home.social

#earthobservation — Public Fediverse posts

Live and recent posts from across the Fediverse tagged #earthobservation, aggregated by home.social.

  1. Antarctic ozone hole reaches 25 million km² as growth accelerates

    The Copernicus Atmosphere Monitoring Service (CAMS) reported that the Antarctic ozone hole expanded rapidly in the first half…
    #NewsBeep #News #Science #CA #Canada #climatechange #climatemodelling #Earthobservation
    newsbeep.com/ca/897816/

  2. Antarctic ozone hole reaches 25 million km² as growth accelerates

    The Copernicus Atmosphere Monitoring Service (CAMS) reported that the Antarctic ozone hole expanded rapidly in the first half…
    #NewsBeep #News #Science #AU #Australia #Climatechange #climatemodelling #Earthobservation
    newsbeep.com/au/893976/

  3. A Scalable Cloud-Native Framework for Satellite Rainfall-Driven Landslide Early Warning in Data-Scarce Regions
    --
    doi.org/10.5194/egusphere-2026 <-- shared technical article
    --
    ee-yunuscool.projects.eartheng <-- shared web map based ALERT, Automated Landslide Early Risk Tracker
    --
    doi.org/10.1186/s40677-022-002 <-- shared paper
    --
    H/T @ Yunus Ali Pulpadan | Assistant Professor at Indian Institute of Science Education and Research (IISER), Mohali
    “[The authors] are excited to share ALERT — the Automated Landslide Early Risk Tracker, a cloud-native framework and web application developed to support near-real-time, impact-based assessment of rainfall-triggered landslide hazards.
    ALERT integrates multiple components within a single scalable framework:
    🛰️ Satellite-based rainfall observations
    🌦️ Operational weather forecasts
    🗺️ Terrain susceptibility information
    📈 Rainfall intensity–duration thresholds
    🏔️ Debris-flow runout modelling
    🏘️ Building exposure
    The goal is to move beyond simply identifying intense rainfall and towards understanding where hazardous conditions may trigger landslides and what may lie in the potential path of downstream impacts. A particular motivation behind ALERT is the challenge of developing landslide early-warning capabilities in data-scarce mountainous regions, where dense rain-gauge networks and operational monitoring infrastructure are often limited. The framework incorporates a catalogue of rainfall thresholds while also allowing users to integrate their own thresholds and susceptibility information, making it adaptable across different climatic and geomorphological settings…”
    #GIS #spatial #mapping #AI #deeplearning #massmovement #landslide #engineeringgeology #water #precipitation #rainfall #earlywarning #remotesensing #earthobservation #global #webmap #dataportal #risk #hazard #infrastructure #building #weather #forecasting #imagery #debrisflow #model #modeling #ALERT #opendata #climate #geology hydrogeomorphology geomorphology public safety global #regional

  4. A Scalable Cloud-Native Framework for Satellite Rainfall-Driven Landslide Early Warning in Data-Scarce Regions
    --
    doi.org/10.5194/egusphere-2026 <-- shared technical article
    --
    ee-yunuscool.projects.eartheng <-- shared web map based ALERT, Automated Landslide Early Risk Tracker
    --
    doi.org/10.1186/s40677-022-002 <-- shared paper
    --
    H/T @ Yunus Ali Pulpadan | Assistant Professor at Indian Institute of Science Education and Research (IISER), Mohali
    “[The authors] are excited to share ALERT — the Automated Landslide Early Risk Tracker, a cloud-native framework and web application developed to support near-real-time, impact-based assessment of rainfall-triggered landslide hazards.
    ALERT integrates multiple components within a single scalable framework:
    🛰️ Satellite-based rainfall observations
    🌦️ Operational weather forecasts
    🗺️ Terrain susceptibility information
    📈 Rainfall intensity–duration thresholds
    🏔️ Debris-flow runout modelling
    🏘️ Building exposure
    The goal is to move beyond simply identifying intense rainfall and towards understanding where hazardous conditions may trigger landslides and what may lie in the potential path of downstream impacts. A particular motivation behind ALERT is the challenge of developing landslide early-warning capabilities in data-scarce mountainous regions, where dense rain-gauge networks and operational monitoring infrastructure are often limited. The framework incorporates a catalogue of rainfall thresholds while also allowing users to integrate their own thresholds and susceptibility information, making it adaptable across different climatic and geomorphological settings…”
    #GIS #spatial #mapping #AI #deeplearning #massmovement #landslide #engineeringgeology #water #precipitation #rainfall #earlywarning #remotesensing #earthobservation #global #webmap #dataportal #risk #hazard #infrastructure #building #weather #forecasting #imagery #debrisflow #model #modeling #ALERT #opendata #climate #geology hydrogeomorphology geomorphology public safety global #regional

