Global mean evapotranspiration (ET) has been increasing in recent decades under climate warming. Yet the magnitude and spatial distribution of ET variation remain highly uncertain. ET changes in different regions are still poorly understood due to limitations in observation records, especially in semiarid regions with undeveloped economic systems and sparse observations. Based on the Priestley‐Taylor Jet Propulsion Laboratory model, ET was estimated over global typical semiarid regions for 1984–2013. All of these regions show a decreasing ET trend, which is opposite to the trend in global mean ET. In particular, North Africa has the fastest decreasing trend, 8.6 mm/year, while South Africa has the slowest decreasing trend, 0.7 mm/year. North America, South America, northern Africa, and Australia have declining trends in ET during both warm and cold seasons, while the Loess Plateau, East Asia, central Asia, and South Africa have declining trends in ET only during warm seasons. Accounting for basic factors controlling ET, three important results are identified: First, atmospheric demand is increasing over all semiarid regions due to climate warming; second, the effect of atmospheric composition and cloud weakening radiation is strengthening over all semiarid regions; and finally, annual precipitation is decreasing over all semiarid regions except for South Africa. Factorial experiments indicate that the remarkable declining trend in relative air humidity forces the decreasing trend in ET over all semiarid regions. These results imply a slowing water cycle in global semiarid regions.

No abstract

There have been few studies conducted on the changes in actual ET over the Loess Plateau, due to the lack of reliable ET data. Based on ET data simulated by the Community Land Model, the present study analyzed the changes in ET over the Loess Plateau. The results showed the domain-average ET to have decreased in the past 31 years, at a rate of 0.78 mm year −1 . ET fluctuated much more strongly in the 1990s than in the 1980s and 2000s, and, apart from in autumn, ET decreased in all seasons. In particular, ET in summer comprised about half of the annual ET trend and had the sharpest trend, dominating the interannual decline. ET also decreased more sharply in the semiarid than semihumid regions. The declining trend of ET was attributed to declining precipitation and air humidity. Locally, the ET trend was closely related to local mean annual precipitation: in areas with precipitation less than 400 mm, ET showed a decreasing trend; in areas with precipitation larger than 600 mm, ET showed an increasing trend; and in areas with precipitation in the range of 400-600 mm could be classified as a transitional zone.

Evapotranspiration (ET) is a critical component in the hydrological cycle. However, its actual values appear to be difficult to obtain, especially in areas in which precipitation has high inter-annual variability. Here, we evaluated eight commonly used ET models in semi-arid and semi-humid areas of China. The order of overall performance from best to worst is as follows: the revised Priestley-Taylor model (PT-JPL, 0.71, 1.65 [18.37%], 4.72 [49.19%]) a , the modified PT-JPL model (M1-PT-JPL, 0.67, À0.68 [7.56%], 3.87 [40.31%]), the Community Land Model (CLM, 0.68, À2.52 [28.01%], 5.10 [53.17%]), the modified PT-JPL model (M2-PT-JPL, 0.63, 0.57 [6.27%], 5.04 [52.52%]), the revised Penman-Monteith model (RS-PM, 0.62, 3.56 [37.40%], 6.11 [63.68%]), an empirical model (Wang, 0.59, À1.04 [11.57%], 5.61 [58.43%]), the advection-aridity model (AA, 0.55, 5.56 [61.78%], 7.45 [77.60%]), and the energy balance model (SEBS, 0.35, 5.11 [56.72%], 9.43 [98.18%]). The performance of all of the models is comparably poor in winter and summer, except for the PT-JPL model, and relatively good in spring and autumn. Because of the vegetation control on ET, the Wang, RS-PM, PT-JPL, M1-PT-JPL, and M2-PT-JPL models perform better for cropland, whereas the AA model, SEBS model and CLM perform better for grassland. The CLM, PT-JPL, and Wang models perform better in semiarid region than in semi-humid region, whereas the opposite is true for SEBS and RS-PM. The AA, M1-PT-JPL, and M2-PT-JPL models perform similarly in semi-arid and semi-humid regions. When considering the inter-annual variability in precipitation, the Wang model has relatively good performance under only some annual precipitation conditions; the performance of the PT-JPL and AA models is reduced under conditions of high precipitation; the two modified PT-JPL models inherited the steady performance of the PT-JPL model and improved the performance under conditions of high annual precipitation by the modification of the soil moisture constraint. RS-PM is more appropriate for humid conditions. CLM and PT-JPL models could be effectively applied to all precipitation conditions because of their good performance across a wide annual precipitation range. a Statistics (model abbreviation, coefficient of determination, bias [relative value], standard deviation [relative value]). HYDROLOGICAL PROCESSES Figure 6. Performance of the selected models in winter, spring, summer, and autumn. The dashed line represents the line bias equal to 0. (a) depicts the coefficient of determination, (b) depicts bias, and (c) depicts standard deviation 4304 Z. YANG ET AL.Figure 7. Performance of the eight ET models under different land cover types in winter, spring, summer, and autumn. (a) depicts the coefficient of determination, (b) depicts bias, and (c) depicts standard deviation 4306 Z. YANG ET AL.Figure 8. Performance of the eight ET models in different climate zones. The column is the average value, and the error bar represents the standard deviation for all of the study sites in the climate region. (...