Estimating mean field residue cover on midwestern soils using satellite imagery.
Estimating mean field residue cover on midwestern soils using satellite imagery.
复制标题
DOI：
10.2134/agronj2007.0249
复制
发表时间：
2009-05
期刊：
影响因子：
2.1
通讯作者：
B. Gelder;A. Kaleita;R. Cruse
中科院分区：
文献类型：
--
作者：
B. Gelder;A. Kaleita;R. Cruse
残留物覆盖的知识是有针对性的保护工作，以减少土壤侵蚀，径流和相关的环境影响的关键;然而，目前还没有一个快速，准确，廉价的方法。以前的研究表明，混合结果检测农作物残留使用Landsat残留指数，但条件一般包括土壤颜色对比度差，新兴植被，或分类残留物覆盖。我们的目标是评估一个新的归一化差异残留指数（NDRI），沿着与其他指数，在2005年和2006年的多个图像日期在中北部爱荷华州黑暗的土壤。采用了一种自动化的方法来划定油田边界。NDRI，使用Landsat波段3和7，表现最好的整体，解释81%的残留物覆盖差异的整体，78%出现前，9%出现后。标准化差异耕作指数（NDTI），使用Landsat波段5和7，也表现良好，解释68%的变化，出苗前86%，出苗后6%。引入经验修正对绿色植被改善指数性能的影响。经过绿色植被校正后，NDTI的表现优于NDRI，解释了67%的变化，而NDTI解释了63%的变化。NDTI也返回了最好的RMSE（0.11）在出苗前的条件下，和0.15后，绿色植被校正。一般来说，利用Landsat Band 7（其中含有土壤中不存在的木质素和纤维素吸收带）的指标，返回最佳残留物检测结果。利用陆地卫星波段4的指数，其中绿色植被的反射率高，难以检测残留物覆盖，特别是在植物出苗后。
Knowledge of residue cover is crucial for targeting conservation efforts to reduce soil erosion, runoff, and associated environmental impacts ; however, a rapid, accurate, inexpensive methodology is not currently available. Previous studies have shown mixed results detecting crop residue using Landsat residue indices, but conditions generally included poor soil color contrast, emergent vegetation, or categorized residue cover. Our objectives were to evaluate a new normalized difference residue index (NDRI), along with other indices, over multiple image dates in 2005 and 2006 on dark soils in north-central Iowa. An automated method for field boundary delineation was used. The NDRI, using Landsat Bands 3 and 7, performed best overall, explaining 81% of the residue cover differences overall, 78% before emergence, and 9% after emergence. The normalized difference tillage index (NDTI), using Landsat Bands 5 and 7, also performed well explaining 68% of the variation overall, 86% before emergence, and 6% after emergence. Introduction of an empirical correction of the influence of green vegetation improved index performance. The NDTI outperformed the NDRI after green vegetation correction, explaining 67% of the variation versus 63%. The NDTI also returned the best RMSE (0.11) under preemergence conditions, and 0.15 after green vegetation correction. Generally, indices utilizing Landsat Band 7, which contain lignin and cellulose absorption bands absent in soil, returned the best residue detection results. Indices utilizing Landsat Band 4, where the reflectance of green vegetation is high, had difficulty detecting residue cover, especially after plant emergence.