Optimization of Soil Sampling Design Based on Road Networks – A Simulated Annealing/Neural Network Algorithm

In this study, the spatial distribution pattern of the roads, historical samples, digital elevation data, and other available resources were incorporated into the design of a soil-sampling scheme to predict the soil organic matter (SOM) of the northern region of Zhongxiang City, Hubei Province, and simulated annealing (SA) was applied to optimize the sampling design. The sampling points determined after optimization were used to establish a multivariate linear regression model to adequately reproduce the intrinsic link between topographic factors and the SOM at 13 different sampling scales in areas nearby the existing roadways in the study area. The topographic factors included slope, plane curvature, profile curvature, topographic wetness index (TWI), stream power index (SPI), and sediment transport index (STI). A multilayer perceptron (MLP) model was also constructed. Comparison of the accuracy of the multivariate linear regression and MLP models demonstrated the feasibility of an optimized soil sampling design based on the road network. With the optimized sampling design, accurate soil-landscape information can be obtained, and its precision is greater than that of the original sampling scheme before optimization. The optimized sampling design obtained reduces sampling costs, increases sampling efficiency, and provides an effective method for obtaining the spatial distribution pattern of organic matter in soils.