An Integrated Yield-Based Methodology for Improving Soil Nutrient Management at a Regional Scale

Abstract

The relationships between crop yield and its selected related impact factors has often been explored using ordinary least squares regression (OLSR). However, this model is non-spatial and non-robust. This study first used stepwise regression to identify the main factors affecting winter wheat yield from twelve potential related factors in Yucheng County, China. Next, robust geographically weighted regression (RGWR) was used to explore the spatially non-stationary relationships between wheat yield and its main impact factors. Then, its modeling effect was compared with that of GWR and OLSR. Last, robust geostatistical analysis was conducted for spatial soil management measures in low-yield areas. Results showed that: (i) three main impact factors on wheat yield were identified by stepwise regression, namely soil organic matter, soil total phosphorus, and pH; (ii) the spatially non-stationary effects of the main impact factors on wheat yield were revealed by RGWR but were ignored by OLSR; (iii) RGWR obtained the best modeling effect (RI = 52.31%); (iv) robust geostatistics obtains a better spatial prediction effect and the low-yield areas are mainly located in the northeast and the middle east of the study area. Therefore, the integrated yield-based methodology effectively improves soil nutrient management at a regional scale.

Soil Sustainability in the Anthropocene

“…The research study of Qu et al [11] used robust geographically weighted regression (RGWR) to explore the spatially non-stationary relationships between wheat yield and its main influencing factors. The authors obtained the best modeling effect (RI = 52.31%).…”