Pith. sign in

REVIEW 4 major objections 6 minor 77 references

Global spatio-temporal downscaling of ERA5 precipitation through generative AI

T0 review · 4 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A Germany-trained generative network turns ERA5's coarse rain into 2 km, 10-minute fields that match radar, in Germany, the US, and Australia.

desk verdict A solid, well-evaluated cGAN for downscaling ERA5 precipitation that earns a serious referee, but the 'global applicability' claim outruns what three regions and seven weeks can support. read the letter →

arxiv 2411.16098 v1 pith:7FCXIQAS submitted 2024-11-22 physics.ao-ph cs.CVcs.LGphysics.geo-ph

Luca Glawion , Julius Polz , Harald Kunstmann , Benjamin Fersch , Christian Chwala This is my paper · ORCID
classification physics.ao-phcs.CVcs.LGphysics.geo-ph PACS 92.60.Ry
keywords downscalingprecipitationgenerativeadversarialnetworkERA5weatherradarextremespatio-temporalensembleuncertainty
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

ERA5's global precipitation record is long and consistent, but its 24 km, hourly grid averages out the small convective cells that cause flash floods, and its extreme-value tail is too weak. The paper argues that a conditional generative adversarial network can fix that: given only ERA5's convective and large-scale rain fields, it produces 2 km, 10-minute rain maps whose spatial structure, motion, and rain-rate distribution (including heavy events) resemble weather radar. The network is trained on German radar alone, yet evaluation on US and Australian radar in 2021 shows it transfers to other climate zones, which is the basis for the claim of global applicability and for the "first global deep-learning spatio-temporal downscaling" label. Because inference takes 0.04 s per patch, the same model can generate large probabilistic ensembles, making the downscaling uncertainty explicit rather than hidden.

What carries the argument

The load-bearing object is the conditional generative adversarial network, with a 3D-convolutional residual generator, a 3D ResNet discriminator, and temporally constant dropout as the stochastic source for ensembles. Inputs are the two ERA5 variables convective and large-scale precipitation; the discriminator compares generated or observed high-resolution video sequences against the same coarse context. Three pieces make the approach work: the UNET-like multiresolution path with skip-crop connections, the adversarial plus ensemble-L1 loss, and an inference-time patch-wise mean-field bias correction that anchors each output to the ERA5 patch average. Around the core network, a patch-stitching pipeline with overlapping domains and linear blending assembles seamless global 0.018-degree fields.

What would settle it

Run the released model on a full year of ERA5 over a region outside the three evaluated countries—say tropical West Africa or monsoon Asia—and compare the downscaled fields to gauge-adjusted radar or dense gauge networks; if the fractions skill score for rain rates above 5 mm/h does not beat interpolation, or the rank histograms turn strongly U-shaped, the global-transferability claim fails.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that one cGAN, trained on 12 years of gauge-adjusted German radar, can disaggregate ERA5 precipitation into fields statistically indistinguishable from radar at 2 km and 10 minutes, even outside its training region. The generator turns a 672 km, 16-hour ERA5 context patch into a 336 km, 8-hour prediction, and the discriminator acts as a learned loss that pushes the output toward realistic structures and intensities. A fixed dropout seed per ensemble member plus an ensemble L1 loss and a mean-field bias correction give a calibrated probabilistic product that keeps ERA5's patch-average rain amount while restoring the missing high-intensity tail. In the three-region 2021 evaluation, the generated fields beat rainFARM and trilinear interpolation on fractions skill score at intense thresholds, closely match radar power spectra and anisotropy, and show only slight underdispersion in rank histograms.

Load-bearing premise

The global claim rests on the assumption that the relationship between ERA5's coarse rain and true fine-scale rain, learned from German radar, stays roughly the same in every climate zone, so a Germany-only training set can serve the whole planet.

Editorial extensions

If this is right

  • Downscaled fields can be generated for the entire globe by patch stitching, including ocean areas where no high-resolution observations exist.
  • Large ensembles are cheap: one patch takes 0.04 s on a single GPU, so uncertainty quantification becomes routine rather than a computational obstacle.
  • The reconstructed heavy-rain tail and realistic cell anisotropy make the output suitable for flood-risk and impact studies that currently avoid ERA5 because it misses extremes.
  • Evaluations in Germany, the US, and Australia show the trained network transfers across climate zones, so a single training set can serve many regions.
  • The method is generic: the same cGAN pipeline can be retargeted to other input datasets and resolutions because it only learns a conditional mapping from coarse to fine fields.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Because the mean-field constraint ties each output to the ERA5 patch average, any regional bias in ERA5 propagates into the downscaled product; users should apply a regional bias correction before interpreting the high-resolution fields.
  • The paper evaluates only one forecast year and only three countries; the strongest test of global transferability would be an independent evaluation over a tropical or monsoon region with very different storm phenomenology, which the paper itself lists as a limitation.
  • The slight underdispersion in rank histograms and the case-study note that ensembles vary more in intensity than in position suggest these fields are safest for probabilistic intensity distributions and less suited to event-by-event matching.
  • A testable extension: train the same network on a second radar network from a convective regime and compare the marginal gains in fractions skill score outside both training domains; if gains are large, the Germany-only training choice was not optimal.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. SpateGAN-ERA5 is a conditional GAN that takes ERA5 convective and large-scale precipitation fields at 24 km and hourly resolution and generates 2 km and 10-minute precipitation fields, trained on gauge-adjusted radar (RADKLIM-YW) over Germany. The model uses 3D convolutional residual blocks, dropout-based ensembles, and a patch-wise mean-field bias correction that preserves the ERA5 patch average. The authors evaluate the model on held-out German data and on US MRMS and Australian radar data from the first week of each month from July to December 2021, comparing against rainFARM and trilinear interpolation. They report FSS, CRPS, rank histograms, RAPSD, and anisotropy measures, and conclude that spateGAN-ERA5 produces realistic spatio-temporal rainfields, well-calibrated ensembles, and strong generalization, suggesting robust global applicability.

