Study On The Refinement Of Spatial Distribution Of Typical Agricultural Climate Factors

In order to cost-effectively achieve the development of intelligent agriculture and precision agriculture and the formulation of new agricultural climate zones in China,it is the necessities to obtain the latest,high-precision and fine spatial distribution data of agroclimatic factors.However,the results of previous researches are not enough accurate and fine,and have larger errors due to improperly choosing or using interpolation methods.At the same time,it was lack of studies on many typical agroclimatic factors.In this study,we focused on methods how to obtain the high spatial resolution and high precision of the distribution data for the typical agroclimatic factors(precipitation,accumulated temperature(>10?)and frost-free period days),using the open data of precipitation and accumulated temperature(>10?)from 1971 to 2010 and of frost-free period data from 1971 to 2000 collected by the National Meteorological Information Center.The optimal data refinement scheme was experimented on the Loess Plateau.The preprocessing and refinement methods were obtained under the different characteristics of the target variables.Through the interpolation zoning,the refinement scheme was applied to the national land area(Do not include Taiwan Island and its affiliated islands and South China Sea Islands which are both Chinese terriyory).The performance and precision of the various interpolation methods were test in different zones,and then making some adjustment on the zoning and interpolating.Finally,a set of data refinement scheme and the national spatial distribution data of three typical agricultural climate factors were obtained.The main conclusions are as follows(1)An effective scheme for agroclimatic factors interpolation at a regional scale:Firstly setting the peripheral buffer to the study area,secondly making the target interpolation variable to conversion form with Logit transformation,next determine the interpolation predictors with stepwise regression,and then fitting and selecting the fittable Semivariogram functions,finally interpolating with a appropriate method according to the target variable characteristics.When interpolating at the national scale,it would be best to make interpolation zoning according to the known distribution characteristics of agroclimatic factors,and then interpolating on each zone.(2)The interlpolation result was effected by the spatial autocorrelation of the target variable and the relationship between the target variable and the environmental factors.When choosing the appropriate interpolation method,you can make an estimation of the interpolation result according to the spatial autocorrelation of the target factor and relationship with the geographical factors.It should be as more as possible used the geographical factors which has a causal relationship with the target interpolation.(3)In this study,Multiple Linear Regression(MLR),Geographically Weighted Regression(GWR),Ordinary Kriging(OK),Multiple Linear Regression Kriging(MLRK)and Geographically Weighted Regression Kriging(GWRK)methods were used to interpolate the three typical agroclimatic factors.Result showed that MLRK and GWRK interpolation methods are better.The spatial interpolation refinement scheme of three agroclimatic factors at the region scale provided by this study can be expanded application,and also can be provided a case and a technical support for other related research.The spatial distribution data of three agroclimatic factors with a high resolution(500 m x 500 m spatial resolution)and a high precision(R2 above 0.9)obtained in this study can be used to provide important basis and high-precision data for regional agricultural regionalization,agricultural land suitability evaluation,regional agricultural industry structure evaluation.