Retrieval of land surface bidirectional reflectivity in the mid-infrared from MODIS channels 22 and 23

This paper proposes a method to retrieve the land surface bidirectional reflectivity rho(b) in the mid-infrared (MIR) from MODIS channels 22 and 23. A split-window-like algorithm was developed to determine the MIR ground brightness temperature T-g(0) without the contribution of the solar direct beam from ground brightness temperatures measured at two adjacent MIR channels. Comparing the actual values of rho(b) with those estimated using our proposed method, the root mean square error (RMSE) was determined to be 0.0022 for the solar zenith angle (SZA) at nadir. Similar results were obtained for other SZAs, indicating that the proposed method can retrieve b accurately. A detailed sensitivity analysis found that the effects of instrumental noise, variations in water vapour content in the atmosphere, and reasonable variations in horizontal visibility on the retrieval of rho(b) were negligible and an error of 1K on the ground brightness temperature caused by the inaccuracy of atmospheric corrections can lead to a maximum error of 0.034 on the retrieved rho(b). We compared the use of radiosonde data and European Centre for Median-range Weather Forecast (ECMWF) atmospheric data to perform atmospheric corrections. The results show that atmospheric corrections can be performed with ECMWF data instead of radiosonde data without losing the accuracy of the retrieved rho(b). MODIS land surface temperature/emissivity product MOD11B1 data were also used to validate T-g(0). Comparison of T-g(0) estimated respectively with the proposed method and with MOD11B1 data showed that the RMSE is less than 1K for cloud-free skies.