substantial. Chapter 3 develops a variable source area (VSA) SWAT (SWAT-VSA) model for a 14 km² watershed, incorporating the three-class TIC to improve hydrologic response unit (HRU) definition. Downscaled and bias-corrected satellite soil moisture data, along with streamflow data, were used for single and multi-objective calibration. Calibration using soil moisture data improved field-scale moisture estimation, while multi-objective calibration enhanced overall model performance. The three-class TIC structure reduced computational cost while effectively representing VSA dynamics. Chapter 4 demonstrated the utility of downscaled and bias corrected satellite soil moisture for streamflow estimation in ungauged watersheds. Three watersheds in the northeastern US with flow monitoring stations were analyzed. Soil moisture-calibrated models, using root mean squared error (RMSE) and SPatial EFficiency (SPAEF) metrics, were compared to streamflow-calibrated and regionalization-based models. While streamflow data and regionalization technique performed better for streamflow estimation, soil moisture calibrated models incorporating spatial metrics in calibration yielded comparable streamflow predictions. Satellite-based calibration offers an objective alternative to traditional regionalization, especially when properly downscaled, bias-corrected, and scaled. This research demonstrates the potential of satellite soil moisture data to improve hydrologic model performance and reduce uncertainty in data-scarce environments, offering a scalable and objective approach to watershed modeling.","abstract_html":"Watershed models are important tools for quantifying hydrologic processes and evaluating water management practices. These models simplify complex systems through parametrization and assumptions, particularly in semi-distributed frameworks like Soil and Water Assessment Tool (SWAT). Clear conceptualization is critical to realize adequate representation of internal hydrologic processes, particularly in saturated excess runoff dominated watersheds where the spatio-temporal dynamics of saturated areas influence both the distributed (soil moisture state, runoff generation, or pollutant export) and integrated (streamflow) watershed responses. While streamflow is commonly used to constrain model parameters due to its availability and integrative nature, the issue of equifinality, where multiple parameter sets yield similar streamflow, but divergent internal process estimates, can lead to significant uncertainty. This dissertation addresses these challenges by introducing new and improved techniques for representing field scale soil moisture patterns and using satellite soil moisture data for model calibration in ungauged and data-scarce watersheds, especially for watersheds where variable source area runoff mechanism dominate. Chapter 2 proposes the topographic index (TI) as a tool to represent the spatial soil moisture patterns. Using unsupervised machine learning on in-situ soil moisture data from a 4.2 ha field, three distinct soil moisture groups were identified. TI values were classified into three groups using various approaches (equal-interval, equal-area, k-means, Fisher) and digital elevation model (DEM) sources (United States Geological Survey (USGS) and drone-based LiDAR) generating topographic index classes (TIC). Performance was evaluated using Spearman's correlation and misclassification rates. Results showed that low-resolution LiDAR and USGS DEMs outperformed high-resolution LiDAR DEMs, with equal-interval classification yielding the best performance. The result showed resampling to a lower resolution improved the performance of LiDAR DEMs, while TICs derived from publicly available USGS DEMs outperformed those from LiDAR DEMs. Among classification approaches equal-interval provided the highest performance. Overall, the result showed a three classes TIC can represent field scale soil moisture pattern and the effect of DEM type, resolution, and classification approach can be substantial. Chapter 3 develops a variable source area (VSA) SWAT (SWAT-VSA) model for a 14 km² watershed, incorporating the three-class TIC to improve hydrologic response unit (HRU) definition. Downscaled and bias-corrected satellite soil moisture data, along with streamflow data, were used for single and multi-objective calibration. Calibration using soil moisture data improved field-scale moisture estimation, while multi-objective calibration enhanced overall model performance. The three-class TIC structure reduced computational cost while effectively representing VSA dynamics. Chapter 4 demonstrated the utility of downscaled and bias corrected satellite soil moisture for streamflow estimation in ungauged watersheds. Three watersheds in the northeastern US with flow monitoring stations were analyzed. Soil moisture-calibrated models, using root mean squared error (RMSE) and SPatial EFficiency (SPAEF) metrics, were compared to streamflow-calibrated and regionalization-based models. While streamflow data and regionalization technique performed better for streamflow estimation, soil moisture calibrated models incorporating