B53C-0558
Investigating Landsat-derived forest evapotranspiration in the Amazon

Friday, 18 December 2015
Poster Hall (Moscone South)
Kul B Khand, Izaya Numata, Jeppe Kjaersgaard and Mark A Cochrane, South Dakota State University, Brookings, SD, United States
Abstract:
Nearly half of annual rainfall in the Amazon rainforest region is returned to the atmosphere through evapotranspiration (ET). However, this land-atmosphere water vapor feedback in Amazonia has been continuously disturbed by anthropogenic influence and climate change such as severe drought events. While forest ET dynamics in the Amazon have been studied from both point estimates (or in-situ measurements) and regional land-surface models as well as coarse-spatial satellite data, finer spatial data is required to address the spatial variability of forest ET associated with both forest disturbances and extreme climate events. We use Landsat-based METRIC (Mapping Evapotranspiration at high Resolution with Internalized Calibration) model to generate high-resolution (30 m) ET products and investigate its potential to characterize local and regional ET behavior by comparison to ET calculated from flux tower data. METRIC estimates actual ET as residual of the surface energy balance and is applied to capture the spatial variability of forest ET. The flux tower data were collected at two sites with different forest types: Para with wet equatorial forest and Rondônia with seasonally dry tropical forest. Our study was conducted on the dry season of the years 2003 and 2005 for Para, and 2000 through 2002 for Rondônia as a function of data availability of both cloud-free Landsat images and meteorological data for METRIC processing. Daily gridded actual ET estimates from METRIC during the dry season were obtained using a cubic spline interpolation of ETrF (fraction of reference ET) values between the satellite image dates and multiplying by daily reference ET. Across the all study years, differences between the daily ET estimates for the selected image dates from METRIC and the flux towers were less than 1.2 mm/day, while on monthly basis, these averaged daily ET differences were much lower (< 0.5 mm). At Para, the correlation (R2) between the daily ET rates from METRIC and the flux towers were 0.75 and 0.51 for the years 2003 and 2005, respectively, whereas the correlation at Rondônia were 0.77, 0.84, and 0.72 for 2000, 2001 and 2003, respectively. These results indicate that METRIC derived ET estimates have good correlation with the flux tower ET measurements.