Classification and predictive vegetation mapping of rangelands

Objectives: Description of vegetation patterns and diversity, analysis of herbaceous and ligneous biomass production, predictive vegetation mapping

Contact Person: Prof. Dr. Andreas Bürkert and Prof. Dr. Eva Schlecht, Institute of Crop Science, University of Kassel, Germany

Email: tropcrops@uni-kassel.de

Scientific staff: Dr. Katja Brinkmann, Dr. Uta Dickhoefer, Dr. Annette Patzelt

Study area: Al Jabal al Akhdar

Duration: 2007-2009

Classification of rangeland vegetation

Rangelands in the rural areas of Oman are valuable wildlife and plant habitats and still form the major source of nutrition for livestock, which constitute a chief basis of rural community income. The rangelands of Al Jabal al Akhdar Mountain belong to the WWF’s Global 200 ecoregion ‘Arabian Highland Woodlands and Shrublands' and are a local centre of plant endemism in the Sultanate of Oman with a particular high species diversity.
We investigated the species compositions and several environmental variables such as browsing rate, sward structure and distance to the settlement for 62 samples (20x30 m) along an altitudinal and a grazing gradient. Classification analysis (two-way cluster analysis and indicator species analysis) and ordination methods (Canonical Variate Analysis) was used to define vegetation types and to identify underlying environmental gradients.

Photo 1. View from the ungrazed Ras al Kabul plateau with Olea europaea and Sideroxylon mascatense in the tree layer and the grasses Fingerhuthia africana and Heteropogon contortus

Our statistical analyses revealed five different vegetation types along the environmental gradients, which were defined on the basis of the dominant species: The Sideroxylon mascatense–Dodonaea viscosa group (Photo 2) on grazed and the Olea europaea–Fingerhuthia africana group (Photo 1) on ungrazed plateau sites at 2 000 m a.s.l., the Ziziphus spina-cristi–Nerium oleander group on wadi sites, and the Moringa peregrina–Pteropyrum scoparium group at 1 200 m a.s.l, and the Acacia gerrardii–Leucas inflata group at 1 700 m a.s.l..

Photo 2. Sideroxylon mascatense - Dodonaea viscosa group
on disturbed places on the Sayq plateau

The plant species richness followed a unimodal distribution along the altitudinal gradient with the highest number of species at an altitude between 1 400 and 1 600 m a.s.l.
The grazing intensity decreased with increasing distance from human settlements and with proximity to the steep, inaccessible slopes at the peripheries. Altogether, 27 % of the species showed a high degree of browsing damage. The highest browsing intensity was found on the Sayq plateau, reflected by the puny development of the ground layer and a severely impoverished species composition (Photo 2). Conservation strategies beyond mere fencing should be worked out in order to maintain wildlife and plant habitats at Al Jabal al Akhdar.

Predictive vegetation mapping

We choose an inductive GIS modelling approach for the predictive vegetation mapping at Al Jabal al Akhdar based on the analysis of field collected data. The environmental data for such a modeling approach were either obtained from existing digital datasets (geology map, grazing intensity) or derived from a Digital Elevation Model (altitude) and a recent LANDSAT satellite image (NDVI).  The spatial information on the vegetation attribute of interest (type) was derived from the vegetation samples and the classification results of a previous analysis (Brinkmann et al, 2008), as well as further field observations (70 additional vegetation sample points and 123 validation plots).
A classification tree analysis was performed with the vegetation sample data set to examine the presence/absence of the vegetation types against six selected spatial environmental variables (topographic location, geology type, NDVI, altitude, distance to the settlement, grazing intensity). The logic contained in the decision rules derived by the statistical analysis was used to build a decision tree classifier for the predictive vegetation mapping.
All spatial information was resampled to a 5m Grid within ArcGIS 9.2 and exported as separate image files for a spatial subset of 10 km². To model the distribution of vegetation types we used the interactive decision tree tool of the remote sensing software ENVI 4.3 and applied the decision rules (Figure 2) to the multiple image dataset. This modeling approach classified 97 % of the plant communities within the learning data set correctly

Figure 2. ENVI’s decision tree classifier for the six rangeland vegetation types and. the oasis agriculture areas of the Jabal al Akhdar (simplified). SD = Sideroxylon mascatense - Dodonaea viscose; OF = Olea europaea - Fingerhuthia africana, AL = Acacia gerrardii - Leucas inflata, MP = Moringa peregrina - Pteropyrum scoparium; ZN = Ziziphus spina-christi - Nerium oleander, JET = Juniperus- Euphedra-Teucrium zone above 2200 m a.s.l

Biomass production of rangelands

Ligneous and herbaceous vegetation on grazed and ungrazed sitesat Al Jabal al Akhdar was analyzed to evaluate the possibilities of improved pasture management to maintain fodder production. Foliar biomass ranged between 3 – 6 t DM/ha on the grazed and ungrazed plateau areas, and reached 41 t DM/ha on mountain slopes and shallow valleys. Herbaceous yields changed over the seasons and contributed < 11% to total biomass yields, stressing the importance of ligneous foliage for livestock nutrition in particular during dry periods. While vegetation composition differed clearly between grazed and ungrazed sites and edible biomass yields were lower on mountain pastures, canopy cover and biomass yields were similar in a 15 year old exclosure and on a naturally ungrazed mountain plateau.
Improved pasture management incorporating traditional grazing practices of farmers may allow to maintain plant stands on grazed pastures while satisfying the needs of livestock keepers in a supplement-added rangeland-based land use system

Publications

	Brinkmann, K., Gebauer, J., Patzelt, A., Schlecht, E. & Buerkert, A. 2007. Vegetation and biodiversity on Jebel al Akhdar. Pride. Al Roya Press and Publishing House, Muscat, Oman, pp. 90-96. [Arabic version pp. 94-98].
	Brinkmann, K, Dickhoefer, U., SChlecht, E. & Buerkert, A. 2008. Classification of rangeland vegetation and modelling of vegetation patterns st the Jabal al Akhdar mountain, Northern Oman. In: Proceedings Multifunctional Grasslands in a Changing World, Vol. I. XXI International Grassland Congress, VIII International Rangeland Congress, 67.
	Brinkmann, K, Patzelt, A., Dickhoefer, U., Schlecht, E. & Buerkert, A. 2009. Vegetation patterns and diversity along an altitudinal and a grazing gradient in the Jabal al Akhdar mountain range of northern Oman. Journal of Arid Environments 73, 1035-1045.
	Brinkmann, K., Patzelt, A., Schlecht, E. & Buerkert, A. 2011. Use of environmental predictors for vegetation mapping in semi-arid mountain rangelands and the determination of conservation hotspots. Journal of Applied Vegetation Science, 14, 17-30.
	Brinkmann, K., Dickhoefer, U., Schlecht, E. & Buerkert, A. 2011. Quantification of aboveground rangeland productivity and anthropogenic degradation on the Arabian Peninsula using Landsat imagery and field inventory data. Remote Sensing of Environment, 115, 465-474.