Pith. sign in

REVIEW 3 major objections 5 minor 72 references

Downscaling with AI reveals the large role of internal variability in fine-scale projections of climate extremes

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read At fine scales, internal variability — not emissions scenario — dominates uncertainty in future precipitation extremes over New Zealand.

desk verdict A credible, large-ensemble emulator study showing internal variability dominates fine-scale precipitation extremes over New Zealand; the headline ratios rest on emulator-generated variance that deserves one more validation step before it is quoted. read the letter →

arxiv 2507.06527 v1 pith:ALSHS5PK submitted 2025-07-09 physics.ao-ph physics.geo-ph

Neelesh Rampal , Peter B. Gibson , Steven C. Sherwood , Laura E. Queen , Hamish Lewis , Gab Abramowitz This is my paper
classification physics.ao-phphysics.geo-ph
keywords internalclimatevariabilitygenerativeadversarialnetworkstatisticaldownscalingprecipitationextremesprojectionuncertaintylargeensemblesNewZealandregionalmodelemulator
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

This paper argues that at the fine spatial scales resolved by regional downscaling, internal climate variability — the unpredictable, chaotic component of the climate system — is the dominant source of uncertainty in future precipitation extremes over New Zealand. Using a generative-AI emulator of a regional climate model, the authors downscale more than 15,000 simulation years from 20 global models, four emissions scenarios, and two large initial-condition ensembles, far more than dynamical downscaling could produce. They find that the forced (predictable) part of future change in precipitation extremes is spatially smoother and locally smaller than change in any individual simulation, and that the variance contributed by internal variability exceeds the variance across emissions scenarios by a factor of four for annual extremes and ten for decadal extremes. The same analysis shows that the variance of fine-scale precipitation extremes grows with warming, roughly tripling by the end of the century, which makes rare local extremes less predictable than coarse-scale projections suggest.

What carries the argument

The central object is a residual generative adversarial network that emulates a regional climate model. A deterministic U-Net first predicts the large-scale, circulation-driven component of daily precipitation and temperature; a GAN is then trained on the residuals between that prediction and the RCM truth, with an adversarial loss and a regional intensity constraint, so the emulator captures the distribution and extremes of the fine-scale fields. Because the emulator runs orders of magnitude faster than the RCM, the authors can generate an ensemble of more than 15,000 simulation years, and the spread within the two SMILEs is used to separate internal variability from model and scenario uncertainty using the variance-decomposition framework of earlier work.

What would settle it

Run the same 40-member ACCESS-ESM1-5 (or CanESM5) initial-condition ensemble through a conventional regional climate model at comparable resolution, compute the spread of RX10y and RX1Day change signals across members, and compare its magnitude and its trend with warming to the emulator's reported 72% and 59% variance increases; a substantially smaller or non-growing RCM spread would falsify the claim that fine-scale internal variability dominates and grows.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that fine-scale projections of precipitation extremes are dominated by irreducible internal variability. At New Zealand's resolved downscaling scales, the signal-to-noise ratio for 10-year precipitation extremes is 0.77 in the emulator versus 1.12 in the driving global model, meaning internal precipitation variability exceeds the climate-change signal at fine scales but not quite at coarse scales. For annual and decadal extremes, internal variability contributes over 50% of total projection variance by 2080–2099, while for temperature extremes it contributes about 11% (annual) and 27% (decadal). The paper also reports that internal-variability variance for precipitation extremes increases by 59% (annual), 72% (10-year) and 78% (20-year) from the historical period to 2080–2099, and that this growth is missed by the common static-variability assumption, which underestimates end-of-century internal variability by more than a factor of two.

Load-bearing premise

The load-bearing premise is that the emulator's fine-scale internal variability — the spread among its ensemble members — faithfully represents the internal variability a real regional climate model would produce, since the paper's headline variance numbers are generated by the emulator itself rather than verified against a full RCM large ensemble.

Editorial extensions

If this is right

  • If correct, individual downscaled projections of future precipitation extremes should not be read as the expected change; the ensemble-mean forced response is smoother and locally smaller than any single member.
  • Detecting forced change in fine-scale precipitation extremes over New Zealand will require ensembles far larger than the three-member RCM set used previously, because internal variability can exceed the signal.
  • Rare 10- and 20-year precipitation extremes will be both more severe and more uncertain than annual extremes, with internal-variability spread growing as the climate warms.
  • Uncertainty assessments that assume internal variability is constant in time will understate end-of-century uncertainty in fine-scale precipitation by more than a factor of two.
  • Temperature extremes remain comparatively more predictable at fine scales, with scenario spread dominating by late century, though internal variability is not negligible.

