Retrieving Middle-Infrared Reflectance Using Physical and Empirical Approaches: Implications for Burned Area Monitoring

A systematic comparison is carried out between retrieved values of middle-infrared (MIR) reflectance by means of the complete radiative transfer equation (RTE) and the simplified algorithm proposed by Kaufman and Remer in 1994 (KR94). The added value to be expected when using RTE is assessed both within and beyond the region where KR94 produces usable estimates of MIR reflectance, paying special attention to their application for discriminating burned areas (BAs) in tropical environments, where KR94 is the most common approach. For large values of land surface temperature (LST) and solar zenith angle (SZA), the retrieval of MIR reflectance based either on RTE or KR94 is an ill-posed problem, i.e., small perturbations due to sensor noise and uncertainties in atmospheric profiles and LST may induce large errors in the retrieved values. It is found that the RTE approach leads to better estimates in virtually all cases, with the exception of high values of LST and SZA, where results from KR94 are also not usable. Impacts on BA discrimination were finally evaluated using Moderate Resolution Imaging Spectroradiometer imagery showing a large fire event in southern Brazil. Synthetic values were generated, assuming a hot tropical environment, and MIR reflectance was retrieved using the two approaches. Whereas retrieved values of MIR reflectance via KR94 did not allow an effective discrimination between burned and unburned areas, those obtained via RTE have shown to be usable for BA monitoring, opening good perspectives for successful applications in hot tropical environments.