Significance. If the results hold, spateGAN-ERA5 is a valuable step for km-scale precipitation downscaling: it is one of the first demonstrations of a single cGAN mapping ERA5 to a 2-km, 10-minute product and transferring to three different continents. The evaluation design is genuinely independent: training (2009-2020), model selection (Jan-Jun 2021), and evaluation (Jul-Dec 2021) are separated in time, and evaluation regions are outside the training area. The paper provides a useful comparison to rainFARM and shows that the generative approach outperforms statistical downscaling in distributional and structural metrics, with a better-calibrated ensemble (CRPS, rank histograms). The computational speed (0.04 s per patch) makes large ensembles feasible. However, the global-generalization claim is stronger than the evidence, and the structural scores are computed on a favorable subset, so the headline claims need to be tempered or supplemented.

major comments (4)
  1. [Abstract; Section 6] The claim that spateGAN-ERA5 demonstrates 'robust global applicability' is not supported by the evidence presented. The evaluation covers three countries (Germany, US, Australia) and only the first week of each month from July to December 2021; most of the world's climate regimes, including tropical monsoon, arid, and polar regimes, are not tested. Section A.3 itself concedes that 'especially extremely strong rain events which rarely occur in Germany can lead to more unrealistic spatial patterns'. Since global applicability is the central headline claim, the authors should either provide additional validation (e.g., in a tropical region or over a longer period) or restrict the claim explicitly to regimes similar to those evaluated.
  2. [Section 7.5.2; Fig. 4] The spatial-structure analysis (RAPSD and anisotropy) is restricted to a subset of the evaluation data with interpolated ERA5 mFSS > 0.2. This selection removes the most challenging cases, for which ERA5 has little skill, and thus biases the structural scores toward favorable outcomes. The paper should report the number/percentage of cases that pass this filter, show that the conclusions are insensitive to the threshold, or analyze the full dataset with a fallback for non-rain cases. This is particularly important for the claim of realistic small-scale patterns.
  3. [Section 7.1; Section A.2.4; Fig. 3] The model's mean-field bias correction preserves the patch-averaged ERA5 precipitation (Section 7.1). Consequently, the predicted rain-rate distribution -- including extremes -- inherits the regional bias of ERA5, which the paper itself documents at ERA5 resolution (Section A.2.4, Fig. A9). The abstract's claim of 'accurate rain rate distribution including extremes' is therefore too strong: Fig. 3b shows overestimation of strong precipitation frequencies for Germany and underestimation for Australia and the US, consistent with the ERA5-radar bias (Table A1). The claim should be qualified as accurate after adjustment to the ERA5 mean, or the extremes should be evaluated relative to the reference product after the same mean-field correction.
  4. [Section A.1.1; Section 7.5.2] The training sample selection (A.1.1) retains only samples with high precipitation totals and a high 66th quantile in both ERA5 and radar, and the model-selection target is rescaled to the ERA5 mean (7.5.1). This design chooses well-paired events, which is sensible for learning, but it means the model has not seen many poorly paired regimes; the subset filter in 7.5.2 compounds this by evaluating structure only on well-simulated cases. The manuscript should quantify how many samples are lost at each stage and discuss how the selection affects the generalization claims.
minor comments (6)
  1. [Section 7.4.1] There is a broken sentence: 'y lack valuable scale-related information [29, 51, 52], excludes oceans and coastal areas and has a higher release latency [53].' The intended subject and verb are missing; please revise.
  2. [Section 2 vs Section 7.2] The training setup is described inconsistently: Section 2 states 'Data-parallel training on 4 A100 GPUs took 3 days', while Section 7.2 states 'data-parallel training on 3 Nvidia A100 GPUs for 4 days.' Please correct.
  3. [Table A1] The table header 'RADKLIM MRMS Australia' is not formatted correctly; 'MRMS' should be 'US (MRMS)' or similar. Also, the BIAS metric in Eq. A13 is defined as (Y - X)/Y, which is the negative of the usual bias (X - Y)/Y; please state the sign convention explicitly.
  4. [Section A.1.3] Typo: 'This enures that' should be 'ensures'.
  5. [Section A.2.2] Typo: 'sspateGAN-ERA5' appears instead of 'spateGAN-ERA5'.
  6. [Section 3] The statement that rainFARM 'fails' is too strong given that rainFARM has lower MAE/RMSE in all regions (Table A1) and only slightly worse CRPS; please moderate the language to 'does not reproduce the small-scale convective structures'.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity; the evaluation is externally benchmarked and the self-citation is non-load-bearing.