  5. A Scalable Cloud-Native Framework for Satellite Rainfall-Driven Landslide Early Warning in Data-Scarce Regions
    --
    doi.org/10.5194/egusphere-2026 <-- shared technical article
    --
    ee-yunuscool.projects.eartheng <-- shared web map based ALERT, Automated Landslide Early Risk Tracker
    --
    doi.org/10.1186/s40677-022-002 <-- shared paper
    --
    H/T @ Yunus Ali Pulpadan | Assistant Professor at Indian Institute of Science Education and Research (IISER), Mohali
    “[The authors] are excited to share ALERT — the Automated Landslide Early Risk Tracker, a cloud-native framework and web application developed to support near-real-time, impact-based assessment of rainfall-triggered landslide hazards.
    ALERT integrates multiple components within a single scalable framework:
    🛰️ Satellite-based rainfall observations
    🌦️ Operational weather forecasts
    🗺️ Terrain susceptibility information
    📈 Rainfall intensity–duration thresholds
    🏔️ Debris-flow runout modelling
    🏘️ Building exposure
    The goal is to move beyond simply identifying intense rainfall and towards understanding where hazardous conditions may trigger landslides and what may lie in the potential path of downstream impacts. A particular motivation behind ALERT is the challenge of developing landslide early-warning capabilities in data-scarce mountainous regions, where dense rain-gauge networks and operational monitoring infrastructure are often limited. The framework incorporates a catalogue of rainfall thresholds while also allowing users to integrate their own thresholds and susceptibility information, making it adaptable across different climatic and geomorphological settings…”
    #GIS #spatial #mapping #AI #deeplearning #massmovement #landslide #engineeringgeology #water #precipitation #rainfall #earlywarning #remotesensing #earthobservation #global #webmap #dataportal #risk #hazard #infrastructure #building #weather #forecasting #imagery #debrisflow #model #modeling #ALERT #opendata #climate #geology hydrogeomorphology geomorphology public safety global #regional

  6. A Scalable Cloud-Native Framework for Satellite Rainfall-Driven Landslide Early Warning in Data-Scarce Regions
    --
    doi.org/10.5194/egusphere-2026 <-- shared technical article
    --
    ee-yunuscool.projects.eartheng <-- shared web map based ALERT, Automated Landslide Early Risk Tracker
    --
    doi.org/10.1186/s40677-022-002 <-- shared paper
    --
    H/T @ Yunus Ali Pulpadan | Assistant Professor at Indian Institute of Science Education and Research (IISER), Mohali
    “[The authors] are excited to share ALERT — the Automated Landslide Early Risk Tracker, a cloud-native framework and web application developed to support near-real-time, impact-based assessment of rainfall-triggered landslide hazards.
    ALERT integrates multiple components within a single scalable framework:
    🛰️ Satellite-based rainfall observations
    🌦️ Operational weather forecasts
    🗺️ Terrain susceptibility information
    📈 Rainfall intensity–duration thresholds
    🏔️ Debris-flow runout modelling
    🏘️ Building exposure
    The goal is to move beyond simply identifying intense rainfall and towards understanding where hazardous conditions may trigger landslides and what may lie in the potential path of downstream impacts. A particular motivation behind ALERT is the challenge of developing landslide early-warning capabilities in data-scarce mountainous regions, where dense rain-gauge networks and operational monitoring infrastructure are often limited. The framework incorporates a catalogue of rainfall thresholds while also allowing users to integrate their own thresholds and susceptibility information, making it adaptable across different climatic and geomorphological settings…”
    #GIS #spatial #mapping #AI #deeplearning #massmovement #landslide #engineeringgeology #water #precipitation #rainfall #earlywarning #remotesensing #earthobservation #global #webmap #dataportal #risk #hazard #infrastructure #building #weather #forecasting #imagery #debrisflow #model #modeling #ALERT #opendata #climate #geology hydrogeomorphology geomorphology public safety global #regional

