基于遥感与随机森林算法的陕西省土壤有机质空间预测

发布时间：2018-08-22 12:03

【摘要】：遥感数据作为反映土壤组成结构及植被生长状况的数据源,借助辅助环境因子的土壤属性预测在数字土壤制图中日益受到重视。论文运用随机森林(Random Forest,RF)算法,基于AWIFS(分辨率56 m)和MODIS(分辨率250 m)遥感数据及501个实测样点数据对陕西省土壤有机质空间分布状况进行预测,并对预测精度进行估算。结果表明陕西省土壤有机质含量以南部的秦岭山地区和大巴山区为最高,土壤有机质含量大于25 g·kg~(-1),黄土高原南部处于中等水平,大部分在16~25 g·kg~(-1)之间,关中平原和汉中低山丘陵区含量偏低,大部分在13~25 g·kg~(-1),而黄土高原北部和风沙滩区含量大部分低于10 g·kg~(-1)。基于AWIFS影像的预测效果要优于MODIS影像,成像日期对有机质预测的影响不大。基于RF模型的土壤有机质预测精度在设定的不同抽样百分比条件下,独立验证数据集的平均误差大部分不超过3 g·kg~(-1),预测值与实测值的相关系数在0.7以上。高程是影响土壤有机质预测的最重要因子,当影像的分辨率降低时,样点分布的地理经纬度和坡度对土壤有机质预测的影响上升,植被因子的影响程度下降。
[Abstract]:Remote sensing data, as a data source reflecting soil composition and vegetation growth, have been paid more and more attention in digital soil mapping with the help of soil attribute prediction of auxiliary environmental factors. Based on AWIFS (resolution 56m) and MODIS (resolution 250m) remote sensing data and 501 measured samples, the spatial distribution of soil organic matter in Shaanxi Province is predicted by Random forest RF algorithm, and the prediction accuracy is estimated. The results showed that the content of soil organic matter was the highest in Qinling Mountains and Daba Mountains in the south of Shaanxi Province. The content of soil organic matter was more than 25 g kg ~ (-1) in the southern part of the Loess Plateau, most of which was between 16 ~ 25 g kg ~ (-1) in the south of Loess Plateau. The contents in Guanzhong Plain and Hanzhong low mountain and hilly area were low, mostly in 1325 g kg-1, while in the north of Loess Plateau and wind-beach area, the content was less than 10g kg-1. The prediction effect of AWIFS image is better than that of MODIS image, and the date of imaging has little effect on organic matter prediction. The prediction accuracy of soil organic matter based on RF model under different sampling percentages, the average error of independently verified data set is less than 3 g kg-1, and the correlation coefficient between the predicted value and the measured value is more than 0.7. Height is the most important factor affecting the prediction of soil organic matter. When the resolution of the image is reduced, the geographical longitude, latitude and slope of the distribution of the sample point increase in the prediction of soil organic matter, and the influence degree of vegetation factor decreases.
【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)~~
【分类号】：S153.621

【参考文献】