full rationale

The paper's central derivation is an empirical cGAN training pipeline: ERA5 CP/LSP patches condition a generator trained on RADKLIM-YW radar, with model selection on a temporally independent 2021 period and evaluation on held-out 2021 weeks in Germany, the US, and Australia. The evaluation is genuinely external: training and evaluation data are separated in time and geography, and the reported scores are computed against radar observations rather than against the model's own fitted outputs. The mean field bias correction (Section 7.1) preserves the ERA5 patch average by design, and the paper explicitly states that the model is therefore 'constrained to the ERA5 precipitation average of the input patch' (Section A.3); this is a stated limitation, not a hidden equivalence. The one self-citation, to the authors' prior spateGAN paper [32], is architectural lineage rather than load-bearing: the present work fully specifies the generator, discriminator, loss function, training data, and evaluation protocol, and no external result needed for the paper's claims is imported solely from that citation. The limitation in Section A.3 that extreme events rare in Germany 'can lead to more unrealistic spatial patterns despite a potentially correct estimate of the amplitude' is an honest concession about generalization, not a circular step; it concerns the strength of the global-applicability extrapolation, which is a correctness-risk issue rather than a derivation that reduces to its inputs. No fitted parameter is renamed as a prediction, and no uniqueness claim is imported from prior self-cited work. Thus the derivation chain is self-contained against external benchmarks, and the appropriate circularity score is minimal.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The scientific contribution is conditional on accepting radar as reference truth, ERA5 CP and LSP as sufficient inputs, and cross-regional stationarity of the learned sub-grid rainfall mapping. No new physical entities are introduced.

free parameters (4)
  • All generator/discriminator weights and biases = Trained on about 20,000 RADKLIM-YW samples from 2009-2020; total parameter count not stated
    The downscaling mapping is learned, not derived, so the paper's claims are about the behavior of these fitted weights. No trained checkpoint is released.
  • Training sample selection thresholds epsilon1 and epsilon2 = epsilon1 = | -450 * epsilon + 4500 |, epsilon2 = | -50 * epsilon' + 500 |, with epsilon, epsilon' drawn from…
    Hand-chosen distributions in Section A.1.1 oversample rainy and well-matched samples, shaping the training distribution and therefore the learned mapping.
  • Dropout probability p = 0.2 at three generator depths
    Manual design choice used to produce stochastic ensemble members and to reconstruct the missing tail of the precipitation distribution.
  • Optimizer and training schedule = Learning rates 1e-4 (generator) and 2e-4 (discriminator), batch size 9, 2e5 adversarial steps
    Hand-picked hyperparameters; the paper notes that extensive hyperparameter optimization was not feasible.
assumptions (4)
  • domain assumption Gauge-adjusted radar products (RADKLIM-YW, MRMS, Australian radar) are treated as ground truth for 2 km/10 minute precipitation.
    All training and evaluation scores compare against radar QPE; radar retrieval errors are inherited by the target distribution.
  • domain assumption ERA5 convective and large-scale precipitation are sufficient conditioning inputs for sub-grid rainfall variability.
    Section 7.4.1 states that additional variables such as wind and temperature did not improve performance; the model assumes all predictable small-scale structure is encoded in CP and LSP plus noise.
  • domain assumption The Germany-learned conditional distribution transfers to other climate zones.
    The global applicability claim rests on this; it is tested in portions of the US and Australia but not globally.
  • domain assumption ERA5 and radar samples are loosely paired in space and time after regridding, so the network can learn a meaningful conditional mapping.
    Sections 7.5.1 and 7.5.2 align ERA5 to radar grids and select overlapping patches; spatial and temporal offsets are treated as part of the stochastic target.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Global spatio-temporal downscaling of ERA5 precipitation through generative AI." pith.science (2026). https://pith.science/paper/7FCXIQAS

@misc{pith2026241116098,
  author       = {Pith},
  title        = {Pith review of: Global spatio-temporal downscaling of ERA5 precipitation through generative AI},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7FCXIQAS}},
  note         = {Machine review of arXiv:2411.16098}
}
read the original abstract

