|Metzger, S; Junkermann, W; Mauder, M; Butterbach-Bahl, K; Trancón y Widemann, B; Neidl, F; Schäfer, K; Wieneke, S; Zheng, X H; Schmid, HP; Foken, T: Spatially explicit regionalization of airborne flux measurements using environmental response functions, Biogeosciences, 10, 2193–2217 (2013), doi:10.5194/bg-10-2193-2013|
The goal of this study is to characterize the sensible (H) and latent (LE) heat exchange for different land covers in the heterogeneous steppe landscape of the Xilin River catchment, Inner Mongolia, China. Eddy-covariance flux measurements at 50–100 m above ground were conducted in July 2009 using a weight-shift microlight aircraft. Wavelet decomposition of the turbulence data enables a spatial discretization of 90 m of the flux measurements. For a total of 8446 flux observations during 12 flights, MODIS land surface temperature (LST) and enhanced vegetation index (EVI) in each flux footprint are determined. Boosted regression trees are then used to infer an environmental response function (ERF) between all flux observations (H, LE) and biophysical (LST, EVI) and meteorological drivers. Numerical tests show that ERF predictions covering the entire Xilin River catchment (~3670 km2) are accurate to <- 18% (1delta). The predictions are then summarized for each land cover type, providing individual estimates of source strength (36Wm-2 < H < 364Wm-2, 46Wm-2 < LE < 425Wm-2) and spatial variability (11Wm-2 < delta H < 169Wm-2, 14Wm-2 < delta LE < 152Wm-2) to a precision of <- 5 %. Lastly, ERF predictions of land cover specific Bowen ratios are compared between subsequent flights at different locations in the Xilin River catchment. Agreement of the land cover specific Bowen ratios to within 12±9% emphasizes the robustness of the presented approach. This study indicates the potential of ERFs for (i) extending airborne flux measurements to the catchment scale, (ii) assessing the spatial representativeness of long-term tower flux measurements, and (iii) designing, constraining and evaluating flux algorithms for remote sensing and numerical modelling applications.