spatial metrics in calibration yielded comparable streamflow predictions. Satellite-based calibration offers an objective alternative to traditional regionalization, especially when properly downscaled, bias-corrected, and scaled. This research demonstrates the potential of satellite soil moisture data to improve hydrologic model performance and reduce uncertainty in data-scarce environments, offering a scalable and objective approach to watershed modeling.","abstract_has_math":false,"creators":["Asfaw, Binyam Workeye"],"institution":"Virginia Tech","degree_name":"Doctor of Philosophy","degree_level":"doctoral","degree_discipline":"Biological Systems Engineering","degree_department":"Biological Systems Engineering","school":null,"contributors":[],"advisors":[],"committee_chairs":["Easton, Zachary"],"committee_members":["Czuba, Jonathan A.","Hession, William Cully","Rippy, Megan A."],"year":2025,"date_issued":"2025-12-22","date_published":"2025-12-22","updated_at":"2026-07-22T22:19:57Z","subjects":["Topographic Indices; Satellite Soil Moisture; Downscaling; Bias Correction; K-means Clustering; Calibration; SWAT model; Parameters; Streamflow; Ungauged"],"languages":["en"],"rights":["Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International"],"rights_urls":["http://creativecommons.org/licenses/by-nc-nd/4.0/"],"identifier_entries":[{"key":"dc:identifier.other","label":"Dc Identifier Other","values":["vt_gsexam:45315"],"render_values":[{"text":"vt_gsexam:45315","href":null,"code":true}]}]},"links":{"outbound_url":"https://hdl.handle.net/10919/140545","outbound_label":"Handle","outbound_source":"dc:identifier.uri"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.committeechair","label":"Committee Chair","values":["Easton, Zachary"]},{"key":"dc:contributor.committeemember","label":"Committee Member","values":["Czuba, Jonathan A.","Hession, William Cully","Rippy, Megan A."]},{"key":"dc:contributor.department","label":"Department","values":["Biological Systems Engineering"]},{"key":"dc:creator","label":"Author","values":["Asfaw, Binyam Workeye"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2025-12-23T09:00:47Z"]},{"key":"dc:date.available","label":"Dc Date Available","values":["2025-12-23T09:00:47Z"]},{"key":"dc:date.issued","label":"Date","values":["2025-12-22"]},{"key":"dc:publisher","label":"Institution","values":["Virginia Tech"]},{"key":"dc:type","label":"Dc Type","values":["Dissertation"]},{"key":"thesis:degree_discipline","label":"Discipline","values":["Biological Systems Engineering"]},{"key":"thesis:degree_level","label":"Degree Level","values":["doctoral"]},{"key":"thesis:degree_name","label":"Degree Name","values":["Doctor of Philosophy"]},{"key":"thesis:institution_name","label":"Thesis Institution Name","values":["Virginia Polytechnic Institute and State University"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Topographic Indices; Satellite Soil Moisture; Downscaling; Bias Correction; K-means Clustering; Calibration; SWAT model; Parameters; Streamflow; Ungauged"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language.iso","label":"Language (ISO)","values":["en"]},{"key":"dc:rights","label":"Dc Rights","values":["Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International"]},{"key":"dc:rights.uri","label":"Rights URI","values":["http://creativecommons.org/licenses/by-nc-nd/4.0/"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.other","label":"Dc Identifier Other","values":["vt_gsexam:45315"]},{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://hdl.handle.net/10919/140545"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["Watershed models are important tools for quantifying hydrologic processes and evaluating water management practices. These models simplify complex systems through parametrization and assumptions, particularly in semi-distributed frameworks like Soil and Water Assessment Tool (SWAT). Clear conceptualization is critical to realize adequate representation of internal hydrologic processes, particularly in saturated excess runoff dominated watersheds where the spatio-temporal dynamics of saturated areas influence both the distributed (soil moisture state, runoff generation, or pollutant export) and integrated (streamflow) watershed responses. While streamflow is commonly used to constrain model parameters due to its availability and integrative nature, the issue of equifinality, where multiple parameter sets yield similar streamflow, but divergent internal process estimates, can lead to significant uncertainty. This dissertation addresses these challenges by introducing new and improved techniques for representing field scale soil moisture patterns and using satellite soil moisture data for model calibration in ungauged and data-scarce watersheds, especially for watersheds where variable source area runoff mechanism dominate. Chapter 2 proposes the topographic index (TI) as a tool to represent the spatial soil moisture patterns. Using unsupervised machine learning on in-situ soil moisture data from a 4.2 ha field, three distinct soil moisture groups were identified. TI values were classified into three groups using