ABSTRACT

    Land cover change associated with agriculture has had an enormous effect on the structure and functioning of  temperate ecosystems. However, the empirical evidence for the impact of land use on ecosystem functioning at the regional scale is scarce. Most of our knowledge on land use impact has been derived  from simulation studies or from small plot experiments. In this article we studied the effects of land use on, 1) the seasonal dynamics and 2) the interannual variability of the Normalized Difference Vegetation Index (NDVI), a variable linearly related to the fraction of the photosynthetically active radiation (PAR) intercepted by the canopy. We also analyzed the relative importance of environmental factors and land use on the spatial patterns of NDVI. We compared three cultivated land cover types against native grasslands. The seasonal dynamics of NDVI was used as a descriptor of ecosystem functioning. To reduce the dimensionality of our data we analyzed  the annual integral (NDVI-I), the date of maximum NDVI (DMAX) and the quarterly average NDVI. These attributes were studied for 7 years and for 346 sites distributed across eastern Colorado (USA).

    Land use did modify  ecosystem functioning at the regional level in eastern Colorado. The seasonal dynamics of NDVI, a surrogate for the fraction of PAR intercepted by the canopy, were significantly altered by agricultural practices.  Land use modified both the NDVI integral and the seasonal dynamics of this spectral index. Despite the variability within land cover categories, land use was the most important factor in explaining regional differences of the NDVI attributes analyzed. Within the range of environmental conditions found in eastern Colorado, land use was more important than mean annual precipitation, mean annual temperature and soil texture in determining the seasonal dynamics of NDVI.
 

Keywords: land use, remote sensing, primary production, grasslands, Normalized Difference Vegetation Index. NOAA/AVHRR.

 INTRODUCTION

    Land use is the component of human-driven global change that has been occurring for the longest period of time. Land cover change associated with agriculture has had an enormous impact on the structure and functioning of  temperate ecosystems.  For example, agricultural practices have reduced soil carbon stocks by 35% in the U.S. Central Grassland Region over the past 50 years (Burke et al. 1989). On a world basis, Houghton (1999) reported carbon losses of 124 Pg between 1850 and 1980. Pielke et al. (1997) and Stohlgren et al. (1998) showed that land use pattern has a critical influence on mesoscale atmospheric processes and hence on local climate. Stohlgren et al. (1998) presented evidence showing that land use changes had reduced summer temperatures in the Front Range area of Colorado (USA) during the last several decades. They also suggested that atmospheric changes associated with the replacement of semiarid grasslands has increased water discharge and modified the dynamics of tree populations in mountain environments. Further effects of land use on climate have been reported by Bonan (1997) and Pan et al. (1999).  Mosier et al. (1991) showed significant changes in trace gas fluxes associated with cultivation of a semiarid grassland: crop production increased the release of CH4 and N2O, two greenhouse gases. The additions of CO2 and other trace gases to the atmosphere may represent significant feedbacks to global climate change, both from an ecological and a human economic perspective (Sala & Paruelo 1997).

    Land use has increased the rate of extinction of species not only by replacing natural ecosystems, but also by changing the disturbance regime. Leach & Givnish (1996) showed that 8 to 60% of the original plant species of remnant prairie sites were lost after 50 years due to landscape fragmentation or fire suppression. Such changes can have local, regional and global consequences, including loss of the soil fertility, soil erosion, reduction of biological diversity, hydrological changes, climatic alteration and modification of the atmospheric composition.

