This file was created by scanning the printed publication. Errors identified by the software have been corrected; however, some errors may remain. GREAT BASIN ANNUAL VEGETATION PATTERNS ASSESSED BY REMOTE SENSING Paul T. Tueller ABSTRACT CIIEATGRASS Cheatgrass-infested wildfire scars are readily identifiable on various scales of aerial photography including Landsat images if the bums are recent. Measurements were made from four Landsat MSS images and 11 NASA high-altitude color infrared aerial photographs of a scale of 1:29,000. Individual Landsat bands and various ratio and coeffwient-based vegetation indices can separate burned from unburned sites using image processing techniques. Several vegetation indices were useful in linear models describing range sites with increasing amounts of cheatgrass. There is ample evidence in the literature that cheatgrass is strongly competitive in the sagebrush/grass vegetation in the Great Basin and Intermountain region. Young and others (1971) state that annual species, like cheatgrass, are highly evolved to occupy a low seral niche caused by fire or grazing. Native species have never evolved to occupy these niches. These niches are not present in pristine big sagebrush-perennial bunchgrass communities, but when a perturbation occurs such as fire, alien annuals readily invade. Part of their success is attributable to their highly developed reproductive system, which makes them tough competitors after they have been established. Cheatgrass, sometimes called downy brome, readily invades a low seral niche created by a disturbance. The presence of plant litter and a rough microtopography are key seedbed characteristics permitting cheatgrass establishment, although success on bare ground is not high (Young and Evans 1972). Even though cheatgrass caryopses are greatly reduced after fire, the surviving caryopses respond dynamically to the released environment potential. The response may include hybridization and recombination. Cheatgrass has the inherent competitive ability to close the new seral community to seedlings of native perennial grasses and annuals, and its dominance predisposes the site to recurring wildfire perturbation and cyclic environmental degradation (Young and Evans 1977). Cheatgrass is able to rapidly occupy the belowground space and actively utilize soil resources competing with native species for soil water. This negatively affects their soil-water status, aboveground biomass production, and root-length density, giving a strong competitive advantage to the cheatgrass, which then dominates the site (Melgoza 1989). Young and others (1969) conclude that the important factors in the dynamics of downy brome populations are: (a) large numbers of viable caryopses are carried from one year to the next in the litter and soil; (b) downy brome caryopses production is density dependent; and (c) the simultaneous germination characteristics of freshly harvested downy brome caryopses can be conditioned environmentally to continuous germination. Wright and Klemmedson (1965) compared the experimental burning of four species of grasses-Sandberg bluegrass CPoa secunda), bottlebrush squirreltail (Sitanion hystrix), needle-and-thread (Stipa comata), and Thurber needlegrass (Stipa thurberiana)-to understand their compatibility with cheatgrass. Sandberg bluegrass and squirreltail proved to be fairly resistant to fire, with squirreltail being damaged only in June and the bluegrass INTRODUCTION In recorded history, many hundreds of wildfires have occurred throughout the Great Basin and Intermountain region. The majority of these have occurred in the sagebrush vegetation that covers an estimated 19.5 million hectares throughout the region (Tueller 1989a). These wildfires vary considerably as to number and acreage involved. Many thousands of hectares have been impacted. The result has generally been the creation of an annual grass type dominated by cheatgrass (Bromus tectorum). The number, extent, and location have not generally been recorded for future reference. This paper considers the potential uses of remote sensing technology for examining the extent of and changes in the annual cheatgrass vegetation types found in the Great Basin. Consideration has been given to the measurement of these sometimes monospecific plant communities where cheatgrass is the