Vegetation dynamics in the southwest of Burkina Faso in response to rainfall variability and land use

Thumbnail Image
April, 2016
Journal Title
Journal ISSN
Volume Title
Assessing vegetation dynamics is relevant for ensuring sustainable development especially in regions where natural vegetation is altered by anthropogenic land use and rainfall variability. This is particularly crucial in the Sudan savannah of West Africa where vegetation dynamics remains poorly understood and is subject to debate. The main objective of this study was to evaluate the response of vegetation dynamics to rainfall variability and land use in the southwest of Burkina Faso. For this purpose, Landsat data (1999, 2006 and 2011) and ancillary data were used to determined changes in Land Use/Cover (LULC) with random forest classifier. Standardised Precipitation Index (SPI) and correlation analysis were used to determine rainfall variability and the relationship between vegetation indices (NDVI and EVI) and rainfall respectively. Lastly, the integration of temporal persistence analysis and Mann-Kendall’s trend test was employed to detect trends in vegetation based on MODIS NDVI 250 m data (2000-2013). Local perceptions have also been determined. The results revealed that multi-temporal LULC classification significantly outperformed mono-temporal data classification. However, combining mono-temporal imagery and ancillary data significantly enhanced the accuracy to the level of multi-temporal classification. In the period 1999-2011, LULC dynamics in the study area was mainly characterised by expansion of agricultural area and bare surface and reduction of woodland and mixed vegetation. In all the decades within the period 1981-2012, the study area was frequently under near normal conditions of rainfall with intermittent occurrence of extreme events. A non-significant increasing rainfall trend was predominant mainly in the periods 1981-2012 and 2001-2012. Vegetation dynamics was found to be strongly related to rainfall, and NDVI was slightly more sensitive to rainfall than EVI. This research also showed that the study area (83.8%) was dominated by inconsistent dynamics of vegetation in the period 2000-2013. Decreasing trajectory (14%) was prominent among the detected trends and was particularly found in agricultural area and also in areas under high and moderate human footprint. Greening trend (2.2%) was observed mainly in woodland and areas less affected by human footprint. Human was identified as the main driver of vegetation trends in the study area. The perception of local population of vegetation trends was in agreement with the remote sensing observation. In general, between 2000 and 2013, the vegetation of the study area was found to have reduced, and this is more because of unsustainable land use than rainfall conditions. These findings call for more sustainable land use management practices in this part of Burkina Faso.
A thesis submitted to the Department of Civil Engineering, College of Engineering, in partial fulfilment of the requirements for the degree of Doctor Of Philosophy in Climate Change and Land Use,