Use of environmental predictors for vegetation mapping in semi-arid mountain rangelands and the determination of conservation hotspots

Abstract

Question: Can we predict the spatial distribution of plant communities in semi‐arid rangelands based on a limited set of environmental variables? Where are priority areas for conservation located? Location: Al Jabal al Akhdar, Sultanate of Oman. Methods: A Classification Tree Analysis (CTA) was used to model the presence/absence of seven rangeland communities and agricultural areas based on seven selected environmental predictor variables. The latter were either obtained from existing digital datasets or derived from a digital elevation model and satellite images, whereas the grazing intensity was spatially modelled with the kernel density estimation technique. The resulting decision rules of a CTA were applied for predictive mapping within the study area (400 km2, resolution of 5 m) by means of ENVI's decision tree classifier. Plant communities of natural rangelands were subsequently evaluated to determine priority areas for nature conservation. Results: Altitude, grazing intensity and landform revealed the highest predictive power. Most of the rangelands were predicted as Sideroxylon–Oleetum. The overall classification accuracy was 89%, whereby agricultural areas and the Ziziphus spina‐christi‐Nerium oleander community at wadi sites had no misclassification. Inaccuracies occurred mainly because of low sample numbers and errors in available maps of predictor variables. The highest rank for nature conservation was observed for the Teucrio‐Juniperetum occupying 20% of the study area. Conclusions: Vegetation mapping using CTA is a valuable tool for rangeland monitoring and identification of key representative areas for nature conservation. An extrapolation of the model used might be feasible to regions adjacent to the central Hajar Mountains.

Predictive Mapping of Dwarf Shrub Vegetation in an Arid High Mountain Ecosystem Using Remote Sensing and Random Forests

“…Studies on species separation mainly use hyperspectral data often under experimental conditions [9]. As to expected, elevation was an important predictor apart from spectral information gained by remote sensing in our study region and in all studies in mountain areas [4,5,8]. Within this area, 79 km 2 (0.9%) are covered by water bodies, of which, the lakes Tus Kul and Sasyk Kul (west of Alichur), and the eastern end of Lake Sarez (in the north-west), as well as Uch Kul (in the south) make the largest portion.…”

“…Compared with other studies about species composition, our results give a considerably accurate picture about the distribution of single species. The other studies on arid and similar mountain ecosystems do not map single species, but species composition or vegetation classes [4,5,7,8], plant richness [47]; however, using similar indices. Studies on species separation mainly use hyperspectral data often under experimental conditions [9].…”

“…Despite the wide application of remote sensing data, arid regions and high mountains are underrepresented in studies, especially those focusing on more detailed aspects, such as species composition or species distribution [4][5][6][7][8]. However, even most of these studies do not work on single species, but mainly on plant communities.…”

Prediction of senescent rangeland canopy structural attributes with airborne hyperspectral imagery

“…Another, perhaps complementary approach includes synoptic mapping of canopy structural attributes with remote sensing-based data (Pickup et al 1994;Toevs et al 2011). Remote sensing techniques have been developed when plants are mostly green to help meet rangeland forage information needs (Tucker et al 1983;Guerschman et al 2009;Brinkmann, Dickhoefer, et al 2011;Brinkmann, Patzelt, et al 2011), but techniques are lacking for senescent canopies postgrowing-season (Tucker 1978;Roberts et al 1993;Roberts et al 1998;Rundquist 2002). The need to assess rangelands following grazer removal in autumn (Bement 1969) and the potential for hyperspectral data applications are clear; however, specific approaches to understanding and modeling senescent mixed-grass prairie need to be evaluated for expansive rangeland landscapes in the northern Great Plains.…”

“…Several groups have investigated how vegetation indices (VIs) derived from airborne hyperspectral data can be modeled to estimate biomass, N content, and leaf area during the growing season (Jacquemoud et al 1995;Kokaly and Clark 1999;Beeri et al 2007). For example, the best predictor for savannah and mixed-grass prairie PV during the growing season was the Normalized Difference Vegetation Index (NDVI) derived from hyperspectral imagery (Gamon et al 1997;Beeri et al 2007;Guerschman et al 2009;Brinkmann, Dickhoefer, et al 2011;Brinkmann, Patzelt, et al 2011). Guerschmann et al (2009 also found that spectral data available from the Moderate Infrared Spectrometer (MODIS), the simple ratio (ρ 2130 /ρ 1640 ), was the optimum VI for estimating fractional cover of NPV.…”