Local tower-based merging of two land evaporation products

Abstract

Abstract. An inverse-variance weighting of two terrestrial evaporation (ET) products from the WACMOS-ET project based on FLUXNET sites is presented. The two ET models, PT-JPL and GLEAM, share a common modeling framework based on the Priestley–Taylor evaporation formulation, although the stress factors to convert from potential to actual evaporation are calculated differently. Both data sets are daily at 25 km resolution and use common input data when possible. The weights are based on the variance of the difference between the tower ET and the modeled ET, and are made dynamic by estimating them using a 61-day running window centered on each day. Seasonal variability in these weights is observed over some stations, but the deviations from the 0.5 value assumed by the arithmetic mean of both products is typically small. Seasonal averaged statistics for three land cover groups confirm that the performance of the weighted product is very close to the arithmetic mean product. This implies that at a large number of stations the tower data was not informative enough to identify distinct error patterns between GLEAM and PT-JPL likely to result in more varying weights. Different factors can be responsible for this behavior. Concerning the model estimates, some of the shared modeling assumptions and the common inputs make the ET estimates from both models rather dependent, thus their errors are expected to be correlated. Regarding the tower data, the 84 selected tower sites are mostly at temperate regions. In these regions the models ET differences are mainly reflecting the variability in Priestley and Taylor estimates of potential ET, which translates in model ET differences smaller than in other places. The large mismatch between the model spatial resolution and the tower footprint is also likely to result again in common error patterns for the model estimates. Overall, this study suggests that merging tower observations and ET products at the time and spatial scales of this study is not straightforward, and that care should be taken regarding the dependence of the products to be merged, the tower spatial representativeness of their measurements at the products resolution, and the nature of the error in both towers and gridded data sets.

Linear Optimal Runoff Aggregate (LORA): A global gridded synthesis runoff product

“…Despite the demonstrated improved predictive capability of multi-model ensemble approaches (Sahoo et al, 2011;Pan et al, 2012;Bishop and Abramowitz, 2013;Mueller et al, 2013;Munier et al, 2014;Aires, 2014;Rodell et al, 2015;Jiménez et al, 2017;Hobeichi et al, 2018;Zhang et al, 2018), very little has been done to utilise this range of model simulations toward improved runoff estimates. This paper implements the weighting and rescaling method introduced in Bishop and Hydrol.…”