    The Central Grassland region of North America covers more than 3.02 x 106 km2 and has been heavily influenced by agriculture (Lauenroth et al. 2000).  As a consequence of this intensive agriculture, the region contains the most important corn and wheat production areas of the  US and Canada. Within the Central Grassland region, the shortgrass steppe type is located on the western edge, characteristically receiving less than 400 mm of precipitation (Lauenroth & Milchunas 1992). The shortgrass steppe was the dominant land cover before European settlement. These grasslands are dominated by two C4 species: Bouteloua gracilis (blue grama) and Buchloë dactyloides (buffalo grass).  Present land-cover  includes a mix of native grasslands, rain fed crops (mainly wheat) and irrigated crops (mainly corn) (Figure 1). Beginning in 1985, some agricultural fields have been converted into pastures under the Conservation Reserve Program (CRP). Under this program landowners are paid to revegetate highly erodible land for a 10 year period (Osborn 1993). The species used in CRP areas of Colorado include primarily C3 grasses and legumes.

    What are the consequences of replacing native grasslands with crops or sowed pastures on ecosystem functioning at the regional scale? How does the replacement of the dominant plant functional type (C4 grasses by C3 grasses) or irrigation modify the exchange of energy and matter of the ecosystem? Empirical evidence for the impact of land use on ecosystem functioning at the regional scale is scarce. Most of our knowledge on land use impact has been derived  from simulation studies (i.e., Burke et al. 1991; Pielke et al. 1997; Stohlgren et al. 1998; Baron et al. 1998; DeFries et al. 1999) or from small plot experiments (i.e. Mosier et al. 1991; Robles & Burke 1997). In this paper, we analyzed the effect of land use on the seasonal dynamics of the Normalized Difference Vegetation Index (NDVI), a variable positively related to the fraction of the photosynthetically active radiation (PAR)  intercepted by the canopy (Dye & Goward 1993; Sellers et al. 1994). The NDVI dynamics have been extensively used as an estimator of primary production (Goward et al.1985; Tucker et al. 1985; Box et al. 1989; Burke et al. 1991; Hobbs 1995;  Paruelo et al. 1997). We focused specifically on the study of the effects of land use on 1) the seasonal dynamics, and 2) the interannual variability of NDVI; and on the analysis of the relative importance of environmental factors and land use on the spatial patterns of NDVI. Our null hypothesis was that land use does not modify ecosystem functioning. We compared three cultivated land cover types to native grasslands. We considered grazed grasslands to approximate native systems, since this area has a long evolutionary history of grazing by large generalist herbivores (Milchunas et al. 1988).
 

 MATERIAL AND METHODS

    We used the seasonal dynamics of the Normalized Difference Vegetation Index (NDVI) calculated from NOAA/AVHRR LAC data as a descriptor of ecosystem functioning.  McNaughton et al. (1989) showed that primary production is an integrative descriptor of ecosystem functioning. It has been shown that NDVI is a powerful tool to describe the dynamics of primary production (Malingreau 1986; Lloyd 1990; Loveland et al. 1991; Fischer 1994a, b; Tucker et al. 1985; Prince 1991; Paruelo et al. 1997). NDVI is computed from the reflectance in channel 1 (red, 580-680 nm) and channel 2 (near infrared, 725-1100 nm) from the AVHRR sensors on board the NOAA satellites ( NDVI  =  (Channel 2 - Channel 1) / (Channel 1 + Channel 2)). The use of NOAA/AVHRR data to characterize vegetation at regional and global scales is receiving a great deal of attention (Running et al. 1995; Nemani and Running 1997; DeFries and Beldwar 2000).
 NDVI data were obtained from the EROS Data Center in Sioux Falls (ND). We obtained biweekly images per year for 7 years (1990-1996). Each image corresponded to the maximum NDVI composites (Holben 1986) of Large Area Coverage (1.1 km) images from the daily orbital passes of the NOAA satellite (Eidenshink 1992). Images were geometrically registered to the Lambert Azimuthal Equal Area map projection.

