physics.ao-phApr 9, 2026

CERBERUS: A Three-Headed Decoder for Vertical Cloud Profiles

Authors: Emily K. deJong, Nipun Gunawardena, Kevin Smalley, Hassan Beydoun, Peter Caldwell

Organizations: Lawrence Livermore National Laboratory, Livermore, CA 94550 USA

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Atmospheric clouds exhibit complex three-dimensional structure and microphysical details that are poorly constrained by the predominantly two-dimensional satellite observations available at global scales. This mismatch complicates data-driven learning and evaluation of cloud processes in weather and climate models, contributing to ongoing uncertainty in atmospheric physics. We introduce CERBERUS, a probabilistic inference framework for generating vertical radar reflectivity profiles from geostationary satellite brightness temperatures, near-surface meteorological variables, and temporal context. CERBERUS employs a three-headed encoder-decoder architecture to predict a zero-inflated (ZI) vertically-resolved distribution of radar reflectivity. Trained and evaluated using ground-based Ka-band radar observations at the ARM Southern Great Plains site, CERBERUS recovers coherent structures across cloud regimes, generalizes to withheld test periods, and provides uncertainty estimates that reflect physical ambiguity, particularly in multilayer and dynamically complex clouds. These results demonstrate the value of distribution-based learning targets for bridging observational scales, introducing a path toward model-relevant synthetic observations of clouds.

Figures & tables

Figure 1: Illustration of the CERBERUS three-headed architecture and neural network operations. Bold text indicates hyperparameters that were optimized.
Figure 2: Statistics of per-sample RMSE (dBZ) for each cloud regime (see Table A.1 ) over the validation set: median (orange), interquartile-range ( IQRIQR ; box), Q1/Q3±1.5IQRQ_{1}/Q_{3}\pm 1.5IQR (whiskers), and outliers (circles). Regime abbreviations defined in Table A.1 .
Figure 3: Test data and predicted reflectivity mean and uncertainty as a function of altitude, ranked according to quantile using altitude-scale energy (analogous to RMSE; see section A.2 ).
Figure 4: Illustration of the time-evolving reflectivity on Jan 29, 2025 (test set): true KAZR data (far left); and CERBERUS model mean and uncertainty (panels 2-3).
Appendix figures & tables10 assets

Supplementary material from the paper’s appendix.

Appendix

Figure A.1: Clear-sky vs cloud detection from GOES and KAZR at the ARM SGP site in 2022, aggregated across altitudes. Only the ”True Positive” data are included in the training and validation datasets.
Figure A.2: Distribution of aggregated target reflectivities from the training and validation KAZR data, aggregated across altitude (blue, orange) and timestamps (all). Blue includes all training/validation data with NaN-reflectivities filled with the -60dBZ detection threshold; orange indicates the PDF with these non-cloudy ( ≤−60\leq-60 dBZ) data removed; and green indicates the PDF of all reflectivities at 10km altitude aggregated across training/validation.
Figure A.3: Continuously ranked probability score (or MAE, for deterministic case) for the ZIB CERBERUS model and the non-ZI beta and deterministic baselines. CRPS is computed independently at each altitude for a given reflectivity measurement-prediction pair, and we report the mean CRPS across the training, validation, and testing datasets, respectively.
Figure A.4: Correlation coefficient R2R^{2} (left), RMSE (middle), and scaled Energy (RMSE) for the ZIB model and baselines. The first two metrics are reported for training, validation, and testing data splits. The Energy / RMSE is reported only for the validation dataset; Energy is scaled by the vector size ( ×1/128\times 1/\sqrt{128} ) to make it directly comparable to the RMSE that is reported for the deterministic model.
Figure A.5: Training and validation loss for the presented ZIB configuration of CERBERUS.
Figure A.6: (a) Receiver operating characteristic (ROC) curve for the classification head π\pi of CERBERUS evaluated on validation data; (b) parity between the true reflectivity values and the mean of the ZIB prediction, aggregated across altitudes and timestamps in the training and validation sets, respectively.
Cloud RegimeSatMAE RMSE in dBZCERBERUS RMSE / Energy in dBZ
(Abbreviation)(SEVIRI prevalence)(CERBERUS prevalence)
Cumulus (Cu)7.0 (0.8%)3.4 / 2.9 (7.8%)
Stratocumulus (Sc)7.5 (1.4%)4.7 / 3.9 (4.1%)
Stratus (St)4.8 (3.0%)6.2 / 5.1 (0.1%)
Altocumulus (AC)8.4 (0.6%)5.6 / 4.7 (9.6%)
Altostratus (AS)9.8 (0.4%)8.1 / 6.5 (10.1%)
Appendix
Table A.1: RMSE (mean across validation set samples), scaled Energy (for CEBERUS), and frequency of various cloud regimes, including results from CERBERUS at the SGP and results reported in Girtsou et al. (2025) for the SatMAE model over the MSG/SEVIRI satellite range. Clear-sky scenes are excluded from this analysis. Overall RMSE is reported across cloud regimes according to raw data weighting and according to SEVIRI cloud regime weighting. These data correspond to Figure 2 .
Figure A.7: Permutation feature importance with the original model loss (negative log likelihood) plotted as a red-dashed line. Features are plotting in order of least important/smallest change to loss, to most important, and are colored according to field type (GOES) or scalar type (meteorology/reanalysis).
Figure A.8: As in Figure 4 , for February 11, 2025 (test set).
Figure A.9: As in Figure 4 , for February 12, 2025 (test set).
More in this topic

