Agricultural Drought Prediction Using Drought Monitoring Data Between 2001 and 2021 Based on Deep Learning Techniques

Due to its geographic location and climate, Central Asia is vulnerable to desertification, land degradation and drought, and much of the land has already been affected. According to the United Nations Development Program report, "Strategic measures to combat desertification in the Republic of Kazakhstan until 2025", Kazakhstan loses about 200 million dollars annually because of desertification, land degradation and drought. According to the Koppen-Geiger classification, more than 80% of the cultivated area is located mainly in the cold semi-arid natural zone. Given that the lion's share is rainfed crops, i.e. agricultural crops are cultivated without artificial irrigation, drought is the main threat of crop failure. Therefore, there is a need for monitoring and forecasting agricultural drought throughout the region. Three remote sensing products were used in this study: Normalized-Difference Vegetation Index (NDVI), precipitation, and soil moisture, which reflect, respectively, the intensity of vegetation and the main soil water supply during the growing season. Agricultural drought was monitored using a modified Combined Deficit Index (mCDI) based on the monthly deficit of NDVI, precipitation, and soil moisture. The digitized wheat yield data for 2020 in the Kostanay region was used to validate the drought monitoring results. Accordingly, very high correlation (R2=0.93) was observed, which indicated the high reliability of mCDI in monitoring agricultural drought over the Republic of Kazakhstan. Furthermore, historical monitoring data was combined in the Temporal Fusion Transformer (TFT) deep learning model to perform agricultural drought predictions for one, three and six months. Agricultural drought prediction for one month demonstrated outstanding results with average Mean Absolute Scaled Error (MASE) of 0.4-0.6 and drought severity class match of around 85%. In the case of three months prediction, drought severity class match was around 45% and the average MASE increased to 1.6, while six months prediction showed insufficient outcome which required further improvement. Overall, this study provides new approaches for agricultural drought assessment and data-driven forecasting that contribute to early warning and mitigation of agricultural drought.