Uncertainty assessment of potential evapotranspiration in arid areas, as estimated by the Penman-Monteith method

The Penman-Monteith (PM) method is the most widely used technique to estimate potential worldwide evapotranspiration. However, current research shows that there may be significant errors in the application of this method in arid areas, although questions remain as to the degree of this estimation error and how different surface conditions may affect the estimation error. To address these issues, we evaluated the uncertainty of the PM method under different underlying conditions in an arid area of Northwest China by analyzing data from 84 meteorological stations and various Moderate Resolution Imaging Spectroradiometer (MODIS) products, including land surface temperature and surface albedo. First, we found that when the PM method used air temperature to calculate the slope of the saturation vapor pressure curve, it significantly overestimated the potential evapotranspiration; the mean annual and July-August overestimation was 83.9 and 36.7 mm, respectively. Second, the PM method usually set the surface albedo to a fixed value, which led to the potential evapotranspiration being underestimated; the mean annual underestimation was 27.5 mm, while the overestimation for July to August was 5.3 mm. Third, the PM method significantly overestimated the potential evapotranspiration in the arid area. This difference in estimation was closely related to the underlying surface conditions. For the entire arid zone, the PM method overestimated the potential evapotranspiration by 33.7 mm per year, with an overestimation of 29.0 mm from July to August. The most significant overestimation was evident in the mountainous and plain nonvegetation areas, in which the annual mean overestimation reached 5% and 10%, respectively; during July, there was an estimation of 10% and 20%, respectively. Although the annual evapotranspiration of the plains with better vegetation coverage was slightly underestimated, overestimation still occurred in July and August, with a mean overestimation of approximately 5%. In order to estimate potential evapotranspiration in the arid zone, it is important that we identify a reasonable parameter with which to calibrate the PM formula, such as the slope of the saturation vapor pressure curve, and the surface albedo. We recommend that some parameters must be corrected when using PM in order to estimate potential evapotranspiration in arid regions.