Mapping topsoil electrical conductivity by a mixed geographically weighted regression kriging: A case study in the Heihe River Basin, northwest China

摘要: Abstract Spatial prediction is an important approach to obtain location-specific values of soil electrical conductivity (EC), which a proxy salinity and for agricultural management in arid semi-arid areas. Linear regression models assume that the relation between EC environmental covariates constant over area be predicted. This problematic at regional scale, some parameters may indeed globally constant, whereas others vary locally. Moreover, model residuals often exhibit spatial dependence, invalidates ordinary least squares linear regression. study examined combination mixed geographically weighted with simple kriging (MGWGK) mapping Heihe River Basin, inland river basin northwest China. We compared performance MGWRK those multiple (MLR), (RK), (GWR), (GWRK) (MGWR). Environmental were developed from information on topography, climate, vegetation, geographic position. A ten-fold cross-validation was applied evaluate predictive accuracy various methods. Soil ranged 0.031 182.100 dS m−1, exhibiting contrasting distribution upper, middle, lower reaches. The method outperformed other effects different revealed by fixed varying MGWRK. nugget-to-sill ratios fitted variogram all fell 25 32%, moderate autocorrelation residuals. Predictive improved MGWRK, as dependence included prediction. When selecting optimal model, should tested see if they are (as MLR) or spatially GWR) semi-varying MGWR), MGWRK).