In the borderlands of the US and Mexico, agriculture, pastoralism, urbanization, industrialization, and other growing anthropogenic disturbances are interacting with long-term climate change and interannual climate variability to accelerate desertification. Recent studies have shown that desertification has no one cause; initiation and propagation of desertification is a consequence of a number of natural and human variables. Both traditional and novel field and remote sensing techniques are vital for understanding the processes involved in desertification, but they are inadequate for regional-scale desertification studies when used alone. This is especially true for studies aimed at forecasting desertification.

Remote sensing of arid regions is difficult and necessitates innovative techniques. Desert plants typically manifest long periods of dormancy interspersed with brief "greenings" associated with storms or seasonal rainfall. During these relatively short productive periods, the characteristic spectral features of desert plants change, as does total vegetation cover. Current long repeat times of Landsat and other present satellite sensors provide insufficient temporal resolution to reliably capture the short, but critical, greening.

Further, arid region vegetation is intrinsically difficult to study remotely because: (1) vegetative cover usually is sparse compared to soil background, (2) soil and plant spectral signatures tend to mix non-linearly, and (3) arid plants tend to lack the strong red edge found in plants of humid regions due to ecological adaptations to harsh desert environment. A very important result of these studies is that conventional vegetative red indices can be unreliable measures of arid region plant cover with potential for over- or underestimation of the actual vegetative cover. Despite these challenges, we believe that there is the potential now and in the EOS-era to measure secular changes in vegetation and eventually to remotely monitor several desertification processes.

Desertification is a compound phenomenon in that there are many small- and large-scale phenomena which contribute to overall landscape degradation. No single research program can treat all desertification phenomena. Therefore, we have chosen to concentrate on several key phenomena which have interpretable remote sensing signatures. In particular, three desertification phenomena have been isolated for study: sand blow-outs and dune formation, grazing gradients, and increases in shrub- over grass-cover.

Sand blow-outs appear in remote sensing images as bright "smudges" in the direction of the prevailing winds, due to the brightness of blown sand in all MSS and TM bands. We believe that this is the result of the concentration at the surface of sand-sized grains in areas of blown sand. These smaller grains, in turn, are thought to be more reflective in the infrared than larger grains of similar mineralogy. This explanation has been invoked elsewhere as a means of discriminating active from inactive sand dunes, even in areas of partial vegetation.

Grazing gradients in ecological or physical parameters occur due to the preferential grazing of livestock in some areas over others. Since grazing livestock concentrate around point sources of water in arid regions, grazing gradients often develop in heavily grazed areas adjacent to water sources. Biomass around point sources of water, therefore, will be reduced by the presence of livestock. This, in itself, does not signal desertification. However, if grazed areas do not recover throughout a normal growing season, some landscape damage may be inferred. Furthermore, the degree to which a grazing gradient is erased throughout a growing season is inversely proportional to the degree of desertification. In remote sensing images, grazing gradients appear as sub-circular areas of anomalous reflectance around water sources relative to surrounding areas.

Increases in shrubs in areas undergoing desertification is widespread in the Southwest US, and may be the result of anthropogenic, climate, or some combined disturbance signal. It is visible in remote sensing images by the replacement of grass-dominated reflectance in a pixel by shrub-dominated reflectance. The appearance of a particular pixel, however, can be expected to vary throughout the year as shrubs green and become dormant, and grasses green, senesce, and die. Linear unmixing of spectra from the AVIRIS instrument using well-calibrated field spectra of pixel constituents may provide an effective tool in the monitoring of the grassland to shrubland conversion which accompanies.

Finally, in an attempt to make greater use of both current and future satellite sensors, we are initiating a program of neural net classification of air- and spaceborne remote sensing using high-performance computing. Although this approach will begin with imaging spectrometer data, we will produce reliable classifiers which can use a wide array of current and future remote sensing data. We also hope to create "smart" classifiers which use field spectra of ground constituents to create increasingly more useful classification schemes.

Since many desertification processes occur as "runaway" phenomena for which there are no known remediation techniques, predicting desertification is vital for the sustainability of human activities in arid regions. A prototype conceptual framework is presented which, upon optimization, will aid in the prediction, and therefore hindrance of runaway desertification processes and will contribute to sustainable land management of arid regions.

We plan to develop Leading Desertification Indicators (LDIs) analogous to the widely-used Leading Economic Indicators (LEIs). Like the LEIs, our LDIs will be comprised of an eclectic group of factors each of which is given a weighting factor adjusted to optimize the predicting power of an LDI. Factors included in the LDIs will involve remote sensing, climate, ecological, and socioeconomic data which, whenever possible, must be easily available and routinely collected.