  7. A Scalable Cloud-Native Framework for Satellite Rainfall-Driven Landslide Early Warning in Data-Scarce Regions
    --
    doi.org/10.5194/egusphere-2026 <-- shared technical article
    --
    ee-yunuscool.projects.eartheng <-- shared web map based ALERT, Automated Landslide Early Risk Tracker
    --
    doi.org/10.1186/s40677-022-002 <-- shared paper
    --
    H/T @ Yunus Ali Pulpadan | Assistant Professor at Indian Institute of Science Education and Research (IISER), Mohali
    “[The authors] are excited to share ALERT — the Automated Landslide Early Risk Tracker, a cloud-native framework and web application developed to support near-real-time, impact-based assessment of rainfall-triggered landslide hazards.
    ALERT integrates multiple components within a single scalable framework:
    🛰️ Satellite-based rainfall observations
    🌦️ Operational weather forecasts
    🗺️ Terrain susceptibility information
    📈 Rainfall intensity–duration thresholds
    🏔️ Debris-flow runout modelling
    🏘️ Building exposure
    The goal is to move beyond simply identifying intense rainfall and towards understanding where hazardous conditions may trigger landslides and what may lie in the potential path of downstream impacts. A particular motivation behind ALERT is the challenge of developing landslide early-warning capabilities in data-scarce mountainous regions, where dense rain-gauge networks and operational monitoring infrastructure are often limited. The framework incorporates a catalogue of rainfall thresholds while also allowing users to integrate their own thresholds and susceptibility information, making it adaptable across different climatic and geomorphological settings…”
    hydrogeomorphology geomorphology public safety global

  8. Australian supersites on international cal/val list

    A map of CEOS-recommended calibration/validation ‘supersites’ around the world. Credit: CEOS Twelve Australian locations have been included in…
    #NewsBeep #News #Science #AU #Australia #Cal/Val #calibration/validation #CEOs #Earthobservation #remotesensing #satellitederiveddata #supersites #tern
    newsbeep.com/au/851262/

  9. Australian supersites on international cal/val list

    A map of CEOS-recommended calibration/validation ‘supersites’ around the world. Credit: CEOS Twelve Australian locations have been included in…
    #NewsBeep #News #Science #AU #Australia #Cal/Val #calibration/validation #CEOs #Earthobservation #remotesensing #satellitederiveddata #supersites #tern
    newsbeep.com/au/851262/

  10. Flood And Landslide Susceptibility Assessment And Multi Hazard Interaction Mapping Using Machine Learning And GIS For Sustainable Settlement Planning In Nepal
    --
    doi.org/10.1007/s44288-026-006 <-- shared paper
    --
    H/T @Narayan Thapa | Earth Data Modeling
    “Nepal lies within an active seismic zone and is influenced by most dynamic climatic systems in the world. It faces compounding floods and landslide threats. Impacts are worst where multi-hazard interactions create spatially linked corridors. Despite frequent co-occurrence, national-scale assessments remain limited. This study presents machine learning and GIS-based approach to map nationwide susceptibility to floods, landslides, and identify their potential interaction zones, and delineate critical multi-hazard flow zones through spatial adjacency analysis. Using Google Earth Engine, the Random Forest model integrates topographic, climatic, environmental, and hydrological datasets to overcome subjective expert-driven methods. The model achieved strong predictive accuracy (AUC: 0.84 for floods, 0.85 for landslides). The results showed 19% of Nepal’s lowlands are medium to very highly susceptible to inundation, threatening approximately 900,000 people and over 3.4 million buildings; whilst in the hilly terrains, 40% is susceptible to slope-failure endangering 200,000 people and about 0.6 million buildings. K-means clustering followed by spatial adjacency analysis identified four spatial zonation: 81% of national area as low-hazard zone, 9% as flood-only zone, 5% as landslide-only zone, and 5% as interaction zones. Critical multi-hazard flow zone covering 7,588 km² represents spatially connected corridors linking interaction zones to downstream flood-prone populated areas, affecting 88 km² built-up land and 1,722 km² cropland. These zones represent susceptibility-based spatial connectivity rather than physically simulated cascading processes. These findings support recommendations for risk-informed land-use planning, resilient infrastructure development and climate adaptation aligned to sustainable development and investment risk screening…”
    #GIS #spatial #mapping #GoogleEarthEngine #MachineLearning #RemoteSensing #GeospatialAI #DisasterRiskReduction #MultiHazard #ClimateAdaptation #climatechange #extremeweather #LandUsePlanning #SustainableDevelopment #InfrastructurePlanning #RiskAssessment #NaturalHazards #Nepal #EarthObservation #HinduKushHimalaya #HKH #HinduKush #Himalayas #risk #hazard #assessment #national #regional #spatialanalysis #spatiotemporal #massmovement #landslide #assessment #mitigation #water #hydrology #flood #flooding #sustainability

