Determining the dynamics of evapotranspiration from fragmented forests under drought in southwestern Amazonia using Landsat imagery

Deforestation in the Amazon has resulted in massive amounts of forest biomass loss and also in extensive forest fragmentation across the region. Fragmented tropical forests are exposed to abrupt environmental changes and experience several biological and ecological changes across distances from forest edges. Extreme droughts in 2005 and 2010 have caused extensive tree mortality across this region. These events may exacerbate edge effects, where already water stressed forest fragments dry more rapidly potentially enabling other disturbances such as forest fire. We analyzed the effects of forest fragmentation and drought on forest evapotranspiration (ET) estimated using the energy balance-based model METRIC with Landsat imagery in Rondônia State in the southwestern Amazon. Forest ET estimates were produced for the dry seasons (June-August) of 2009-2011 thus including the 2010 drought event and pre- and post-event periods. METRIC ET data were combined with forest edge data with edge distances of 100m, 300m, 500m, 1000m, 5000m and >5000m (core forest), generated from Landsat land cover maps for spatiotemporal analysis of forest ET. METRIC ET estimates had an agreement with flux tower ET data from the field of R2 = 0.72. Within the study time period, the 2010 drought year showed the lowest average ET from core forest (2.5mm/day), followed by 2011 (3.0mm/day) and 2009 (3.6mm/day) in the month of August, the mid dry season, while no significant differences were noted among three study years earlier in the dry seasons. In terms of edge effects, the major changes in forest ET occur up to 300 m from the forest edges, with ET decreasees of 30 % at 100 m as compared to further distances. The magnitude of edge-related ET changes became even greater during August of the drought year (2010) and the post-drought year (2011). Annual (drought and non-drought) and seasonal (June-August) forest ET variations were highly significant (p<0.001), while the impact of distance from edge on forest ET was significant only in the drought year (p<0.05).