| Traditional field plot techniques do not extrapolate well over large areas of sparse, diverse vegetation. This research was conducted to determine if Landsat TM data, the individual TM bands, and TM based ratio indices coupled with a multivariate statistical analysis can be used to effectively monitor or measure subtle changes in desert vegetation.;Forty-one field sites in north central Nevada were selected to develop a Landsat Thematic Mapper based methodology for mapping changes in high desert vegetation. A multiple linear regression analysis and a canonical correlation analysis were used to determine the best subsets of dependent variables to predict percent cover for each cover type. Multiple linear regression analysis showed low R-squared values, ranging from 0.382 to 0.6717. The p-values were significant, ranging from 0.035 to 0.0001. In the canonical correlation analysis all p-values were significant, less than 0.05, with an r-squared value of 86%.;An ordination analysis was also used, which separated the field sites into 14 groups. An analysis of variance was applied to these groups, the means were compared resulting in 6 groups that showed a significant difference from one another. These groups appear to follow a gradient from high total bare ground and low cover to an area of low total bare ground and a high percentage of cover. It is suggested that this approach to monitoring arid lands could detect a change in cover indicative of desertification. |