The spatial and temporal distribution of precipitation has a significant impact on human lives by determining freshwater resources and agricultural yield, but also rainfall-driven hazards like flooding or landslides. While the ERA5 reanalysis dataset provides consistent long-term global precipitation information that allows investigations of these impacts, it lacks the resolution to capture the high spatio-temporal variability of precipitation. ERA5 misses intense local rainfall events that are crucial drivers of devastating flooding - a critical limitation since extreme weather events become increasingly frequent. Here, we introduce spateGAN-ERA5, the first deep learning based spatio-temporal downscaling of precipitation data on a global scale. SpateGAN-ERA5 uses a conditional generative adversarial neural network (cGAN) that enhances the resolution of ERA5 precipitation data from 24 km and 1 hour to 2 km and 10 minutes, delivering high-resolution rainfall fields with realistic spatio-temporal patterns and accurate rain rate distribution including extremes. Its computational efficiency enables the generation of a large ensemble of solutions, addressing uncertainties inherent to the challenges of downscaling. Trained solely on data from Germany and validated in the US and Australia considering diverse climate zones, spateGAN-ERA5 demonstrates strong generalization indicating a robust global applicability. SpateGAN-ERA5 fulfils a critical need for high-resolution precipitation data in hydrological and meteorological research, offering new capabilities for flood risk assessment, AI-enhanced weather forecasting, and impact modelling to address climate-driven challenges worldwide.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

77 extracted references · 61 canonical work pages

  1. [1]

    Gherardi, L. A. & Sala, O. E. Effect of interannual precipitation variability on dryland productivity: A global synthesis. Global Change Biology 25, 269– 276 (2019). URL https://onlinelibrary.wiley.com/doi/abs/10.1111/gcb.14480. eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1111/gcb.14480

    work page doi:10.1111/gcb.14480 2019

  2. [2]

    & Wenz, L

    Kotz, M., Levermann, A. & Wenz, L. The effect of rainfall changes on economic production. Nature 601, 223–227 (2022). URL https://www.nature.com/articles/ s41586-021-04283-8. Publisher: Nature Publishing Group

    work page 2022

  3. [3]

    K., Gerber, J

    Ray, D. K., Gerber, J. S., MacDonald, G. K. & West, P. C. Climate variation explains a third of global crop yield variability. Nature Communications 6, 5989 (2015). URL https://www.nature.com/articles/ncomms6989. Publisher: Nature Publishing Group

    work page 2015

  4. [4]

    Zhang, W., Zhou, T. & Wu, P. Anthropogenic amplification of precipitation variability over the past century. Science 385, 427–432 (2024). URL https:// 18 www.science.org/doi/10.1126/science.adp0212. Publisher: American Association for the Advancement of Science

    doi:10.1126/science.adp0212 2024

  5. [5]

    & Obled, C

    Berne, A., Delrieu, G., Creutin, J.-D. & Obled, C. Temporal and spatial resolution of rainfall measurements required for urban hydrology. Journal of Hydrology 299, 166–179 (2004). URL https://www.sciencedirect.com/science/article/pii/ S0022169404003634

    work page 2004

  6. [6]

    Ochoa-Rodriguez, S. et al. Impact of spatial and temporal resolution of rainfall inputs on urban hydrodynamic modelling outputs: A multi-catchment investiga- tion. Journal of Hydrology 531, 389–407 (2015). URL https://www.sciencedirect. com/science/article/pii/S0022169415003856

    work page 2015

  7. [7]

    & Tamm, T

    Tamm, O., Saarem¨ ae, E., Rahkema, K., Jaagus, J. & Tamm, T. The inten- sification of short-duration rainfall extremes due to climate change – Need for a frequent update of intensity–duration–frequency curves. Climate Ser- vices 30, 100349 (2023). URL https://www.sciencedirect.com/science/article/ pii/S2405880723000109

    work page 2023

  8. [8]

    Kidd, C. et al. So, How Much of the Earth’s Surface Is Covered by Rain Gauges? Bulletin of the American Meteorological Society 98, 69–78 (2017). URL https://journals.ametsoc.org/view/journals/bams/98/1/bams-d-14-00283. 1.xml. Publisher: American Meteorological Society Section: Bulletin of the American Meteorological Society

    work page 2017

Show all 77 references
  1. [9]

    Guo, H. et al. Inter-Comparison of High-Resolution Satellite Precipitation Prod- ucts over Central Asia. Remote Sensing 7, 7181–7211 (2015). URL https: //www.mdpi.com/2072-4292/7/6/7181. Number: 6 Publisher: Multidisciplinary Digital Publishing Institute

    2015

  2. [10]

    Guo, H. et al. Early assessment of Integrated Multi-satellite Retrievals for Global Precipitation Measurement over China. Atmospheric Research 176- 177, 121–133 (2016). URL https://www.sciencedirect.com/science/article/pii/ S0169809516300370

    2016

  3. [11]

    Bai, L., Shi, C., Li, L., Yang, Y. & Wu, J. Accuracy of CHIRPS Satellite-Rainfall Products over Mainland China. Remote Sensing 10, 362 (2018). URL https: //www.mdpi.com/2072-4292/10/3/362. Number: 3 Publisher: Multidisciplinary Digital Publishing Institute

    2018

  4. [12]

    Hersbach, H. et al. The ERA5 global reanalysis. Quarterly Jour- nal of the Royal Meteorological Society 146, 1999–2049 (2020). URL https://onlinelibrary.wiley.com/doi/abs/10.1002/qj.3803. eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1002/qj.3803

    2020 doi

  5. [13]

    Gebrechorkos, S. H. et al. Global-scale evaluation of precipitation datasets for hydrological modelling. Hydrology and Earth System Sciences 28, 3099–3118 19 (2024). URL https://hess.copernicus.org/articles/28/3099/2024/. Publisher: Copernicus GmbH

    2024

  6. [14]

    Tarek, M., Brissette, F. P. & Arsenault, R. Evaluation of the ERA5 reanalysis as a potential reference dataset for hydrological modelling over North America. Hydrology and Earth System Sciences 24, 2527–2544 (2020). URL https://hess. copernicus.org/articles/24/2527/2020/. Publ...

