Atmospheric Water Vapor Pressure over Land Surfaces: A Generic Algorithm with Data Input Limited to Air Temperature, Precipitation and Geographic Location

Summary

A lack of information for surface water vapor pressure (WVP) represents a major impediment to model-assisted ecosystem analysis for understanding plant-environment interactions or for projecting biospheric responses to global climate change. This paper reports on a generic algorithm that captures global variation in monthly WVP. The algorithm solves WVP in terms of reduction from saturation WVP as a negative exponential function of potential evapotranspiration; the reduction rate per unit potential evapotranspiration in turn varies with monthly precipitation and a series of variables that distinguish local climate regimes. Data input to the algorithm is limited to monthly air temperature and precipitation, plus latitude, longitude and elevation. The algorithm is specified through regression fitting to monthly climate normal data from 852 stations around the world. It accounts for 96% of the variance in the WVP data, with a root mean square error of 0.17 kPa, or 12% of the data mean. The algorithm closely reproduces five-year sequential monthly WVP data for each of five selected United States locations representitative of diverse climate regimes: the average error generally falls within ±12% of the data mean, and the absolute error within ±0.2 kPa. Its projections also compare favorably against the WVP output from the General Circulation Models for temperature and precipitation conditions under the scenario of a doubled atmospheric carbon-dioxide concentration: the two fall within ±10% of each other for 75% of a total 264 data cases, or within ±20% for 94% of the cases. These statistics suggest that the spatially-based algorithm is useful for projecting temporal variation in WVP, and for extrapolative applications beyond the fluctuation range of present climate.