Publications

Wang, YJ; Zhou, J; Tang, WB; Ma, J; Wang, ZW; Wang, SF (2025). A Time Series Method With Physically Guided Selection of Surface Indicators for Passive Microwave Brightness Temperature Swath Gap-Filling. IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, 63, 5300615.

Abstract
Passive microwave brightness temperature (PMW BT) images acquired by PMW imagers onboard polar-orbit satellites suffer from large observations missing between adjacent orbits due to the swath width of images, i.e., PMW BT swath gaps, limiting the spatiotemporal integrity and application potential of PMW BT-generated products. Here, we propose a gap-filling method [i.e., physical indicators-guided CNN-LSTM (PICL)] for PMW BT images by physically guided selection of surface indicators with CNN-LSTM model, which is suitable for special underlying surfaces (e.g., seasonal permafrost and snow) using only BT data to generate spatially gapless PMW BT images. The core of PICL is to use the CNN-LSTM model to capture the relationship of BT time series, thereby filling the missing BT values via historical BT data. PICL is applied to 7, 10, 18.7, 36, and 89 GHz frequencies of Advanced Microwave Scanning Radiometer 2 (AMSR2) for the Tibetan Plateau (TP). Validation results show good accuracy of the PICL filled BT, with the root-mean-squared error (RMSE) ranging from 1.28 to 2.43 K (<89 GHz), and the accuracy decreases as the frequency increases. The reconstructed BT images agree well with the original AMSR2 BT images and show no obvious boundary effect. PICL also has a good ability in capturing the temporal trends and discontinuities caused by snow and seasonal permafrost. PICL only requires historical BT before the missing moment, highlighting its feasibility to be extended to other satellite PMW imagers. It enables the generation of spatially seamless products such as all-weather land surface temperature (LST) and soil moisture.

DOI:
10.1109/TGRS.2025.3534825

ISSN:
1558-0644