COMPARISON OF THE VEGETATION INDICES TO DETECT THE TROPICAL RAIN FOREST CHANGES USING BREAKS FOR ADDITIVE SEASONAL AND TREND (BFAST) MODEL

Abstract

Remotely  sensed  vegetation  indices  (VI)  such  as  the  Normalized  Difference Vegetation Index (NDVI) are increasingly used as a proxy indicator of the state and condition of  the  land  cover/vegetation,  including  forest.  However,  the  Enhanced  Vegetation  Index (EVI)  on  the  outcome  of  forest  change  detection  has  not  been  widely  investigated.  We compared the influence of using EVI and NDVI on the number and time of detected changes by applying Breaks for Additive Seasonal and Trend (BFAST), a change detection algorithm. We  used  MODIS  16-day  NDVI  and  EVI  composite  images  (April  2000-April  2012)  of  three pixels  (pixels  352,  378,  and  380)  in  the  tropical  peat  swamp  forest  area  around  the  flux tower of  Palangka Raya, Central Kalimantan.  The results  of  BFAST method were compared to  the  Normalized  Difference  Fraction  Index  (NDFI)  maps  and  the  maps  were  validated  by the  hotspot  of  the  Infrastructure  and  Operational  MODIS-Based  Near  Real-Time  Fire(INDOFIRE).  Overall,  the  number  and  time  of  changes  detected  in  the  three  pixels  differed with both time series data  because of the  data quality due to the cloud cover.  Nonetheless, we  found  that  EVI  is  more  sensitive  than  NDVI  for  detecting  abrupt  changes  such  as  the forest fires of August 2009-October 2009 that occurred in our study area and it was verified by  the  NDFI  and  the  hotspot  data.  Our  results  demonstrated  that  the  EVI  for  forest monitoring in the tropical peat swamp forest area which is covered by intense cloud cover is better  than  that  NDVI.  Nonetheless,  further  research  with  improving  spatial  resolution  of satellite images for application of NDFI is highly recommended.