Reading between the lines

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

  • The emulator's stochastic layer draws a single random noise vector per day; testing whether multiple draws change the variance statistics would directly probe whether the reported internal-variability growth is a property of the climate or of the generator.
  • The fourfold and tenfold ratios are domain-specific; the same decomposition applied to other mid-latitude, orographically influenced regions would show whether New Zealand is a best-case or worst-case baseline for fine-scale predictability.
  • The finding that downscaling itself can alter the forced signal by an amount comparable to internal variability suggests that emulator (or RCM) structural uncertainty should be included alongside model, scenario, and internal variability in future decomposition studies.
  • If the variance-growth result generalizes, adaptation planning for local rainfall extremes should be framed around risk-based ensemble ranges rather than single-model 'most likely' projections.
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

3 major / 5 minor

Summary. The paper presents a generative-AI (residual GAN) emulator for dynamical downscaling of daily precipitation and temperature over New Zealand, trained on a single CCAM RCM simulation driven by ACCESS-CM2. The emulator is applied to 20 CMIP6 GCMs, four SSPs, and two SMILEs (ACCESS-ESM1-5, CanESM5), generating more than 15,000 simulation years. The authors evaluate the emulator's historical climatology against observations and out-of-sample RCM simulations, and its climate-change signals against two unseen RCMs. They then use the emulator-downscaled ensembles to decompose uncertainty in annual and decadal extremes into model, scenario, and internal-variability components. The headline findings are that internal variability dominates model and scenario uncertainty for fine-scale precipitation extremes, that its variance grows with warming (59-78% increases for ACCESS-ESM1-5 by 2080-2099), and that it exceeds scenario variance by fourfold for annual and tenfold for decadal extremes. The paper also reports that internal-variability spread is larger at fine scales than at GCM scales and that individual ensemble members are poor guides to the forced response.

Significance. If the findings hold, they have substantial significance for regional climate risk assessment and for the design of downscaling ensembles. The paper demonstrates a promising and computationally efficient approach to generating very large downscaled ensembles, and it quantifies, at fine scales, the irreducible uncertainty from internal variability that is often underestimated in coarse-resolution studies. The evaluation against out-of-sample RCMs and observations, the use of two SMILEs, and the public availability of code and data are notable strengths. The central claim, however, rests on the emulator's ability to reproduce not only mean and extreme-change signals but also the magnitude, spatial pattern, and future scaling of fine-scale internal variability. That aspect is not directly validated against a dynamical RCM large ensemble, which is the key risk to the headline quantitative results.

major comments (3)
  1. [Methods, 'Generating a Large Ensemble of Climate Projections'] The statement that 'For each day, a single GAN prediction is generated from a randomized noise vector, rather than using an ensemble as in previous studies, since variability was found to be minimal for the statistics explored here' is load-bearing for the paper's central claims, but the supporting analysis is not shown. The internal-variability variance, the fourfold/tenfold ratios, and the 59-78% variance increases (Figure 5) are all computed from ensemble spread that includes the stochastic GAN output. If the single-draw approximation suppresses or inflates the stochastic component, those numbers could be emulator artifacts. Please provide the evidence that variability across multiple GAN noise draws is negligible for the variance and rare-extreme statistics used (e.g., RX1Day, RX10y, TXx, TX10y), for both historical and future periods. If the sensitivity is not negligible, the ensemble should be constructed with multiple stochastic draws per day, or the uncertainty in the variance estimates should be quantified and propagated into the decomposition.
  2. [Results, 'Emulator Performance in Historical and Future Climates' and 'Internal Variability in Future Climate…] The headline results concern the magnitude, spatial pattern, and future growth of fine-scale internal variability, yet these are never directly validated against a dynamical RCM SMILE. The emulator is evaluated against observations and RCMs for historical climatology and mean/extreme change signals (Figure 1; Supplementary S32-S35), and against historical interannual variability or two SMILEs only qualitatively (Supplementary S25-S27). The comparison with raw GCM spread (Figures 2 and 5) is suggestive but cannot rule out that the emulator generates structurally plausible but quantitatively wrong fine-scale variance, especially for rare decadal extremes. Please provide a quantitative validation of emulator-generated internal variability against a dynamical RCM ensemble with multiple initial conditions. If no RCM SMILE is available, a direct comparison using the CCAM n=6 ensemble (Figure 3) or another multi-member RCM dataset for the historical period, including the variance of RX10y/TX10y statistics, would materially address this concern. Absent this, the paper overstates the confidence that can be placed in the central claim.
  3. [Methods, 'Fractional Uncertainty in Climate Projections'] The model uncertainty component is 'defined as the variance across GCM-specific forced response fits, averaged over scenarios,' where each GCM is represented by a single realization (except for the two SMILEs). The forced-response fit is a second-degree polynomial to smoothed decadal means, which reduces but does not eliminate the influence of internal variability in each GCM's trajectory. As a result, the 'model' variance likely contains a substantial contribution from sampling noise due to internal variability, and the decomposition may double-count internal variability when the SMILE-based internal-variance estimate is added separately. The authors acknowledge this issue for the static method but not for their default decomposition. Please quantify the sensitivity: for the two SMILEs, compute model variance both from single fits of individual members and from fits of the ensemble mean, and report how the fractional contributions change. If the double-counting is material, the interpretation of 'model uncertainty dominates' versus 'internal variability dominates' could shift.
minor comments (5)
  1. [Abstract] The phrase 'Internal variability spread is larger at fine scales that at the coarser scales simulated in climate models' contains a typo: 'that' should be 'than'.
  2. [Methods, 'Fractional Uncertainty in Climate Projections'] The sentence 'We use absolute precipitation anomalies as a pose to previous studies using percentage changes' contains a typo: 'as a pose to' should be 'as opposed to'.
  3. [Figure 3 caption] The symbol 'A indicates the percentage of the area with sign agreement' is unclear; define the symbol explicitly in the caption or label it 'Area agreement' in the figure.
  4. [Results, 'Uncertainty decomposition at fine scales'] In Figure 4 and its description, 'land-averaged' and 'land average' are used interchangeably; please standardize the terminology to avoid confusion between averaging over land points before versus after computing extreme statistics.
  5. [References] The Harrington et al. reference has malformed author formatting ('Harrington, L. J., , Peter B., G., , Nicolas, F., , Hamish, L., , Dave, F., , and Rosier, S. M.'). Please correct.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central internal-variability findings emerge from an emulator independently benchmarked against observations and out-of-sample RCMs, and the self-citations are not load-bearing.

