{"id":{"repo_id":"calpoly","oai_identifier":"oai:digitalcommons.calpoly.edu:theses-5009"},"canonical_url":"https://search.ndltd.org/etd/calpoly/oai:digitalcommons.calpoly.edu:theses-5009","repository":{"repo_id":"calpoly","name":"Cal Poly","base_url":"https://digitalcommons.calpoly.edu/do/oai/"},"display":{"title":"Multimodal Machine Learning for Soil Burn Severity Mapping Across California Wildfires","abstract":"Accurate mapping of soil burn severity (SBS) is critical for post-fire watershed management, erosion risk assessment, and ecological recovery planning, yet traditional field-based approaches remain costly, time-intensive, and spatially limited. This thesis presents a machine learning pipeline for wall-to-wall SBS classification across California wildfires using multi-sensor satellite imagery, terrain derivatives, and bioclimatic covariates. Field-collected SBS observations (n = 2,180) from 52 wildfires occur- ring between 2013 and 2025, sourced from the U.S. Forest Service and CAL FIRE, were used to train and evaluate multiple classification architectures within a Google Earth Engine and Google Cloud-based prediction framework. After upsampling the unburned class to address class imbalance, the training dataset included n = 3,204 observations. Among the evaluated models, XGBoost trained on expert-engineered spectral change features achieved the highest cross-validated performance (Cohen’s Kappa = 0.61), outperforming neural architectures including a CNN, MLP, and hybrid CNN+MLP fusion model, suggesting that carefully designed spectral indices capture the dominant signal in soil burn severity classification more eﬀectively than spatially learned representations at this dataset scale. The XGBoost model is also validated using a LOFO protocol where one fire is held out each train/test split for generalizable, ecologically valid results (Cohen’s Kappa = 0.44). The pipeline produces spatially continuous soil burn severity maps across all 52 study fires, with an interactive visualization tool under development at the University of California, Davis to provide land managers with accessible access to these outputs. Together, this thesis demonstrates that accurate landscape-scale soil burn severity mapping is achievable without expensive field campaigns, oﬀering a reproducible framework and multi-fire benchmark dataset for future research in fire-prone regions.","abstract_html":"Accurate mapping of soil burn severity (SBS) is critical for post-fire watershed management, erosion risk assessment, and ecological recovery planning, yet traditional field-based approaches remain costly, time-intensive, and spatially limited. This thesis presents a machine learning pipeline for wall-to-wall SBS classification across California wildfires using multi-sensor satellite imagery, terrain derivatives, and bioclimatic covariates. Field-collected SBS observations (n = 2,180) from 52 wildfires occur- ring between 2013 and 2025, sourced from the U.S. Forest Service and CAL FIRE, were used to train and evaluate multiple classification architectures within a Google Earth Engine and Google Cloud-based prediction framework. After upsampling the unburned class to address class imbalance, the training dataset included n = 3,204 observations. Among the evaluated models, XGBoost trained on expert-engineered spectral change features achieved the highest cross-validated performance (Cohen’s Kappa = 0.61), outperforming neural architectures including a CNN, MLP, and hybrid CNN+MLP fusion model, suggesting that carefully designed spectral indices capture the dominant signal in soil burn severity classification more eﬀectively than spatially learned representations at this dataset scale. The XGBoost model is also validated using a LOFO protocol where one fire is held out each train/test split for generalizable, ecologically valid results (Cohen’s Kappa = 0.44). The pipeline produces spatially continuous soil burn severity maps across all 52 study fires, with an interactive visualization tool under development at the University of California, Davis to provide land managers with accessible access to these outputs. Together, this thesis demonstrates that accurate landscape-scale soil burn severity mapping is achievable without expensive field campaigns, oﬀering a reproducible framework and multi-fire benchmark dataset for future research in fire-prone regions.","abstract_has_math":false,"creators":["Checker, Sanjana"],"institution":null,"degree_name":"MS in Computer Science","degree_level":null,"degree_discipline":"Computer Science","degree_department":null,"school":null,"contributors":["Jonathan Ventura","Computer Science","College of Engineering"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":null,"date_issued":"","date_published":null,"updated_at":"2026-11-19T07:12:58Z","subjects":["Soil burn severity","wildfire","remote sensing","machine learning","XGBoost","Google Earth Engine","Applied Statistics","Artificial Intelligence and Robotics","Natural Resources Management and Policy"],"languages":[],"rights":[],"rights_urls":[],"identifier_entries":[],"person_identifier_entries":[]},"links":{"outbound_url":"https://digitalcommons.calpoly.edu/theses/3393","outbound_label":"Repository record","outbound_source":"dc:identifier","record_url":"https://digitalcommons.calpoly.edu/theses/3393","record_label":"Repository record","record_kind":"record","record_source":"dc:identifier","download_url":null,"download_label":null,"download_source":null,"access_status":null,"embargo_until":null,"access_note":null},"source_record":{"url":"https://digitalcommons.calpoly.edu/do/oai/?verb=GetRecord&metadataPrefix=oai_etdms&identifier=oai%3Adigitalcommons.calpoly.edu%3Atheses-5009","prefix":"oai_etdms"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Jonathan