Impact Of Climate Change And Extremes On The Growth And Yield Of Winter Wheat In Xinjiang

Climate is the natural environment on which mankind depends on,and it is also an important basic resource for sustainable economic and social development.However,the ongoing impacts of climate change will increase the risk winter wheat production.Sustainably improving winter wheat production is urgently needed to meet population growth and people's living demand with the rapid economic development.How to use high-tech methods to mass-produce and reflect the impact of climate change and extreme events on winter wheat yield on a large scale,and understand the relationship between climate change and extreme events and winter wheat yield is vital for assessing the sustainability of the agricultural production.Based on collecting meteorology,soil,crop,and geographic data from 20 Xinjiang agro-meteorological sites over 1981?2017,the study investigated the spatiotemporal variation characteristics of climatic variables and extreme events during the growth period duration of winter wheat(from September to July of the following year)and the correlation betweeen winter wheat growth-yields and climatic variables or extreme events.Then based on linear regression and nonlinear regression methods,it established regression models between climatic variables or extreme events and wheat growth-yields,and we found the best quantitative relationships between yield-related indices and the key climatic variables or extreme events.Further study used RZWQM2 model to obtain parameters of the three winter wheat varieties(early variety,medium variety,and late variety)and simulated the growth period,the maximum leaf area index,the aboveground biomass in mature period,yield and water use efficiency of winter wheat for 15 meteorological sites at the historical period(1981?2020)and the future period(2021?2060 and 2061?2100).Further study analyzed the influence of climatic variables on winter wheat yield in the future,and finally made a comparison between the three machine learning methods(decision tree,random forest,and neural network)and linear regression method,and the climate extreme index with the highest contribution was selected.The main conclusions are as follows:(1)Winter wheat yield was closely related to average temperature,precipitation and relative humidity.In 1981?2017,the climate at most sites changed to wetter,warmer,less sunshine,lower wind speed and lower humidity.The growth period of winter wheat showed a shortening trend,while the number of grains per spike and the aboveground biomass and yield showed an obvious increasing trend.The pearson correlation coefficient(r)between the growth-yields and the climatic variables ranges from–0.66 to 0.54,and they had low correlations at most sites.The maximum R²_adj of the regression model for plant height,growth period duration,1000-kernel weight,kernel number per ear,above-ground biomass and yield are 0.41,0.29,0.27,0.50,0.41 and 0.40,respectively,indicating that climatic variables explained 27%?50%variability of winter wheat yield.For the R²_adj value of the regression model,most sites showed multivariate nonlinear regression>multivariate linear regression>single linear regression.(2)The order of the influence of extreme temperature index on winter wheat yield was:TX90p>FD>SU=TXx=TNn.TNn?TXx?SU and TX90p showed an upward trend at8,18,20,and 4 sites,respectively,1,5,11,and 0 showed a significant upward trend over1981?2017.FD all showed a downward trend,and 3 sites showed a significant downward trend.These 5 extreme temperature indexes indicated that the climate in Xinjiang had been warming in the past 30 years.The r between the wheat growth-yields and the extreme temperature index ranged from–0.59 to 0.67.The maximum R²_adj of the regression model for plant height,growth period duration,1000-kernel weight,kernel number per ear,above-ground biomass and yield are 0.58,0.46,0.29,0.44,0.30 and 0.45,indicating that extreme temperature explained 29?58%variability of winter wheat yield.(3)The order of extreme precipitation index on winter wheat yield changes was:R95p>R10>RX5day>SDII=CDD>RX1day.SDII,RX1day,RX5day and R10 of most sites showed an upward trend,while R95p and CDD showed a downward trend in most sites over 1981?2017.The r between the yield component index and the extreme precipitation index ranged from–0.59 to 0.67.The maximum Radj2 of the regression model for plant height,growth period duration,1000-kernel weight,kernel number per ear,above-ground biomass and yield are 0.58,0.36,0.58,0.27,0.38 and 0.45,indicating that extreme temperature explained 27%?58%variability of winter wheat yield.(4)Increasing temperature and decreasing radiation in the future would have a negative impact on winter wheat production in Xinjiang,while changing in CO₂concentration will have a positive impact.The RZWQM2 was calibrated and verified with historical data for three varieties of early,medium and late maturity,showing good performance.Compared with the historical period,during the winter wheat growing season in Xinjiang,except for radiation,the lowest temperature,highest temperature and precipitation were all increasing trends at each site under SSP2-4.5 and SSP5-8.5 scenarios.For the future SSP2-4.5 and SSP5-8.5 scenarios,the flowering and maturity time of winter wheat will be shortened,and the trends of the three varieties are roughly the same.Additionally,the maximum leaf area index,maximum above-ground biomass and yield of winter wheat simulated by RZWQM2 would be increased.The difference between the three varieties is small,and the SSP5-8.5 scenario increases more than the SSP2-4.5 scenario,and 2061?2100increases more than 2021?2060.The order of water use efficiency is medium>early>late.(5)The random forest could better predict the yield in the four methods,TNn,TXn and R10 were the most important extreme indexes.Compared with the historical period,the extreme temperature indexes of TNn,TNx,TXn,TXx,SU,TX90 and TN90 would increase in the two future periods and under the two scenarios of SSP2-4.5 and SSP5-8.5.FD,ID,TX10,TN10 and SU were in a decreasing trend,and DTR was mainly in decreasing trend.The extreme precipitation indexes of PRCPTOT,RX1day,RX5day,R95p and R10 were increasing,and CDD was decreasing.The trend of SSP5-8.5 is greater than SSP2-4.5 scenario.Based on the establishing model of extreme temperature index and yield,random forest would predict yield better among the four methods.Using the four methods established by the extreme precipitation index,the random forest could also better predict the yield among the four methods.Among the three varieties,the random forest method can explain the priority order of early>medium>late.For the three varieties,TNn and TXn were the most important extreme temperature indexes.and R10 and R95p were the most important extreme precipitation index on production.This study could provide a reference for dealing with the impact of extreme weather events on winter wheat yield in Xinjiang.