full rationale

I walked the derivation chain from GCM inputs through the residual-GAN emulator to the uncertainty decomposition. The headline quantities, such as the fourfold/tenfold variance ratios and the 59-78% increases in precipitation variance, are computed as ensemble statistics of emulator-downscaled SMILE outputs; they are not fitted parameters and are not defined in terms of the conclusions. The emulator itself is validated in this paper against the VCSN observational dataset (Figure 1a,b) and against two RCMs not used in training (EC-Earth3 and NorESM2-MM) for mean and extreme climate-change signals (Figure 1c-e; Supplementary S32-S35), so the central claim does not reduce to a self-citation chain. The methodological statement in the Methods that one GAN noise draw per day suffices because 'variability was found to be minimal for the statistics explored here' is an unsupported assertion and a validation gap, but it is not a case where a predicted quantity is defined in terms of the input or where a fitted parameter is renamed as a prediction. Similarly, the absence of a large dynamical RCM SMILE for directly validating the future magnitude of fine-scale internal variability is a correctness and robustness concern, not a circularity. The Hawkins-Sutton style variance decomposition is standard and externally established, and the internal-variability estimate is taken from SMILE member spread rather than from the emulator's own loss function. No step was found where Eq. X = Eq. Y by construction, and the self-citations to Rampal et al. (2024a, 2025) describe architecture and training choices that are re-evaluated here rather than invoked as an unexamined uniqueness theorem. Overall, the paper's central claims have independent empirical content and are not forced by definition or by self-citation.

Assumptions & free parameters 5 free parameters · 6 assumptions · 0 invented entities

The central conclusions rest on a trained deep-learning model (millions of fitted weights), on the Hawkins-Sutton decomposition with its independence assumption, on the assumption that a single stochastic draw per day does not affect variance and extremes, and on two SMILEs being representative of internal variability. No new physical entities are introduced.

free parameters (5)
  • Residual GAN and U-Net weights = 3.5 million trainable parameters
    Trained on 140 years of CCAM/ACCESS-CM2 output; the fine-scale variability and extremes that support the central claim are outputs of this fitted model.
  • GAN adversarial loss weight lambda_adv = 0.01
    Chosen in prior work; controls balance of MSE, adversarial, and intensity-constraint losses in the residual GAN.
  • Intensity constraint pooling window = 20 pixels (about 350 km)
    Regional pooled maxima over 20-pixel windows, adapted from a global maximum in prior work, to capture regional extreme precipitation.
  • Polynomial fit degree for forced response = second degree, 1950-2100
    Used to estimate forced response in the Hawkins-Sutton variance decomposition; choice affects residual internal-variability estimates.
  • Training hyperparameters = learning rates 2e-4 (GAN) and 7e-4 (CNN), batch 32, 200 epochs
    Manual choices from prior studies; not fitted to target result but determine emulator behavior.
assumptions (6)
  • domain assumption Model, scenario, and internal variability uncertainties are independent and their covariance is negligible at fine scales.
    Methods, 'Fractional Uncertainty in Climate Projections': decomposition follows Hawkins and Sutton (2009) and Lehner et al. (2020); paper acknowledges the assumption is 'not strictly valid' but cites Yip et al. (2011) that covariance impact is minimal.
  • domain assumption The emulator, trained in the perfect framework on CCAM forced by ACCESS-CM2, generalizes to other GCMs, including rare 10-year extremes.
    Central to applying the emulator to 20 GCMs; validated against only two unseen GCM-driven RCMs (EC-Earth3, NorESM2-MM) and against observations.
  • ad hoc to paper A single stochastic GAN draw per day is representative for the variance and extreme statistics studied.
    Methods, 'Generating a Large Ensemble of Climate Projections': 'For each day, a single GAN prediction is generated from a randomized noise vector... since variability was found to be minimal for the statistics explored here.' The evidence for this claim is not shown in the main text.
  • domain assumption The two SMILEs (ACCESS-ESM1-5, CanESM5) represent the range of internal variability across models.
    Internal variability is estimated as the average spread from two SMILEs; GCM-dependent differences in internal variability are acknowledged by the paper.
  • domain assumption No bias correction of GCM inputs is appropriate for change-signal analysis.
    Methods: 'We did not apply bias correction to the GCM inputs to be consistent and comparable with conventional numerical RCMs.' This assumes biases do not distort fine-scale change signals.
  • standard math Second-degree polynomial fits capture the forced response of extremes.
    Used in the Hawkins-Sutton decomposition; a standard smoothing and trend assumption for separating forced response from internal variability.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Downscaling with AI reveals the large role of internal variability in fine-scale projections of climate extremes." pith.science (2026). https://pith.science/paper/ALSHS5PK