Remote Sensing

Remote sensing uses satellite, aerial, and ground-based observations to measure and map Earth and other planetary surfaces.

See all 528 papers

Explore similar work

CardsList
Jul 18, 2026

C3DIR: A Deep Learning 3-Dimensional Cloud Property Retrieval Scheme for Passive Satellite Imagers

We develop the Cloud 3-Dimensional Imager Retrieval (C3DIR), a deep learning model that estimates 3-D cloud properties for multiple passive satellite imagers trained to match retrievals from the Earth Cloud Aerosol and Radiation Explorer(EarthCARE) ACM-CAP product. This work is aimed towards moving AI/ML 3-D cloud algorithms closer towards operational use. C3DIR predicts the occurrence water content of ice, cloud liquid, and rain along the imager line-of-sight and uses a voxel-level collocation approach to account for the misaligned viewing geometries of passive imagers and active profiling instruments. This precise collocation methodology allows for constructing vertical profiles using voxels contained by multiple imager pixels to facilitate comparisons with active profiling instruments. Qualitative case studies show that C3DIR can accurately depict multiple distinct overlapping cloud layers, albeit with some smoothing. Quantitative evaluations illustrate that C3DIR overall excels at hydrometeor detection which intuitively tends to be a function of water content. However, detection of voxels classified as liquid cloud remains difficult due to the their small geometric thickness, finer horizontal scale, and frequent tendency to be obscured or embedded within ice clouds. In general, water content estimation is reasonably accurate, yielding the best results in ice clouds but uncertainties remain for liquid and rain water content. Column-integrated water paths are in tighter agreement with EarthCARE. Comparisons with the algorithms underpinning current NOAA operational products highlight several areas where C3DIR may offer improvement. Overall, these results demonstrate the potential for C3DIR to provide flexible 3-D output depicting vertically resolved cloud structure which can offer broader utility for aviation applications, numerical weather modeling, and climate research.

Charles H. White, Yoo-Jeong Noh, John M. Haynes +1

Cooperative Institute for Research in the Atmosphere, Colorado State University, Fort Collins, CO. · Electrical and Computer Engineering, Colorado State University, Fort Collins, CO.

Remote Sensing
AskPDF
May 31, 2026

Cross-Modal Hierarchical Fusion for from Multi-Sensor Ground Observation

Dense volumetric reconstruction of cloud microphysical fields from sparse ground-based instruments remains an open problem, largely because the available measurements are heterogeneous in both modality and spatial coverage. We present AtmoFuseNet, a framework that fuses multi-view sky camera imagery with millimeter-wave cloud radar and ceilometer observations to produce 4D (three spatial dimensions plus time) estimates of cloud state and wind. The method operates in three stages: a cross-modal hierarchical aggregation module that combines image feature pyramids with instrument-derived vertical profiles through layer-wise cross-attention; a conditional variational refinement module that maps the resulting volume to physically consistent microphysical fields under differentiable radar and image forward models; and a correlation-based motion estimator that recovers per-voxel 3D wind vectors from consecutive volumetric reconstructions. On collocated observations from a semi-arid site, AtmoFuseNet reaches 0.026 g m^-3 liquid water content MAE and 1.18 m s^-1 wind speed MAE, improving over existing retrieval baselines. Ablation experiments isolate the contribution of each module.

Xinze Zhang

University of Southern California, Los Angeles, CA 90007, USA

Image ReconstructionRemote SensingMultimodal Sensor Fusion
AskPDF
Aug 26, 2026

Tropospheric temperature and humidity profile retrieval from Meteosat Flexible Combined Imager based on deep learning

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.

Alejandro Salgueiro, Johannes Rausch, Julie Thérèse Villinger +1

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

Remote Sensing
AskPDF