Tropospheric temperature and humidity profile retrieval from Meteosat Flexible Combined Imager based on deep learning
Authors: Alejandro Salgueiro, Johannes Rausch, Julie Thérèse Villinger, Angela Meyer
Organizations: Department of Geoscience and Remote Sensing, Delft University of Technology, 2628 CN Delft, The Netherlands · School of Engineering and Computer Science, Bern University of Applied Sciences, 2502 Biel/Bienne, Switzerland · Meteomatics AG
The Meteosat Third Generation (MTG) Flexible Combined Imager (FCI) offers new opportunities for tropospheric temperature and humidity profiling, at higher spatio-temporal resolutions and expanded spectral coverage relative to its predecessor. Vertically resolved retrievals from broadband imagers are inherently challenging, and operational retrieval algorithms typically rely on numerical weather prediction (NWP) background fields to compensate for limited infrared spectral resolution, reducing the retrievals' independence. We develop a spatially aware deep learning framework to retrieve all-sky tropospheric temperature and humidity profiles from FCI, without forecast profiles as input. A Residual U-Net that exploits spatial context across all 16 FCI channels was trained on 14 months of collocated FCI observations and CERRA reanalysis targets over Europe. Validated against independent radiosondes, retrieved temperatures show biases below 0.4 K and standard deviations of 1.5-1.9 K. Retrieved relative humidity standard deviations range from 12-20 %, compared to 9-19 % for CERRA. Performance degrades modestly under clouds, with standard deviation increases below 0.4 K and 3 % RH beneath cloud tops despite limited direct radiative information. Ablation experiments show that spatial context improves retrievals, with the largest gains below cloud tops. Feature sensitivity analysis indicates broad consistency with FCI bands' established radiative transfer characteristics. Visible and near-infrared channels contribute despite not being commonly used in physics-based profile inversions. These results demonstrate that spatially aware deep learning models can extract statistically reliable tropospheric profiles from geostationary imager observations, independent of NWP forecast fields, enabling more rapid autonomous monitoring of the atmosphere.
Figures & tables
Figure 1: (a) True-colour RGB composite constructed from the FCI visible channels (0.4, 0.5, and 0.6 μ m). The red rectangle indicates the study domain. (b) Plate carree projection of the study domain map showing the locations of available IGRA radiosonde stations and satellite zenith angle. Values exceeding 75∘ are masked to exclude large viewing geometries.
Figure 2: Residual U-Net with squeeze and excitation and dilated bottleneck blocks. In blue are the components with trainable parameters, in green the downsample and upsample components, in orange the main blocks and the rest of the components in gray.
Figure 3: Test set bias and STD of profiles of (a) temperature, (b) specific humidity, and (c) relative humidity against radiosonde measurements in all sky conditions for CERRA, the U-Net, the 1x1-Net, the ancillary-only U-Net and ERA5 climatology. The subpanel (d) shows the total number of valid radiosonde measurements from all stations (Figure 1 b) used for each pressure level.
Figure 4: Test set bias and STD of retrieved profiles of (a) temperature, (b) specific humidity, and (c) relative humidity against radiosonde measurements for above and below clouds with the U-Net and 1x1-Net. Blue and green lines represent the errors below cloud while orange and red above the cloud top. The subpanel (d) shows the total number of valid radiosonde measurements used for each cloud condition.
Figure 5: Bias and STDs of our retrieved profiles for the first and second test sets against radiosondes. Blue: main test set (DJF 2024/25), using the rolling-split test days (Figure S5). Orange: second test set (DJF 2025/26), using all hourly retrievals.
Figure 6: Spatial validation of retrieved atmospheric profiles against CERRA reanalysis data at 300 hPa on October 23, 2024 at 0600 UTC. Panels (a-f) illustrate the full spatial domain, and panels (g-l) display the target zoomed region marked by the solid black rectangle. Columns compare retrieved (a,g) and target (b,h) temperature, retrieved (c,i) and target (d,j) specific humidity, alongside contextual satellite inputs: FCI Water Vapor channel 7.3 μ m (e, k) and Cloud Top Pressure (f, l). The black dashed line indicates the cross-section track analyzed in Figure 8 .
Figure 7: Same as Figure 6 for the 900 hPa pressure level. Areas where the measurements are below pressure levels are masked in black.
Figure 8: Vertical cross-section comparison between retrievals and CERRA for temperature (a-b), specific humidity (c-d), and relative humidity (e-f) along a selected transect. The columns show retrievals (left) and CERRA fields (right). Purple lines indicate the cloud top pressure along the transect.
Figure 9: Radially averaged dimensional variance spectral density of (a) temperature and (b) specific humidity as a function of spatial wavelength (in pixels) for selected pressure levels. Solid lines represent the U-Net retrieval, and dashed lines correspond to the CERRA reference targets. Vertical dashed lines indicate equivalent physical distances based on the average grid spacing of 2.78 km/pixel. Technical details of the spectral computation are provided in the supporting information Text S5.
Figure 10: Test set spatial RMSE distribution of retrieved atmospheric profiles evaluated against CERRA reanalysis. (a-c) Temperature and (d-f) relative humidity across three pressure levels. Contour lines added for satellite zenith angles of 60, 65, 70 and 75°.
Figure 11: Feature importance of the (a) temperature and (b) relative humidity profiles as the difference in RMSE compared to retrievals without substitution.
