Crop Yield Prediction as an Early Warning Tool for Drought and Food Security Disasters
Timely and reliable crop yield prediction is a critical component of early warning systems for drought and food security disasters, particularly in climate-vulnerable regions such as Afghanistan. The country is highly exposed to extreme weather events, and winter wheat plays a central role in national food security. Accurately capturing the influence of irrigation on crop productivity is therefore essential for anticipatory food security planning.
In this study, we evaluate Earth Observation (EO) based yield prediction models for winter wheat in Afghanistan by distinguishing between irrigation-sensitive and irrigation-insensitive predictors. EO datasets were grouped accordingly, and an irrigated area mask was applied to isolate signals from irrigated croplands. To enhance model robustness and reduce noise from inter annual variability, a first-difference approach was applied to both yield and predictor time series. The irrigation-sensitive model incorporates vegetation indices and biophysical parameters (NDVI, LAI, and FAPAR) along with surface and root-zone soil moisture from GLEAM (Global Land Evaporation Amsterdam Model), while the irrigation-insensitive model relies on precipitation, reference evapotranspiration, aridity index, and soil moisture from FLDAS ( FEWS NET Land Data Assimilation System).
Winter wheat yields were predicted from January through May, revealing that forecasts generated in February and March, approximately four months before harvest, were the most accurate. The combined vegetation-and-precipitation model achieved the lowest prediction error (RMSE ≈ 0.30 mt/ha), outperforming models that relied solely on irrigation-sensitive or irrigation-insensitive predictors. Results demonstrate the potential of EO-driven yield forecasting as an effective early warning tool for drought and food security monitoring. By providing reliable seasonal yield estimates well ahead of harvest, such models can support proactive decision-making and targeted interventions in regions where irrigation plays a critical role in buffering climatic shocks.
Dr. Barnali Das
Assistant Professor at Kansas State University
Barnali Das is Assistant Professor of Geography and Director of Natural Resources and Environmental Sciences Secondary Major at Kansas State University. She received her bachelor’s degree in Geography from University of Calcutta in 2009. She earned her master’s degree in 2011 and doctorate in 2022, both in geography, from University of Pune. She has also Diploma in Remote Sensing and GIS from Indian Institute of Remote Sensing (IIRS), Indian Space Research Organisation (ISRO). Her PhD work won the Best Scientific Story Award in 2022, a national award across all scientific and technical fields organized by the Augmenting Writing Skills for Articulating Research (AWSAR) program of the Department of Science and Technology (DST). Dr. Das worked as Postdoctoral research scholar at Smithsonian Tropical Research Institute (STRI) in Panama and at University of California Santa Barbara (UCSB) before joining K-State in 2025. Her work focused on how climate change and shoreline dynamics affect mangroves in Central America as a postdoctoral researcher at the STRI. At UCSB her work utilizes earth observation products, remote sensing, satellite-derived biophysical parameters, and machine learning methods to understand how climate extremes affect crops in food insecure countries. Earlier in her career, she worked with the India Meteorological Department on the Forecasting Agricultural Output using Space, Agro-meteorology, and Land-based Observations (FASAL) project, where she applied geospatial and agro-meteorological approaches to operational crop yield forecasting.