Apple Yield Forecast And Trend Analysis Under The Influence Of Meteorological Conditions At A County Scale

The application of technologies such as big data and artificial intelligence in agriculture is conducive to accelerate the transformation of agricultural production mode,assisting production decision-making,and promoting quality and efficiency in agricultural production.China is the world's largest producer and consumer of apple,and apple yield forecasting is an important part of the apple big data platform and an essential link in apple industry system,so how to make accurate yield forecasts based on meteorological data is a vital issue in the development of the apple industry.Since apple yields often fluctuate greatly due to meteorological changes,accurate forecasting of recent yields and long-term meteorological yield trend analysis based on meteorological data are of great significance to assist apple industry decisions and optimize planting structure.Based on the meteorological data of the main apple producing areas and the apple yield as the research object,this paper analyzes the relationship between meteorological factors and apple yield,and constructs a set of multi angle and multi time dimension analysis and prediction model.The specific contents are as follows:Firstly,an apple yield prediction model based on meteorological mean data is built.Apple yield is closely related to meteorological factors,and building a prediction model with meteorological factors as variables is an important means to achieve yield prediction.However,the existing single model is limited by prediction theory and often responds well in specific regions.In this paper,an ensemble apple yield forecasting model is constructed by analyzing historical meteorological data and apple yield data.The apple yield is decomposed into trend yield and meteorological yield by trend analysis method.Then the key meteorological factors that have a great impact on yield are screened based on distance correlation coefficients for monthly average meteorological data.In the end,support vector machine regression,multiple linear regression and decision tree regression are used as the base models to form the integrated learning model.In the verification of four main appleproducing counties in Shandong,the results show that the accuracy of the ensemble learning model is better than that of the single model,and its average relative error is between 3% and 4.5%,and the root mean square error is between 1.5 and 2.6,the model shows good prediction effect and strong stability in different regions.Secondly,an analysis model of apple yield change based on extreme meteorological data is built.Extreme meteorological factors,such as freezing during flowering,which could cause large shadow fluctuations in apple yield.While monthly averaging of meteorological data often obscures short-term extreme meteorological factors,leading to a decrease in yield prediction accuracy.In this paper,we take the optimal meteorological conditions for different phenological periods of apples as the starting point and disaster-prone meteorological conditions as the reference,extract 11 meteorological factors,and separate meteorological yields for disaster-prone and disaster-less areas using cubic spline interpolation and HP filtering methods respectively,to construct an analytical model based on AdaBoost ensemble classification of the impact of extreme meteorological factors on apple yield.The validation is carried out in two main production counties located in Shandong Province and Shaanxi Province,and the results show that the prediction accuracy of the Ada Boost integrated classification model is above 0.85,and the results could be used as a supplement to the short-term apple yield prediction model.Finally,an analysis model of apple meteorological yield change trend based on future meteorological data is built.As human activities cause global warming,the suitable growing area for apples shifted northward,and apple yield in Shandong also showed certain change trends.This paper is based on the 2007-2099 data simulated by the RegCM4.6 climate model run with four input parameter models and two emission models RCP2.6 and RCP8.5 as initial parameters.Taking Shandong and 11 major producing counties in the region as examples,an analysis model based on long-term and short-term memory network is constructed to analyze the future trend of apple production in Shandong.The results show that: with the passage of time,the temperature shows an obvious rising trend under different conditions,and the precipitation fluctuates within a certain range;the meteorological yield in Shandong shows a decreasing trend;the meteorological yield of the main producing counties in the eastern Shandong mainly shows a decreasing trend,the meteorological yield of the main producing counties in the central and southwestern Shandong shows a trend of flat or slightly increasing.