    Using a Colorado land use database, we identified 346 sites that corresponded to four different land use classes: irrigated crops (I) (50 sites), non-irrigated crops (NI) (129 sites), rangelands (RL) (96 sites) and Conservation Reserve Program fields (CRP) (71 sites) (Figure 1). Sites corresponded to areas large enough to minimize misregistration errors. The proportion of eastern Colorado occupied by the four land covers was 7.8%, 20.3%, 63.14% and 5.0% for irrigated crops, non-irrigated crops, rangelands  and Conservation Reserve Program fields. The land use database was constructed by digitizing county maps of land use provided by the Natural Resource Conservation Service (NRCS) in 1985. We manually updated these maps to accurately represent new Conservation Reserve Program lands and other land use changes by incorporating the most recent records from each county NRCS office. The allocation of land among the different land cover types showed almost no change throughout the period studied. Each site was at least 400 ha (4 pixels in the NDVI image) to allow for a proper localization within the images. For each site we obtained the NDVI values for the 21 images of each of the 7 years available. Our basic data set consisted, then, of a matrix of 21 variables times 7 years and 346 sites (2422 total observations).   We performed discriminant analysis on this matrix, grouping the cases by land cover (4 groups). We used the Mahalanobis (Afifi & Clark 1990) distance to identify differences in the seasonal dynamics of NDVI among land uses and years. From the seasonal curves of NDVI, we derived 5 attributes that describe basic aspects of ecosystem function: the fraction of the total intercepted PAR (IPAR) [the annual integral of NDVI (NDVI-I)] and the seasonality of the fraction of PAR intercepted [the day of maximum NDVI (DMAX) and the NDVI average for three periods of the year (Q_1: Julian dates 1-153, late winter-spring; Q_2: Julian dates 154-237, summer; Q_3: Julian dates 238-365, fall-early winter)] (Figure 2). Analyses were performed both for individual years and the 7-year averages.

    Eastern Colorado is an almost ideal area to study the impact of land use because of the homogeneity of the potential vegetation (a shortgrass steppe) and the smooth environmental gradient. We characterized the physical environment of each site from the mean annual precipitation (MAP), mean annual temperature (MAT) and the percentage of clay of the soils (CLY) (Figure 1). Environmental data were interpolated from existing databases for the US Central Grassland region compiled by the Shortgrass Steppe Long Term Ecological Research project. For each of the 5 attributes of the NDVI curves we used the three environmental variables as covariates in an ANCOVA. To capture differences in the seasonal dynamics, we also studied the differences among sites in the quarterly NDVI average using MANOVA.
 

 RESULTS & DISCUSSION

Effect of landuse on the seasonal dynamics of NDVI

    The analysis of the 7-year average of the seasonal curves of NDVI (Discriminant Analysis) showed  significant differences among the four land use categories studied (Wilks’s Lambda test: 0.24981,  F (63,7158)=67.214, p<0.001). In the hyper space defined by the 21 dates, irrigated crops showed the farthest distance from any other land use category (Table 1). The most distinct land cover types were the two crop systems (Figure 3, Table 1). The analysis of each individual year showed similar distances among land use categories as the average curve (results not shown).

    The analysis of the NDVI integral (NDVI-I)  and the date of maximum NDVI (DMAX) also showed differences among land cover types (Figure 4). Land use categories differed in the integral and timing of the NDVI peak (Figures 3 and 4).  The average NDVI integral was 46% and 7% higher in irrigated and rain fed crops respectively than in grasslands (Figure 4a) (ANCOVA, F=139, d.f. 3, 339, p<0.001, see Table 2). CRP areas and rangelands did not differ significantly in their NDVI-I (Figure 4a).   Land use had an important effect on the seasonality of PAR absorption (Figure 3). On average, NDVI peaked more than 20 days later in irrigated fields than in native grasslands (Figure 4b) (ANCOVA, F=71,  d.f. 3, 339, p<0.001, see Table 2). Maximum NDVI occurred 20 days earlier in non-irrigated fields than in grasslands (Figure 4b) (p<0.001). CRP areas did not differ significantly from grasslands in the timing of the NDVI peak.

