Description: Estimation of high-resolution terrestrial evapotranspiration (ET) from Landsat data is important in many climatic, hydrologic, and agricultural applications, as it can help bridging the gap between existing coarse-resolution ET products and point-based field measurements. However, there is large uncertainty among existing ET products from Landsat that limit their application. This study presents a simple Taylor skill fusion (STS) method that merges five Landsat-based ET products and directly measured ET from eddy covariance (EC) to improve the global estimation of terrestrial ET. The STS method uses a weighted average of the individual ET products and weights are determined by their Taylor skill scores (S). The validation with site-scale measurements at 206 EC flux towers showed large differences and uncertainties among the five ET products. The merged ET product exhibited the best performance with a decrease in the averaged root-mean-square error (RMSE) by 2–5 W/m2 when compared to the individual products. To evaluate the reliability of the STS method at the regional scale, the weights of the STS method for these five ET products were determined using EC ground-measurements. An example of regional ET mapping demonstrates that the STS-merged ET can effectively integrate the individual Landsat ET products. Our proposed method provides an improved high-resolution ET product for identifying agricultural crop water consumption and providing a diagnostic assessment for global land surface models.

Description: Remote Sensing, Vol. 9, Pages 1140: Satellite-Derived Spatiotemporal Variations in Evapotranspiration over Northeast China during 1982–2010 Remote Sensing doi: 10.3390/rs9111140 Authors: Lilin Zhang Yunjun Yao Zhiqiang Wang Kun Jia Xiaotong Zhang Yuhu Zhang Xuanyu Wang Jia Xu Xiaowei Chen Evapotranspiration (ET) is a critical process for the climate system and water cycles. However, the spatiotemporal variations in terrestrial ET over Northeast China over the past three decades calculated from sparse meteorological point-based data remain large uncertain. In this paper, a recently proposed modified satellite-based Priestley–Taylor (MS–PT) algorithm was applied to estimate ET of Northeast China during 1982–2010. Validation results show that the square of the correlation coefficients (R2) for the six flux tower sites varies from 0.55 to 0.88 (p < 0.01), and the mean root mean square error (RMSE) is 0.92 mm/d. The ET estimated by MS–PT has an annual mean of 441.14 ± 18 mm/year in Northeast China, with a decreasing trend from southeast coast to northwest inland. The ET also shows in both annual and seasonal linear trends over Northeast China during 1982–2010, although this trend seems to have ceased after 1998, which increased on average by 12.3 mm per decade pre-1998 (p < 0.1) and decreased with large interannual fluctuations post-1998. Importantly, our analysis on ET trends highlights a large difference from previous studies that the change of potential evapotranspiration (PET) plays a key role for the change of ET over Northeast China. Only in the western part of Northeast China does precipitation appear to be a major controlling influence on ET.

Description: Remote Sensing, Vol. 9, Pages 1326: MODIS-Based Estimation of Terrestrial Latent Heat Flux over North America Using Three Machine Learning Algorithms Remote Sensing doi: 10.3390/rs9121326 Authors: Xuanyu Wang Yunjun Yao Shaohua Zhao Kun Jia Xiaotong Zhang Yuhu Zhang Lilin Zhang Jia Xu Xiaowei Chen Terrestrial latent heat flux (LE) is a key component of the global terrestrial water, energy, and carbon exchanges. Accurate estimation of LE from moderate resolution imaging spectroradiometer (MODIS) data remains a major challenge. In this study, we estimated the daily LE for different plant functional types (PFTs) across North America using three machine learning algorithms: artificial neural network (ANN); support vector machines (SVM); and, multivariate adaptive regression spline (MARS) driven by MODIS and Modern Era Retrospective Analysis for Research and Applications (MERRA) meteorology data. These three predictive algorithms, which were trained and validated using observed LE over the period 2000–2007, all proved to be accurate. However, ANN outperformed the other two algorithms for the majority of the tested configurations for most PFTs and was the only method that arrived at 80% precision for LE estimation. We also applied three machine learning algorithms for MODIS data and MERRA meteorology to map the average annual terrestrial LE of North America during 2002–2004 using a spatial resolution of 0.05°, which proved to be useful for estimating the long-term LE over North America.