Determining agricultural drought for spring wheat with statistical models in a semi-arid climate

Abstract

Agricultural drought frequently occurs and results in major grain yield loss in semi-arid climate region, but determining it is difficult. This study was conducted to determine agricultural drought for spring wheat (Triticum aestivum L.) in the western Loess Plateau of China. Several statistical models were established and evaluated by long-term data, including soil water in soil layer of 50 cm depth at sowing day, air temperature, precipitation, pan evaporation during spring wheat growing season, and two groups of spring wheat yield (one from field experiments during 1987-2011 and the other from statistical Bureau during 1980-2013). Even though each of water supply factors, precipitation during growing season and the soil water at sowing day, could separately explain no more than 30% variation of the yield, both of them could explain > 55% yield variation under dry condition. Average air temperature and precipitation during growing season that displayed two apparent yield categories (drought and normal) could be used to determine agricultural drought by pattern recognition when years with the soil water at sowing day of >98.4 mm were eliminated. Based on long-term meteorological data and the relationship between soil water at sowing day and yield under different growing season moisture conditions, the probability of agricultural drought occurrence in Dingxi for spring wheat was speculated, which nearly corresponds with the observational data during 1980-2013.