Application of BP - ANN model in evaluation of soil quality in the arid area, northwest China

Abstract

Soil quality evaluation is an effective way to improve ecological environment quality of soil and to perfect management system as well as keep its productivity of the soil sustainable. However, in arid areas, the comprehensive study of soil quality has been paid little attention. This paper aims to classify and evaluate comprehensive soil quality which combined heavy metal contamination with soil nutrients in the arid area, and to explore the natural environmental controlling factors of soil quality under regional differences. Taking Hexi area, a typical arid area in northwest China, as the study area, the comprehensive evaluation model of soil quality based on back propagation artificial neural network (BP - ANN) was proposed. The results showed that the accuracies of the validated models were both more than 97 %. The soil quality in Hexi area was divided into six classes: IA, IB, IC, IIA, IIB, IIC (The number (I and II) represents the degree of heavy metal pollution, the larger the number, the higher the pollution degree; the letter (A, B, and C) represents the nutrient level, the latter the letter, the more barren the nutrient). Most of the samples were classified as IB and IC, indicating that the soil quality was good with mild heavy metal contamination and poorer nutrients. Group analysis was used to explore the impact of natural environmental factors on soil quality under different environmental characteristics through linear regression modeling. The results showed that elevation, annual mean cumulative precipitation, slope and topographic wetness index were the environmental controlling factors of the soil quality for all samples. Besides, the other important factors affecting soil quality in group 1 were: annual mean temperature and soil type. For group 2 it was LUCC (Land Use Cover Change). This study is of great significance to classify and define the soil quality, to maintain and improve agricultural production and soil ecological health in arid areas.