@misc{pith2026250706527,
  author       = {Pith},
  title        = {Pith review of: Downscaling with AI reveals the large role of internal variability in fine-scale projections of climate extremes},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ALSHS5PK}},
  note         = {Machine review of arXiv:2507.06527}
}
read the original abstract

The computational cost of dynamical downscaling limits ensemble sizes in regional downscaling efforts. We present a newly developed generative-AI approach to greatly expand the scope of such downscaling, enabling fine-scale future changes to be characterised including rare extremes that cannot be addressed by traditional approaches. We test this approach for New Zealand, where strong regional effects are anticipated. At fine scales, the forced (predictable) component of precipitation and temperature extremes for future periods (2080--2099) is spatially smoother than changes in individual simulations, and locally smaller. Future changes in rarer (10-year and 20-year) precipitation extremes are more severe and have larger internal variability spread than annual extremes. Internal variability spread is larger at fine scales that at the coarser scales simulated in climate models. Unpredictability from internal variability dominates model uncertainty and, for precipitation, its variance increases with warming, exceeding the variance across emission scenarios by fourfold for annual and tenfold for decadal extremes. These results indicate that fine-scale changes in future precipitation are less predictable than widely assumed and require much larger ensembles to assess reliably than changes at coarser scales.

Figures

Figures reproduced from arXiv: 2507.06527 by the authors.

Figure 1
Figure 1. Performance, for historical climatology and future land-averaged climate change signals (SSP3-7.0), of an RCM emulator vs. RCM ground truth. (a,b): grid-cell level RMSE of downscaled historical (1986-–2005) daily maximum temperature (TXx, DJF Tasmax) and precipitation (Rx1day, DJF Pr) climatologies against the VCSN gridded observational dataset. (c-e): Climate-change signals, i.e., difference between the future (208… view at source ↗
Figure 2
Figure 2. Internal variability of 10-year extreme precipitation and temperature, at fine-scales (GAN) vs. from the original GCM. Climate change signals for extreme precipitation (RX10y, a–b) and extreme temperature (T X10y, c–d) from the GAN-downscaled ACCESS-ESM1-5 ensemble (top row in each block) vs. interpolation from the GCM (bottom row in each block). Columns show the ensemble mean, the wettest/hottest and dri￾est/coldes… view at source ↗
Figure 3
Figure 3. Spread in fine-scale extreme precipitation projections from a large ensemble: a comparison of GCM, RCM, and emulator. End of century (2080–2099) climate change signals (%) for the SSP3-7.0 scenario relative to the historical climatology. (a–b): Ensemble-mean climate change signal in RX10y (a) and RX1Day (b) across four simulations—raw ACCESS-ESM1-5, GAN downscaled CMIP6 (20 members), GAN downscaled ACCESS￾ESM1-5 (40… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Temporal and spatial variability of uncertainty decomposition for extreme temperature and precipi￾tation. (a-b) spatial patterns of fractional uncertainty for temperature (a) and precipitation extremes (b), from model spread (left column), scenario spread (middle), and…
Figure 5
Figure 5. Figure 5: Temporal evolution of internal variability variance for precipitation extremes (RX10y, RX20y, RX1Day) in the raw vs GAN-Downscaled ACCESS-ESM1-5 ensemble. Here, internal variability variance is computed annually (no decadal smoothing as in Figure 4d), comparing GAN (gr…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

72 extracted references · 66 canonical work pages

  1. [1]

    E., Lenderink, G., van Meijgaard, E., and van den Hurk, B

    Aalbers, E. E., Lenderink, G., van Meijgaard, E., and van den Hurk, B. J. J. M. (2018). Local-scale changes in mean and heavy precipitation in Western Europe , climate change or internal variability? Climate Dynamics , 50(11):4745--4766

    work page 2018

  2. [2]

    Addison, H., Kendon, E., Ravuri, S., Aitchison, L., and Watson, P. A. (2024). Machine learning emulation of precipitation from km-scale regional climate simulations using a diffusion model. arXiv:2407.14158 [physics]

    arXiv 2024

  3. [3]

    J., Cannon, A

    Annau, N. J., Cannon, A. J., and Monahan, A. H. (2023). Algorithmic Hallucinations of Near - Surface Winds : Statistical Downscaling with Generative Adversarial Networks to Convection - Permitting Scales . arXiv:2302.08720 [physics]

    arXiv 2023

  4. [4]

    S., Rampal, N., and Gibson, P

    Bailie, T., Koh, Y. S., Rampal, N., and Gibson, P. B. (2024). Quantile- Regression - Ensemble : A Deep Learning Algorithm for Downscaling Extreme Precipitation . Proceedings of the AAAI Conference on Artificial Intelligence , 38(20):21914--21922. Number: 20

    work page 2024

  5. [5]

    Bartók, B., Wild, M., Folini, D., Lüthi, D., Kotlarski, S., Schär, C., Vautard, R., Jerez, S., and Imecs, Z. (2017). Projected changes in surface solar radiation in CMIP5 global climate models and in EURO - CORDEX regional climate models for Europe . Climate Dynamics , 49(7):2665--2683

    work page 2017

  6. [6]