  11. Flood And Landslide Susceptibility Assessment And Multi Hazard Interaction Mapping Using Machine Learning And GIS For Sustainable Settlement Planning In Nepal
    --
    doi.org/10.1007/s44288-026-006 <-- shared paper
    --
    H/T @Narayan Thapa | Earth Data Modeling
    “Nepal lies within an active seismic zone and is influenced by most dynamic climatic systems in the world. It faces compounding floods and landslide threats. Impacts are worst where multi-hazard interactions create spatially linked corridors. Despite frequent co-occurrence, national-scale assessments remain limited. This study presents machine learning and GIS-based approach to map nationwide susceptibility to floods, landslides, and identify their potential interaction zones, and delineate critical multi-hazard flow zones through spatial adjacency analysis. Using Google Earth Engine, the Random Forest model integrates topographic, climatic, environmental, and hydrological datasets to overcome subjective expert-driven methods. The model achieved strong predictive accuracy (AUC: 0.84 for floods, 0.85 for landslides). The results showed 19% of Nepal’s lowlands are medium to very highly susceptible to inundation, threatening approximately 900,000 people and over 3.4 million buildings; whilst in the hilly terrains, 40% is susceptible to slope-failure endangering 200,000 people and about 0.6 million buildings. K-means clustering followed by spatial adjacency analysis identified four spatial zonation: 81% of national area as low-hazard zone, 9% as flood-only zone, 5% as landslide-only zone, and 5% as interaction zones. Critical multi-hazard flow zone covering 7,588 km² represents spatially connected corridors linking interaction zones to downstream flood-prone populated areas, affecting 88 km² built-up land and 1,722 km² cropland. These zones represent susceptibility-based spatial connectivity rather than physically simulated cascading processes. These findings support recommendations for risk-informed land-use planning, resilient infrastructure development and climate adaptation aligned to sustainable development and investment risk screening…”
    #GIS #spatial #mapping #GoogleEarthEngine #MachineLearning #RemoteSensing #GeospatialAI #DisasterRiskReduction #MultiHazard #ClimateAdaptation #climatechange #extremeweather #LandUsePlanning #SustainableDevelopment #InfrastructurePlanning #RiskAssessment #NaturalHazards #Nepal #EarthObservation #HinduKushHimalaya #HKH #HinduKush #Himalayas #risk #hazard #assessment #national #regional #spatialanalysis #spatiotemporal #massmovement #landslide #assessment #mitigation #water #hydrology #flood #flooding #sustainability