dominant or codominant species on both small and large areas. Lengthy ecotones have been formed. These have been evaluated by interpretation and measurements from aerial photographs and the measurement of these same attributes on satellite-derived computer compatible data with image processing. Procedures for monitoring successional trajectories are discussed. The spectral signatures associated with burned sagebrush or pinyon/juniper plant communities have been examined. Paper presented at the symposium on Ecology. Management. and Restoration of Intermountain Annual Rangelands, Boise, ID, May 18-22, 1992. Paul T. Tueller is Professor of Range Ecology. Department of Range, Wildlife and Forestry, University of Nevada. Reno. 126 not at all. Needle-and-thread appeared to be the most susceptible to fire. Burning in June and July caused extreme damage to needle-and-thread, but in August it was relatively resistant. Thurber needlegrass had a similar response as needle-and-thread but less severe. Season of burn determined the extent of damage to both needlegrass species, and size of plant became increasingly important in determining damage in late summer. DeFlon (1986) describes positive aspects of cheatgrass. Cheatgrass is one of the few grasses that will invade and flourish in the alkaline soils associated with the marginal rainfall of the Great Basin. He goes on to say that it should not be grazed in the summer, but in the winter it can be grazed when it is more palatable and nutritious than crested wheatgrass. He feels the case of cheatgrass as a grazeable species should be reexamined. Raison (1979) has drawn several conclusions relative to the management of controlled burns. Effects of fire vary with each ecological situation, and currently insufficient data exist to allow accurate prediction of its full long-term effects on many ecosystems that are regularly burned Furthermore, there needs to be an integrated regional study of the long-term effects of fire on plant communities; regularly burned areas need to be established and maintained so that fire ecology can be better researched and understood. Remote sensing technology has a role to play in this endeavor. parameters such as vegetation biomass, species composition, phenology, species cover, and vegetation changes (plant succession) based on Landsat TM data (Tueller 1989). Most vegetation indices have been developed under humid conditions. In arid zones the ability to interpret spectral information based on vegetation indices is restricted because of the usual small portions of green phytomass and generally low percentages of shrub cover. In semi-arid regions with large areas of bare soil (50-100 percent) most ratio-based indices are known to be adversely affected by differences in soil brightness (Elvidge and Lyon 1985; Huete 1986). Differing spectral responses on changing soil types significantly limit the accuracy of vegetation estimates (Heilman and Boyd 1986; Jackson 1983). Huete and Jackson (1987) found both ratio and orthogonal vegetation indices to be unreliable for detecting green phytomass of range canopies. Site-specific soil lines and coefficients can reduce the soil background influence and thus improve the estimate of greenness. Huete (1988) developed a soil-a(ljusted vegetation index (SAVI) to minimize the soil brightness spectral influence. Ratio-based indices include the Ratio Vegetation Index (RVI), the Normalized Difference Vegetation Index (NDVI), the Transformed Normalized Difference Vegetation Index (TNDVI), the Modified Normalized Difference Vegetation Index (MNDVI), the Soil A(ljusted Vegetation Index (SAVI), and the Transformed Soil Adjusted Vegetation Index (TSAVI). Paltridge and Barber (1988) applied a modified normalized difference vegetation index (MNDVI) (NIR-1.2*RED)/ (NIR+RED). They used a 1.2 factor to a(ljust the NDVI to arid rangeland conditions in Australia. N-space indices include the Greenness Vegetation Index (GVI), the Soil Brightness Index (SBI), and the Perpendicular Vegetation Index (PVI). Jackson (1983) gives a clear example of how to calculate the coefficients of n-space vegetation indices, GVI, and SBI using the Gram-Schmitt orthogonalization process (Frieberger 1960) to obtain vectors. Four linear combinations of Landsat MSS bands known as "brightness," "greenness," "yellowness," and "nonsuch" originate from a four-band Landsat MSS orthogonal linear combination as described by Kauth and Thomas (1976). The two-dimensional Perpendicular Vegetation Index (PVI) measures the orthogonal distance of a pixel back to a pre-established soil line (Richardson and Wiegand 1977). The farther a pixel is away from the soil line, the higher the green vegetation contribution to the pixel. We found that the PVI is highly correlated to shrub cover on salt desert and sagebrush-dominated rangelands (r =0.91). Other indices to consider include the WDVI (Weighted Difference Vegetation Index) (Clevers 1989) and the TSAVI (Transformed Soil Adjusted Vegetation Index) (Baret and Guyot 1991). Corrections for atmospheric variability are important to normalize spectral data recorded from high altitudes. Atmospheric corrections can be made for satellite data by using linear regressions of light and dark standard reflectance targets on the ground (Marsh and Lyon 1980). Path radiance corrections can increase the classification accuracy of the data (Kowalik and others 1983). Elvidge and Lyon (1985) also describe a useful method for defining the atmospheric influence on digital counts for the TM bands. REMOTE SENSING Remote sensing techniques hold considerable promise for the inventory and monitoring of natural resources in arid regions. They have the potential to integrate various systems for rangeland management (Tueller 1989b) leading to more efficient utilization of forage resources from rangelands by wildlife species as well as domestic livestock and wild horses. However, there has been a fundamental reason why applications have not been forthcoming-a significant lack of information concerning basic spectral characteristics of rangeland/aridland vegetation and soils. Without such basic information relevant to the sensors that are now available to rangeland resource managers, it has been and will remain difficult to develop the desired applications. Soil substrate plays a significant role in vegetation assessments from spectral data. Similar plant communities can appear different if they have different soil types. Vegetation biomass tends to be overestimated on dark substrates and underestimated on light substrates (Eividge and Lyon 1984). The typical spectra from rangeland scenes are closely correlated with the soil line developed by Kauth and Thomas (1976). Huete and Jackson (1987) found that senesced grass and weathered litter significantly altered the spectral response of the range vegetation canopy in the first four Thematic Mapper wavebands and thus seriously hamper tn.e use of spectral vegetation indices in assessing green phytomass levels. However, pixel modeling may allow phytomass estimations (Huete 1986). Data transformations involving such things as differences, sums, and ratios both with the red and infrared spectral space can be evaluated to determine if such data manipulation procedures will have predictor value for 127 Table 1-landsat TM bands acquired and vegetation indices computed in this study A modified AVHRR vegetation index was found to be related to "grassland" fuel-moisture content in a study near Victoria, Australia. The AVHRR system as a whole proved extremely valuable for monitoring potential fire danger areas of the State. The relationship was based on the dominant control of leaf moisture status on satelliteobserved vegetation indices (Paltridge and Barber 1988). Yue-Hong and others (1990) compared spatial weighting functions and found that the contiguity weight can be used to define the spatial term of neighborhood effects on wildfire. Data overlays of multiple GIS layers derive the explanatory variables for modeling the distribution of wildfires from logistic regressions. The model improves when the spatial term is included. Their results suggest that neighborhood effects are a primary factor in the distribution of wildfires. Burned and unburned grass canopies had distinctly different diurnal surface radiative temperatures, and measurements of surface energy balance components revealed a difference in partitioning of the available energy between the two canopies, which resulted in the difference of their measured surface temperatures. Additionally, the timing of measurements and topographic conditions affect the magnitude of the difference in surface temperatures (Asrar and others 1988). A geographic information system (GIS) can be used to identify and analyze not only the type and amount of change on a burn site, but also those classes which did not change. A GIS restricts change analysis to the fireaffected area by masking out unaffected vegetation through overlay operations. Furthermore, the flexibility afforded by a GIS allows additional data sets from preceding or succeeding dates to be added, creating a database for the study of long-term change within the study site (Jakubauski and others 1990). TM1 Thematic Mapper Band1 TM2 Thematic Mapper Band2 TM3 Thematic Mapper Band3 TM4 Thematic Mapper Band4 Shrub Cover Percent BATE Cover Percent DVI = Difference Vegetation Index RVI = Ratio Vegetation Index TNDVI = Transformed Normalized Difference Vegetation Index Bl == Normalized Difference Vegetation Index PVI Perpendicular Vegetation Index MNDVI = Modified Normalized Vegetation Index SBI Soil Brightness Index GVI = Greenness Vegetation Index SAVI = Soil Adjusted Vegetation Index TSAVI = Transformed Soil Adjusted Vegetation Index WDVI = Weighted Difference Vegetation Index amounts of cheatgrass mixed in the stand were sampled. The data set consisted of ground measurements of cover and both reflectance and radiance data from a 20-pixel sample from the same ground locations of western Nevada shrub-dominated range sites. The 20 radiance values for each TM band were used to· compute several vegetation indices as described earlier in the paper. For this paper I have examined only the linear relationships with ground-measured dominant shrub (usually sagebrush) cover, bare ground, and cheatgrass cover as the dependent variable against a number of independent variables (vegetation indices). A third data set consisted of interpreting and studying four Landsat MSS images and 11 NASA high-altitude frames of a scale of 1:29,000. These were analyzed by photo interpretation techniques. For these scenes we identified and measured each bum in the scene (most were in the sagebrush vegetation) and determined the acreage and length of perimeter of the bum. METHODS Three data sets were developed for this study. First we examined 10 burns in the sagebrush or pinyon/juniper woodland vegetation from three Landsat subscenes in western and central Nevada. Homogeneous areas were selected for burned and unburned sites at these 10 burn locations. From these homogeneous sites 20 pixels of Landsat Thematic Mapper data were obtained for analysis. Data for each of the TM bands (one through five and seven) except the thermal band were summarized (table 1). Then 11 vegetation indices were computed for each site based on TM Band four (Near Infrared) and TM Band three (red). Formulas for these vegetation indices are listed in table 2. No ground data were available for these 10 bum-unburn comparisons. The various vegetation indices were chosen and interpreted based on their potential to distinguish between burned and unburned vegetation sites during June. Means and standard deviations were calculated along with a Student's t statistic in order to compare burned with unburned sites. A second data set was analyzed as part of another study we are working on designed to model pixels for their various components based on various vegetation indices. Ten unburned shrub-dominated range sites in western Nevada (or they had not been recently burned) with differing Table 2-Vegetation indices used in this study Vegetation Index RVI 1 NDVI TNDVI MNDVI PVI SBI GBI DVI WDVI SAVI TSAVI 1 128 Formula TM4/TM3 TM4-TM3/TM4+TM3 ~ (TM4-TM3/TM4+TM3)+0.5 TM4-1.20*TM3/TM4+TM3 ~ (Rgg5-TM3)2+(Rgg7-TM4)2 (0.664*TM3)+(0.747*TM4) (-o.747*TM3)+(0.664*TM4) TM4-TM3 TM4-0.08*TM3 TM4-TM3/(TM4+TM3+0.5) • (1.0+0.5) a1 (TM4-a1 *TM3-b1 )/ [a1-TM4+TM3-a1*b1 +0.08(1 +a12)] For variable names see table 1. RESULTS AND DISCUSSION Table 4-A comparison of Landsat TM bands as simple comparative signatures for burned and unburned pixel samples from the Fort Sage study site in western Nevada p Site: Fort Sage Mean S.D. S.E. T Increasing amounts of cheatgrass are usually associated with site degradation by overgrazing, wildfire, or other potential perturbations. Burns that are only a few years old are reasonably easy to identify on the Landsat Thematic Mapper images displayed on the image processing monitor. Tables 3 and 4 