Advancing Soil Moisture Prediction Using Satellite and UAV-based Imagery with Machine Learning Models
Crop yields in Pakistan are significantly lower than their potential, primarily due to limited water availability and the reliance on rotation water delivery instead of demand-based water supply. The absence of spatially explicit information on water stress at the farm level further constrains overa...
| Autores principales: | , , , , , |
|---|---|
| Formato: | Journal Article |
| Lenguaje: | Inglés |
| Publicado: |
Springer
2025
|
| Materias: | |
| Acceso en línea: | https://hdl.handle.net/10568/178364 |
| _version_ | 1855518907583954944 |
|---|---|
| author | Hussain, S. Arshad, M. Cheema, Muhammad Jehanzeb Masud Qamar, M. U. Wajid, S. A. Daccache, A. |
| author_browse | Arshad, M. Cheema, Muhammad Jehanzeb Masud Daccache, A. Hussain, S. Qamar, M. U. Wajid, S. A. |
| author_facet | Hussain, S. Arshad, M. Cheema, Muhammad Jehanzeb Masud Qamar, M. U. Wajid, S. A. Daccache, A. |
| author_sort | Hussain, S. |
| collection | Repository of Agricultural Research Outputs (CGSpace) |
| description | Crop yields in Pakistan are significantly lower than their potential, primarily due to limited water availability and the reliance on rotation water delivery instead of demand-based water supply. The absence of spatially explicit information on water stress at the farm level further constrains overall crop productivity. Therefore, it is essential to map soil moisture availability and monitor moisture stress to enhance the efficiency of water delivery at the district level and promote precision in on-farm irrigation application. To continuously monitor soil moisture availability and identify moisture stress hotspots, peanut crops were cultivated during the 2021 and 2022 growing seasons at PMAS-Arid Agriculture University Research Center, Koont Farm, Rawalpindi. Experimental trials were conducted in both irrigated (drip irrigation) and rainfed fields to address soil moisture variability in the context of precision irrigation management. Soil moisture monitoring was performed using a combination of proximal soil moisture sensors, satellite data (Landsat 8/9 and Sentinel-2), and Unmanned Aerial Vehicles (UAVs) equipped with multispectral sensors. Satellite data and UAV imagery were processed to calculate soil moisture indices, including the Normalized Difference Water Index (NDWI), Moisture Vegetation Index (MVI), Water Stress Index (WSI), and Drought Stress Water Index (DSWI-4). Ground-truth data, including in situ soil moisture measurements at 15 cm depth and crop yield observations, were recorded for validation. A machine learning (ML) model, Random Forest (RF), was employed to accurately predict soil moisture content at 15 cm depth. The spatial maps were generated using satellite data and UAV-based imagery to identify the specific areas experiencing moisture stress. Among the moisture stress indices, WSI demonstrated a strong positive correlation with soil moisture (R² = 0.95 in 2022 and 0.85 in 2021) in the drip irrigated field. The RF model predicted soil moisture with a high accuracy (R² = 0.97 to 0.99 and RMSE = 0.00) using UAV-based moisture indices as input parameters. Furthermore, the improved surveillance of moisture stress enabled the identification of hotspot areas, allowing for the targeted implementation of protective measures for precision irrigation at the farm level. |
| format | Journal Article |
| id | CGSpace178364 |
| institution | CGIAR Consortium |
| language | Inglés |
| publishDate | 2025 |
| publishDateRange | 2025 |
| publishDateSort | 2025 |
| publisher | Springer |
| publisherStr | Springer |
| record_format | dspace |