  12. Flood And Landslide Susceptibility Assessment And Multi Hazard Interaction Mapping Using Machine Learning And GIS For Sustainable Settlement Planning In Nepal
    --
    doi.org/10.1007/s44288-026-006 <-- shared paper
    --
    H/T @Narayan Thapa | Earth Data Modeling
    “Nepal lies within an active seismic zone and is influenced by most dynamic climatic systems in the world. It faces compounding floods and landslide threats. Impacts are worst where multi-hazard interactions create spatially linked corridors. Despite frequent co-occurrence, national-scale assessments remain limited. This study presents machine learning and GIS-based approach to map nationwide susceptibility to floods, landslides, and identify their potential interaction zones, and delineate critical multi-hazard flow zones through spatial adjacency analysis. Using Google Earth Engine, the Random Forest model integrates topographic, climatic, environmental, and hydrological datasets to overcome subjective expert-driven methods. The model achieved strong predictive accuracy (AUC: 0.84 for floods, 0.85 for landslides). The results showed 19% of Nepal’s lowlands are medium to very highly susceptible to inundation, threatening approximately 900,000 people and over 3.4 million buildings; whilst in the hilly terrains, 40% is susceptible to slope-failure endangering 200,000 people and about 0.6 million buildings. K-means clustering followed by spatial adjacency analysis identified four spatial zonation: 81% of national area as low-hazard zone, 9% as flood-only zone, 5% as landslide-only zone, and 5% as interaction zones. Critical multi-hazard flow zone covering 7,588 km² represents spatially connected corridors linking interaction zones to downstream flood-prone populated areas, affecting 88 km² built-up land and 1,722 km² cropland. These zones represent susceptibility-based spatial connectivity rather than physically simulated cascading processes. These findings support recommendations for risk-informed land-use planning, resilient infrastructure development and climate adaptation aligned to sustainable development and investment risk screening…”
    #GIS #spatial #mapping #GoogleEarthEngine #MachineLearning #RemoteSensing #GeospatialAI #DisasterRiskReduction #MultiHazard #ClimateAdaptation #climatechange #extremeweather #LandUsePlanning #SustainableDevelopment #InfrastructurePlanning #RiskAssessment #NaturalHazards #Nepal #EarthObservation #HinduKushHimalaya #HKH #HinduKush #Himalayas #risk #hazard #assessment #national #regional #spatialanalysis #spatiotemporal #massmovement #landslide #assessment #mitigation #water #hydrology #flood #flooding #sustainability

  13. Flood And Landslide Susceptibility Assessment And Multi Hazard Interaction Mapping Using Machine Learning And GIS For Sustainable Settlement Planning In Nepal
    --
    doi.org/10.1007/s44288-026-006 <-- shared paper
    --
    H/T @Narayan Thapa | Earth Data Modeling
    “Nepal lies within an active seismic zone and is influenced by most dynamic climatic systems in the world. It faces compounding floods and landslide threats. Impacts are worst where multi-hazard interactions create spatially linked corridors. Despite frequent co-occurrence, national-scale assessments remain limited. This study presents machine learning and GIS-based approach to map nationwide susceptibility to floods, landslides, and identify their potential interaction zones, and delineate critical multi-hazard flow zones through spatial adjacency analysis. Using Google Earth Engine, the Random Forest model integrates topographic, climatic, environmental, and hydrological datasets to overcome subjective expert-driven methods. The model achieved strong predictive accuracy (AUC: 0.84 for floods, 0.85 for landslides). The results showed 19% of Nepal’s lowlands are medium to very highly susceptible to inundation, threatening approximately 900,000 people and over 3.4 million buildings; whilst in the hilly terrains, 40% is susceptible to slope-failure endangering 200,000 people and about 0.6 million buildings. K-means clustering followed by spatial adjacency analysis identified four spatial zonation: 81% of national area as low-hazard zone, 9% as flood-only zone, 5% as landslide-only zone, and 5% as interaction zones. Critical multi-hazard flow zone covering 7,588 km² represents spatially connected corridors linking interaction zones to downstream flood-prone populated areas, affecting 88 km² built-up land and 1,722 km² cropland. These zones represent susceptibility-based spatial connectivity rather than physically simulated cascading processes. These findings support recommendations for risk-informed land-use planning, resilient infrastructure development and climate adaptation aligned to sustainable development and investment risk screening…”
    #GIS #spatial #mapping #GoogleEarthEngine #MachineLearning #RemoteSensing #GeospatialAI #DisasterRiskReduction #MultiHazard #ClimateAdaptation #climatechange #extremeweather #LandUsePlanning #SustainableDevelopment #InfrastructurePlanning #RiskAssessment #NaturalHazards #Nepal #EarthObservation #HinduKushHimalaya #HKH #HinduKush #Himalayas #risk #hazard #assessment #national #regional #spatialanalysis #spatiotemporal #massmovement #landslide #assessment #mitigation #water #hydrology #flood #flooding #sustainability

