18 March 2025 in IEEE Transactions on Geoscience and Remote Sensing
The potential applications of high spatiotemporal resolution Land Surface Temperature (LST) products are extensive. However, the trade-off between spatial and temporal resolutions of remote sensing data has significantly constrained the availability of such LST products. Existing spatiotemporal fusion methods appear to encounter certain limitations. This article proposes a two-stage hierarchical spatiotemporal fusion network (THSTNet) to fuse MODIS LST products and Landsat LST products. THSTNet adopts the shift windows (Swin) transformer architecture and employs a two-stage process to reconstruct the fine-resolution image at the target time. The innovation within THSTNet is multifaceted: it combines spatiotemporal mapping with deep learning, enhancing the extraction of global information for fusion results; leveraging self-attention computation, it adopts a two-stage structure to improve the understanding of intricate LST changes; additionally, it incorporates a texture converter module aimed at enhancing spatial details within the reconstruction results. Validation of the model's predictions was conducted using actual images and ground observations, affirming the high reliability of THSTNet's predictive outcomes. Compared with two traditional methods (STARFM and ESTARFM) and four deep learning based methods (EDCSTFN, GANSTFM, STTFN, and MSNet), THSTNet demonstrated superior performance [average RMSE is below 1.3K, and average SSIM is 0.939]. The prediction results of THSTNet also maintain high consistency with ground observations [average RMSE is 2.3K, and average R2 is 0.9]. (Code will be available at https://github.com/HuPengHua2021/THSTNet).
22 January 2025 in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Passive microwave (PMW) remote sensing is increasingly employed for generating seamless all-weather land surface temperature (LST) data due to its ability to penetrate cloud cover and capture the actual surface conditions underneath. Existing passive microwave retrieval methods often utilize large amounts of remote sensing data, overlooking the fact that redundant data can increase computational and time costs, reduce model interpretability, and may negatively impact accuracy. In this paper, we proposed a PMW-LST retrieval method that integrates CatBoost-Based adaptive feature selection. First, we categorized the data into six groups based on the underlying surface types and data view time. Next, for each group, we ranked the feature sets according to their importance and employed the Recursive Feature Elimination (RFE) method for feature selection. Finally, the optimized feature sets were used in the CatBoost algorithm to construct the PMW-LST retrieval model. We compared the accuracy of the proposed method with the Holmes, multi-channel, and Random Forest algorithms. Results showed that the proposed method had lowest RMSE, with the value of 3.28K (1.95K), 2.69K(1.65K), and 3.71K (2.22K) on grassland, cropland, and barren land at daytime (nighttime), respectively. The verification at sites in Heihe river basin shows that the ubRMSE ranges from 1.73K to 4.48K at daytime and 2.71K to 3.19K at nighttime under clear-sky conditions, and from 1.83K to 5.23K at daytime and 2.77K to 3.93K at nighttime under cloudy-sky conditions. These results indicate the proposed method achieves higher accuracy in generating seamless all-weather LST data.
11 July 2025 in IEEE Transactions on Geoscience and Remote Sensing
Land surface temperature (LST) is a crucial parameter linked to water and carbon cycles. However, due to the wide swath widths of polar-orbiting satellites, the temporal inconsistencies in each pixel observation lead to significant LST variations, resulting in temporal incomparability in the current LST products acquired from these satellites. Therefore, a physical method based on the surface energy balance and Bowen ratio (SEB-BR) was developed to correct the temporal effects on the LST. The advantage of the SEB-BR approach lies in its ability to establish a physical connection between various physical parameters and the LST, quantifying each parameter’s contribution to LST variation and thereby mechanically linking physical variables that contribute to LST variation. A diurnal temperature cycle (DTC) model named GOT01-dT and random forest regression (RFR) were introduced for comparison. Results showed that the SEB-BR model achieved higher accuracy with an average RMSE (bias) of 2.85 (1.26) K, outperforming the RFR model (2.96 (1.36) K) and GOT01-dT model (2.94 (0.39) K). The SEB-BR method also demonstrated better agreement with in-situ LST across various local solar time differences, particularly for temporal differences between 0.8–0.9 hours, where the RMSE decreased significantly from 3.54 K to 2.63 K. The improvements in accuracy across all seasons, reflected in the RMSE reductions from 3.56 (2.08) K to 3.39 (1.80) K, were significant. These results highlight the effectiveness of the SEB-BR approach in correcting temporal effects and its potential for large-scale temporal normalization, offering strong interpretability and spatiotemporal extensibility.
16 April 2026 in IEEE Transactions on Geoscience and Remote Sensing
Land surface temperature (LST) is a key parameter in the land surface energy cycle. However, due to cloud contamination, LST products derived from thermal infrared remote sensing exhibit significant missing areas. Although many studies have attempted to reconstruct missing areas using reanalysis data, such as the land component of the fifth generation of European ReAnalysis (ERA5 Land), existing methods neglect the temporal differences between MODIS and reanalysis data, suffering from several limitations in applicability. Therefore, a two-stage surface energy balance-guided graph neural network (TSSEB-GNN) was developed, with the innovation of coupling the network with physical mechanisms. First, the physical factors influencing LST were determined based on surface energy balance theory and the Bowen ratio. Second, the stage one model based on GNN module was designed to generate an initial gap-free LST. Third, the second stage model fuses the initial LST by adding sparse and dense self-attention mechanisms and incorporating twelve physical factors to capture spatial and temporal differences. The results show that the TSSEB-GNN model achieves a root mean square error (RMSE) of 2.56–3.45 K under clear-sky conditions and 3.39–3.63 K under cloudy-sky conditions. Compared with existing methods, TSSEB-GNN not only provides higher accuracy but is also less affected by long-term gaps in MODIS LST. Ablation experiments demonstrate that the TSSEB-GNN model achieves the best performance by accounting for both temporal and spatial differences. The results highlight the effectiveness of the TSSEB-GNN, its interpretability, improved applicability, and greater potential for large-scale applications.
1 January 2026 in IEEE Transactions on Geoscience and Remote Sensing