Forecasting Soybean Yield in Agricultural Regions of the Russian Far East Using Remote Sensing Data

Abstract

Soybeans are the main agricultural crop in the southern part of the Russian Far East. Predicting soybean yield at the regional level is an important task that contributes to the planning of acreage and assessment of risks in fulfilling contractual obligations. Methods based on satellite data have recently begun to be used to solve this problem. To assess the soybean yield at the municipal level, a regression model was built, where the maximum normalized difference vegetation index (NDVI) value for arable land in the district and the number of days with an average daily temperature of more than 10 °C (D) were used as independent variables. The regression model was built using data from 2010 to 2018 for six municipal districts belonging to four regions of the Russian Federation: the Amur Region, Primorskiy and Khabarovsk Territories, and the Jewish Autonomous Region. It was found that the maximum NDVI of arable land in these areas occurred between the 30th and 33rd calendar weeks (late July–mid-August); the D value varied slightly and ranged from 83 to 90 days. The estimation of the method accuracy showed that the mean absolute percentage error (MAPE) of the model was in the range 4.8–9.7%; the root mean square error (RMSE) was 0.05–0.15 t/ha. Using the created model, the soybean yield for 2019 was estimated. The forecast error for the three areas not affected by the 2019 flood did not exceed 10.5%. For areas with flood events, error ranged from 11.7 to 22.0%.