Change Detection in Protected Landscapes for Adaptive Protected Area Management A Practical Geoinformatics Approach

Abstract

The Department of Wildlife Conservation (DWC), the entity which is entrusted with themandate to conserve fauna and flora of the country manages 14% of total land area of SriLanka, which are designated as Protected Areas (PA) under the Fauna and Flora ProtectionOrdinance (FFPO). Conservation of biodiversity and non destructive uses of wildliferesources especially for recreationl purposes within the PAs is a challanging task for theDWC. Horton Plains National Park is a one such PA which support higher diversity ofmontane fauna and flora in the country. The PA is at present also subjected some habitatchanges, which are taking place naturally. Managing such a sensitive landscape requrestimely and accuarare scientific information regarding the species richness of the area andtheir spatial and temporal distribution patterns. It is also important to understand the extent ofhabitat change, where it occurs and what habitats are converted into a different habitat.

In conventional wildlife management approach thematic management plans are prepared fora ten year time frame. The review interval of such a plan is generally 5 years. However, in SriLankan context, due to the nature of the dynamics in protected areas and the taskenvironment, DWC uses a concept called “adaptive management” where the reviewing andadjustments of the activities are done frequently depending on the previous periods’experiences, usually less than five years. In this context temporal change detection of thehabitats is a task which should be carried out as frequent as possible

This study focuses on developing classified vegetation map for two time periods (1998 and2008) and quntify the temporal area variation of the vegetation.Vegetation map wasdeveloped using IRS LISSA III images. The vegetation classes were classfied intoUndisturbed forest, Disturbed forest, Ecotone, Carpet Grass, Tussocky grass, and DwarfBamboo. The methodology was developed intergrating Normalize Difference VegitationIndex (NDVI) and Tasseled Cap Transformation. Finally the landscape of the PA wasclassified into different vegetation types using pixel based unsupervised classificationalgorithum. User and producer accuracies were calculated with the field observation data.

The results clearly indicate that all the vegetation types within the PA have changed between1998 and 2008. Open forest cover increased from 23.56% to 28.41%, Marsh / dwarf bamboocover was increased from 5.35% to 8.76% during the period of ten years. It was also notedthat the dense forest areas are reduced from 48.19% to 47.52%, carpet grass dominant areareduced from 2.12% to 1.02% and tussock grass cover was reduced from 20.78% to 13.63%.In general forest cover of the PA has increased from 71.75% to 75.93% and grass cover(carpet and tussock) was reduced from 28.25% to 24.07% between 1998 and 2009 period.

With the user accuracies of 83% in 1998 and 82% in 2008, and producer accuracy of 87% in1998 and 80% 2008 it can be concluded that the methodology adopted in the study issufficient in accuracy for practical usage. Hence, technique proposed in this study could bepracticed periodically to detect the vegetation changes quantitatively and effectively for themanagement activities of the park.