Bush Encroachment Mapping for Africa: Multi-Scale Analysis with Remote Sensing and GIS

ZEF - Center for Development Research University of Bonn, Discussion Paper No. 218

51 Pages Posted: 12 Jul 2016

See all articles by Valerie Graw

Valerie Graw

University of Bonn - Center for Development Research (ZEF)

Carsten Oldenburg

University of Bonn - Center for Development Research (ZEF)

Olena Dubovyk

University of Bonn - Center for Development Research (ZEF)

Date Written: July 2016

Abstract

Bush Encroachment (BE) describes a global problem severely affecting savanna ecosystems in Africa. Invasive species and woody vegetation spread out in areas where they are not naturally occurring and suppress endemic vegetation, mainly grasses. Livestock is directly affected by decreasing grasslands and inedible invasive species which are a result of the process of BE. For many small scale farmers in developing countries livestock represents a type of insurance particular in times of crop failure and droughts. Among that, BE is also becoming an increasing problem for crop production. Studies on the mapping of BE have so far only focused on smaller regions using high-resolution data and aerial photography. But they rarely provide information that goes beyond the local or national level. In our project, we aimed at a continental-wide assessment of BE. For this, we developed a process chain using a multi-scale approach to detect woody vegetation for the African continent. The resulted map was calibrated with field data provided by field surveys and experts in Southern and Eastern Africa. Supervised classification linked field data of woody vegetation, known as BE, to the respective pixel of multi-scale remote sensing data. The regression technique was based on random forests, a machine learning classification and regression approach programmed in R. Hotpots of woody vegetation were further overlaid with significant increasing Normalized Difference Vegetation Index (NDVI) trends which can refer to BE. Secondly, the probability of BE occurrence based on possible identified causes such as fire occurrence, mean annual precipitation rates, soil moisture, cattle density and CO2 emissions was analyzed. By this, possible areas for BE occurrence based on their preconditions and risk factors were identified.

This approach includes multiple datasets derived from earth observation data to detect BE – a severe and ongoing global problem – at the continental level. Within the study's duration of seven months, a method to upscale field data to a larger level could be developed. Nevertheless, improvement is needed to provide a reliable continental map on BE. Especially the integration of more field data will be needed which is currently under consideration. The identification of woody vegetation and the probability of its occurrence can help to prevent further ecosystem degradation. Moreover, sustainable land management strategies in these areas can be focused to support pastoralists and their livelihoods in rural areas.

Keywords: bush encroachment, remote sensing, multi-scale analysis, probability map, random forests, regression trees

Suggested Citation

Graw, Valerie and Oldenburg, Carsten and Dubovyk, Olena, Bush Encroachment Mapping for Africa: Multi-Scale Analysis with Remote Sensing and GIS (July 2016). ZEF - Center for Development Research University of Bonn, Discussion Paper No. 218. Available at SSRN: https://ssrn.com/abstract=2807811 or http://dx.doi.org/10.2139/ssrn.2807811

Valerie Graw (Contact Author)

University of Bonn - Center for Development Research (ZEF) ( email )

Walter-Flex_Str. 3
Bonn, 53113
Germany

Carsten Oldenburg

University of Bonn - Center for Development Research (ZEF) ( email )

Walter-Flex-Str. 3
Bonn, NRW 53113
Germany

Olena Dubovyk

University of Bonn - Center for Development Research (ZEF) ( email )

Walter-Flex-Str. 3
Bonn, NRW 53113
Germany

Register to save articles to
your library

Register

Paper statistics

Downloads
106
Abstract Views
704
rank
260,755
PlumX Metrics