摘要
基于地统计方法的土壤属性制图通常需要大量的采样与实验室测定。本研究提出利用可见光近红外（visible-nearinfrared spectroscopy,VNIR）光谱技术测定替代实验室测定,并与地统计方法相结合预测土壤质地的空间变异。通过建立砂粒（〉0.02 mm）,粉粒（0.002-0.02 mm）,黏粒（〈0.002 mm）含量的VNIR光谱预测模型,将模型预测得到的质地数据和建模点实测质地数据一同用于地统计分析和Kriging插值制图。以江苏北部黄淮平原地区为案例的研究结果表明,砂粒、粉粒、黏粒含量的预测值和实测值的均方根误差（RMSE）分别为8.67%、6.90%3、.51%,平均绝对误差（MAE）分别为6.46%、5.60%、3.05%,显示了较高的预测精度。研究为快速获取平原区土壤质地空间分布提供了新的可能的途径。
Digital soil mapping methods based on Geo-statistics often needs a large number of samples.This study investigated a method that integrating geostatistics and visible-near-infrared（VNIR） spectroscopy to estimate soil texture of a plain area in Jiangsu province.Predictive models between soil texture（sand,silt and clay content） and VNIR spectroscopy were established.Soil texture data of training samples and those obtained from the predicted models were used for mapping soil texture using ordinary kriging method.A validation dataset produced the estimates of error for the predicted maps of sand,silt and clay expressed as RMSE with the values of 8.67%,6.90%,and 3.51%,respectively.This study demonstrates the possibility to potentially map regional soil texture variation digitally with considerable success.
出处
《土壤通报》
CAS
CSCD
北大核心
2012年第2期257-262,共6页
Chinese Journal of Soil Science
基金
江苏省基础研究计划(BK2008058)
中国科学院知识创新工程重要方向性项目(KZCX2-YW-409)资助