    Baño-Medina, J., Iturbide, M., Fernández, J., and Gutiérrez, J. M. (2024). Transferability and explainability of deep learning emulators for regional climate model projections: Perspectives for future applications. Artificial Intelligence for the Earth Systems . Section: Artificial Intelligence for the Earth Systems

    work page 2024

  7. [7]

    and Hodges, K

    Bengtsson, L. and Hodges, K. I. (2019). Can an ensemble climate simulation be used to separate climate change signals from internal unforced variability? Climate Dynamics , 52(5):3553--3573

    work page 2019

  8. [8]

    L., López-Zurita, C

    Blanusa, M. L., López-Zurita, C. J., and Rasp, S. (2023). Internal variability plays a dominant role in global climate projections of temperature and precipitation extremes. Climate Dynamics , 61(3):1931--1945

    work page 2023

Show all 72 references
  1. [9]

    Boé, J., Mass, A., and Deman, J. (2023). A simple hybrid statistical–dynamical downscaling method for emulating regional climate models over Western Europe . Evaluation , application, and role of added value? Climate Dynamics , 61(1):271--294

    2023

  2. [10]

    B., Stuart, S., Broadbent, A

    Campbell, I., Gibson, P. B., Stuart, S., Broadbent, A. M., Sood, A., Pirooz, A. A. S., and Rampal, N. (2024). Comparison of three reanalysis-driven regional climate models over New Zealand : Climatology and extreme events. International Journal of Climatology , n/a(n/a). \_epr...

    2024 doi

  3. [11]

    Chadwick, R., Coppola, E., and Giorgi, F. (2011). An artificial neural network technique for downscaling GCM outputs to RCM spatial scale. Nonlinear Processes in Geophysics , 18(6):1013--1028. Publisher: Copernicus GmbH

    2011

  4. [12]

    K.-H., and Takbash, A

    Chapman, S., Syktus, J., Trancoso, R., Thatcher, M., Toombs, N., Wong, K. K.-H., and Takbash, A. (2023). Evaluation of Dynamically Downscaled CMIP6 - CCAM Models Over Australia . Earth's Future , 11(11):e2023EF003548. \_eprint: https://agupubs.onlinelibrary.wiley.com/doi/pdf/1...

    2023 doi

  5. [13]

    C., Xie, S.-P., and Dai, X

    Dai, A., Fyfe, J. C., Xie, S.-P., and Dai, X. (2015). Decadal modulation of global surface temperature by internal climate variability. Nature Climate Change , 5(6):555--559. Publisher: Nature Publishing Group

    2015

  6. [14]

    de Vries, I., Sippel, S., Zeder, J., Fischer, E., and Knutti, R. (2024). Increasing extreme precipitation variability plays a key role in future record-shattering event probability. Communications Earth & Environment , 5(1):1--15. Publisher: Nature Publishing Group

    2024

  7. [15]

    B., Ault, T., Delworth, T

    Deser, C., Lehner, F., Rodgers, K. B., Ault, T., Delworth, T. L., DiNezio, P. N., Fiore, A., Frankignoul, C., Fyfe, J. C., Horton, D. E., Kay, J. E., Knutti, R., Lovenduski, N. S., Marotzke, J., McKinnon, K. A., Minobe, S., Randerson, J., Screen, J. A., Simpson, I. R., and Tin...

    2020

  8. [16]

    Deser, C., Phillips, A., Bourdette, V., and Teng, H. (2012). Uncertainty in climate change projections: the role of internal variability. Climate Dynamics , 38(3):527--546

    2012

  9. [17]

    S., Alexander, M

    Deser, C., Phillips, A. S., Alexander, M. A., and Smoliak, B. V. (2014). Projecting North American Climate over the Next 50 Years : Uncertainty due to Internal Variability . Journal of Climate . Section: Journal of Climate

    2014

  10. [18]

    Di Luca, A., de Elía, R., and Laprise, R. (2013). Potential for added value in temperature simulated by high-resolution nested RCMs in present climate and in the climate change signal. Climate Dynamics , 40(1):443--464

    2013

  11. [19]

    P., Di Luca, A., Grose, M

    Di Virgilio, G., Evans, J. P., Di Luca, A., Grose, M. R., Round, V., and Thatcher, M. (2020). Realised added value in dynamical downscaling of Australian climate change. Climate Dynamics , 54(11-12):4675--4692

    2020

  12. [20]

    Doury, A., Somot, S., and Gadat, S. (2024). On the suitability of a convolutional neural network based RCM -emulator for fine spatio-temporal precipitation. Climate Dynamics

    2024

  13. [21]

    Doury, A., Somot, S., Gadat, S., Ribes, A., and Corre, L. (2022). Regional climate model emulator based on deep learning: concept and first evaluation of a novel hybrid downscaling approach. Climate Dynamics , 60(5):1751--1779

    2022

  14. [22]

    Fischer, E. M. and Knutti, R. (2016). Observed heavy precipitation increase confirms theory and early models. Nature Climate Change , 6(11):986--991. Publisher: Nature Publishing Group

    2016

  15. [23]

    B., Broadbent, A

    Gibson, P. B., Broadbent, A. M., Stuart, S. J., Lewis, H., Campbell, I., Rampal, N., Harrington, L. J., and Williams, J. (2025). Downscaled CMIP6 future climate projections for New Zealand : climatology and extremes. Weather and Climate Extremes , page 100784

    2025

  16. [24]

    B., Rampal, N., Dean, S

    Gibson, P. B., Rampal, N., Dean, S. M., and Morgenstern, O. (2024a). Storylines for Future Projections of Precipitation Over New Zealand in CMIP6 Models . Journal of Geophysical Research: Atmospheres , 129(5):e2023JD039664. \_eprint: https://agupubs.onlinelibrary.wiley.com/doi...