  14. Flood And Landslide Susceptibility Assessment And Multi Hazard Interaction Mapping Using Machine Learning And GIS For Sustainable Settlement Planning In Nepal
    --
    doi.org/10.1007/s44288-026-006 <-- shared paper
    --
    H/T @Narayan Thapa | Earth Data Modeling
    “Nepal lies within an active seismic zone and is influenced by most dynamic climatic systems in the world. It faces compounding floods and landslide threats. Impacts are worst where multi-hazard interactions create spatially linked corridors. Despite frequent co-occurrence, national-scale assessments remain limited. This study presents machine learning and GIS-based approach to map nationwide susceptibility to floods, landslides, and identify their potential interaction zones, and delineate critical multi-hazard flow zones through spatial adjacency analysis. Using Google Earth Engine, the Random Forest model integrates topographic, climatic, environmental, and hydrological datasets to overcome subjective expert-driven methods. The model achieved strong predictive accuracy (AUC: 0.84 for floods, 0.85 for landslides). The results showed 19% of Nepal’s lowlands are medium to very highly susceptible to inundation, threatening approximately 900,000 people and over 3.4 million buildings; whilst in the hilly terrains, 40% is susceptible to slope-failure endangering 200,000 people and about 0.6 million buildings. K-means clustering followed by spatial adjacency analysis identified four spatial zonation: 81% of national area as low-hazard zone, 9% as flood-only zone, 5% as landslide-only zone, and 5% as interaction zones. Critical multi-hazard flow zone covering 7,588 km² represents spatially connected corridors linking interaction zones to downstream flood-prone populated areas, affecting 88 km² built-up land and 1,722 km² cropland. These zones represent susceptibility-based spatial connectivity rather than physically simulated cascading processes. These findings support recommendations for risk-informed land-use planning, resilient infrastructure development and climate adaptation aligned to sustainable development and investment risk screening…”

  15. 🌍 My #ML4EO2026 talk focused on spatial machine learning and prediction-domain adaptive evaluation: defining the prediction domain, adapting validation, and weighting evaluation by deployment location properties.

    Slides: jakubnowosad.com/ml4eo2026/

    #SpatialML #MachineLearning #RSpatial #EarthObservation

  16. 🌍 My talk focused on spatial machine learning and prediction-domain adaptive evaluation: defining the prediction domain, adapting validation, and weighting evaluation by deployment location properties.

    Slides: jakubnowosad.com/ml4eo2026/

  17. 🌍 My #ML4EO2026 talk focused on spatial machine learning and prediction-domain adaptive evaluation: defining the prediction domain, adapting validation, and weighting evaluation by deployment location properties.

    Slides: jakubnowosad.com/ml4eo2026/

    #SpatialML #MachineLearning #RSpatial #EarthObservation

  18. 🌍 My #ML4EO2026 talk focused on spatial machine learning and prediction-domain adaptive evaluation: defining the prediction domain, adapting validation, and weighting evaluation by deployment location properties.

    Slides: jakubnowosad.com/ml4eo2026/

    #SpatialML #MachineLearning #RSpatial #EarthObservation

  19. 🌍 My #ML4EO2026 talk focused on spatial machine learning and prediction-domain adaptive evaluation: defining the prediction domain, adapting validation, and weighting evaluation by deployment location properties.

    Slides: jakubnowosad.com/ml4eo2026/

    #SpatialML #MachineLearning #RSpatial #EarthObservation

  20. Improving Forest Loss Mapping In Nepal Using Landtrendr Time-Series And Machine Learning
    --
    doi.org/10.1016/j.rsase.2025.1 <-- share paper
    --
    “HIGHLIGHTS:
    • ViT-based forest mask, multispectral ensemble LandTrendr and terrain shadow mask.
    • District-level RF/XGBoost model training with expert-weighted validation.
    • Outperformed GFC and REDD + AI benchmarks in accuracy and F1 performance.
    • RF excelled in High Mountains/Himalayas; XGBoost in the lower Mountain regions.
    • NBR contributed the most; snow-impacted forest loss uncertainty was observed..."
    #Forestdisturbance #forest #disturbance #remotesensing #LandTrendr #workflow #timeseries #ViT #RF #XGBoost #GEE #Nepal #ForestNepal #spatial #GIS #mapping #earthobservation #landsat #Himalayas #mountains #alpine #vegetation #AI #multispectral #monitoring #spatialanalysis #spatiotemporal #loss #change #machinelearning #NDR #conservation #planning #policy #mitagion #ecology #Karnali #Bagmati, #Darchula #Siwalik #GlobalForestChange #Degradation

