Evaluating the Effects of Subpixel Heterogeneity on Pixel Average Fluxes

Abstract

Radiometric temperature observations T R (φ) at a sensor view angle φ are routinely available from weather satellites such as the Geostationary Orbiting Environmental Satellite (GOES) and provide a unique spatially distributed boundary condition for surface energy balance modeling at regional scales. Reliable flux estimates over heterogeneous surfaces have been obtained using two-source models that implicitly account for differences between T R (φ) and aerodynamic temperature, T O , by considering separately the contributions of soil/substrate and vegetation to T R (φ) observations and to the turbulent fluxes. A simple two-source energy balance model developed to use T R (φ) observations has been applied successfully to a wide range of vegetation cover conditions at the field scale. However, its application with course resolution weather satellite data (i.e., pixel resolution≳1 km) will invariably result in errors in pixel-averaged heat flux estimation for surfaces with significant variability in vegetation cover and stress conditions. Indeed, with the highest resolution of satellite T R (φ) data∼100 m, subpixel heterogeneity will still be significant for many landscapes, especially arid and semiarid areas. Uncertainty in flux estimation due to significant subpixel heterogeneity is examined using the simplified two-source model with T R (φ) inputs from simulations using a detailed plant-environment model (Cupid) under six different “homogeneous” surface conditions commonly found in semiarid and arid regions and under high and low winds. These surface types are comprised of shrub and tall riparian vegetation, high and low canopy cover, wet and dry surface soil moisture state, and stressed versus unstressed vegetation condition. From six homogeneous surface conditions defined by vegetation type, cover, surface moisture and stress, four mixed-pixel cases were constructed, each containing two contrasting surface types. Significant or unacceptable errors (i.e., ∼50 W m −2 ) in pixel-average heat fluxes are found in all four mixed-pixel cases, but the significant errors primarily occur when the fraction of the extreme surface condition (e.g., riparian wetland) comprises between ∼20% and 80% of the mixed-pixel. Additionally, the results are influenced by the wind conditions with a higher wind speed tending to reduce errors. This preliminary analysis suggests that when there is a significant discontinuity in surface conditions, particularly under low winds, the subpixel variability in energy fluxes will likely cause unacceptable errors in two-source model predictions. However, daytime wind speeds are typically >2 m s −1 and the resolution of T R (φ) observations from weather satellites are relatively coarse (i.e., ∼5–10 km), which means riparian areas are likely to comprise less than 10% of a pixel. Both of these factors are likely to reduce errors in heat flux predictions at these large spatial scales caused by using pixel-average inputs. The Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) instrument proposed for NASA's Earth Observing System (EOS) has 90 m resolution. This will be useful for evaluating the impact of subpixel variability on flux predictions with coarser resolution data more routinely available from weather satellites.