Integrating Remotely Sensed and Field-monitored Soil Moisture Data for High Resolution Agricultural Drought Monitoring

Agricultural drought monitoring using high resolution soil moisture information is particularly useful for management of precision agriculture and drought early warning studies. The soil moisture products obtained from different active microwave remote sensing satellites may not be appropriate for local-level drought studies due to their limited spatial resolution. This study aims to accurately estimate surface soil moisture (SSM) by utilizing high-resolution multispectral imagery available from the Landsat 8 OLI (Optical Land Imager) mission for monitoring agricultural droughts using soil water deficit index (SWDI). The study demonstrates that using the Landsat-derived SWDI at a spatial resolution of 30 m, and at bimonthly scale, can provide drought information for use in precision irrigation especially at watershed-scale. The red, green, near infrared (NIR), and short-wave infrared (SWIR) bands of Landsat 8 after atmospheric and geometric correction, are utilized for estimating SSM in this study, by considering popular vegetation indices such as normalized difference vegetation index (NDVI), normalized difference moisture index (NDMI), and normalized difference water index (NDWI), as inputs. Field-monitored soil moisture data available for an agricultural watershed in eastern India during 2016-2017 are utilized for SSM model development. The SSM estimation model is developed using conventional linear regression and artificial neural networks (ANN) models. The conventional linear regression algorithm gave correlation coefficient (R) of 0.60 and mean square error (MSE) of 0.012 cm3/cm3. Whereas, the machine learning-based ANN model has performed SSM estimation with R and MSE of 0.67 and 0.011 cm3/cm3 respectively. Further, the study utilized SSM based on the ANN technique for estimation of SWDI at 30 m resolution for long-term drought monitoring over the study watershed. Based on the computed SWDI, seasonal variations in agricultural drought patterns are also evaluated for the study area.Keywords: Agricultural Drought Monitoring, Surface Soil Moisture, Remote Sensing, Landsat-8, Vegetation Indices, Machine Learning