    Multiple Analysis of Variance (MANOVA) of the seasonal NDVI averages also indicated a significant effect of land use on the seasonal dynamics of NDVI ((Wilks’s Lambda test: 0.20578 , F (9,827)=84.121, p<0.001). During late winter and spring (Q_1), native grasslands showed the lowest NDVI (Figure 5). CRP areas and native rangelands presented a higher average NDVI than non-irrigated crops during the summer (Q_2). During the last part of the year (Q_3), CRP areas, native grasslands and non-irrigated crops showed the same average NDVI (Figure 5). Irrigated crops showed the highest NDVI during the three portions of the year (Figure 5, p<0.001). The analysis of the NDVI-I, DMAX and seasonal NDVI averages for each of the seven years available showed the same pattern as the average values presented here.

    The major change in the dynamics of NDVI was associated with irrigation. On average, adding water increased the NDVI-I, a measure of the amount of PAR intercepted by the canopy by 47% compared to native grasslands (Figure 4a). The replacement of grasslands by wheat-fallow systems produced a modest increase in the NDVI integral (7%) (Figure 4a). Of course, this increase is not only due to the replacement of the dominant species and irrigation. Large amounts of nitrogen are added to crops annually. Even under rain fed conditions nitrogen significantly increases carbon gains in this area. Lauenroth et al. (1978) observed an increase of up to 57% in primary production in the shortgrass steppe when nitrogen was added. In these rangelands, adding water and nitrogen increased aboveground net primary production by a factor of almost 10.

    Using a simulation approach, Baron et al. (1998) reported, for the South Platte basin in Colorado, Wyoming and Nebraska, a 237% and a 20% increase in net photosynthesis when grasslands are replaced by irrigated and non-irrigated crops respectively. Baron et al.’s (1998) simulation analysis reported that net carbon gain increased by a factor of 4 due to the replacement of grasslands by irrigated crops in a particular watershed of the area. Estimates of aboveground net primary production (ANPP) for wheat-fallow systems in eastern Colorado and Kansas showed that wheat is more productive than grasslands in areas with mean annual precipitation lower than 570 mm (Lauenroth et al. in press). The observed differences in ANPP between wheat and natural grasslands for Colorado counties receiving 320 mm.yr-1 is 154 g.m-2 (a 142%)  CRP areas did not differ significantly from grasslands in total light interception.

    Our approach ignored completely the differences in belowground allocation between crops and grasslands. Milchunas and Lauenroth (1992) reported for the shortgrass steppe similar values for above and belowground net primary production. In contrast, the belowground net primary production for crops is a small fraction of ANPP (less than 20%) (Fischer 1983, Anderson  1988, Eghball & Maranville 1993). Knowing the amount of carbon allocated belowground is crucial to estimating the carbon balance of the different land cover types.
Effect of land use on the interannual variability of NDVI

    Irrigation significantly reduced the interannual variability of both the NDVI-I and the date of maximum NDVI (Figure 6). We used the Coefficient of Variation (CV=standard deviation/mean) to characterize the interannual variability of the NDVI integral (NDVI-I) and the date of maximum NDVI (DMAX). Irrigated crops displayed a significantly lower interannual variability (ANOVA, F=16, d.f. 3,342, p<0.001) than the other land use categories (Figure 6a). The interannual variability of the day of maximum NDVI  also differed among land cover types (ANOVA, F=19, d.f. 3,342, p<0.001) (Figure 6b). Irrigated crops showed the lowest CV of DMAX and non-irrigated crops the highest. Summer was the least variable season and the spring the most variable for all the land use categories (Figure 7). For every portion of the year, irrigated crops displayed the lowest interannual variability.

