Predictive Mapping Of Soil Subtype Distribution Based On Soil-Landscape Relationship

With the rapid development of precise agriculture and great attention to land resources and environment, detailed degree of present soil information in our country can not satisfy the needs of some fields. Traditional national soil survey will consume much material resource, financial resources and human resources. The new digital soil survey technology takes dokychaev’s theory of soil forming factors as theoretical basis and uses 3S technology and modern math methods to establish quantitative soil landscape model and analyze similarity of soil type space distribution. Then reasoning graph is made based on the above, making it become main research direction of the field. The paper takes Ming He county in Qinghai province as an example and makes reasoning graph by analyzing soil type and corresponding environment factors based on digital digging technology. The results are the following:(1) The paper takes dokychaev’s theory of soil forming factors as theoretical basis. It chooses geology, vegetation, land use and land cover, annual mean temperature, average annual precipitation, DEM and other environmental factors such as slope, exposure, profile curvature, plan curvature and topographic wetness index to establish environmental factors database.(2) Taking 500 m as a unit to make automatic grid stationing; requiring that each pattern spot be more than 5 and each subclass more than 20 points; to pattern spots which can not meet the requirements, 11 environmental elements shall be selected to gain sample set in the research region. After dealing with the sample data, C5.0 decision tree method to calculate soil-environment relationship and use method of continuous removing and classifying data to enhance the classification precision to 98.95%. By analyzing relationship between soil and single environmental factors and concluding relationship between soil type and environmental factors group, rules of decision tree generating has been integrated. Finally, rule set of soil-environment relationship in the research region has been acquired.(3) This study uses fuzzy reasoning method based on rules to calculate soil similarity. Using rules to set subjection curve between each soil subclass and single environmental factors. Hence, rules of each soil type are established. Similarity graphs of 14 soil subclass are obtained by integrating corresponding environmental factors to each soil subclass. Soil subclass distribution graph is gained finally by hardening treatment with it. By comparing with the soil graph, the results present high similarity but soil subclass distribution in the reasoning graph is scattered. Use field sample data to examine precision of digital soil subclass graph in the research region by reasoning and the results show that a part of subclass produces error dividing because of gradual change and alternativity in space. However, the total classification precision has reached 60%. In the limited test of field samples, the precision has gained expected result. Therefore, using the method to make reasoning graph can not only present continuous change of soil in the space to overcome graph expression model of traditional soil graph based on polygon, but also can investigation time and expenses in the soil survey and graph to enhance efficiency of soil survey.