    2020

  7. [15]

    Nearing, G. et al. Global prediction of extreme floods in ungauged water- sheds. Nature 627, 559–563 (2024). URL https://www.nature.com/articles/ s41586-024-07145-1. Publisher: Nature Publishing Group

    2024

  8. [16]

    & Wenz, L

    Kotz, M., Levermann, A. & Wenz, L. The economic commitment of climate change. Nature 628, 551–557 (2024). URL https://www.nature.com/articles/ s41586-024-07219-0. Publisher: Nature Publishing Group

    2024

  9. [17]

    Lenton, T. M. et al. Quantifying the human cost of global warming. Nature Sustainability 6, 1237–1247 (2023). URL https://www.nature.com/articles/ s41893-023-01132-6. Publisher: Nature Publishing Group

    2023

  10. [18]

    Lam, R. et al. GraphCast: Learning skillful medium-range global weather forecasting (2023). URL http://arxiv.org/abs/2212.12794. ArXiv:2212.12794 [physics]

    2023 arXiv

  11. [19]

    Pathak, J. et al. FourCastNet: A Global Data-driven High-resolution Weather Model using Adaptive Fourier Neural Operators (2022). URL https://arxiv.org/ abs/2202.11214v1

    2022 arXiv

  12. [20]

    Lang, S. et al. AIFS – ECMWF’s data-driven forecasting system (2024). URL http://arxiv.org/abs/2406.01465. ArXiv:2406.01465 [physics]

    2024 arXiv

  13. [21]

    Kochkov, D. et al. Neural general circulation models for weather and climate. Nature 1–7 (2024). URL https://www.nature.com/articles/s41586-024-07744-y. Publisher: Nature Publishing Group

    2024

  14. [22]

    A., Simmons, A., Vamborg, F

    Lavers, D. A., Simmons, A., Vamborg, F. & Rodwell, M. J. An evaluation of ERA5 precipitation for climate monitoring. Quarterly Journal of the Royal Meteorological Society 148, 3152–3165 (2022). URL https://onlinelibrary.wiley.com/doi/abs/10.1002/qj.4351. eprint: https://online...

    2022 doi

  15. [23]

    Volosciuk, C., Maraun, D., Semenov, V. A. & Park, W. Extreme Precipitation in an Atmosphere General Circulation Model: Impact of Horizontal and Vertical Model Resolutions. Journal of Climate 28, 1184–1205 (2015). URL https:// journals.ametsoc.org/view/journals/clim/28/3/jcli-d...

    2015

  16. [24]

    Seo, D. J. Real-time estimation of rainfall fields using radar rainfall and rain gage data. Journal of Hydrology 208, 37–52 (1998). URL https://www.sciencedirect. com/science/article/pii/S0022169498001413

    1998

  17. [25]

    G., Keenan, T

    Sun, X., Mein, R. G., Keenan, T. D. & Elliott, J. F. Flood estimation using radar and raingauge data. Journal of Hydrology 239, 4–18 (2000). URL https: //www.sciencedirect.com/science/article/pii/S0022169400003504

    2000

  18. [26]

    A., Semenov, V

    Aleshina, M. A., Semenov, V. A. & Chernokulsky, A. V. A link between surface air temperature and extreme precipitation over Russia from station and reanalysis data. Environmental Research Letters 16, 105004 (2021). URL https://dx.doi. org/10.1088/1748-9326/ac1cba. Publisher: I...

    2021 doi

  19. [27]

    Ben-Bouallegue, Z. et al. The rise of data-driven weather forecasting (2023). URL http://arxiv.org/abs/2307.10128. ArXiv:2307.10128 [physics]

    2023 arXiv

  20. [28]

    Tabari, H. et al. Local impact analysis of climate change on precipitation extremes: are high-resolution climate models needed for realistic simulations? Hydrology and Earth System Sciences 20, 3843–3857 (2016). URL https://hess. copernicus.org/articles/20/3843/2016/. Publishe...

    2016

  21. [29]

    & Samimi, C

    Zandler, H., Haag, I. & Samimi, C. Evaluation needs and temporal per- formance differences of gridded precipitation products in peripheral mountain regions. Scientific Reports 9, 15118 (2019). URL https://www.nature.com/ articles/s41598-019-51666-z. Publisher: Nature Publishing Group

    2019

  22. [30]

    R., Livneh, B., Barsugli, J

    Bjarke, N. R., Livneh, B., Barsugli, J. J., Pendergrass, A. G. & Small, E. E. Evaluating Large-Storm Dominance in High-Resolution GCMs and Observations Across the Western Contiguous United States.Earth’s Future 12, e2023EF004289 (2024). URL https://onlinelibrary.wiley.com/doi/...

