基于多分辨率遥感数据与随机森林算法的土壤有机质预测研究

发布时间：2018-05-06 16:45

本文选题：多分辨率遥感数据 + 随机森林　；参考：《土壤学报》2016年02期

【摘要】：遥感数据已经在数字土壤制图中得到广泛应用,并且可以一定程度上提高土壤属性预测的精度。本文以榆阳区的黄土丘陵和风沙滩地两种地貌区为例,利用不同分辨率的专题制图仪(thematic mapper,TM)、先进宽视场传感器(advanced wide field sensor,AWIFS)和中等分辨率成像仪(Moderate resolution imaging spectroradiometer,MODIS)的遥感影像数据(分辨率分别为30 m、56 m和250 m)和基于高级热量散射和反射辐射仪全球数字高程模型(advanced spaceborne thermal emission and reflection radiometer global digital elevation model,ASTER GDEM)的地形衍生数据,结合其他影响土壤有机质分布的辅助因子,用随机森林算法(random forest,RF)对表层土壤有机质进行模拟预测,并通过实测数据的百分比抽样对预测结果进行了验证。结果表明,在榆阳区的黄土丘陵区,基于TM数据的土壤有机质预测结果较好;在风沙滩地区,基于AWIFS数据的土壤有机质预测结果较好。基于RF的土壤有机质预测在榆阳区的黄土丘陵区结果较好,三个分辨率下的平均绝对误差在1.27~1.57 g kg-1之间,在风沙滩地区预测精度较低,平均绝对误差在1.46~2.08 g kg-1之间。高程、地理位置和植被是影响黄土丘陵区土壤有机质预测的主要因素,在风沙滩地区,植被、高程和离水源地的距离是影响有机质预测的主要因素。可见,在地貌相对简单的地区进行土壤有机质含量的预测时可以使用较低分辨率的数据代替较高分辨率的数据,同时,RF算法在复杂地貌区的土壤有机质预测更有效。
[Abstract]:Remote sensing data have been widely used in digital soil mapping and can improve the precision of soil attribute prediction to some extent. In this paper, the loess hilly and wind-sand beach areas in Yuyang district are taken as examples. Using thematic cartography with different resolutions, advanced wide field sensor AWIFSs and moderate resolution imaging spectroradiorometerMODIS-based remote sensing image data (resolutions of 30mmt and 250m, respectively) and based on advanced thermal scattering (AHRS), advanced wide-field sensor (AWAF) and moderate resolution imaging spectroadior (MODISs) are used in this paper. Topographic derived data of advanced spaceborne thermal emission and reflection radiometer global digital elevation model (ASTER GDEM). Combined with other auxiliary factors which affect the distribution of soil organic matter, the random forest algorithm random forestfr is used to simulate and predict the surface soil organic matter, and the predicted results are verified by percentage sampling of measured data. The results show that the prediction results of soil organic matter based on TM data are better in the loess hilly area of Yuyang area, and the results of soil organic matter prediction based on AWIFS data are better in wind-beach area. The results of soil organic matter prediction based on RF are better in the loess hilly area of Yuyang area. The average absolute error is between 1.27 and 1.57 g kg-1 under three resolutions, and the average absolute error is between 1.46 and 2.08 g kg-1 in wind-beach area. Height, geographical location and vegetation are the main factors that affect the prediction of soil organic matter in loess hilly region. In wind-sand area, vegetation, height and distance from water source are the main factors that influence the prediction of organic matter. It can be seen that the lower resolution data can be used instead of the higher resolution data in the prediction of soil organic matter content in areas with relatively simple geomorphology, and the RF algorithm is more effective in predicting soil organic matter in complex geomorphological areas.
【作者单位】：西北农林科技大学资源环境学院;农业部西北植物营养与农业环境重点实验室;
【基金】：国家科技基础性工作专项项目(2014FY110200A08)资助~~
【分类号】：S153.621

【相似文献】