The Characterization of Soil Properties to Develop “Soil Management/Mapping Units” Using High-Resolution Remotely Sensed Data Sets

The intent of this research was to assess the possible use of high resolution remotely sensed hyperspectral and multispectral data to characterize soil types, specifically focusing on organic matter content, in an associative manner with the results obtained from traditional Order 1 and Order 2 soil surveys. A chi-square analysis indicated a strong association between soil type and organic matter content. A Cramer’s V analysis (of a supervised classification) indicated a stronger relationship between the Order 1 and organic matter. However, when an unsupervised classification scheme was applied to the aerial imagery, again using Cramer’s analysis, the Order 2 out-performed the Order 1. This superior performance was due in part to the grouping of multi-band spectral response patterns into statistically separable clusters. A One-Way ANOVA analysis indicated that all soils were significantly different in the Order 2 survey for both the hyperspectral and the multispectral data sets. However, the Order 1 results show the ITD sensor more successfully grouping the darker soils than did the ATLAS which grouped the lighter soils. A linear discriminate analysis (LDA) demonstrates that the computer classification of images more successfully assessed the Order 2 survey than the Order 1. Again it is worth noting that the LDA also grouped the soils in a similar manner as did the ANOVA in that the ITD sensor grouped the darker soils and the ATLAS sensor grouped the lighter soils. This sensor preference is another significant secondary finding of this study. Despite the subjective nature of the soil mapping exercise and the use of un-calibrated data sets, high resolution imagery was able to differentiate different soil mapping scales. Even though associations were relatively low statistically, this study supports the hypothesis that high resolution imagery, although limited by its two-dimensional capabilities, can be effectively used as a predictive tool, although with the current technological limits, the imagery cannot serve as a surrogate for more traditional soil surveys