Spatial Variability And Comprehensive Evaluation Of Soil Chemical Properties In Small Watershed Of Mountainous Areas Of Southern Ningxia

The mountainous area of South Ningxia is located on the northwestern edge of the Loess Plateau and is a typical hilly and gully area.The climatic and topographical conditions had caused serious soil erosion,resulting in uneven distribution of nutrients in space,and severe loss of nutrients,affecting the restoration of local vegetation.In recent years,vegetation reconstruction,ecological restoration and other projects had been gradually carried out in the mountainous area of Southern Ningxia,fundamentally preventing the deterioration of regional ecological environment.As the most basic properties of soil resources,soil nutrients,enzyme activities and heavy metals strongly affected the process of vegetation restoration and succession,and determined the quality of soil.Different soil properties were affected by the surrounding environment and human activities in the same area,and different degrees of influence led to their different spatial variability characteristics.In this paper,the spatial heterogeneity of soil chemical properties in the small watershed of Ningnan were studied by using mathematical statistics and geostatistics based on the measured soil data and multi-source environmental variable data of 260 sampling points.The ordinary Kriging,regression Kriging,geographically weighted regression Kriging,BP neural network and random forest method were used to establish the spatial prediction model of regional soil nutrients.In order to provide a theoretical basis for vegetation and ecological reconstruction in the comprehensive management of small watershed in smountainous area of Southern Ningxia,and ensure the smooth progress of the construction of a harmonious ecological environment,the cloud model was used to calculate the soil comprehensive index of the study area,and the comprehensive index was spatialized.The main results are as follows.(1)Except for alkali hydrolyzable nitrogen,the rest of soil nutrients were of moderate variation.The best fitting models of total phosphorus and available phosphorus were spherical model and Gaussian model respectively,and the best fitting models of other nutrients were exponential model.The spatial distribution characteristics of total phosphorus and total potassium were similar,gradually increasing from the middle to the periphery.The spatial distribution of alkaline hydrolyzed nitrogen had obvious point source characteristics.Except for high value in some points,other large areas were low.The available phosphorus was high in the west,low in the East,and low in the West Potassium is high in the middle and low around.(2)The mean value of soil urease activity was 0.36 mg/g,and the mean value of soil invertase activity was 2.56 mg/g.the coefficients of variation of urease and invertase were 88.89%and 55.56%,respectively,showing moderate variability.The best fitting model of urease was exponential model,and that of invertase was Gaussian model.The high value area of urease mainly concentrated in the north and southwest of the study area,the low value area mainly concentrated in the central,northwest and northeast.The high value area of invertase mainly concentrated in the north and the east,and the low value area was mainly distributed in the west.(3)Except for Ni,the average value of all heavy metal elements were higher than the background value of Ningxia soil.The pollution index showed that there was a slight to moderate potential ecological risk in the study area.The source analysis showed that the main sources of heavy metals in soil were natural parent material,agricultural activities,mining development and transportation.The spatial distribution of Ni,Zn and Mn was similar,the high values were concentrated in the northwest,central and southwest,the low values were mainly distributed in the northeast,Southeast and West.The high values of Pb were in the north,central and southwest,the low values were distributed in the West and northeast.Cd high values were distributed in the center,east and southwest,low values were distributed in the northeast,and Cu high values were banded.The high Cr values were mainly distributed in the West and southeast of the study area.(4)Five methods were used to build soil nutrient spatial prediction models and evaluated the accuracy.In general,the prediction accuracy of GWRK,RF and BP neural network methods were relatively high.The best prediction model of quick-acting nutrients were BP neural network,while the application effect of OK and RK methods were poor.(5)Based on the analysis of cloud model method,the percentage of soil nutrient environment in the total study area was as follows:high(0.62%),high(10%),average(31.71%),poor(27.37%),poor(30.3%).The western of the study area was the best soil nutrient environment.