various approaches (equal-interval, equal-area, k-means, Fisher) and digital elevation model (DEM) sources (United States Geological Survey (USGS) and drone-based LiDAR) generating topographic index classes (TIC). Performance was evaluated using Spearman's correlation and misclassification rates. Results showed that low-resolution LiDAR and USGS DEMs outperformed high-resolution LiDAR DEMs, with equal-interval classification yielding the best performance. The result showed resampling to a lower resolution improved the performance of LiDAR DEMs, while TICs derived from publicly available USGS DEMs outperformed those from LiDAR DEMs. Among classification approaches equal-interval provided the highest performance. Overall, the result showed a three classes TIC can represent field scale soil moisture pattern and the effect of DEM type, resolution, and classification approach can be substantial. Chapter 3 develops a variable source area (VSA) SWAT (SWAT-VSA) model for a 14 km² watershed, incorporating the three-class TIC to improve hydrologic response unit (HRU) definition. Downscaled and bias-corrected satellite soil moisture data, along with streamflow data, were used for single and multi-objective calibration. Calibration using soil moisture data improved field-scale moisture estimation, while multi-objective calibration enhanced overall model performance. The three-class TIC structure reduced computational cost while effectively representing VSA dynamics. Chapter 4 demonstrated the utility of downscaled and bias corrected satellite soil moisture for streamflow estimation in ungauged watersheds. Three watersheds in the northeastern US with flow monitoring stations were analyzed. Soil moisture-calibrated models, using root mean squared error (RMSE) and SPatial EFficiency (SPAEF) metrics, were compared to streamflow-calibrated and regionalization-based models. While streamflow data and regionalization technique performed better for streamflow estimation, soil moisture calibrated models incorporating spatial metrics in calibration yielded comparable streamflow predictions. Satellite-based calibration offers an objective alternative to traditional regionalization, especially when properly downscaled, bias-corrected, and scaled. This research demonstrates the potential of satellite soil moisture data to improve hydrologic model performance and reduce uncertainty in data-scarce environments, offering a scalable and objective approach to watershed modeling."]},{"key":"dc:description.abstractgeneral","label":"General Abstract","values":["Watershed models are used to represent and quantify the movement and exchange of water in the landscape and between land and atmosphere. They are important tools which aid scientists to evaluate alternative water management practices. Watershed models are developed using simplifying assumptions to reduce the complexity of the system they represent and, hence, require training in order to emulate the behavior of the system. However, trained models often fall short in representing parts of the system behavior. For example, models trained on measured streamflow may produce a range of alternative model setups that can provide satisfactory streamflow estimation but divergent representation of soil moisture conditions. Scientists work to tackle this challenge, often in data-scarce conditions. This research work addresses these problems using new and improved approaches to represent model components and remotely acquired satellite soil moisture data for improved model skill in estimating streamflow and soil moisture in data-scarce watersheds. Chapter 2 illustrates the use of terrain properties for simplified representation and estimation of soil moisture variability across the landscape. This requires information on the extent of spatial variability of soil moisture and its simplification. A mathematical and statistical approach is employed to identify homogeneous soil moisture units using soil moisture measurements across a landscape. Estimating terrain properties as it relates to soil moisture is not a rudimentary task as the derived information depends on the source and detail of topographic data used. Moreover, derived terrain properties need to be classified into groups to represent a corresponding number of homogeneous soil moisture units, which also depends on choice of classification technique. Using in-situ soil moisture measurements, this study found three homogenous soil moisture units may be representative of field scale soil moisture variability and highlighted the effect of using different sources and scales of topographic data as well as classification techniques on the use and performance of derived terrain properties. Chapter 3 develops a watershed model informed on the findings in chapter 2 for improved estimation of soil moisture variability. The watershed model was trained using streamflow and satellite soil moisture individually and in conjunction. This is particularly challenging in small watersheds as satellite soil moisture data have pixel size much bigger than the size of a small watershed. To overcome this challenge and disaggregate the