    2024 doi

  17. [25]

    B., Stuart, S., Sood, A., Stone, D., Rampal, N., Lewis, H., Broadbent, A., Thatcher, M., and Morgenstern, O

    Gibson, P. B., Stuart, S., Sood, A., Stone, D., Rampal, N., Lewis, H., Broadbent, A., Thatcher, M., and Morgenstern, O. (2024b). Dynamical downscaling CMIP6 models over New Zealand : added value of climatology and extremes. Climate Dynamics

    2024

  18. [26]

    S., and Bates, G

    Giorgi, F., Brodeur, C. S., and Bates, G. T. (1994). Regional Climate Change Scenarios over the United States Produced with a Nested Regional Climate Model . Journal of Climate , 7(3):375--399. Publisher: American Meteorological Society Section: Journal of Climate

    1994

  19. [27]

    Glawion, L., Polz, J., Kunstmann, H., Fersch, B., and Chwala, C. (2023). spateGAN : Spatio - Temporal Downscaling of Rainfall Fields Using a cGAN Approach . Earth and Space Science , 10(10):e2023EA002906. \_eprint: https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1029/2023EA002906

    2023 doi

  20. [28]

    Goodfellow, I., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., Courville, A., and Bengio, Y. (2014). Generative Adversarial Nets . In Advances in Neural Information Processing Systems , volume 27. Curran Associates, Inc

    2014

  21. [29]

    Gulrajani, I., Ahmed, F., Arjovsky, M., Dumoulin, V., and Courville, A. C. (2017). Improved Training of Wasserstein GANs . In Advances in Neural Information Processing Systems , volume 30. Curran Associates, Inc

    2017

  22. [30]

    J., , Peter B., G., , Nicolas, F., , Hamish, L., , Dave, F., , and Rosier, S

    Harrington, L. J., , Peter B., G., , Nicolas, F., , Hamish, L., , Dave, F., , and Rosier, S. M. (2025). On the procurement of physical risk assessments for climate-related disclosures: guidance from a climate science perspective. Journal of the Royal Society of New Zealand , 0...

    2025

  23. [31]

    S., Gregory, J

    Hawkins, E., Smith, R. S., Gregory, J. M., and Stainforth, D. A. (2016). Irreducible uncertainty in near-term climate projections. Climate Dynamics , 46(11):3807--3819

    2016

  24. [32]

    and Sutton, R

    Hawkins, E. and Sutton, R. (2009). The Potential to Narrow Uncertainty in Regional Climate Predictions . Bulletin of the American Meteorological Society , 90(8):1095--1108

    2009

  25. [33]

    and Sutton, R

    Hawkins, E. and Sutton, R. (2011). The potential to narrow uncertainty in projections of regional precipitation change. Climate Dynamics , 37(1-2):407--418

    2011

  26. [34]

    A., Domínguez, M., and Rodríguez-Camino, E

    Hernanz, A., García-Valero, J. A., Domínguez, M., and Rodríguez-Camino, E. (2022). A critical view on the suitability of machine learning techniques to downscale climate change projections: Illustration for temperature with a toy experiment. Atmospheric Science Letters , 23(6)...

    2022 doi

  27. [35]

    B., Edwards, N

    Holden, P. B., Edwards, N. R., Garthwaite, P. H., and Wilkinson, R. D. (2015). Emulation and interpretation of high-dimensional climate model outputs. Journal of Applied Statistics , 42(9):2038--2055

    2015

  28. [36]

    J., Addison, H., Doury, A., Somot, S., Watson, P

    Kendon, E. J., Addison, H., Doury, A., Somot, S., Watson, P. A. G., Booth, B. B. B., Coppola, E., Gutiérrez, J. M., Murphy, J., and Scullion, C. (2025). Potential for machine learning emulators to augment regional climate simulations in provision of local climate change inform...

    2025

  29. [37]

    J., Fischer, E

    Kendon, E. J., Fischer, E. M., and Short, C. J. (2023). Variability conceals emerging trend in 100yr projections of UK local hourly rainfall extremes. Nature Communications , 14(1):1133. Number: 1 Publisher: Nature Publishing Group

    2023

  30. [38]

    Lafferty, D. C. and Sriver, R. L. (2023). Downscaling and bias-correction contribute considerable uncertainty to local climate projections in CMIP6 . npj Climate and Atmospheric Science , 6(1):1--13. Publisher: Nature Publishing Group

    2023

  31. [39]

    B., Giguère, M., Brissette, F., Turcotte, R., Braun, M., and Scinocca, J

    Leduc, M., Mailhot, A., Frigon, A., Martel, J.-L., Ludwig, R., Brietzke, G. B., Giguère, M., Brissette, F., Turcotte, R., Braun, M., and Scinocca, J. (2019). The ClimEx Project : A 50- Member Ensemble of Climate Change Projections at 12-km Resolution over Europe and Northeaste...