  21. Improving Forest Loss Mapping In Nepal Using Landtrendr Time-Series And Machine Learning
    --
    doi.org/10.1016/j.rsase.2025.1 <-- share paper
    --
    “HIGHLIGHTS:
    • ViT-based forest mask, multispectral ensemble LandTrendr and terrain shadow mask.
    • District-level RF/XGBoost model training with expert-weighted validation.
    • Outperformed GFC and REDD + AI benchmarks in accuracy and F1 performance.
    • RF excelled in High Mountains/Himalayas; XGBoost in the lower Mountain regions.
    • NBR contributed the most; snow-impacted forest loss uncertainty was observed..."
    #Forestdisturbance #forest #disturbance #remotesensing #LandTrendr #workflow #timeseries #ViT #RF #XGBoost #GEE #Nepal #ForestNepal #spatial #GIS #mapping #earthobservation #landsat #Himalayas #mountains #alpine #vegetation #AI #multispectral #monitoring #spatialanalysis #spatiotemporal #loss #change #machinelearning #NDR #conservation #planning #policy #mitagion #ecology #Karnali #Bagmati, #Darchula #Siwalik #GlobalForestChange #Degradation

  22. Improving Forest Loss Mapping In Nepal Using Landtrendr Time-Series And Machine Learning
    --
    doi.org/10.1016/j.rsase.2025.1 <-- share paper
    --
    “HIGHLIGHTS:
    • ViT-based forest mask, multispectral ensemble LandTrendr and terrain shadow mask.
    • District-level RF/XGBoost model training with expert-weighted validation.
    • Outperformed GFC and REDD + AI benchmarks in accuracy and F1 performance.
    • RF excelled in High Mountains/Himalayas; XGBoost in the lower Mountain regions.
    • NBR contributed the most; snow-impacted forest loss uncertainty was observed..."
    #Forestdisturbance #forest #disturbance #remotesensing #LandTrendr #workflow #timeseries #ViT #RF #XGBoost #GEE #Nepal #ForestNepal #spatial #GIS #mapping #earthobservation #landsat #Himalayas #mountains #alpine #vegetation #AI #multispectral #monitoring #spatialanalysis #spatiotemporal #loss #change #machinelearning #NDR #conservation #planning #policy #mitagion #ecology #Karnali #Bagmati, #Darchula #Siwalik #GlobalForestChange #Degradation

  23. Improving Forest Loss Mapping In Nepal Using Landtrendr Time-Series And Machine Learning
    --
    doi.org/10.1016/j.rsase.2025.1 <-- share paper
    --
    “HIGHLIGHTS:
    • ViT-based forest mask, multispectral ensemble LandTrendr and terrain shadow mask.
    • District-level RF/XGBoost model training with expert-weighted validation.
    • Outperformed GFC and REDD + AI benchmarks in accuracy and F1 performance.
    • RF excelled in High Mountains/Himalayas; XGBoost in the lower Mountain regions.
    • NBR contributed the most; snow-impacted forest loss uncertainty was observed..."
    #Forestdisturbance #forest #disturbance #remotesensing #LandTrendr #workflow #timeseries #ViT #RF #XGBoost #GEE #Nepal #ForestNepal #spatial #GIS #mapping #earthobservation #landsat #Himalayas #mountains #alpine #vegetation #AI #multispectral #monitoring #spatialanalysis #spatiotemporal #loss #change #machinelearning #NDR #conservation #planning #policy #mitagion #ecology #Karnali #Bagmati, #Darchula #Siwalik #GlobalForestChange #Degradation

  24. Improving Forest Loss Mapping In Nepal Using Landtrendr Time-Series And Machine Learning
    --
    doi.org/10.1016/j.rsase.2025.1 <-- share paper
    --
    “HIGHLIGHTS:
    • ViT-based forest mask, multispectral ensemble LandTrendr and terrain shadow mask.
    • District-level RF/XGBoost model training with expert-weighted validation.
    • Outperformed GFC and REDD + AI benchmarks in accuracy and F1 performance.
    • RF excelled in High Mountains/Himalayas; XGBoost in the lower Mountain regions.
    • NBR contributed the most; snow-impacted forest loss uncertainty was observed..."
    ,