information into smaller pixel sizes, a technique which employee a statistical and mathematical approach was applied. The resulting data was then scaled to match the range of soil moisture variation observed from in situ soil moisture measurements. The developed model showed the use of satellite soil moisture data, especially in conjunction with streamflow data, improved the model's streamflow and soil moisture estimation performance. Chapter 4 repurposed the approach in chapter 3 for use in watersheds which does not have streamflow measurement stations. The result showed the use of satellite soil moisture data as alternative to streamflow based direct and indirect model training approaches provided adequate model training for streamflow estimation in streamflow scarce watersheds. The performance of the soil moisture based model training was especially satisfactory when model training techniques which capture the soil moisture information varying both across space and in time were included."]},{"key":"dc:description.degree","label":"Dc Description Degree","values":["Doctor of Philosophy"]},{"key":"dc:format.medium","label":"Dc Format Medium","values":["ETD"]},{"key":"dc:title","label":"Title","values":["Enhanced Soil Moisture and Streamflow Estimation in Ungauged or Data-Scarce Watersheds"]}]}],"canonical_facts":{"dc:contributor.committeechair":["Easton, Zachary"],"dc:contributor.committeemember":["Czuba, Jonathan A.","Hession, William Cully","Rippy, Megan A."],"dc:contributor.department":["Biological Systems Engineering"],"dc:creator":["Asfaw, Binyam Workeye"],"dc:date.accessioned":["2025-12-23T09:00:47Z"],"dc:date.available":["2025-12-23T09:00:47Z"],"dc:date.issued":["2025-12-22"],"dc:description.abstract":["Watershed models are important tools for quantifying hydrologic processes and evaluating water management practices. These models simplify complex systems through parametrization and assumptions, particularly in semi-distributed frameworks like Soil and Water Assessment Tool (SWAT). Clear conceptualization is critical to realize adequate representation of internal hydrologic processes, particularly in saturated excess runoff dominated watersheds where the spatio-temporal dynamics of saturated areas influence both the distributed (soil moisture state, runoff generation, or pollutant export) and integrated (streamflow) watershed responses. While streamflow is commonly used to constrain model parameters due to its availability and integrative nature, the issue of equifinality, where multiple parameter sets yield similar streamflow, but divergent internal process estimates, can lead to significant uncertainty. This dissertation addresses these challenges by introducing new and improved techniques for representing field scale soil moisture patterns and using satellite soil moisture data for model calibration in ungauged and data-scarce watersheds, especially for watersheds where variable source area runoff mechanism dominate. Chapter 2 proposes the topographic index (TI) as a tool to represent the spatial soil moisture patterns. Using unsupervised machine learning on in-situ soil moisture data from a 4.2 ha field, three distinct soil moisture groups were identified. TI values were classified into three groups using various approaches (equal-interval, equal-area, k-means, Fisher) and digital elevation model (DEM) sources (United States Geological Survey (USGS) and drone-based LiDAR) generating topographic index classes (TIC). Performance was evaluated using Spearman's correlation and misclassification rates. Results showed that low-resolution LiDAR and USGS DEMs outperformed high-resolution LiDAR DEMs, with equal-interval classification yielding the best performance. The result showed resampling to a lower resolution improved the performance of LiDAR DEMs, while TICs derived from publicly available USGS DEMs outperformed those from LiDAR DEMs. Among classification approaches equal-interval provided the highest performance. Overall, the result showed a three classes TIC can represent field scale soil moisture pattern and the effect of DEM type, resolution, and classification approach can be substantial. Chapter 3 develops a variable source area (VSA) SWAT (SWAT-VSA) model for a 14 km² watershed, incorporating the three-class TIC to improve hydrologic response unit (HRU) definition. Downscaled and bias-corrected satellite soil moisture data, along with streamflow data, were used for single and multi-objective calibration. Calibration using soil moisture data improved field-scale moisture estimation, while multi-objective calibration enhanced overall model performance. The three-class TIC structure reduced computational cost while effectively representing VSA dynamics. Chapter 4 demonstrated the utility of downscaled and bias corrected satellite soil moisture for streamflow estimation in ungauged watersheds. Three watersheds in the northeastern US with flow monitoring stations were analyzed. Soil moisture-calibrated models, using root mean squared error (RMSE) and SPatial EFficiency (SPAEF) metrics, were compared to streamflow-calibrated and regionalization-based models. While streamflow data