    2019

  32. [40]

    and Deser, C

    Lehner, F. and Deser, C. (2023). Origin, importance, and predictive limits of internal climate variability. Environmental Research: Climate , 2(2):023001. Publisher: IOP Publishing

    2023

  33. [41]

    M., Brunner, L., Knutti, R., and Hawkins, E

    Lehner, F., Deser, C., Maher, N., Marotzke, J., Fischer, E. M., Brunner, L., Knutti, R., and Hawkins, E. (2020). Partitioning climate projection uncertainty with multiple large ensembles and CMIP5 /6. Earth System Dynamics , 11(2):491--508. Publisher: Copernicus GmbH

    2020

  34. [42]

    Leinonen, J., Nerini, D., and Berne, A. (2021). Stochastic Super - Resolution for Downscaling Time - Evolving Atmospheric Fields With a Generative Adversarial Network . IEEE Transactions on Geoscience and Remote Sensing , 59(9):7211--7223. Conference Name: IEEE Transactions on...

    2021

  35. [43]

    Y., Zepeda-Núñez, L., Schneider, T., Anderson, J., and Sha, F

    Lopez-Gomez, I., Wan, Z. Y., Zepeda-Núñez, L., Schneider, T., Anderson, J., and Sha, F. (2025). Dynamical-generative downscaling of climate model ensembles. Proceedings of the National Academy of Sciences , 122(17):e2420288122. Publisher: Proceedings of the National Academy of...

    2025

  36. [44]

    L., , and Smith, L

    Machete, R. L., , and Smith, L. A. (2016). Demonstrating the value of larger ensembles in forecasting physical systems. Tellus A: Dynamic Meteorology and Oceanography , 68(1):28393. Publisher: Taylor & Francis \_eprint: https://doi.org/10.3402/tellusa.v68.28393

    2016 doi

  37. [45]

    Maher, N., Lehner, F., and Marotzke, J. (2020). Quantifying the role of internal variability in the temperature we expect to observe in the coming decades. Environmental Research Letters , 15(5):054014. Publisher: IOP Publishing

    2020

  38. [46]

    Maher, N., Milinski, S., and Ludwig, R. (2021). Large ensemble climate model simulations: introduction, overview, and future prospects for utilising multiple types of large ensemble. Earth System Dynamics , 12(2):401--418. Publisher: Copernicus GmbH

    2021

  39. [47]

    M., Chandler, R

    Maraun, D., Wetterhall, F., Ireson, A. M., Chandler, R. E., Kendon, E. J., Widmann, M., Brienen, S., Rust, H. W., Sauter, T., Themeßl, M., Venema, V. K. C., Chun, K. P., Goodess, C. M., Jones, R. G., Onof, C., Vrac, M., and Thiele-Eich, I. (2010). Precipitation downscaling und...

    2010 doi

  40. [48]

    Martel, J.-L., Mailhot, A., Brissette, F., and Caya, D. (2018). Role of Natural Climate Variability in the Detection of Anthropogenic Climate Change Signal for Mean and Extreme Precipitation at Local and Regional Scales . Journal of Climate . Section: Journal of Climate

    2018

  41. [49]

    A., Arblaster, J

    Meehl, G. A., Arblaster, J. M., and Branstator, G. (2012). Mechanisms Contributing to the Warming Hole and the Consequent U . S . East – West Differential of Heat Extremes . Journal of Climate . Section: Journal of Climate

    2012

  42. [50]

    Miralles, O., Steinfeld, D., Martius, O., and Davison, A. C. (2022). Downscaling of Historical Wind Fields over Switzerland Using Generative Adversarial Networks . Artificial Intelligence for the Earth Systems , 1(4). Publisher: American Meteorological Society Section: Artific...

    2022

  43. [51]

    and Osindero, S

    Mirza, M. and Osindero, S. (2014). Conditional Generative Adversarial Nets . arXiv:1411.1784 [cs, stat]

    2014 arXiv

  44. [52]

    G., Knutti, R., Lehner, F., Deser, C., and Sanderson, B

    Pendergrass, A. G., Knutti, R., Lehner, F., Deser, C., and Sanderson, B. M. (2017). Precipitation variability increases in a warmer climate. Scientific Reports , 7(1):17966. Publisher: Nature Publishing Group

    2017

  45. [53]

    and Ludwig, R

    Poschlod, B. and Ludwig, R. (2021). Internal variability and temperature scaling of future sub-daily rainfall return levels over Europe . Environmental Research Letters , 16(6):064097. Publisher: IOP Publishing

    2021

  46. [54]

    B., Sherwood, S., and Abramowitz, G

    Rampal, N., Gibson, P. B., Sherwood, S., and Abramowitz, G. (2024a). On the Extrapolation of Generative Adversarial Networks for downscaling precipitation extremes in warmer climates. Geophysical Research Letters , 51(e2024GL112492). arXiv:2409.13934

    2024 arXiv

  47. [55]

    B., Sherwood, S., Abramowitz, G., and Hobeichi, S

    Rampal, N., Gibson, P. B., Sherwood, S., Abramowitz, G., and Hobeichi, S. (2025). A Reliable Generative Adversarial Network Approach for Climate Downscaling and Weather Generation . Journal of Advances in Modeling Earth Systems , 17(1):e2024MS004668. \_eprint: https://onlineli...

