Landsat TM data were used to characterize annual changes in vegetation abundance from 1984 to 1997. Vegetation abundance information is extracted from the imagery by assuming every pixel is a mixture of vegetation, soil, and shade; using this mixture model, the fraction of vegetation in each pixel is calculated. This method, called Spectral Mixture Analysis (SMA) is a developing approach that is gaining acceptance in the remote sensing community. More commonly used methods of measuring vegetation abundance from remotely sensed data have been shown to be affected by the background soil color and to saturate at high vegetation abundance. We have shown, however, that SMA has neither of these properties and is highly correlated with field measures of percent live cover.

Figure 8 shows how SMA-derived percent live cover measurements compare to field measurements at the IO1 monitoring site. A seasonal cycle is apparent in the field measurements that were taken more frequently than the satellite measurements. However, for each August measurement between 1991 and 1996 (all of which are concurrent between data sets), the measurements are nearly identical. Error bars, shown in the lower left-hand corner of Figure 8, were derived from our understanding of the method (for SMA) and repeat measurements (for field data).

Figure 8. Percent live cover during the last 15 years as measured in the field and from satellite imagery for the IO1 monitoring site. Field data were collected several times a year, thus detecting seasonal variation, while the satellite data show just late summer of each year.

Figure 9. Changes in percent live cover from one year to the next as measured from satellite images (Spectral Mixture Analysis) and in the field show excellent correlation. Heavy and light lines represent one and two standard deviations from the best fit line, respectively.

We have accurately located all of the 33 permanent monitoring sites in each yearly satellite image and have extracted the corresponding live cover estimates. These data can be summarized by comparing each satellite measurement with its concurrent field measurement, as in Figure 8. In addition, our research is concerned with year-to-year changes in percent live cover; therefore, we have subtracted from each measurement the value from the previous year. These respective changes, in the satellite-derived and field-derived data sets, have been plotted against each other in Figure 9. The lines drawn on the plot represent distances of 4.0% and 8.0% from the best fit line and show the precision of this method at the 65% and 95% confidence limits.

Figures 8 and 9 demonstrate the strength of this approach. We have shown that we can measure percent live cover with an uncertainty of just +/-4.0. Our data sets cover the entire Owens Valley, allowing for regional monitoring of vegetation. These data can now be compared with ICWD's GIS data and maps for depth to groundwater, precipitation, soil type, and vegetation community information.

In the future, we hope the results of the remote sensing study will contribute to the development of a tool to predict the effects of drought and groundwater pumping on Owens Valley vegetation on a regional scale.