Energy exchange and evapotranspiration over two temperate semi-arid grasslands in North America

Abstract

The seasonal and interannual variability in surface energy exchange and evapotranspiration (E) of two temperate semi-arid grasslands in southeastern Arizona, USA, were investigated using multi-year (2004–2007) eddy covariance measurements. The study sites, a post-fire site (AG) and an unburned site (KG), received 43–87% of the annual precipitation (P) during the North American Monsoon season (July–September), with the lowest values in the drought years of 2004 and 2005. Irrespective of the differences in temperature, surface albedo, vegetation cover and soil characteristics both sites responded similarly to changes in environmental conditions. The seasonal and interannual variations in the partitioning of net radiation to turbulent fluxes were mainly controlled by P through the changes in soil water content (θ) and vegetation growth. Drastic changes in albedo, vegetation growth and energy fluxes occurred following the onset of the monsoon season in July. During the dry or cool periods of autumn, winter and spring, the sensible heat flux was the largest component of the energy balance, whereas latent heat flux dominated during the warm and wet periods of summer. The dry-foliage Priestley–Taylor coefficient (α) declined when θ in the 0–15cm soil layer dropped below 0.08m3m−3 at AG, and 0.09m3m−3 at KG, respectively. The July–September average of dry-foliage surface conductance, α and E, reached their lowest values in 2004 at AG and in 2005 at KG. During July–September, monthly E was linearly correlated to the monthly mean θ and the broadband normalized difference vegetation index (NDVI), whereas during May–June the relationship between NDVI and E was not significant. Annual E varied from 264 to 322mm at AG and from 196 to 284mm at KG with the lowest value during the severe drought year at the site. July–September E had positive correlation with total P, the mean NDVI and the number of growing season days during that period. Annual P explained more than 80% of the variance in annual E.