Ventura","Computer Science","College of Engineering"]},{"key":"dc:creator","label":"Author","values":["Checker, Sanjana"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.available","label":"Dc Date Available","values":["2026-06-11T07:00:00Z"]},{"key":"dc:date.created","label":"Dc Date Created","values":["2026-06-01T07:00:00Z"]},{"key":"thesis:degree_discipline","label":"Discipline","values":["Computer Science"]},{"key":"thesis:degree_name","label":"Degree Name","values":["MS in Computer Science"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Soil burn severity","wildfire","remote sensing","machine learning","XGBoost","Google Earth Engine","Applied Statistics","Artificial Intelligence and Robotics","Natural Resources Management and Policy"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["https://digitalcommons.calpoly.edu/theses/3393"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["Accurate mapping of soil burn severity (SBS) is critical for post-fire watershed management, erosion risk assessment, and ecological recovery planning, yet traditional field-based approaches remain costly, time-intensive, and spatially limited. This thesis presents a machine learning pipeline for wall-to-wall SBS classification across California wildfires using multi-sensor satellite imagery, terrain derivatives, and bioclimatic covariates. Field-collected SBS observations (n = 2,180) from 52 wildfires occur- ring between 2013 and 2025, sourced from the U.S. Forest Service and CAL FIRE, were used to train and evaluate multiple classification architectures within a Google Earth Engine and Google Cloud-based prediction framework. After upsampling the unburned class to address class imbalance, the training dataset included n = 3,204 observations. Among the evaluated models, XGBoost trained on expert-engineered spectral change features achieved the highest cross-validated performance (Cohen’s Kappa = 0.61), outperforming neural architectures including a CNN, MLP, and hybrid CNN+MLP fusion model, suggesting that carefully designed spectral indices capture the dominant signal in soil burn severity classification more eﬀectively than spatially learned representations at this dataset scale. The XGBoost model is also validated using a LOFO protocol where one fire is held out each train/test split for generalizable, ecologically valid results (Cohen’s Kappa = 0.44). The pipeline produces spatially continuous soil burn severity maps across all 52 study fires, with an interactive visualization tool under development at the University of California, Davis to provide land managers with accessible access to these outputs. Together, this thesis demonstrates that accurate landscape-scale soil burn severity mapping is achievable without expensive field campaigns, oﬀering a reproducible framework and multi-fire benchmark dataset for future research in fire-prone regions."]},{"key":"dc:title","label":"Title","values":["Multimodal Machine Learning for Soil Burn Severity Mapping Across California Wildfires"]}]}],"canonical_facts":{"dc:contributor":["Jonathan Ventura","Computer Science","College of Engineering"],"dc:creator":["Checker, Sanjana"],"dc:date.available":["2026-06-11T07:00:00Z"],"dc:date.created":["2026-06-01T07:00:00Z"],"dc:description.abstract":["Accurate mapping of soil burn severity (SBS) is critical for post-fire watershed management, erosion risk assessment, and ecological recovery planning, yet traditional field-based approaches remain costly, time-intensive, and spatially limited. This thesis presents a machine learning pipeline for wall-to-wall SBS classification across California wildfires using multi-sensor satellite imagery, terrain derivatives, and bioclimatic covariates. Field-collected SBS observations (n = 2,180) from 52 wildfires occur- ring between 2013 and 2025, sourced from the U.S. Forest Service and CAL FIRE, were used to train and evaluate multiple classification architectures within a Google Earth Engine and Google Cloud-based prediction framework. After upsampling the unburned class to address class imbalance, the training dataset included n = 3,204 observations. Among the evaluated models, XGBoost trained on expert-engineered spectral change features achieved the highest cross-validated performance (Cohen’s Kappa = 0.61), outperforming neural architectures including a CNN, MLP, and hybrid CNN+MLP fusion model, suggesting that carefully designed spectral indices capture the dominant signal in soil burn severity classification more eﬀectively than spatially learned representations at this dataset scale. The XGBoost model is also validated using a LOFO protocol where one fire is held out each train/test split for generalizable, ecologically valid results (Cohen’s Kappa = 0.44). The pipeline produces spatially continuous soil burn severity maps across all 52 study fires, with an interactive visualization tool under development at the University of California, Davis to provide land managers with accessible access to these outputs. Together, this thesis demonstrates that accurate landscape-scale soil burn severity mapping is achievable without expensive field campaigns, oﬀering a reproducible framework and multi-fire benchmark dataset for future research in fire-prone regions."],"dc:identifier":["https://digitalcommons.calpoly.edu/theses/3393"],"dc:subject":["Soil burn severity","wildfire","remote sensing","machine learning","XGBoost","Google Earth Engine","Applied Statistics","Artificial Intelligence and Robotics","Natural Resources Management and Policy"],"dc:title":["Multimodal Machine Learning for Soil Burn Severity Mapping Across California Wildfires"],"thesis:degree_discipline":["Computer Science"],"thesis:degree_name":["MS in Computer Science"]},"updated_at":"2026-11-19T07:12:58Z"}