Cascading Impacts of Longwave Radiation Uncertainty on Modeled Snowmelt and Summer Evapotranspiration at Mountain Research Sites

Abstract:

Atmospheric longwave radiation is a key source of energy for surface hydrological processes such as snowmelt and evapotranspiration. Many modeling applications estimate this energy flux using empirical parameterizations for atmospheric emissivity that vary with surface conditions (e.g., temperature, humidity) and assumed cloud cover. Given that a wide variety of such empirical parameterizations exists, many modeling studies apply different approaches, thereby inducing systematic differences in the simulated surface energy balance. In snow-dominated, mountain basins, sites that observe longwave radiation are rare, and thus it is difficult to discriminate which methods and parameters are most representative. This uncertainty in longwave radiation cascades in complex ways to the modeled hydrology of mountain regions, influencing not only the magnitude of energy available for snowmelt and evapotranspiration, but also the timing of the growing season and the synchrony between local water and energy cycles.

Using observations at three well-instrumented mountain sites in maritime, intermountain, and continental climates, we examine how longwave uncertainty propagates through modeled snow processes and summer evapotranspiration. We represent longwave radiation with an ensemble of 400 different approaches for calculating atmospheric emissivity (based on 20 clear sky and 20 cloud correction methods). We find that the daily range typically exceeds 100 W m^-2 at all three sites, and that the variability attributed to the cloud-correction is typically greater than or comparable to that of the clear sky method. This longwave uncertainty has significant impacts on modeled mid-winter melt occurrence, peak snow water equivalent (typically +/- 25%), surface energy feedbacks, and snow disappearance timing (1-2 months). We quantify how longwave radiation specifically propagates through each of these components and test the sensitivity of four different snow models to longwave uncertainty.