省域尺度土壤有机质空间分布的神经网络法预测

土壤有机质空间分布预测方法研究对指导省域尺度下土壤有机质空间插值模型选取和精度优化具有重要意义。以江西省为例,利用BP神经网络模型与普通克里金结合的方法(BPNN-OK)、RBF神经网络模型与普通克里金结合的方法(RBFNN-OK)以及普通克里金法(OK)3种方法,预测省域尺度下耕地表层(0~20 cm)土壤有机质的空间分布。16 109个土壤样点随机分成12 887个建模样点,3 222个测试样点。结果表明:在省域尺度下,BPNN-OK法、RBFNN-OK法较OK法在土壤有机质空间预测精度上有较大提升,三者的预测精度为BPNN-OK>RBFNN-OK>OK。BPNN-OK法对土壤有机质预测结果的均方根误差、平均绝对误差、平均相对误差较OK法分别降低28.66%、30.71%、34.76%,RBFNN-OK法较OK法分别降低27.76%、29.74%、33.71%。在省域尺度下,神经网络模型与普通克里金结合的方法能很好地捕捉土壤有机质的复杂空间变异关系。研究结果可指导江西省土壤有机质空间插值模型选取。

基于遥感与随机森林算法的陕西省土壤有机质空间预测

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。

Regression-Kriging of Soil Organic Matter Using the Environmental Variables Derived from MODIS and DEM

[Objective] The objective of this project was to evaluate and compare spa- tial estimation accuracy by ordinary kriging and regression kriging with MODIS data, predicting SOM contents using limited available data in Shimen County, Hunan Province, China. [Method] Terrain parameters （derived from DEM） and Normalized differential vegetation index （NDVI）, Land surface temperature （LST） （derived from MODIS data） were used as auxiliary data to predict the SOM spatial distribution. The mean error （ME） and mean square error （RMSE） were adopted to validate the SOM prediction accuracy. The descriptive statistics and data transformation were conducted by using computer technology. [Result] Regression kriging with terrain and remotely sensed data was superior to ordinary kriging in the case of limited available samples; even the linear relationship between environmental variables and SOM content was moderate. The accuracy assessment showed that the regression kriging method combining with environmental factors obtained a lower mean predication error and root mean square prediction error. The relative improvement was 6.03% compared with ordinary kriging. [Conclusion] Remotely sensed data such as MODIS im- age have the potential as useful auxiliary variables for improving the precision and reliability of SOM prediction in the hilly regions.

Soil Organic Matter Mapping by Decision Tree Modeling

Based on a case study of Longyou County, Zhejiang Province, the decision tree, a data mining method, was used to analyze the relationships between soil organic matter (SOM) and other environmental and satellite sensing spatial data. The decision tree associated SOM content with some extensive easily observable landscape attributes, such as landform, geology, land use, and remote sensing images, thus transforming the SOM-related information into a clear, quantitative, landscape factor-associated regular syst…