Grid-by-grid Drought Risk Assessment Of Winter Wheat Based On Remote Sensing And Model Assimilation

Drought is one of the most serious natural disasters and has profound impacts on global agricultural production and ecological environment.Accurately drought conditions monitoring combined remote sensing and crop models at large regional scales can reveal crop water stress response mechanisms,which can also estimate crop yields under drought conditions.It is important for ensuring food security,providing decision support,and promoting agricultural intelligence and informatization.Crop model is a powerful tool in monitoring crop growth process under water stress at a single point.However,facing the non-homogeneous surface conditions and the difficulty in obtaining input parameters required by crop model makes it limited in monitoring drought on a large regional scale.In contrast,the introduction of large-scale,multi-temporal remote sensing inversion information to re-estimate sensitive parameters and initial conditions required by crop models at the regional scale gives the crop models an advantage in regional simulations.Based on the current research on remote sensing drought monitoring,crop model development,and data assimilation techniques at home and abroad,this study seeks new methods for regional crop model construction and accumulates valuable working experience for developing new methods.This study will carry out in-depth research on two scientific problems,which are drought process simulation of winter wheat and its key factors,drought monitoring model construction by coupling remote sensing and crop growth model,and propose effective solutions to drought.Through field observation experiments combined with crop model point simulation,we will clarify the mechanism of crop water stress,couple multi-source remote sensing data with crop growth model,use remote sensing quantitative inversion,data fusion and assimilation,computer simulation and other technical means to build a regional model for comprehensive agricultural drought monitoring.The region model aims to simulate regional drought development process,identify key factors affecting drought development process,explore the role of each influencing factor,and assess the crop tolerance mechanism.The model is designed to simulate the development of drought,identify key factors affecting the development of drought,explore the mechanism of each influencing factor,assess the drought tolerance time of crops,estimate the drought loss yield,and achieve effective monitoring of agricultural drought.In this paper,winter wheat in Hebei Province was selected as the research object,and a winter wheat irrigation water gradient control experiment was conducted in Langfang City,Hebei Province to collect and observe key crop data and environmental data to identify key factors affecting drought.Sensitivity analysis,parameter localization and point simulation were conducted on the WOFOST model using the winter wheat data observed in the field to investigate the mechanism of water stress action in winter wheat.Based on the experience of point simulation,a grid-by-grid regional database construction method is proposed,and a grid-by-grid regional crop growth model is constructed based on the database.The machine learning parameter automatic optimization algorithm and remote sensing model assimilation algorithm are coupled into the grid-by-grid regional crop growth model,aiming to improve the regional model simulation accuracy and reduce model uncertainty,and to simulate the winter wheat growth process in Hebei Province from 2011 to 2020.Based on the simulation results of the grid-by-grid regional crop growth model,we simulated drought scenarios for different fertility periods,graded the drought severity of winter wheat by the distance level yield reduction rate,and proposed a progressive drought risk warning method.The main results were as follows:(1)By analyzing the water stress field trial data of winter wheat in Langfang,Hebei Province,winter wheat suffered from water stress and environmental temperature and moisture conditions were closely related,and the correlation between environmental variables was significant,such as significant negative correlation between air temperature and air pressure,significant positive correlation between water vapor pressure and temperature,and significant positive correlation between soil moisture and air temperature.The differences in soil moisture,canopy temperature,spectral morphology and biological yield of winter wheat under drought stress were significant compared with those of winter wheat under non-stress conditions.(2)The results of field point simulations showed that the WOFOST model better simulated the growth process of winter wheat under water stress conditions and correctly described the growth and development of winter wheat and LAI changes.The aboveground biomass RMSE of winter wheat ranged from 128.55 to 264.17 kg/ha,and NRMSE ranged from 0.25 to 0.41%,which was well simulated.Soil moisture RMSE was 0.017～0.20 m~3/m~3 and NRMSE was 21.30～22.07%,which were well simulated.(3)Based on the raster spatial database,a grid-by-grid regional crop model construction method based on machine learning and remote sensing assimilation is proposed.The simulation results of various models were compared,and the simulation accuracy of the grid-by-grid regional crop growth model was the best with R~2 of 0.96 and RMSE of 225.73 kg/ha under the simulation strategy of"SCE-UA+4Dvar";(4)The drought risk assessment of winter wheat was carried out with the grid-by-grid regional crop growth model.After winter wheat entered the flowering stage,the response to drought began to become sensitive,and the response to drought was most sensitive during the filling stage.The winter wheat growing area in the northern part of Langfang City,Hebei Province,had the most severe yield reduction in winter wheat in 2013 and 2018,which were drought years.The progressive simulation empowers the grid-by-grid crop model to predict the future winter wheat growth and the occurrence of drought events,to give early warning of impending agricultural drought,and the early warning results have good uniformity with the real drought situation;(5)The drought risk assessment results indicate that the period starting from mid-late April to May is the high incidence of winter wheat drought in Hebei Province,which should focus on field moisture conditions in April,pay attention to climatic precipitation conditions,and scientifically replenish water for irrigation to prevent drought events and ensure food security.