show the results from two of the 10 sites, a fire on the Gund Ranch in central Nevada and one on Fort Sage Mountain in western Nevada. Statistical comparisons with each band comparing the burned and unburned sites easily record significant differences using a simple t test. While it is easy to record the difference using such a procedure, not much is really learned about the spectral characteristics of these rather simple ecosystems. Rather there is a need to carefully do some pixel modeling in order to determine the information content of the individual pixels or pixel samples for the site. Data comparing these same 10 burns using several vegetation indices described earlier show that some indices show significant differences and some do not. Data from four of these indices, RVI, GVI, NDVI, and TNDVI, showed potential for describing differences among burned and unburned sites (table 5). In this case the RVI and the PVI were not successful in differentiating among the burned and unburned sites. However, the NDVI and the TNDVI were useful vegetation indices in this regard. These two indices among several others seem to be sensitive to the small amounts of vegetation reflectance from these cold desert sites. For predicting the level of cheatgrass amount on a site I refer the reader to the second data set that was not completed on burned and unburned sites (fig. 1). The adjustedR2 value of0.9661 for this data set is somewhat suspect since two seeming outliers accounted for the strength of the relationship. Further sites with cheatgrass will be required to determine the real value of this relationship. However, cheatgrass tends to be abundant on a site or not abundant depending on disturbance, and TM1 Burned Unburned Mean S.D. S.E. T p TM1 Burned Unburned 79.85 73.10 2.796 1.483 0.6252 .3317 9.54 9.54 0.0000 .0000 42.50 35.50 1.638 .607 .3663 .1357 17.92 17.92 .0000 .0000 Burned Unburned 74.65 59.00 2.207 .858 .4935 .1919 29.55 29.55 .0000 .0000 TM4 Burned Unburned 72.50 54.05 1.960 .887 .4383 .1983 38.35 38.35 .0000 .0000 TM5 Burned Unburned 147.60 121.50 1.932 2.089 .4321 .4672 41.01 41.01 .0000 .0000 0.3293 .4672 10.06 10.06 0.0000 .0000 Burned Unburned 40.35 36.70 .7452 1.302() .1666 .2911 10.88 10.88 .0000 .0000 TM3 Bumed Unburned 66.50 57.75 1.0000 2.6730 .2236 .5977 13.71 13.71 .0000 .0000 TM4 Bumed Unburned 67.95 58.75 .8256 1.0700 .1846 .2392 30.44 30.44 .0000 .0000 TM5 Bumed Unburned 147.50 126.80 1.6700 2.9610 .3735 .6620 27.17 27.17 .0000 .0000 the two outliers reflect disturbed sites with high cheatgrass cover. The percentage of the dominant shrub on our 10 westem Nevada study sites was predictable based on a combination of the two vegetation indices, the NDVI and MNDVI, giving an acljusted R 2 of 0.8066. A prediction based on TM4A and TM2A gave an even higher acljusted R 2 (0.9002) value. It is gratifying to find that a reasonable estimation of the dominant shrub on these cold desert sites can be made with Landsat data alone. The increased acljusted R2 based on ground reflectance of the site-dominant shrub in bands two and five is difficult to explain and requires further research. The third data set involved the photo interpretation of several burn sites in the sagebrush grass. Photo interpretation criteria used to identify these sites include such factors as lighter tones associated with the burned area when compared with the darker tones of the sagebrush vegetation and the irregular shape of the burns. Some of the burns were bounded by tracks made by heavy equipment used in the suppression process. Table 6 and table 7 summarize these data. For the 11 NASA high-flight color infrared photographs we found that the burn size varied from 26.8 acres to 113.2 acres. Perimeters for these burns varied from 0.9 to 10.98 miles in length. Ratios of size to perimeter were from 30/1 to 138/1. In other words for each mile of perimeter there were from 30 to 138 acres of burned ground. Each of the 11 frames represented about 50,000 acres. The recent burns identified and studied represented only 1.9 percent of the total area. The four Landsat MSS images that we studied represented approximately 6,400,000 acres per frame. With an average of 16,446 acres of recently burned ground per frame the area burned represented only 0.25 percent of the total. With periodic analysis of new imagery, monitoring