Quantifying vegetation change in semiarid environments: Precision and accuracy of spectral mixture analysis and the Normalized Difference Vegetation Index

Because in situ techniques for determining vegetation abundance in semiarid regions are labor intensive, they usually are not feasible for regional an alyses. Remotely sensed data provide the large spatial scale necessary, but their precision and accuracy in determining vegetation abundance and its c hange through time have not been quantitatively determined. In this paper t he precision and accuracy of two techniques, Spectral Mixture Analysis (SMA ) and Normalized Difference Vegetation Index (NDVI) applied to Landsat TM d ata, are assessed quantitatively using high-precision in situ data. In Owen s Valley, California we have 6 years of continuous field data (1991-1996) f or 33 sites acquired concurrently with six cloudless Landsat TM images. The multitemporal remotely sensed data were coregistered to within 1 pixel, ra diometrically intercalibrated using temporally invariant surface features, and geolocated to within 30 m. These procedures facilitated the accurate lo cation of field-monitoring sites within the remotely sensed data. Formal un certainties in the registration, radiometric alignment, and modeling were d etermined. Results show that SMA absolute percent live cover (%LC) estimate s are accurate to within +/-4.0%LC and estimates of change in live cover ha ve a precision of +/-3.8%LC. Furthermore, even when applied to areas of low vegetation cover the SMA approach correctly determined the sense of change (i.e., positive or negative) in 87% of the samples. SMA results are superi or to NDVI, which, although correlated with live cover, is not a quantitati ve measure and showed the correct sense of change in only 67% of the sample s. (C) Elsevier Science Inc., 2000.