    2024 doi

  23. [31]

    L., Thompson, E

    Bytheway, J. L., Thompson, E. J., Yang, J. & Chen, H. Evaluation of the RainF ARM Statistical Downscaling Technique Applied to IMERG over Global Oceans Using Passive Aquatic Listener In Situ Rain Measurements. Journal of Hydrometeorology 24, 2351–2367 (2023). URL https://journ...

    2023

  24. [32]

    & Chwala, C

    Glawion, L., Polz, J., Kunstmann, H., Fersch, B. & Chwala, C. spate- GAN: Spatio-Temporal Downscaling of Rainfall Fields Using a cGAN Approach. Earth and Space Science 10, e2023EA002906 (2023). URL https://onlinelibrary.wiley.com/doi/abs/10.1029/2023EA002906. eprint: https://o...

    2023 doi

  25. [33]

    & Berne, A

    Leinonen, J., Nerini, D. & Berne, A. Stochastic Super-Resolution for Downscal- ing Time-Evolving Atmospheric Fields With a Generative Adversarial Network. 21 IEEE Transactions on Geoscience and Remote Sensing 59, 7211–7223 (2021). URL https://ieeexplore.ieee.org/document/92465...

    2021

  26. [34]

    & Ban, N

    Serifi, A., G¨ unther, T. & Ban, N. Spatio-Temporal Downscaling of Climate Data Using Convolutional and Error-Predicting Neural Networks. Frontiers in Climate 3 (2021). URL https://www.frontiersin.org/article/10.3389/fclim.2021.656479

    2021

  27. [35]

    Harris, L., McRae, A. T. T., Chantry, M., Dueben, P. D. & Palmer, T. N. A Generative Deep Learning Approach to Stochastic Downscaling of Precipitation Forecasts. Journal of Advances in Modeling Earth Systems 14, e2022MS003120 (2022). URL https://onlinelibrary.wiley.com/doi/abs...

    2022 doi

  28. [36]

    & Rasp, S

    Price, I. & Rasp, S. Increasing the accuracy and resolution of precipitation fore- casts using deep generative models. arXiv:2203.12297 [cs, stat] (2022). URL http://arxiv.org/abs/2203.12297. ArXiv: 2203.12297

    2022 arXiv

  29. [37]

    Winterrath, T. et al. Radar climatology (RADKLIM) version 2017.002; gridded precipitation data for Germany: Radar-based quasi gauge-adjusted five-minute precipitation rate (YW) (2018). URL https://opendata.dwd.de/climate environment/CDC/help/landing pages/ doi landingpage RADK...

    2018

  30. [38]

    Rebora, N., Ferraris, L., Hardenberg, J. v. & Provenzale, A. RainF ARM: Rainfall Downscaling by a Filtered Autoregressive Model. Journal of Hydrometeorology 7, 724–738 (2006). URL https://journals.ametsoc.org/view/journals/hydr/7/4/ jhm517 1.xml. Publisher: American Meteorolog...

    2006

  31. [39]

    v., Provenzale, A

    D’Onofrio, D., Palazzi, E., Hardenberg, J. v., Provenzale, A. & Calmanti, S. Stochastic Rainfall Downscaling of Climate Models. Journal of Hydrometeorol- ogy 15, 830–843 (2014). URL https://journals.ametsoc.org/view/journals/hydr/ 15/2/jhm-d-13-096 1.xml. Publisher: American M...

    2014

  32. [40]

    Kendon, E. J. et al. Do Convection-Permitting Regional Climate Models Improve Projections of Future Precipitation Change?Bulletin of the American Meteorolog- ical Society 98, 79–93 (2017). URL https://journals.ametsoc.org/view/journals/ bams/98/1/bams-d-15-0004.1.xml. Publishe...

    2017

  33. [41]

    Bell, T. L. A space-time stochastic model of rainfall for satellite remote-sensing studies. Journal of Geophysical Research: Atmospheres 92, 9631–9643 (1987). 22 URL https://onlinelibrary.wiley.com/doi/abs/10.1029/JD092iD08p09631. eprint: https://onlinelibrary.wiley.com/doi/pd...

    1987 doi

  34. [42]

    Crane, R. K. Space-time structure of rain rate fields. Jour- nal of Geophysical Research: Atmospheres 95, 2011–2020 (1990). URL https://onlinelibrary.wiley.com/doi/abs/10.1029/JD095iD03p02011. eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1029/JD095iD03p02011

    1990 doi

  35. [43]

    Zick, S. E. & Matyas, C. J. A Shape Metric Methodology for Studying the Evolv- ing Geometries of Synoptic-Scale Precipitation Patterns in Tropical Cyclones. Annals of the American Association of Geographers 106, 1217–1235 (2016). URL https://www.tandfonline.com/doi/full/10.108...

    2016

  36. [44]

    & Lovejoy, S

    Schertzer, D. & Lovejoy, S. Physical modeling and analysis of rain and clouds by anisotropic scaling multiplicative processes. Journal of Geophysical Research: Atmospheres 92, 9693–9714 (1987). URL https://onlinelibrary.wiley.com/doi/abs/10.1029/JD092iD08p09693. eprint: https:...