| spelling | CGSpace1783642025-11-28T05:49:01Z Advancing Soil Moisture Prediction Using Satellite and UAV-based Imagery with Machine Learning Models Hussain, S. Arshad, M. Cheema, Muhammad Jehanzeb Masud Qamar, M. U. Wajid, S. A. Daccache, A. soil water content satellite imagery unmanned aerial vehicles machine learning irrigation efficiency Crop yields in Pakistan are significantly lower than their potential, primarily due to limited water availability and the reliance on rotation water delivery instead of demand-based water supply. The absence of spatially explicit information on water stress at the farm level further constrains overall crop productivity. Therefore, it is essential to map soil moisture availability and monitor moisture stress to enhance the efficiency of water delivery at the district level and promote precision in on-farm irrigation application. To continuously monitor soil moisture availability and identify moisture stress hotspots, peanut crops were cultivated during the 2021 and 2022 growing seasons at PMAS-Arid Agriculture University Research Center, Koont Farm, Rawalpindi. Experimental trials were conducted in both irrigated (drip irrigation) and rainfed fields to address soil moisture variability in the context of precision irrigation management. Soil moisture monitoring was performed using a combination of proximal soil moisture sensors, satellite data (Landsat 8/9 and Sentinel-2), and Unmanned Aerial Vehicles (UAVs) equipped with multispectral sensors. Satellite data and UAV imagery were processed to calculate soil moisture indices, including the Normalized Difference Water Index (NDWI), Moisture Vegetation Index (MVI), Water Stress Index (WSI), and Drought Stress Water Index (DSWI-4). Ground-truth data, including in situ soil moisture measurements at 15 cm depth and crop yield observations, were recorded for validation. A machine learning (ML) model, Random Forest (RF), was employed to accurately predict soil moisture content at 15 cm depth. The spatial maps were generated using satellite data and UAV-based imagery to identify the specific areas experiencing moisture stress. Among the moisture stress indices, WSI demonstrated a strong positive correlation with soil moisture (R² = 0.95 in 2022 and 0.85 in 2021) in the drip irrigated field. The RF model predicted soil moisture with a high accuracy (R² = 0.97 to 0.99 and RMSE = 0.00) using UAV-based moisture indices as input parameters. Furthermore, the improved surveillance of moisture stress enabled the identification of hotspot areas, allowing for the targeted implementation of protective measures for precision irrigation at the farm level. 2025-11-15 2025-11-28T05:44:12Z 2025-11-28T05:44:12Z Journal Article https://hdl.handle.net/10568/178364 en Limited Access Springer Hussain, S.; Arshad, M.; Cheema, M.J. M.; Qamar, M. U.; Wajid, S. A.; Daccache, A. 2025. Advancing Soil Moisture Prediction Using Satellite and UAV-based Imagery with Machine Learning Models. Earth Systems and Environment. (Online first). doi: https://doi.org/10.1007/s41748-025-00761-5 |
| spellingShingle | soil water content satellite imagery unmanned aerial vehicles machine learning irrigation efficiency Hussain, S. Arshad, M. Cheema, Muhammad Jehanzeb Masud Qamar, M. U. Wajid, S. A. Daccache, A. Advancing Soil Moisture Prediction Using Satellite and UAV-based Imagery with Machine Learning Models |
| title | Advancing Soil Moisture Prediction Using Satellite and UAV-based Imagery with Machine Learning Models |
| title_full | Advancing Soil Moisture Prediction Using Satellite and UAV-based Imagery with Machine Learning Models |
| title_fullStr | Advancing Soil Moisture Prediction Using Satellite and UAV-based Imagery with Machine Learning Models |
| title_full_unstemmed | Advancing Soil Moisture Prediction Using Satellite and UAV-based Imagery with Machine Learning Models |
| title_short | Advancing Soil Moisture Prediction Using Satellite and UAV-based Imagery with Machine Learning Models |
| title_sort | advancing soil moisture prediction using satellite and uav based imagery with machine learning models |
| topic | soil water content satellite imagery unmanned aerial vehicles machine learning irrigation efficiency |
| url | https://hdl.handle.net/10568/178364 |
| work_keys_str_mv | AT hussains advancingsoilmoisturepredictionusingsatelliteanduavbasedimagerywithmachinelearningmodels AT arshadm advancingsoilmoisturepredictionusingsatelliteanduavbasedimagerywithmachinelearningmodels AT cheemamuhammadjehanzebmasud advancingsoilmoisturepredictionusingsatelliteanduavbasedimagerywithmachinelearningmodels AT qamarmu advancingsoilmoisturepredictionusingsatelliteanduavbasedimagerywithmachinelearningmodels AT wajidsa advancingsoilmoisturepredictionusingsatelliteanduavbasedimagerywithmachinelearningmodels AT daccachea advancingsoilmoisturepredictionusingsatelliteanduavbasedimagerywithmachinelearningmodels |