Determination of the most important meteorological parameters affecting the yield and biomass of barley and winter wheat using the random forest algorithm

Abstract

Considering the very important role of climatic parameters on the yield and biomass of plants (especially in rain-fed agriculture), in this research, using data series of 9 stations during 1968–2017 the influence of 6 climatic parameters including the average of annual maximum and minimum temperature (Max-T and Min-T), the average of annual sunshine (Su-Sh), the average of annual relative humidity (H), the average of annual wind speed (Wi) and the average annual precipitation (P) on the yield and biomass of winter wheat (YWW and BWW) and barley (YB and BB) as two important and strategic species to provide the human food and the livestock feed in Iran, was investigated and prioritized. To assess the effectiveness rate of the climatic parameters on the response variables (YWW, BWW, YB and BB), the random forest algorithm (RF) was used. The results indicated that the RF algorithm had a good ability to predict the response variables because (1) the linear regression between simulated and predicted YWW, BWW, YB and BB using the AquaCrop model and the RF algorithm (respectively) had no difference with perfect reliable line (Y = X) in 0.05 or 0.01 significant levels and (2) R2 between simulated and predicted YWW, BWW, YB and BB was significant at 0.01 levels. According to the results, the P, Wi and Min-T parameters were the most influential climatic parameters on the YWW (respectively), the Min-T, Wi and P parameters were the most influential parameters on the YB and the Wi and Min-T parameters were the most influential parameters on the BWW and BB. On the other hand, in all stations, the Su-Sh and Hu were the least influential parameters on the YWW, BWW, YB and BB.