    Paruelo and Lauenroth (1998) and Jobbagy et al. (in press) reported a decrease in the interannual variability of the NDVI integral as water availability increases for native grasslands of North American and semiarid steppes of Patagonia.  Irrigation may contribute to decoupling plant production and precipitation, generating a more stable system. Non-irrigated crops showed a similar interannual variability to grassland and CRP areas. In this case the predominant cultivation system (wheat-fallow) may contribute to increase interannual variability. Our data cannot separate fallow and wheat areas because the fields, due to the strip cropping, are always smaller than the size of our sites (400 ha). The date of maximum NDVI was between 2 and 3 times more variable than the integral of the NDVI. Paruelo and Lauenroth (1998) found the same pattern for unmodified areas across the Central Grassland region of US.

    All the land use categories displayed the highest interannual variability during late winter and spring (Q_1) (Figure 7). In this period the difference between irrigated crops and the other land use categories was lower than during the rest of the year. During late winter and early spring, water is not the main determinant of light interception and temperature exerts an important control over the start of the growing season (Paruelo and Lauenroth, 1998).
Environmental vs land use controls of NDVI dynamics

    Sites corresponding to different land use categories did not differ significantly in mean annual precipitation (MAP) or mean annual temperature (MAT) (ANOVA, n.s). MAP ranged between 272 and 494 mm and MAT between 7 and 12.5 ?C. These values corresponded to the full range of spatial climatic variability observed in eastern Colorado (Figure 1). Irrigated crops were found on finer textured soils than the rest of the land covers. Irrigated crops sites showed a significantly higher clay content in their soils (24.98%) than the other categories (19.27%, 17.20%, 18.29% respectively for non-irrigated crops, grasslands and CRP areas) (F 12.54, d.f. 3,342, p<0.0001). Land use was more important than any environmental variable in accounting for spatial differences in the integral of the NDVI, the date of maximum NDVI or the quarterly average NDVI. When MAT, MAP or clay content (CLY) were included in an ANCOVA their effect on the NDVI attributes was always lower than the effect of  land use.(Table 2).

    At the regional scale the interannual changes of the mean NDVI-I of grasslands, non-irrigated crops and CRP sites were highly correlated (Table 3). This indicates a strong environmental control of the temporal changes of the NDVI-I of the three land cover types operating at a regional scale. Water availability has been referred to as the most frequent control on carbon gains in the shortgrass steppe, both in time and space (Sala et al. 1988; Lauenroth & Sala 1992). Non-irrigated crops always had a higher NDVI-I than grassland and CRP areas, but the interannual changes in NDVI-I are similar to those observed in grasslands and CRP fields. The average NDVI of irrigated crops was not correlated with the NDVI in grasslands suggesting that irrigation decouples this system from the environmental factors that operate on grasslands, CRP areas and non-irrigated crops.
 

 CONCLUSIONS

    We rejected our null hypothesis. We found that land use did modify ecosystem functioning at the regional level in eastern Colorado. The seasonal dynamics of the NDVI and the annual integral, a good estimator of the fraction of light intercepted by the canopy, was significantly altered by agricultural practices (Figure 3). Despite the variability within land cover categories in the NDVI integral and seasonality (Figure 4), land use was the most important factor in explaining regional differences of the NDVI attributes analyzed (Table 2). Within the range of environmental conditions found in eastern Colorado, land use was more important than mean annual precipitation, mean annual temperature and soil texture in determining the seasonal dynamics of NDVI and hence carbon gains. The replacement of plant communities dominated by C4 species (the shortgrass steppe) by C3 species, either totally (wheat) or partially (CRP areas), significantly increased the NDVI during the first quarter of the year (Figure 5).  Considering the proportion of eastern Colorado occupied by the different land covers considered, land use increased the NDVI integral  9.5%, 2.2% and 5.4% during the first, second and third quarter respectively.

    Our results represent one of the few pieces of empirical evidence of the impact of land use and land-cover change on ecosystem functioning at the regional scale.  Irrigated agriculture transformed the shortgrass steppe of eastern Colorado into a system that, based on the dynamics of NDVI, resembles a tallgrass prairie in Minnesota (Paruelo and Lauenroth, 1995). Colorado non-irrigated agricultural fields had NDVI dynamics similar to a sagebrush steppe of eastern Oregon (Paruelo and Lauenroth, 1995).
 