and regionalization technique performed better for streamflow estimation, soil moisture calibrated models incorporating spatial metrics in calibration yielded comparable streamflow predictions. Satellite-based calibration offers an objective alternative to traditional regionalization, especially when properly downscaled, bias-corrected, and scaled. This research demonstrates the potential of satellite soil moisture data to improve hydrologic model performance and reduce uncertainty in data-scarce environments, offering a scalable and objective approach to watershed modeling."],"dc:description.abstractgeneral":["Watershed models are used to represent and quantify the movement and exchange of water in the landscape and between land and atmosphere. They are important tools which aid scientists to evaluate alternative water management practices. Watershed models are developed using simplifying assumptions to reduce the complexity of the system they represent and, hence, require training in order to emulate the behavior of the system. However, trained models often fall short in representing parts of the system behavior. For example, models trained on measured streamflow may produce a range of alternative model setups that can provide satisfactory streamflow estimation but divergent representation of soil moisture conditions. Scientists work to tackle this challenge, often in data-scarce conditions. This research work addresses these problems using new and improved approaches to represent model components and remotely acquired satellite soil moisture data for improved model skill in estimating streamflow and soil moisture in data-scarce watersheds. Chapter 2 illustrates the use of terrain properties for simplified representation and estimation of soil moisture variability across the landscape. This requires information on the extent of spatial variability of soil moisture and its simplification. A mathematical and statistical approach is employed to identify homogeneous soil moisture units using soil moisture measurements across a landscape. Estimating terrain properties as it relates to soil moisture is not a rudimentary task as the derived information depends on the source and detail of topographic data used. Moreover, derived terrain properties need to be classified into groups to represent a corresponding number of homogeneous soil moisture units, which also depends on choice of classification technique. Using in-situ soil moisture measurements, this study found three homogenous soil moisture units may be representative of field scale soil moisture variability and highlighted the effect of using different sources and scales of topographic data as well as classification techniques on the use and performance of derived terrain properties. Chapter 3 develops a watershed model informed on the findings in chapter 2 for improved estimation of soil moisture variability. The watershed model was trained using streamflow and satellite soil moisture individually and in conjunction. This is particularly challenging in small watersheds as satellite soil moisture data have pixel size much bigger than the size of a small watershed. To overcome this challenge and disaggregate the information into smaller pixel sizes, a technique which employee a statistical and mathematical approach was applied. The resulting data was then scaled to match the range of soil moisture variation observed from in situ soil moisture measurements. The developed model showed the use of satellite soil moisture data, especially in conjunction with streamflow data, improved the model's streamflow and soil moisture estimation performance. Chapter 4 repurposed the approach in chapter 3 for use in watersheds which does not have streamflow measurement stations. The result showed the use of satellite soil moisture data as alternative to streamflow based direct and indirect model training approaches provided adequate model training for streamflow estimation in streamflow scarce watersheds. The performance of the soil moisture based model training was especially satisfactory when model training techniques which capture the soil moisture information varying both across space and in time were included."],"dc:description.degree":["Doctor of Philosophy"],"dc:format.medium":["ETD"],"dc:identifier.other":["vt_gsexam:45315"],"dc:identifier.uri":["https://hdl.handle.net/10919/140545"],"dc:language.iso":["en"],"dc:publisher":["Virginia Tech"],"dc:rights":["Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International"],"dc:rights.uri":["http://creativecommons.org/licenses/by-nc-nd/4.0/"],"dc:subject":["Topographic Indices; Satellite Soil Moisture; Downscaling; Bias Correction; K-means Clustering; Calibration; SWAT model; Parameters; Streamflow; Ungauged"],"dc:title":["Enhanced Soil Moisture and Streamflow Estimation in Ungauged or Data-Scarce Watersheds"],"dc:type":["Dissertation"],"thesis:degree_discipline":["Biological Systems Engineering"],"thesis:degree_level":["doctoral"],"thesis:degree_name":["Doctor of Philosophy"],"thesis:institution_name":["Virginia Polytechnic Institute and State University"]},"updated_at":"2026-07-22T22:19:57Z"}