Regional evapotranspiration and pan evaporation: complementary interactions and long-term trends across the conterminous United States

Abstract

Long-term observations of pan evaporation and water budget-derived evapotranspiration across the conterminous United States provide the first observational evidence supporting the hypothesis of a complementary relationship in regional evapotranspiration, in terms both of the evaporation rates themselves and of long-term trends in their component dynamics. The conjectured relationship has now become an observational fact. To establish a baseline for the study of climate change and/or variability, a complementary relationship model estimates spatially distributed actual evapotranspiration across the conterminous US on a monthly basis over a recent 42-year period. This is used to examine two advective calibration approaches, and trends in actual evapotranspiration and its components as to their direction, magnitude, statistical significance, and spatial distributions. In observations of trends in ETa and in its two component trends—the radiative energy and regional advective dynamics—it is shown that, contrary to previous conclusions that have been predicated on questionable and restrictive assumptions over near-continental scales, trends in the components must be examined concurrently within the context of the complementary relationship to explain trends in regional ETa. It is further shown that only by examining spatially concurrent observations at smaller spatial scales can good conclusions be drawn about the "strength" of the complementary relationship, specifically, and ET trends in general. The two most problematic data sets used in the analysis are improved: solar radiation, which suffers from the effects of local topography; and pan evaporation, which bears the effects of anthropogenic heterogeneities inherent in a variously sourced data set. A procedure is developed to mitigate the confounding influence of rugged terrain on the analysis of the short-wave radiative balance, producing a long-term, high-resolution, topographically corrected net radiation data set. Twelve years of missing diffuse radiation data are replicated based on their historical relationships to coincident, contemporaneous direct normal and global radiation. A monthly topographic correction factor is derived to account for the incidence of direct solar radiation on arbitrarily oriented ground surfaces at any latitude throughout the diurnal and annual cycles. The factor is applied to spatially interpolated surfaces of monthly direct solar radiation which, when added to surfaces of diffuse radiation, provide the total incident solar radiation input to an existing energy budget. This yields the net surface radiation that may then applied in evapotranspiration models. Pan evaporation observations are gathered from two data sources for 228 pans across the conterminous US toward an examination of long-term trends in annual and warm-season totals. The data are characterized by their incompleteness and, more importantly, non-homogeneity that, unless accounted for, can introduce spurious biases into analyses of long-term trends. However, what scant metadata are available are elliptical. The need to retain climatically driven trends after homogenization requires a technique that resolves understandings of both the physical dynamics and the statistical properties of the data by combining objective rigor with subjective judgment. Using the t-test to indicate statistically significant abrupt shifts in each pan's time-series, 172 pans are adjusted at a total of 326 abrupt data-shifts, adjusting 43% of the annual data and 55% of the warm-season data. Comparing trend results from pre- and post-adjustment data across all pans, some differences are noted in the details, but they are not together significant enough to change the conclusions of the trend analysis. Pan evaporation has decreased at 64% of the year-round pans in the conterminous US over the past half-century. The so-called "Pan Evaporation Paradox" is shown to be no more than a manifestation of the natural complementarity between actual and potential evapotranspiration. An examination of trends in the radiative energy and regional advective components of basin-derived actual evapotranspiration shows that both components must be considered together to explain the relationship between actual and potential evapotranspiration. Actual evapotranspiration is modeled using a regional, seasonal Advection-Aridity approach to create a spatially distributed, monthly time-series for a 42-year period at a 5- km resolution over the conterminous US. Formulations of both dynamics in the evaporative process are improved with respect to the applicability of the model across large topographic and climatic variations. The radiative input is the aforementioned topographically corrected data set. The advective input is improved by analysis of two regional calibrations of the wind function: first such that modeled actual evapotranspiration matches basin-derived evapotranspiration at 655 basins across the conterminous US; second such that potential evapotranspiration matches point observations of pan evaporation across the southern tier of states. Each calibration invokes different assumptions and limitations on its applicability in the temporal and spatial domains. The parameter sets of the derived wind functions are similar in value, but the first is noisier while the second bears less significant functional relationships to wind speed. The modeled annual evapotranspiration data are verified against observed water budget-derived actual evapotranspiration. The basin-derived calibration of the wind function performs the best, while the pan evaporation-based calibration under-estimates evapotranspiration. In purely statistical terms, the basin-derived calibration is preferred, but the performances of both calibrations bear functional relationships to the precipitation, basin-derived evapotranspiration, and wind speed in the areas of application. In terms of long-term trends over the modeling period WY 1953-1994, for the conterminous US as a whole, a 4.2% decrease in modeled annual actual evapotranspiration is observed for the basin-derived calibration, a trend significant at the 69% confidence level according to the Mann-Kendall test. Over the southern tier of states, a 3.1% decrease in modeled annual actual evapotranspiration is observed for the basin-derived calibration (significant at 62%), and a 2.1% decrease in modeled annual actual evapotranspiration for the pan evaporation-derived calibration, (significant at 47%). Reducing the spatial scale of trend-analyses—from the continental US through the nested component 18 Water Resource Regions and the further-nested 334 Accounting Units to the 655 relatively undisturbed basins across the continental US—allows for clearer identification of areas with significant trends, and the breakdown into component dynamics shows that trends in actual evapotranspiration can be determined to originate in either energy or water fluxes, or both.