基于CSLE模型和抽样单元法的县域土壤侵蚀估算方法对比

为提高县域尺度地块(栅格)土壤侵蚀模数估算的准确性,以河北省怀来县为例,基于CSLE模型,分别采用全域覆盖计算和4%密度抽样单元推算方法对全县土壤侵蚀进行计算和对比分析。结果表明:全域覆盖计算比4%抽样单元推算水土流失面积大59.0 km~2,相对差异达12.94%。全域覆盖计算可实现空间全覆盖,更准确地反映县域水土流失空间分布特点,适用于中、小尺度土壤侵蚀定量计算,但需要较高精度和全面的数据源保证;抽样单元推算适用于流域、区域等大尺度土壤侵蚀估算,但结果受抽样方法、抽样密度、外推或插值方法等因素影响较大。应进一步加强遥感解译准确性、侵蚀因子精度等对CSLE全域覆盖计算结果影响的研究,完善模型参数数据库,率定因子值,实现参数本地化。

Comparison of soil erosion estimation methods at county scale based on CSLE Model and sampling unit

In recent years, it is a significant basic work in soil and water conservation to implement dynamic quantitative monitoring of soil and water loss in countries and key areas nationwide. Soil erosion calculation based on CSLE model can realize quantitative monitoring of soil and water loss. From the perspective of the spatial scope, it is divided into two types as global coverage calculation and sampling unit estimation. It is of great significance to analyze and evaluate the difference and applicability of the two methods for dynamic monitoring of soil and water loss. In this study, Huailai county of Hebei province was taken as an example to evaluate the soil erosion by global coverage calculation and 4% sampling unit estimation. The differences of the two methods were analyzed by data base, factor calculation, soil erosion modulus calculation, soil erosion assessment results, etc., leading to the discussion of the advantages and disadvantages of two methods in the process of the quantitative estimation of soil erosion at the county scale. Our results showed that the area of soil and water loss calculated by global coverage calculation was 59.0 km2 larger than the result based on 4% sampling unit estimation, and the relative difference between the two methods was 12.94 %. By global coverage calculation, the soil erosion area of farm land, garden land, forest land, grass land and others accounted for 30.3%, 9.1%, 24.2%, 56.5% and 4.7% of the land use area, respectively, while by 4% sampling unit estimation, they accounted for 22.8%, 16.3%, 19.8%, 48.1% and 3.0%, respectively. In addition, by global coverage calculation, , the area of farm land, garden land, forest land, grass land and others accounted for 16.0 %, 21.8 %, 31.1 %, 18.4 %, and 12.7 % of the county''s land use, respectively, while they accounted for 14.7 %, 10.4 %, 45.3 %, 15.8 %, and 13.8 %, respectively, by 4% sampling unit estimation. Our results also showed that the difference in terrain data and engineering measured data had minor effect on the calculation results of the two methods, while the difference in land use and vegetation cover data affected more. In the global coverage calculation method, it was easily to increase chances of misjudging shrubbery as grassland, and failure to calculate the coverage of undergrowth vegetation under the garden land and forest land, which caused the garden B factor to be underestimated and the woodland B factor to be overestimated, and the proportion of soil erosion from garden land was low, while the proportion of soil erosion from forest land was high. Quantitative estimation of soil erosion at county scale could be achieved by global coverage calculation and sampling unit estimation, but the former was 12.94% larger than the latter, which was related to the accuracy of interpretation of land use and vegetation coverage, the accuracy of factor calculation, and the determination and localization of model parameters. The global coverage calculation method could reflect the spatial distribution characteristics of soil erosion more accurately, which was suitable for the quantitative estimation of soil erosion at medium and small scales, demanding a higher accuracy and comprehensive data source assurance. The sampling unit estimation was applicable to the estimation of large-scale soil erosion in river basins, regions, etc., but the results were greatly influenced by factors such as sampling methods, sampling density, and extrapolation or interpolation methods. The parameter localization should be implemented gradually in the follow-up study, focusing on the impact of the CSLE model estimation results by factor accuracy and remote sensing interpretation accuracy, improvement of factor parameter database, and factor value rating.