Conditioning of NLDAS, NARR and arid weather station data to improve their representation of well-watered (reference) environments associated with irrigated agriculture

Valuable high-resolution gridded atmospheric data sets are produced by the NASA Land Data Assimilation Systems (NLDAS) and the North American Regional Reanalysis (NARR) systems using atmospheric models coupled with land surface models (LSM). Evolving uses of these data sets include forecasting of future evaporative demands by irrigated agriculture to improve water management, assessing historic time series of water consumption and demands over large regions, and assessing impacts of future climate projections. One challenge with these gridded sets is that the water balance used for driving the surface partitioning of energy is generally not informed on irrigation and associated ET so that more energy is imparted to sensible heat flux, soil heat flux and thermal radiation that occurs in irrigated settings. As a consequence, meteorological data generated by these models tends to represent that occurring under ambient, rain-fed conditions, with near surface temperature (T) measurements elevated by as much as 5 C, and vapor pressure reduced by up to one-half, as compared to measurements made over an evaporating surface. When the data are entered into reference ET equations for estimating crop water requirements in irrigated agriculture, the data in the gridded NLDAS and NARR data sets tend to overestimate ET anticipated in an irrigated environment by up to 30%.

A theoretical blending height-profiling procedure has been developed for “conditioning” the ambient meteorological data from the gridded data systems or from ‘arid' weather station environments so that data represents near surface air properties for the same climate, but over an evaporating, “reference” surface. Following conditioning, an “equilibrium reference ET” can be scaled using crop coefficients or other parameters to estimate actual ET for a variety of vegetation. The conditioning completes the feedback processes between surface evaporation and near surface air properties.

Comparisons have been made with non-conditioned and conditioned data collected at automated weather stations located, generally, in well-watered agricultural environments. Over 60 stations from 17 Western U.S. states have been compared to over 5 to 25 30 year periods following stringent QA/QC of station data. Results indicate that unconditioned NLDAS and NARR data consistently over estimate irrigation water requirements due to increased air temperature, reduced dew point temperature, and increased windspeed when compared to station data collected in well-watered environments. These results highlight the important point that if unconditioned NLDAS and NARR data are used to compute reference ET and crop ET, significant over estimation will occur.