|Kustas, William - Bill|
|La Loggia, Goffredo|
Submitted to: Remote Sensing of Environment
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 6/7/2012
Publication Date: 9/1/2012
Publication URL: http://handle.nal.usda.gov/10113/59972
Citation: Cammalleri, C., Anderson, M.C., Ciraolo, G., D'Urso, G., Kustas, W.P., La Loggia, G., Minacapilli, M. 2012. Practical applications of the remote sensing-based two-source algorithm for mapping surface energy fluxes without in-situ air temperature observations. Remote Sensing of Environment. 124:502-515. Interpretive Summary: Remote sensing techniques for mapping evapotranspiration often use satellite-derived maps of land-surface temperature as the key boundary condition. The difference between the surface and near-surface air temperature controls the heat energy flux convected from the surface to the atmosphere, while the remainder of the energy available at the land-surface is presumed to be used in evaporating water. Therefore, evapotranspiration estimates from such models can be very sensitive to errors in the assumed surface-to-air temperature gradient. It is beneficial to build remote sensing models that minimize this sensitivity, because often accurate air temperature data are unavailable. In this paper, three variants of a surface energy balance algorithm are compared to ascertain which might be most useful for applications where meteorological observations are sparse or unavailable. The study determined that a simple method requiring no air temperature data performed reasonably over an agricultural landscape in Sicily (Italy). In this method, it was assumed that the coldest areas in the land-surface temperature map were evaporating at the potential rate. A representative air temperature over the modeling regions was back-calculated based on this assumption and used as an upper boundary for flux computations at all other pixels in the scene. This simple approach may be a useful means for estimating evapotranspiration in data-sparse areas where well-watered fields exist somewhere in the satellite imaging scene.
Technical Abstract: The two-source energy balance (TSEB) model uses remotely sensed maps of land-surface temperature (LST) along with local air temperature estimates at a nominal blending height to model heat and water fluxes across a landscape, partitioned between dual sources of canopy and soil. For operational implementation of the TSEB, however, it is often difficult to obtain representative air temperature data that are consistent with the LST retrievals, which may themselves have residual errors due to atmospheric and emissivity corrections. To address this issue, two different strategies in applying the TSEB model without requiring local air temperature data were tested over a typical Mediterranean agricultural area using a set of resolution multi-spectral airborne remote sensing images. Alleviating the need for accurate local air temperature data as input, these two approaches internally estimate the surface-to-air temperature gradient that drives the sensible heat flux. The two approaches include: 1) a scene-based internal calibration (TSEB-IC) procedure that estimates air temperature over a well-watered and fully vegetated pixel in the LST image, and 2) a disaggregation scheme (DisALEXI) that uses air temperature estimates from a coupled TSEB-atmospheric boundary layer model of atmosphere-land exchange (ALEXI). A comparison of the air temperatures modeled by TSEB-IC and DisALEXI with in-situ weather station observations shows good agreement, with average differences on the order of 1 K, comparable with the uncertainties in the remotely sensed surface temperature maps. Surface fluxes estimated by each method agree well with micro-meteorological measurements acquired over an olive orchard within the aircraft imaging domain. In comparison with fluxes generated with TSEB using local measurements of air temperature, instantaneous fluxes from these alternative methods show good spatial agreement, with differences of less than 10 W/m2 across the domain. Finally, a sensitivity analysis of the three models, performed by introducing artificial errors into the model inputs, demonstrates that the DisALEXI and TSEB-IC approaches are relatively insensitive to errors in absolute surface temperature calibration, while turbulent fluxes from TSEB applications using local air temperature measurements show sensitivity of approximately 30 W/m2 per degree temperature perturbation. This highlights the value of internal estimation of the surface-to-air temperature gradient within a surface energy balance framework.