would provide good data on the amount of burned ground possibly on an every 2 or 3 year basis. On these four frames the ratios of acreage to perimeter in miles varied from 4111 to 83/1. TM2 Burned Unburned 1.4730 2.0890 TM2 Table 3-A comparison of Landsat TM bands as simple comparative signatures for burned and unburned pixel samples from the Gund Ranch study site in central Nevada Site: Gund 74.80 69.05 TM3 129 Table ~elected vegetation index values for 10 northern Nevada burned and unburned range sltes1 RVI x B u 0.461 .444 .457 .491 .504 .466 .505 .583 .512 .695 .512 .024 0.463 .4n .505 .445 .462 .525 .668 .797 .475 .586 .540 .036 t p PVI B 5.381 9.938 6.583 3.183 7.619 3.920 6.734 16.542 2.821 10.189 7.287 1.307 -o.91 0.3873 ""NDVI u 4.573 8.226 10.119 .275 5.975 6.628 14.974 19.734 3.644 13.597 8.n4 1.864 TNDVI B u B u 0.010 .054 .019 .015 .025 .005 .026 .199 .022 .064 .044 .018 0.008 .064 .056 .044 .023 .033 .1n .281 .008 .130 .083 .028 0.715 .745 .511 .697 .725 .704 .725 .836 .692 .751 .710 .026 0.713 .751 .746 .675 .725 .725 .823 .884 .701 .794 .754 .020 -1.39 '0.1979 -2.45 0.0350 -1.85 0.0976 8. burned; U • unburned. 1 50 information about burns in the sagebrush/grass resulting in annual grassland. The difficulty extends to the concept of extending the signature to different sites (spatially) and different dates (temporally) and still have a signature based on a vegetation index or some other satellite dataderived index, that can describe specified environmental parameters and follow changes in them for monitoring purposes. This will require enhanced procedures for assessing the quantitative aspects of pixel components for the sites to be studied and monitored. Also, analysis and Actual =16.83-1.01 GVI + 65.73 NDVI-35.03 TNDVI R 2 =.9661 40 w t- a: m 30 (ij :I ~ 20 10 Table 6-Bum aaeage to bum perimeter miles and the ratio of aaeage to perimeter for wildfires in northern Nevada of 11 NASA color infrared photographs (panoramic)1 0, 0 10 20 30 40 Frame No. 50 Estimated BATE Figure 1-The linear relationship between actual cheatgrass and estimated cheatgrass for 10 western Nevada shrub-dominated range sites. CONCLUSIONS Perimeter PM 180 26.8 .9 19.78:1 162 85.8 4.7 18.26:1 111 106 64.4 333.8 1.4 8.0 46.0:1 41.73:1 88 50 648. 1,132. 579. 33.5 391. 5.5 8.2 4.5 .9 5.5 117.82:1 138.05:1 128.67:1 37.22:1 85.0:1 37 30 344.6 911. 4.6 10.98 74.91:1 82.97:1 Ex 4,549.9 413.63 55.18 5.0 71:1 85 81 72 Efforts directed toward using satellite remote sensing data to measure burned sites in the sagebrush/grass vegetation will eventually be routine data inputs to resource management and GIS. Also, such data can be used to monitor changes in rangelands resulting from management or from various perturbations. Satellite radiance data alone can provide information relative to burned vs. unburned sagebrush/grass sites in the Great Basin and Intermountain region. However, the difficult task is the business of establishing satellitebased signatures that can be used to extract meaningful Acre/ Acreage x 1 Scale-1:29,000; date: June 21, 1979. 130 Plant community Pinyon-Juniper/ Sagebrush Pinyon-Juniper/ Sagebrush Sagebrush Sage/Mountain Mahogany Sagebrush Sagebrush Sagebrush Sagebrush Pinyon-Juniper/ Sage Sagebrush Salt Desert Shrub Table 7-Total average total burn perimeter and the ratio of perimeter to acreage for wildfires in northern Nevada on four MSS images from 1973' Total bum acreage Frame Ruby Mountains Santa Rosa Coils Creek Black Rock Desert b i 26,344.5 12,840.1 15,206.1 11 1393.1 65,783.9 16,446.0 Total perimeter miles 317.59 202.384 365.274 210.028 b 1,095.3 i 273.8 Huete, A. R. 1988. A soil adjusted vegetation index (SAVI) Remote Sensing of Environment. 25: 295-309. Huete, A. R.; Jackson, R. D. 1987. Suitability of spectral indices for evaluating vegetation characteristics on arid rangelands. Remote Sensing of Environment. 23: 213-232. Jackson, R. D. 1983. Spectral indices in n-space. Remote Sensing of Environment. 13: 409-421. Jakubauskas, M. E.; Lulla, K. P.; Mausel, P. W. 1990. Assessment of vegetation change in a fire altered forest landscape. Photogrammetric Engineering and Remote Sensing. 