 
 

 ACKNOWLEDGMENTS: Jerry Steenson, Martha Coleman and Maria C. Giallorenzi provided important technical support. This work was supported by a grant from  the Environmental Protection Agency, the NSF-funded Shortgrass Steppe Long Term Ecological Research Project, Fundacion Antorchas and the Inter American Institute for Global Change Studies (ISPIII-077).

Afifi A, Clark V (1990) Computer-aided Multivariate Analysis: New York, Chapman & Hall, 505 p.

Ansderson E.L. (1988) Tillage and N fertilization effects on maize root growth and root:shoot ratio. Plant and Soils ,108, 245-251.

Baron J, Hartman M, Kittel T, Band L, Ojima D, Lammers R (1998) Effects of land cover, water redistribution and temperature on ecosystem processes in the South Platte Basin: Ecological Applications, 8, 1037-1051.

Burke IC, Kittel TGF, Lauenroth WK, Snook P, Yonker CM, Parton WJ (1991) Regional Analysis of the Central Great Plains: Bioscience, 41, 685-92.

Burke IC, Yonker CM, Parton WJ, Cole CV, Flach K, Schimel DS (1989) Texture, climate, and cultivation effects on soil organic content in U.S. grassland soils: Soil Science Society of America Journal, 53, 800-5.

Dye DG, Goward SN (1993) Photosynthetically active radiation absorbed by global land vegetation in August 1984: International Journal  of  Remote Sensing, 14, 3361-3364.

Eghball B, Maranville J (1993) Root development and nitrogen influx of corn genotypes grown under combined drought and nitrogen stresses. Agronomy Journal , 85, 147-152.

Eidenshink JC (1992) The 1990 conterminous US AVHRR data set: Photogrammetric Engineering and Remote Sensing, 58, 809-13.

Fischer RA (1983) Wheat. In: Potential productivity of field crops under different environments, pp 129-154. International Rice Research Institute, Manile, Phillipines.

Fischer A (1994a) A model for the seasonal variations of vegetation indices in coarse resolution data and its inversion to extract crop parameters: Remote Sensing of Environment, 48, 220-30.

Fischer A (1994b) A simple model for the temporal variations of NDVI at regional scale over agricultural countries. Validation with ground radiometric measurements: International Journal of
Remote Sensing, 15, 1421-46.

Holben B (1986) Characteristics of maximum value composite images from temporal AVHRR data: International Journal of Remote Sensing, 7, 1417-34.

Jobbágy EG, Sala OE, Paruelo JM (1999) Patterns and controls of primary production in the patagonian steppe: a remote sensing approach: Ecology, in press.

Lauenroth W, Burke I, Paruelo J (2000) Patterns of production of winter wheat and native grasslands in the Central Grassland region of the United States: Ecosystems in press.

Lauenroth WC, Dodd JL, Sims PL (1978) The effects of water and nitrogen induced stress on plant community structure in a semiarid grassland: Oecologia, 36, 211-22.

Lauenroth WK, Milchunas DG (1992), Short-Grass Steppe. In: Ecosystems of the World: Natural Grasslands, Volume 8A: (ed Coupland RT), pp 183-226.  Elsevier, Amsterdam.

Lauenroth WK, Sala OE (1992) Long-term forage production of North American shortgrass steppe: Ecological Applications, 2, 397-403.

Lauenroth W K, Burke IC, Gutmann MP (2000) The structure and function of ecosystems in the central North American grassland region. Great Plains Research in press

Leach MK, Givnish TJ (1996) Ecological determinants of species loss in remnant prairies: Science, 273, 1555-8.

Lloyd D (1990) A phenological classification of terrestrial vegetation cover using shortwave vegetation index imagery: International Journal of Remote Sensing, 11, 2269-79.

