Spatial Distribution Characters of Potato Nutrient Use Efficiency and Quantitative Influencing Factors in China Using Random Forest Model

Abstract

Improving potato nutrient use efficiency (NUE) is instrumental in addressing food security and environmental sustainability challenges. Currently, potato NUE has a large spatial variation due to the differences in agricultural management practices (AMP), soil properties, and climate. However, the explanatory variables of potato NUE have not been comprehensively elucidated because of technical and methodological restraints. Therefore, this study aimed to characterize the spatial distribution of potato NUE (partial factor productivity abbreviated as PFP, and partial nutrient balance abbreviated as PNB) for nitrogen (N), phosphorus (P), and potassium (K) in 19 provinces of China, quantify the major explanatory variables by random forest model, and analyze the spatial variation of model uncertainty with quantile regression forest. Results indicated that the N and P use efficiency in northeast China was the highest, while K use efficiency in the northwest was the highest. These NUEs were the lowest in south China. The prediction models had a good performance, among which the PFP had better performance than PNB. The relative importance of explanatory variables was AMP > topography > climate > soil > crop > economy. Among them, the first two types of variables explained around 80% of the variance in NUE models, except the PNB of the N model (76%). NUE model uncertainty was higher in northwest China (e.g., Shaanxi, Gansu, Ningxia) and southwest China (e.g., Hubei). In this case, model prediction might have a higher performance by regions, despite fewer data in each model. This study will have far-reaching implications for policymakers toward understanding the spatial variation of potato NUE and more efforts should be undertaken in terms of AMP and topography, which is of great significance for achieving food safety and alleviating environmental pollution.

Keywords: partial factor productivity, partial nutrient balance, explanatory variables, model uncertainty, potato, China