Variable
IASI
ABI
FCI U-Net
Temperature bias
-1 to 0.2 K
-0.4 to 0.2 K
-0.3 to 0.1 K
Temperature STD
1.0 to 2.0 K
0.7 to 2.0 K
1.5 to 1.9 K
Specific humidity bias
-0.1 to 0.2 g/kg
-
-0.03 to 0.13 g/kg
Specific humidity STD
0.0 to 1.2 g/kg
-
0.04 to 1.21 g/kg
Relative humidity bias
-
0 to 16 %
1 to 25 %
Relative humidity STD
-
11 to 15 %
12 to 20 %
Table 1: Comparison of our FCI retrievals with IASI and ABI bias and STDs against radiosondes. IASI and ABI bias and STDs were approximated: for IASI from Figures 30-31 of the IASI level 2 validation report ( EUMETSAT, 2021b ) , and for ABI from Schmit et al. (2019) validation report for GOES-16 Figure 2.
Hyperspectral infrared observations are an important data source for numerical weather prediction (NWP) because they provide rich information on the vertical structure of atmospheric temperature and humidity. However, most existing deep learning methods mainly focus on one-way retrieval from radiances to atmospheric profiles, while the reverse radiance simulation process and the consistency between atmospheric state space and radiance observation space are insufficiently considered. In this study, we propose SIMBA, a unified bidirectional retrieval-forward simulation framework for FY-4A GIIRS hyperspectral infrared radiance modeling toward NWP applications. The framework jointly performs atmospheric profile retrieval and radiance reconstruction, introduces a cycle-consistency constraint to strengthen the coupling between the two processes, and employs a bidirectional Mamba state-space module to capture long-range dependencies along pressure levels. Using collocated FY-4A GIIRS observations and ERA5 reanalysis data, the proposed method is evaluated for temperature retrieval, specific humidity retrieval, long-wave radiance reconstruction, and medium-wave radiance reconstruction. Experimental results show that SIMBA outperforms several representative deep learning baselines across both retrieval and reconstruction tasks, while ablation experiments confirm the contribution of the bidirectional design and cycle-consistency mechanism. These results demonstrate that the proposed framework is effective for joint atmospheric profile retrieval and hyperspectral infrared radiance modeling, and suggest potential for future Jacobian-related analysis and NWP-oriented extensions.
Jingdong Shen, Fu Wang*, Qifeng Lu +4
School of Computer Science and Engineering, Sichuan University of Science & Engineering, Yibin 644000, China · State Key Laboratory of Severe Weather Meteorological Science and Technology, CMA Earth System Modeling and Prediction Centre and Key Laboratory of Earth System Modeling and Prediction, China Meteorological Administration, Beijing, 100081
Accurate Surface Solar Irradiance (SSI) estimation is increasingly important for photovoltaic energy monitoring and forecasting. The recently introduced Meteosat Third Generation (MTG) satellite constellation provides imaging data with higher spatial resolution compared to the Meteosat Second Generation (MSG) satellite constellation, but its benefits for machine-learning-based SSI retrieval have not been well established. In this work, we introduce a multi-imager and multi-resolution convolutional neural network architecture for 10-minute SSI retrieval over Northern Europe (Estonia) using MSG/SEVIRI and MTG/FCI satellite imagery together with solar-geometry and clear-sky irradiance features. Model performance is evaluated against ground-based pyranometer measurements from eight Estonian meteorological stations using site-based cross-validation and multiple training seeds. Model performance is also compared with the SARAH-3 physics-based satellite SSI product. The hybrid SEVIRI-FCI model significantly outperformed the SEVIRI-only model under overcast and cloudy conditions, reducing RMSE by 8.2 W m−2 and 5.7 W m−2, respectively. However, under partly cloudy or clear skies, no statistically significant difference in RMSE was observed between the SEVIRI-FCI hybrid and the SEVIRI-only models. Compared with physics-based SARAH-3, the hybrid model yielded skill scores of 35 % under overcast conditions, 21 % under cloudy conditions, and 20 % overall. Furthermore, both models underperformed SARAH-3 in clear-sky conditions. These results show that higher-resolution MTG/FCI imagery improves CNN-based SSI retrieval when clouds dominate irradiance variability, but also indicate that higher spatial resolution alone is insufficient to address clear-sky limitations in machine-learning-based SSI retrieval.
Gordei Pribõtkin, Piia Post, Velle Toll
Institute of Computer Science University of Tartu STACC OÜ Tartu, Estonia · Centre for Climate Research, Institute of Physics University of Tartu Tartu, Estonia
Accurate regional near-surface temperature forecasting is fundamental to short-range weather services and downstream risk assessment. Existing deep learning-based regional forecasters commonly produce a fixed set of future frames on a prescribed grid, limiting their use when forecast products must be evaluated at query-dependent lead times or display resolutions. To overcome these fixed-output constraints, we formulate regional T2M forecasting as query-conditioned continuous spatiotemporal temperature field evaluation and propose the Continuous Spatiotemporal Temperature Forecaster (CSTF), a neural field that turns forecast lead time and output resolution into explicit queries when evaluating 2-m temperature (T2M). Specifically, CSTF first encodes multivariable ERA5 histories into latent meteorological states and then decodes T2M as a coordinate-based field. Accordingly, spatial location, forecast lead time, and output resolution are introduced as queries, enabling standard hourly forecasts, intermediate lead-time diagnostics, and resolution-controllable outputs within a unified field-evaluation framework. Furthermore, to maintain coherence across flexible field queries, we design spatial-gradient, temporal-difference, and scale-consistency objectives that regularize regional thermal structures, lead-wise evolution, and cross-resolution agreement. Experiments on the Southeast China 0-6 h ERA5-Land benchmark demonstrate that CSTF achieves the best aggregate deterministic skill, including a 17.0 percent reduction in Bias, with global-scope diagnostics further illustrating flexible lead-time and resolution-controllable inference.
Chunlei Shi, Jiong Wang, Yi-Lin Wei +4
Department of Automation, Southeast University, Nanjing 210096, China