56:371-377. Kauth, R. J.; Thomas, G. S.1976. The tasseled cap-a graphic description of the spectral-temporal development of agricultural crops as seen by Landsat. In: Proceedings of the symposium on machine processing of remotely sensed data; Purdue University: 4B-41--4B-51. Kowalik, W. S.; Lyon, R. J.P.; Switzer, P. 1983. The effects of additive radiance terms of ratios of Landsat data. Photogrammetric Engineering and Remote Sensing. 48:659-669. Marsh, S. E.; Lyon, R. J. P. 1980. Quantitative relationships of near-surface spectra to landsat radiometric data. Remote Sensing of Enviro~ent. 10: 241-261. Paltridge, G. W.; Barber, J. 1988. Monitoring grassland dryness and fire potential in Australia with NOAA/ AVHRR data. Remote Sensing of Enviro~ent. 25: 381-394. Raison, R. J. 1979. Modification of the soil enviro~ent by vegetation fires, with particular reference to nitrogen transformations: a review. Plant and Soil. 51: 73-108. Richardson A J.; Wiegand, C. L.1977. Distinguishing vegetation from soil background information. Photogrammetric Engineering and Remote Sensing. 43: 1541-1552. Tueller, P. T. 1989a. Sagebrush dominated vegetation of the Great Basin. In: Proceedings selected papers at the 38th annual meeting of the Society for Range Management: 24-30. Tueller, P. T. 1989b. Remote sensing technology for rangeland management applications. Journal of Range Management. 42(6): 442-453. Wright, H. A; Klemmedson, J. 0. 1965. Effect of fire on the bunchgrasses of the sagebrush-grass region in southem Idaho. Ecology. 46: 680-688. Yue-Hong, Chou; Minnich, R. A; Salazer, L. A; Power, J.D.; Dezzani, R. J.1990. Spatial autocorrelation of wildfire distribution in the Idyllwild Quadrangle, San Jacinto Mountain, CA. Photogrammetric Engineering andRemote Sensing. 56: 1507-1513. Young, J. A; Evans, R. A 1977. Population dynamics after wildfires in sagebrush grasslands. Journal of Range Management. 31:283-289. Young, J. A.; Evans, R. A. 1972. Downy brome-intruder in the plant succession of big sagebrush communities in the Great Basin. Journal of Range Management. 26:410-415. Young, J. A; Evans, R. A.; Eckert, R. E., Jr. 1969. Population dynamics of downy brome. Weed Science. 17: 20-26. Young, J. A.; Evans, R. A; Major, J. 1971. Alien plants in the Great Basin. Journal of Range Management. 25: 194-201. Ratio acreage/PM 82.951:1 63.444:1 41.629:1 54.246:1 i 60.57:1 '0.25 percent of the area burned. interpretation of remote sensing systems with higher spectral resolution may be helpful. Predictions of changes in shrub cover, perennial grass cover, annual grass cover, bare ground, and other surface soil parameters with satellite-derived remote sensing data will allow better management of these cold desert ecosystems. Establishment of descriptive signatures based on satellite data alone may be the most useful. However, the signatures for describing scene components may be improved in some cases by acquiring ground-based reflectance data to include in the model. REFERENCES Asrar, G.; Harris, T. R.; Lapitan, R. L.; Cooper, D. I. 1988. Radiative surface temperatures of the burned and unburned areas in a tall prairie. Remote Sensing of Enviro~ent.24:447-'57 Baret, F.; Guyot, G. 1991. Potentials and limits of vegetation indices for LAI and APAR assessment. Remote SensingofEnviro~ent. 35: 161-173. Clevers, J. G. P. W! 1989. The application of a weighted infrared-red vegetation index for estimating leaf area index by correcting for soil moisture. Remote Sensing ofEnviro~ent. 29: 25-37. DeFlon, J. G. 1986. The case for cheatgrass. Rangelands. 8: 14-17. Elvidge, C. D:; Lyon, R. J. P. 1984. Influence of substrate variation on the assessment of green biomass. Remote Sensing of Enviro~ent. 17: 265-279. Evans, R. A; Young, J. A 1977. Weed control-revegetation systems for big sagebrush-downy brome rangelands. Journal of Range Management. 30: 331-336. Freiberger, W. F. 1960. The international dictionary of applied mathematics. Princeton, NJ: Van Nostrand. 412 p. Gladwell, D. R. 1982. Application of reflectance spectrometry to clay mineral determination in geological materials using portable radiometers. In: Proceedings; international symposium on remote sensing of enviro~ent. Vol. 1. Fort Worth, TX; December 1982:6-10. Heilman, J. L.; Boyd. 1986. Soil background effects on the spectral response of a three-component rangeland scene. Remote Sensing of Environment. 19: 127-137. Huete, A R. 1986. Separation of soil-plant spectral mixtures by factor analysis. Remote Sensing of Environment. 19:237-251. 131