Multispectral data for mapping soil texture: possibilities and limitations.
暂无分享,去创建一个
Soil maps derived from random or grid-based sampling schemes are often an important part of precision crop
management. Sampling and soil analysis to derive such maps require a large investment of both time and money. Aerial
photos have been used as a soil mapping aid for years. Studies have shown such an approach can be useful for defining
management units in precision farming, but these studies are often limited to a single field, not an entire farming
operation. In this study, multispectral airborne [green, red, near infrared (NIR), and thermal] and satellite (SPOT and
Landsat TM) data were used to derive soil textural class maps for 350 ha of a 770 ha research and demonstration farm in
Maricopa, Arizona. These maps were compared to soil textural analysis results from samples in the top 30 cm of the soil
profile at an approximate grid spacing of 120 m. Differences in tillage, residue, soil moisture, etc. between fields limited
the accuracy of spectral classification procedures when applied across the entire study area. However, using spectral
classification procedures on a field-by-field basis, it was possible to map areas of soil textural class with reasonable
accuracy. These results are specific to the study area and may not apply at other locations due to the numerous factors
that can contribute to a soil’s spectral response. Classification procedures were also used with vegetation present over the
study area later in the season. Resulting vegetation classes may be helpful in deciding if soil classes impact crop
development enough to warrant different management practices.