Grey Modeling Approach In Regional Agricultural Drought Losses Assessment

As a natural disaster,drought is one of the most serious meteorological disasters that affects human society because of its high frequency,wide coverage and long duration.In recent years,its frequency of occurrence has been increasing and the scope of its impact has been expanding,which has severely restricted the stable development of social productivity and led to the increasingly serious challenge of national food security and water supply safety.Actively changing traditional concepts and emergency management methods to combat drought scientifically will help ensure high,stable and increased crop yields.The formation of regional agricultural droughts is influenced by both the natural ecological environment and the human social environment,in which the relationships of numerous influencing factors are complex.In order to accurately identify the influencing factors of drought loss and reasonably predict the disaster situation,this study takes the regional agricultural drought loss as the research object and grey system theory and method as the main tools,then constructs the grey time-lag correlation analysis model,time-lag multivariate grey prediction model and grey clustering assessment model respectively,finally applies them to the identification of the influencing factors of regional agricultural drought loss,disaster loss prediction and assessment to help the government and decision-making departments.The study aims to provide theoretical support for the government and decision-making authorities to implement disaster prevention and mitigation measures.The specific contents and results of the study are as follows:(1)A time lag correlation model for dynamic analysis is constructed to address the problem of time lag relationship between regional agricultural drought loss influencing factors and the main system behavior series,which leads to unequal data information among the series and distorted decision-making results.First,the main system behavior sequence is clarified and its influencing factors are analyzed and relevant data are collected.The grey correlation matrix is constructed by selecting the appropriate subsequence length.Secondly,the elements of the matrix are stored using special vectors,and the optimal time lag value that can objectively reflect the time lag characteristics is solved with the maximum mean value as the optimization objective.Finally,the model is applied to the time lag analysis of regional agricultural drought grain yield losses and the influencing factors,which determines the relationship between their prior,contemporaneous and lagged values.A comparative analysis of the time lags at different sub-series lengths is also conducted to further verify the rationality of the parameter selection and the validity of the model.The study can provide data support for the subsequent multivariate prediction of regional agricultural drought losses.(2)A time-lagged cumulative multivariate grey ATDGPM(1,N)prediction model based on dynamic analysis is constructed to address the time-lagged cumulative effect of each driver on the behavior of the main system in regional agricultural drought loss prediction and the nonlinear relationships existing between the sequences of drivers.Firstly,based on the results of the time-lag correlation analysis,the power index is optimally solved by using a particle swarm optimization algorithm.Secondly,it is shown that the DGM(1,N),DGPM(1,N)and ATDGM(1,N)models are all special forms of the model for different values of the parameters.Finally,the numerical experimental results show that the ATDGPM(1,N)model can better describe the time-lagged non-linear relationship between the sequence of system behaviors and the sequence of drivers,thus effectively improve the model accuracy;when applying the model to the simulation and prediction of grain yield in Henan Province,it can be obtained that the simulation and prediction accuracy of the ATDGPM(1,N)model is much higher than that of the DGPM(1,N)model and the GM(1,N)model,which further verifies the validity and feasibility of the model.(3)Aiming at the characteristics of greyness and uncertainty of regional agricultural drought loss assessment index information,a grey clustering-based regional agricultural drought loss assessment model is proposed.Firstly,the model adopts a grey fixed-weight clustering method to determine the turning point of the clustering likelihood function in terms of the mean and standard deviation of the indicator information.Secondly,the information cluster grey correlation model is used to eliminate the correlation existing between the indicator information and solve for the indicator weights.Finally,the disaster case of agricultural drought in Henan Province from 2006 to 2019 is used as an assessment example.Four assessment indicators,namely the rate of drought-affected crop area,the rate of crop loss due to drought,the rate of direct economic loss due to agricultural drought disaster and the proportion of rural population with drinking water difficulties due to drought,are selected to cluster all samples.The disaster assessment results are obtained,which confirms the feasibility of the model.