    1987 doi

  37. [45]

    String of Beads

    Pegram, G. G. S. & Clothier, A. N. High resolution space–time modelling of rainfall: the “String of Beads” model. Journal of Hydrology 241, 26–41 (2001). URL https://www.sciencedirect.com/science/article/pii/S0022169400003735

    2001

  38. [46]

    Hess, P., Dr¨ uke, M., Petri, S., Strnad, F. M. & Boers, N. Physically constrained generative adversarial networks for improving precipitation fields from Earth system models. Nature Machine Intelligence 4, 828–839 (2022). URL https: //www.nature.com/articles/s42256-022-00540-...

    2022

  39. [47]

    & Hutter, F

    Loshchilov, I. & Hutter, F. Decoupled Weight Decay Regularization (2019). URL http://arxiv.org/abs/1711.05101. ArXiv:1711.05101

    2019 arXiv

  40. [48]

    Bell, B. et al. The ERA5 global reanalysis: Preliminary extension to

  41. [49]

    Direct 4D-Var Assimilation of NCEP Stage IV Radar and Gauge Pre- cipitation Data at ECMWF

    Lopez, P. Direct 4D-Var Assimilation of NCEP Stage IV Radar and Gauge Pre- cipitation Data at ECMWF. Monthly Weather Review 139, 2098–2116 (2011). URL https://journals.ametsoc.org/view/journals/mwre/139/7/2010mwr3565.1. xml. Publisher: American Meteorological Society Section: ...

    2011

  42. [50]

    & Mitchell, K

    Lin, Y. & Mitchell, K. E. The NCEP stage II/IV hourly precipitation analyses: Development and applications. Proceedings of the 19th Conference Hydrology, American Meteorological Society, San Diego, CA, USA 10 (2005). 23

    2005

  43. [51]

    Chen, Y. et al. Spatial performance of multiple reanalysis precipitation datasets on the southern slope of central Himalaya. Atmospheric Research 250, 105365 (2021). URL https://www.sciencedirect.com/science/article/pii/ S0169809520313028

    2021

  44. [52]

    & Franc´ es, F

    Gomis-Cebolla, J., Rattayova, V., Salazar-Gal´ an, S. & Franc´ es, F. Evaluation of ERA5 and ERA5-Land reanalysis precipitation datasets over Spain (1951–2020). Atmospheric Research 284, 106606 (2023). URL https://www.sciencedirect.com/ science/article/pii/S0169809523000030

    2023

  45. [53]

    & Peng, J

    Xu, J., Ma, Z., Yan, S. & Peng, J. Do ERA5 and ERA5-land precipitation estimates outperform satellite-based precipitation products? A comprehensive comparison between state-of-the-art model-based and satellite-based precipita- tion products over mainland China. Journal of Hydr...

    2022

  46. [54]

    Bandhauer, M. et al. Evaluation of daily precipitation analyses in E- OBS (v19.0e) and ERA5 by comparison to regional high-resolution datasets in European regions. International Journal of Climatology 42, 727–747 (2022). URL https://onlinelibrary.wiley.com/doi/abs/10.1002/joc....

    2022 doi

  47. [55]

    Beck, H. E. et al. MSWEP V2 Global 3-Hourly 0.1 ° Precipitation: Method- ology and Quantitative Assessment. Bulletin of the American Meteorological Society 100, 473–500 (2019). URL https://journals.ametsoc.org/view/journals/ bams/100/3/bams-d-17-0138.1.xml. Publisher: American...

    2019

  48. [56]

    Winterrath, T. et al. Erstellung einer radargest¨ utzten Niederschlagskli- matologie. Tech. Rep. 251, Deutscher Wetterdienst (2017). URL https://www.dwd.de/DE/leistungen/pbfb verlag berichte/pdf einzelbaende/ 251 pdf.pdf? blob=publicationFile&v=2

    2017

  49. [57]

    Zhang, J. et al. Multi-Radar Multi-Sensor (MRMS) Quantitative Precipitation Estimation: Initial Operating Capabilities. Bulletin of the American Meteo- rological Society 97, 621–638 (2016). URL https://journals.ametsoc.org/view/ journals/bams/97/4/bams-d-14-00174.1.xml. Publis...

    2016

  50. [58]

    Smith, T. M. et al. Multi-Radar Multi-Sensor (MRMS) Severe Weather and Aviation Products: Initial Operating Capabilities. Bulletin of the American Meteorological Society 97, 1617–1630 (2016). URL https://journals.ametsoc.org/ view/journals/bams/97/9/bams-d-14-00173.1.xml. Publ...

