The Sahel region has often been acclaimed as one of the “hot spots” of environmental change. Degradation of environmental conditions was accelerated by droughts and an overall decrease in precipitation. The resulting loss of woody vegetation cover was often considered as irreversible desertification. Recent findings, based on small-scaled analyses of satellite images, show an increase of vegetation greenness since the mid-1980s. However, it often remains unclear if this is a return to pre-drought conditions or a transformation of land cover. This study uses remote sensing techniques, supplemented by ground truth data to compare pre-drought woody vegetation and land cover with the current situation on the Dogon Plateau and the Seno Plains in Mali in a 3600 km² study area. High resolution panchromatic Corona imagery (1.8 m) of December 1967 and multispectral RapidEye imagery (6.5 m) of December 2011 form the basis of this regional scaled study. The feature extraction and classification operations included in ERDAS Imagine Objective are used in an object-oriented approach in combination with spectral properties to analyse the datasets and map millions of individual trees and large shrubs for 1967 and 2011.
Land cover maps are created for 1967 and 2011 at a resolution of 20 m. An unsupervised classification method is used for the Corona images and a supervised classification for the RapidEye images. The two main classes selected are „sparse woody vegetation“ and „dense woody vegetation“. The densely vegetated areas are mostly areas of dense woody vegetation, which have not been deforested for cultivation, or also areas which have been laid fallow for extended periods of time and are now covered by shrubbery and grass. Groups of large trees within cropland areas are also included in this class. Sparsely vegetated areas are usually used for agricultural purposes and include cultivated, fallow and grazing areas.
All individuals of trees inside a 1 ha pixel are converted to a point and counted to quantify and map the tree density in 1967 and 2011. Polygons larger 225 m² are divided by this figure to approximate the actual number of trees and shrubs represented by the single feature. Figures 4 and 5 show a case study area. According to the prevailing land cover change from dense to sparse vegetation, an overall decrease of tree density can be observed. This results in a loss of natural bushland and a spreading of degraded areas on the plateau. Agricultural land in the immediate surroundings of villages see an increase of tree density, mainly on the primary fields which are fertilized and protected.
Our results show, that neither the desertification paradigm nor the greening paradigm can be generalized in the Sahel. Rather spatial variations of changes exist; the explanations for these are equally manifold. Figure 6 demonstrates, that both greening and degradation are present in the whole study area over a period of 50 years. The main causative factor for change in tree cover and density proves to be anthropogenic. Human induced land-cover change corresponds well to tree cover change in that an increase is observed on historic primary fields and a decrease mapped in areas where the dense bushland areas of 1967 have been converted to secondary cropping fields. Furthermore, many areas of the plateau are now degraded, which is often indirectly, if not, directly related to the intense droughts of the 1970s and 80s. On the other hand, the awareness and knowledge of the advantages gained when protecting the environment, i.e. ensuring the sustainable use of trees on farmland, has increased among local inhabitants. This has led to a strong increase of woody vegetation, particularly in the immediate surroundings of settlements. The number of features extracted in the Corona images is roughly four times greater than the number extracted from the RapidEye images. The reverse is true concerning the average area of the features, mainly due to the different pixel size. Thus, there is an obvious dilemma in comparing these maps quantitatively. However, although the quantitative change may not be entirely correct, the trend certainly is.
EGU poster: EGU_2013_spiekermann_small
Spiekermann, R., Brandt, M. & C. Samimi (2013): Using high resolution imagery to detect woody vegetation and land-cover change over 50 years in the Sahel of Mali. Geophysical Research Abstracts, Vol. 15, EGU2013-11937, EGU General Assembly 2013.
master thesis: Spiekermann 2013