基于相似推理法的土壤类型制图研究

Research on soil types mapping based on similarity inference method

关键词:  数字土壤制图  土种制图  相似推理法  第三次全国土壤普查技术规程规范

Key words:  digital soil mapping  mapping of soil species  similarity inference method  technical specifications of the Third National Soil Survey

摘要:

【目的】 获取准确的土壤—环境关系是保证土壤类型制图精度的关键。文章期望利用相似推理法这一基于成土环境条件进行土壤类型推理的技术，对有土属图但无土种图地区进行土种制图可行性研究。【方法】 该研究选取湖南省邵东市牛马司镇为研究区，参考第三次全国土壤普查的土壤类型制图方法，尝试在有二普土属图、但无土种图的地区，进行耕地、园地、林地和草地土属更新制图和耕地土种推理制图的研究。以二普土属图为底图，运用相似推理法，结合土地利用类型现状图、高空间分辨率数字正射影像和地形因子等数据更新土属图。之后结合240个耕地质量评价土壤调查样点的pH、有机质、土层厚度3个理化性质数据，完成耕地土种推理制图，并将85个土壤调查样点的土壤类型信息更新至园林草地土属图与耕地土种图中样点所在的图斑。【结果】 利用62个土壤调查样点评价制图结果的精度：土属图精度为68%，kappa系数为0.68；耕地土种图精度为61%，kappa系数为0.59。精度评价结果不一致的主要原因分为母质不同、落于母质交界处、地类变更、pH不一致和土壤质地不一致五大类。【结论】 基于相似推理法的土壤类型制图结果具有良好的精度水平，证明此方法在有土属图但无土种图地区进行耕地土种制图和非耕地更新土属图都具有可行性。

Abstract:

[Purpose] Obtaining accurate soil-environment relationship is the key to ensuring the accuracy of soil types mapping. This article expects to use the similarity reasoning method，a technique for soil type inference based on soil-forming environmental conditions，to conduct a feasibility study on soil types mapping in areas with soil genus maps but no soil species maps.[Method] In this study，Niumasi Town，Shaodong City，Hunan Province，was selected as the study area. With reference to the soil types mapping method of the Third National Soil Survey，soil genus updating mapping and soil types inference mapping were conducted for arable land，garden land，forest land and grassland in areas where the soil genera maps of the Second National Soil Survey were available but the soil species maps were not. Taking the Second National Soil Survey map as the base map，the soil genus map was updated by applying the similarity inference method and combining data such as the current land use type map，the high spatial resolution digital orthophoto map（DOM）and the topographic factor. Combined with the physical and chemical property data of pH，organic matter and soil thickness of 240 soil survey sample points for the arable land quality evaluation project，the study completed the inference mapping of the arable land soil species. Finally，the soil type information of 85 soil survey sample points was updated to the plots where the sample points were located in the soil genera map of garden grassland and the soil species map of arable land.[Result] The accuracy of the mapping results was evaluated using 62 soil survey sample points. The accuracy of the soil genus map was 68% with a kappa coefficient of 0.68. The accuracy of the soil species map of arable land was 61% with a kappa coefficient of 0.59. The main reasons for the inconsistent precision assessment results could be grouped into five major categories：different parent materials，sample points located at the boundary of parent materials，changes in land use types，inconsistent pH values，and inconsistent soil texture.[Conclusion] The results of soil type inference mapping in Niumasi Town have maintained a relatively good level. It proves that this method has the ability to predict arable land soil species and update soil genera maps in soil type inference mapping in areas with soil genera maps but without soil species maps.