Loveland TR, Merchant JW, Ohlen DO, Brown JF (1991) Development of a land cover characteristics database for the conterminous US: Photogrametric Engineering and Remote  Sensing, 57, 1453-63.

Malingreau JP (1986) Global vegetation dynamics, satellite observations over Asia: International Journal of Remote Sensing, 7, 1121-46.

McNaughton SJ, Oesterheld M, Frank DA, Williams KJ (1989) Ecosystem-level patterns of primary productivity and herbivory in terrestrial habitats: Nature, 341, 142-4.

Milchunas DG, Lauenroth WK (1992) Carbon Dynamics and Estimates of Primary Production by Harvest, C-14 Dilution, and C-14 Turnover, Ecology, 73, 593-607.

Milchunas DG, Sala OE, Lauenroth WK (1988) A generalized model of the effects of grazing by large herbivores on grassland community structure: The American Naturalist, 132, 87-106.

Monteith J L, (1981) Climatic variation and the growth of crops: Quarterly Journal of the Royal Meteorological Society, 107, 749-74.

Mosier A, Schimel D, Valentine D, Bronson K, Parton W (1991) Methane and nitrous oxide fluxes in native, fertilized and cultivated grasslands: Nature, 350, 330-2.

Nemani R, Running S (1997) Land cover characterization using multitemporal red, near-IR, and thermal-IR data from NOAA/AVHRR: Ecological Applications, 7, 79-90.

Osborn T (1993) The conservation reserve program: status, future, and policy options: Journal of Soil and Water Management, 48, 272-279.

Paruelo JM, Epstein HE, Lauenroth WK, Burke IC (1997) ANPP estimates from NDVI for the Central Grassland region of the US: Ecology, 78, 953-8.

Paruelo JM, Lauenroth WK (1995) Regional patterns of NDVI in North American shrublands and grasslands:  Ecology, 76, 1888-98.

Paruelo JM, Lauenroth WK (1998) Interannual variability of the NDVI curves and their climatic controls in North American shrublands and grasslands: Journal of Biogeography, 25, 721-733.

Pielke RA, Lee TJ, Copeland JH, Eastman JL, Ziegler CL, Finley CA (1997) Use of USGS-provided data to improve weather and climate simulations: Ecological Applications, 7, 3-21.

Prince SD (1991) A model of regional primary production for use with coarse resolution satellite data: International Journal of Remote Sensing, 12, 1313-30.

Robles M, Burke I (1997) Legume, grass and Conservation Reserve Programs effects on soil organic matter: Ecological Applications, 7, 345-357.

Running SW, Loveland TR, Pierce LL, Nemani RR, Hunt ER (1995) A remote sensing based vegetation classification logic for global land cover analysis: Remote Sensing of Environment, 51, 39-48.

Sala OE, Parton WJ, Joyce LA, Lauenroth WK (1988) Primary Production of the central grassland region of the United States: Ecology, 69, 40-5.

Sala OE, Paruelo JM (1997) Ecosystem services in grasslands, Ecosystem services: Washington, DC, Island Press.

Sellers PJ, Tucker CJ, Collatz GJ, Los SO, Justice CO, Dazlich DA, Randall DA (1994) A global 1o by 1o NDVI data set fro climate studies. Part 2: The generation of global fields of terrestrial biophysical parameters from the NDVI: International Journal of Remote Sensing, 15, 3519-45.

Stohlgren T, Chase T, Pielke R, Kittel T, Baron J (1998) Evidences that local land use practices influence regional climate, vegetation and stream flow patterns in adjacent natural areas: Global Change Biology, 4, 495-504.

Tucker CJ, Townshend JR, Goff TE (1985) African land-cover classification using satellite data: Science, 227, 369-75.
 

2. Department of Forest Science. Colorado State University. Fort Collins, CO 80523, USA.

3. Department of Rangeland Ecosystem Science. Colorado State University, Fort Collins, CO 80523, USA up