Publications

Zhang, Q; Cheng, J; Liang, SL (2018). Deriving high-quality surface emissivity spectra from atmospheric infrared sounder data using cumulative distribution function matching and principal component analysis regression. REMOTE SENSING OF ENVIRONMENT, 211, 388-399.

Abstract
The Atmospheric Infrared Sounder (AIRS) provides limited hyperspectral thermal infrared (TIR) emissivity data for the retrieval of critical land surface and climate parameters in environmental research. However, the AIRS land surface emissivity (LSE) data lack accuracy, resulting in low-quality data retrieval, particularly for the lower boundary layer. In this study, a practical and effective method is proposed to derive high-accuracy AIRS LSE data and continuous emissivity spectra in the TIR range of 8-14.5 mu m. The AIRS LSE is first rescaled to the Moderate Resolution Imaging Spectroradiometer (MODIS) LSE using cumulative distribution function (CDF) matching, and then the emissivity spectra are recovered from the rescaled AIRS LSE using principal component analysis (PCA) regression. The results show that resealing the AIRS LSE significantly reduced the bias and root mean square (RMS) error in the study area of Africa and the Arabian Peninsula, and PCA regression successfully recovered the emissivity spectra in the 8-14.5 mu m range from the resealed AIRS LSE. At two validation sites in the Namib and Kalahari deserts of southern Africa, the biases of the resealed AIRS LSE at three hinge points are 0.62% and 0.61%, respectively, and the biases of the recovered AIRS LSE spectra in the 8-12 mu m TIR range are 0.53% and 0.56%, respectively. Variations in land cover homogeneity and the accuracy of the MODIS LSE are the critical factors impacting the final accuracy of the rescaled AIRS LSE and the recovered emissivity spectra.

DOI:
10.1016/j.rse.2018.04.033

ISSN:
0034-4257