Remote sensing based vegetation extraction and change detection in the National Park of Niokolo-Koba in southeast of Senegal

M. L. NDIAYE, V. B. Traore, A.T. Diaw

Abstract


Natural vegetation plays a vital role in the balance of the earth's environment, thanks to its multiple functions as purifier, producer and protector. And the monitoring of vegetation in Sahel regions with Remote Sensing data has become increasingly important over the past decade because it is linked to variation in agricultural production and climate change with implications for wildlife management and tourism. Recent advances in Remote Sensing applied make it possible to better characterize vegetation, but also monitoring its spatio-temporal dynamics. This study aims to extract vegetation cover in the National Park of Niokolo-Koba and analyze its changes over a period of 45 years. In doing so, four Landsat images recorded on 1972 (MSS), 1986 (TM), 2000 (ETM+), 2005 (ETM+), 2011 (TM) and 2014 (OLI_TRIS) were used to minimize change detection error introduced by seasonal differences. Images were geometrically and radiometrically corrected. Normalized difference vegetation index (NDVI) is chosen to extract and classify vegetation. Post-classification comparison was applied to Landsat multi-temporal imagery to determine ‘from–to’ change information derived from the classifications maps. The overall accuracy obtained for single classifications is 81.72% (1972), 84.32% (1986), 89.49% (2002) and 93.50% (2016). And for change results, the overall accuracy calculated through the no-change / change error matrix is considered good, with 74.90%, 78.90 and 80.72% for the 1972-1986, 1986-2002 and 2002-2016 images pairs, respectively. These results made it possible to identify the spatial and temporal dynamic of the vegetation within the Park between 1972 and 2016. A spatial regression of the gallery forest and woodland savannah between 1972-1986 and 1986-2002 was observed, to the detriment of the Shrubland and grassy savanna. Nevertheless, between 2002 and 2016 the situation is reversed, notably with an upward trend of the forests gallery and shrub savannah classes. There is also an alternating phase of progression and regression of wetlands. A high correlation (r² > 0.60) is established between each class of vegetation and rainfall variability. This study is part of an integrated management approach to protected areas in Senegal. It is a first approach to monitoring the vegetation of the Niokolo-Koba National Park by using remote sensed data and helps to understand the trends observed in the dynamics of forest resources.


Keywords


Remote sensing, NDVI, Change detection, Protected areas, Niokolo-Koba National park

Full Text:

PDF




ISSN: 2509-2065

If plagiarism is detected more than 30% before/during the peer review process, the submission may be rejected.