    2025 doi

  48. [56]

    B., Sood, A., Stuart, S., Fauchereau, N

    Rampal, N., Gibson, P. B., Sood, A., Stuart, S., Fauchereau, N. C., Brandolino, C., Noll, B., and Meyers, T. (2022). High-resolution downscaling with interpretable deep learning: Rainfall extremes over New Zealand . Weather and Climate Extremes , 38:100525

    2022

  49. [57]

    B., Baño-Medina, J., Abramowitz, G., Beucler, T., González-Abad, J., Chapman, W., Harder, P., and Gutiérrez, J

    Rampal, N., Hobeichi, S., Gibson, P. B., Baño-Medina, J., Abramowitz, G., Beucler, T., González-Abad, J., Chapman, W., Harder, P., and Gutiérrez, J. M. (2024b). Enhancing Regional Climate Downscaling through Advances in Machine Learning . Artificial Intelligence for the Earth ...

    2024

  50. [58]

    and Braun, M

    Rondeau-Genesse, G. and Braun, M. (2019). Impact of internal variability on climate change for the upcoming decades: analysis of the CanESM2 - LE and CESM - LE large ensembles. Climatic Change , 156(3):299--314

    2019

  51. [59]

    Rummukainen, M. (2016). Added value in regional climate modeling. WIREs Climate Change , 7(1):145--159. \_eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1002/wcc.378

    2016 doi

  52. [60]

    Schwarzwald, K., Poppick, A., Rugenstein, M., Bloch-Johnson, J., Wang, J., McInerney, D., and Moyer, E. J. (2021). Changes in Future Precipitation Mean and Variability across Scales . Journal of Climate . Section: Journal of Climate

    2021

  53. [61]

    C., Cole, J

    Swart, N. C., Cole, J. N. S., Kharin, V. V., Lazare, M., Scinocca, J. F., Gillett, N. P., Anstey, J., Arora, V., Christian, J. R., Hanna, S., Jiao, Y., Lee, W. G., Majaess, F., Saenko, O. A., Seiler, C., Seinen, C., Shao, A., Sigmond, M., Solheim, L., von Salzen, K., Yang, D.,...

    2019

  54. [62]

    L., Schär, C., Lüthi, D., and Kjellström, E

    Sørland, S. L., Schär, C., Lüthi, D., and Kjellström, E. (2018). Bias patterns and climate change signals in GCM - RCM model chains. Environmental Research Letters , 13(7):074017. Publisher: IOP Publishing

    2018

  55. [63]

    Tait, A., Henderson, R., Turner, R., and Zheng, X. (2006). Thin plate smoothing spline interpolation of daily rainfall for New Zealand using a climatological rainfall surface. International Journal of Climatology , 26(14):2097--2115. \_eprint: https://rmets.onlinelibrary.wiley...

    2006 doi

  56. [64]

    Tait, A., Sturman, J., and Clark, M. (2012). An assessment of the accuracy of interpolated daily rainfall for New Zealand . Weather and Climate

    2012

  57. [65]

    and McGregor, J

    Thatcher, M. and McGregor, J. L. (2009). Using a Scale - Selective Filter for Dynamical Downscaling with the Conformal Cubic Atmospheric Model . Monthly Weather Review , 137(6):1742--1752. Publisher: American Meteorological Society Section: Monthly Weather Review

    2009

  58. [66]

    von Trentini, F., Leduc, M., and Ludwig, R. (2019). Assessing natural variability in RCM signals: comparison of a multi model EURO - CORDEX ensemble with a 50-member single model large ensemble. Climate Dynamics , 53(3):1963--1979

    2019

  59. [67]

    R., Lehner, F., Pendergrass, A

    Wood, R. R., Lehner, F., Pendergrass, A. G., and Schlunegger, S. (2021). Changes in precipitation variability across time scales in multiple global climate model large ensembles. Environmental Research Letters , 16(8):084022. Publisher: IOP Publishing

    2021

  60. [68]

    Yip, S., Ferro, C. A. T., Stephenson, D. B., and Hawkins, E. (2011). A Simple , Coherent Framework for Partitioning Uncertainty in Climate Predictions . Journal of Climate , 24(17):4634--4643

    2011

  61. [69]

    L., and Held, I

    Zhang, R., Delworth, T. L., and Held, I. M. (2007). Can the Atlantic Ocean drive the observed multidecadal variability in Northern Hemisphere mean temperature? Geophysical Research Letters , 34(2):2006GL028683

    2007

  62. [70]

    Zhang, W., Furtado, K., Zhou, T., Wu, P., and Chen, X. (2022). Constraining extreme precipitation projections using past precipitation variability. Nature Communications , 13(1):6319. Publisher: Nature Publishing Group

    2022

  63. [71]

    Zhang, W., Zhou, T., and Wu, P. (2024). Anthropogenic amplification of precipitation variability over the past century. Science , 385(6707):427--432. Publisher: American Association for the Advancement of Science

    2024

  64. [72]

    A., Law, R

    Ziehn, T., Chamberlain, M. A., Law, R. M., Lenton, A., Bodman, R. W., Dix, M., Stevens, L., Wang, Y.-P., and Srbinovsky, J. (2020). The Australian Earth System Model : ACCESS - ESM1 .5. Journal of Southern Hemisphere Earth Systems Science , 70(1):193--214. Publisher: CSIRO PUBLISHING

    2020

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.