Research On Spatial Heterogeneity Of Wheat Protein Quality And Its Influencing Factors

With the settlement of subsistence problems,wheat quality has become a research hotspot in recent years.Grain protein content(GPC)is an important indicator of wheat quality.In previous studies,usually the empirical models of wheat GPC were built using remote sensing or agronomic data ignoring the effect of spatial location and spatial nonstationarity of sampling sites on wheat GPC.Given this,this paper mainly aims to study the effects of meteorological factors and agronomic parameters of nitrogen on GPC,and to build a GPC geographical weighted regression model.This paper also aims to study the spatial heterogeneity of wheat quality,and to put forward a reliable basis for high precision inversion of wheat GPC by adding geometric coordinate information to detect the spatial nonstationarity among sampling points.This paper based on the wheat test in the years 2008-2015 at the national precision agriculture demonstration and the year 2009 in Beijing suburbs.The ground spectra,agronomic parameters,remote sensing images and meteorological data of the study areas were used as data sources,and a geographical weighted regression algorithm was introduced to study the spatial heterogeneity and influencing factors of GPC.The main contents of this study were:Selecting optimal nitrogen agronomic parameters of wheat GPC by building a "Vegetation index(?)-Agronomic parameter(AP)-GPC" model;Obtaining the meteorological factors,which selected by routine analysis method,using spatial interpolation method;Building the geometric weighted regression model of wheat GPC,and the spatial nonstationarity of each independent variable was discussed;Comparing the inversion accuracy of the global regression model and geographic weighted regression model.The main research results were as follows:(1)A comprehensive "VI-AP-GPC" model was built to reflect the optimal agronomic nitrogen parameters of GPC.Analysis of the correlation between the spectral information and agronomic parameters,and the correlation between agronomic parameters and GPC were carried out.Five spectral indices which had highest correlation were chosen to invert the agronomic parameters of nitrogen using the SVR method,and the comprehensive model was built coupled with a wheat GPC regression model using agricultural parameters and an optimal one was selected by screening the best agronomic parameters of nitrogen with highest inversion accuracy.The results showed that the PNC not only had the highest association with wheat GPC,but also was the optimal variable to inverse GPC.(2)The spectral response function was used to transform the hyperspectral spectra into multispectral spectra to build the inversion models of LNC and PNC.The results showed that the linear relationship between the real multispectral data and the transformed multispectral data was embedded in the model,and can solve the overestimate or underestimate problem of model building to get a higher validation accuracy.(3)Effects of meteorological factors on GPC.Through the spatial interpolation analysis of ARCGIS,the meteorological data of sampling points were obtained by the spatial interpolation module in ArcGIS.The correlation analysis method was used to analyze the effects of meteorological factors on GPC.The results showed that there was a positive correlation between the temperature difference and GPC,while the precipitation and light were negatively correlated with GPC at the regional level from wheat flowering stage to mature stage.(4)Combined with geographically weighted regression,the univariate and multivariate and regression models of GPC were built,respectively.At the same time,the inversion precision of geographically weighted regression and general regression were analyzed.The result showed that there existed spatial variabilities caused by change of location from a single variable to four variables which used for model construction.The accuracy of geographically weighted regression was higher than general regression model.Before the emerging of over-fitting phenomenon,the precision of geographically weighted regression was constantly rising with the increase of the number of independent variables,while the evaluation indexes of AICc and RSS decreased sharply and R2 increased significantly with four independent variables.Therefore,geographically weighted regression coupled with geographic coordinate information could detect spatial nonstationarity effectively and improve the inversion accuracy effectively.