Model-based drought indicators improve the reliability of crop yield simulations with a statistical model in Poland

Abstract

Various inputs can be selected to establish a robust crop yield simulation based on statistical models. Typically, weather variables such as precipitation, temperature, relative humidity, etc., are used as inputs in these models. It is well known that drought is a major limiting factor for crop yield in Central Europe, as manifested in recent years. This study aimed to assess whether adding model-based drought indicators derived from a nationally calibrated and validated process-based agro-hydrological model (Soil and Water Assessment Tool - SWAT) could help increase the predictive power of crop yield prediction. The secondary objective was to assess future projections of crop yield. We considered two drought indicators: the Standardized Precipitation Index (SPI) and the Soil Moisture Index (SMI) with the following accumulation periods: 1970-2019. The ABSOLUT v1.2 (Assessing Best-predictive Sets fOr multiple Linear regressions throUgh exhaustive Testing) model was applied for the prediction of yield of major crops in Poland: winter wheat, spring barley, potatoes, sugar beet, and maize for 16 provinces of the country for the time period 1999-2019. ABSOLUT v1.2 is an adaptive algorithm that utilizes correlations between time-aggregated weather variables and crop yields for yield simulation. Future yield projections were derived based on bias-corrected EURO-CORDEX simulations driven by two Representative Concentration Pathways (RCPs), RCP4.5 and 8.5, corresponding to the radiative forcing levels of 4.5 W/m−2 and 8.5 W/m−2 in the year 2100, respectively. Our results indicate that incorporating drought indicators as predictors in statistical crop yield simulations slightly enhances the reliability of yield prediction in Poland. Projected crop yields reveal that in western parts of Poland, crop yields could experience a decrease of 8%, but in eastern parts, crop yields remain mostly unchanged.