    2016

  51. [59]

    & Velasco-Forero, C

    Seed, A., Curtis, M. & Velasco-Forero, C. AURA - Operational Radar Rainfields 3 (2022). URL https://pid.nci.org.au/doi/f0493 5520 4121 7835. 24

    2022

  52. [60]

    & Sharma, A

    Chumchean, S., Seed, A. & Sharma, A. Correcting of real-time radar rainfall bias using a Kalman filtering approach. Journal of Hydrology 317, 123–137 (2006). URL https://www.sciencedirect.com/science/article/pii/S0022169405002647

    2006

  53. [61]

    Pulkkinen, S. et al. Pysteps: an open-source Python library for probabilistic pre- cipitation nowcasting (v1.0). Geoscientific Model Development 12, 4185–4219 (2019). URL https://gmd.copernicus.org/articles/12/4185/2019/. Publisher: Copernicus GmbH

    2019

  54. [62]

    Paszke, A. et al. PyTorch: An Imperative Style, High-Performance Deep Learning Library (2019). URL http://arxiv.org/abs/1912.01703. ArXiv:1912.01703

    2019 arXiv

  55. [63]

    Goodfellow, I. et al. Ghahramani, Z., Welling, M., Cortes, C., Lawrence, N. & Weinberger, K. Q. (eds) Generative Adversarial Nets . (eds Ghahra- mani, Z., Welling, M., Cortes, C., Lawrence, N. & Weinberger, K. Q.) Advances in Neural Information Processing Systems , Vol. 27 (Cu...

    2014

  56. [64]

    & Efros, A

    Isola, P., Zhu, J.-Y., Zhou, T. & Efros, A. A. Image-to-Image Translation with Conditional Adversarial Networks. 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 5967–5976 (2017). ISSN: 1063-6919

    2017

  57. [65]

    & Raftery, A

    Gneiting, T. & Raftery, A. E. Strictly Proper Scoring Rules, Prediction, and Estimation. Journal of the American Statistical Association 102, 359–378 (2007). URL http://www.tandfonline.com/doi/abs/10.1198/016214506000001437

    2007 doi

  58. [66]

    Assessing the spatial and temporal variation in the skill of precip- itation forecasts from an NWP model

    Roberts, N. Assessing the spatial and temporal variation in the skill of precip- itation forecasts from an NWP model. Meteorological Applications 15, 163–169 (2008). URL https://onlinelibrary.wiley.com/doi/abs/10.1002/met.57. eprint: https://onlinelibrary.wiley.com/doi/pdf/10....

    2008 doi

  59. [67]

    Roberts, N. M. & Lean, H. W. Scale-Selective Verification of Rainfall Accumu- lations from High-Resolution Forecasts of Convective Events. Monthly Weather Review 136, 78–97 (2008). URL https://journals.ametsoc.org/view/journals/ mwre/136/1/2007mwr2123.1.xml. Publisher: America...

    2008

  60. [68]

    Harris, D., Foufoula-Georgiou, E., Droegemeier, K. K. & Levit, J. J. Multiscale Statistical Properties of a High-Resolution Precipitation Forecast. Journal of Hydrometeorology 2, 406–418 (2001). URL https://journals.ametsoc.org/view/ journals/hydr/2/4/1525-7541 2001 002 0406 m...

    2001

  61. [69]

    & Pegram, G

    Sinclair, S. & Pegram, G. G. S. Empirical Mode Decomposition in 2-D space and time: a tool for space-time rainfall analysis and nowcasting. Hydrology and Earth System Sciences 11 (2005). 25

    2005

  62. [70]

    & Talagrand, O

    Candille, G. & Talagrand, O. Evaluation of probabilistic prediction systems for a scalar variable. Quarterly Journal of the Royal Meteorological Society 131, 2131–2150 (2005). URL https://onlinelibrary.wiley.com/doi/abs/10.1256/qj.04

    2005 doi

  63. [71]

    eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1256/qj.04.71

    doi

  64. [72]

    Foresti, L. et al. A quest for precipitation attractors in weather radar archives. Nonlinear Processes in Geophysics 31, 259–286 (2024). URL https://npg. copernicus.org/articles/31/259/2024/. Publisher: Copernicus GmbH. 26 Appendix A Supplementary information A.1 Further metho...

    2024

  65. [73]

    Hamill, T. M. Interpretation of Rank Histograms for Verifying Ensemble Fore- casts. Monthly Weather Review 129, 550–560 (2001). URL https://journals. ametsoc.org/view/journals/mwre/129/3/1520-0493 2001 129 0550 iorhfv 2.0. co 2.xml. Publisher: American Meteorological Society S...

    2001

  66. [75]

    X and y do not contain missing values

  67. [76]

    P h,w,t X > ε1 and P h,w,t y > ε1, where ε1 = | −450ε + 4500| and where ε is drawn from Lognormal(0, 1)

  68. [77]

    The distribution of the thresholds ε1 and ε2 is shown in Fig

    The 66th quantiles of the pixel values inX and y exceed ε2, where ε2 = |−50ε′+500| and where ε′ is drawn from Lognormal(0, 1). The distribution of the thresholds ε1 and ε2 is shown in Fig. A1 and roughly reflects the inverse probability of drawing samples that match the given ...

    2021

  69. [1950]

    URL https://onlinelibrary.wiley.com/doi/abs/10.1002/qj.4174

    Quarterly Journal of the Royal Meteorological Society 147, 4186–4227 (2021). URL https://onlinelibrary.wiley.com/doi/abs/10.1002/qj.4174. eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1002/qj.4174

    2021 doi

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.