CESAR: A Convolutional Echo State AutoencodeR for High-Resolution Wind Forecasting

Matthew Bonas; Paola Crippa; Paolo Giani; Stefano Castruccio

REVIEW 4 major objections 5 minor 68 references

CESAR: A Convolutional Echo State AutoencodeR for High-Resolution Wind Forecasting

T0 review · 4 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read CESAR, a convolutional autoencoder wired to an echo state network, forecasts high-resolution wind speed and power up to 17% better than the best alternative methods on a sub-kilometer simulation over Riyadh.

desk verdict A useful CAE+ESN synthesis for wind forecasting, but the headline improvement hangs on a single 24-hour test window and the paper's own reconstruction equation is inconsistent with its reported results. read the letter →

arxiv 2412.10578 v1 pith:EHCZTTQC submitted 2024-12-13 stat.AP stat.ML

Matthew Bonas , Paolo Giani , Paola Crippa , Stefano Castruccio This is my paper · ORCID

classification stat.APstat.ML

keywords convolutionalautoencoderechostatenetworkwindspeedforecastingspatio-temporalstatisticsuncertaintyquantificationhigh-resolutionWRFsimulationrenewableenergyhierarchicalstate-spacemodel

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 introduces CESAR, a forecasting model that compresses high-resolution wind fields with a convolutional autoencoder and then models the evolution of the compressed features in time with an echo state network. The paper frames this two-block neural architecture as a nonlinear hierarchical spatio-temporal statistical model, which lets it produce calibrated uncertainty intervals as well as point forecasts. On a sub-kilometer WRF simulation over Riyadh, Saudi Arabia, CESAR forecasts wind speed with a median one-step mean squared error of $19.21\times 10^{-1}\,(\mathrm{m/s})^2$, a 16.9% improvement over an LSTM and larger gains over ARIMA and persistence, and it reconstructs the spatial field 97.6% better than PCA. The stated goal is to provide a computationally affordable forecasting template for planned wind farms in regions that lack operational high-resolution weather products.

What carries the argument

The central object is the CESAR pipeline itself: a convolutional autoencoder with three encoder layers of filters $\{32,64,128\}$, stride 2, and LeakyReLU activations that reduces each spatial field to a latent representation, followed by a deep echo state network whose fixed random recurrent weights satisfy the echo state property (spectral radius below one) and whose readout weights are learned. The load-bearing decomposition is the two-step state-space split, in which the CAE plays the observation equation and the ESN plays the state equation, and the iterative forecasting scheme that feeds predictions back as inputs for the next lead time. The same decomposition licenses the ensemble uncertainty quantification: dropout on the CAE weights for spatial intervals, and independent draws of the ESN's spike-and-slab weight matrices for temporal intervals.

What would settle it

Re-fit CESAR, LSTM, ARIMA, and persistence with a rolling-origin protocol over all 240 hours and over a second WRF run, and check whether CESAR's median one-step MSE of $19.21\times 10^{-1}\,(\mathrm{m/s})^2$ remains below LSTM's $23.12\times 10^{-1}\,(\mathrm{m/s})^2$; if the gap closes or reverses outside the reported 24-hour window, the central claim fails.

Watch

Extended reading notes

Core claim

CESAR treats the data as the observation equation of a state-space model, with a deep convolutional autoencoder as the nonlinear observation function and an echo state network as the nonlinear state transition. The CAE is trained first, by reconstruction mean squared error, to map each $256\times 256$ wind field into a small feature vector; the ESN is then trained on those features to produce iterative multi-step forecasts. The paper's central empirical claim is that this combination outperforms ARIMA, persistence, and LSTM on both a 2D Burgers' equation benchmark and the Riyadh WRF simulation, with the largest reported advantage in the real-data case: median one-step forecast MSE of $19.21\times 10^{-1}\,(\mathrm{m/s})^2$ against $23.12\times 10^{-1}$ for LSTM, and spatial reconstruction MSE of $0.17\times 10^{-1}$ against $7.00\times 10^{-1}$ for PCA. Uncertainty is quantified by dropout ensembles for the spatial block and spike-and-slab weight ensembles for the temporal block, with empirical coverage within about 2 percentage points of nominal levels on the wind data.

Load-bearing premise

The load-bearing premise is that one 24-hour test window from a single 10-day WRF summer simulation is representative enough to rank forecasting methods, and that the autoencoder's final layer uses the output function needed to reach the reported reconstruction error.

Editorial extensions

If this is right

On the Riyadh WRF simulation, CESAR's median one-step wind-speed forecast MSE is $19.21\times 10^{-1}\,(\mathrm{m/s})^2$, a 16.9% improvement over the LSTM baseline and 34.0% over ARIMA, so the model would be the best of the tested methods for short-horizon wind forecasting at planned turbine sites.
The CAE spatial compression is much better than linear reductions on this data, with reconstruction MSE of $0.17\times 10^{-1}$ versus $7.00\times 10^{-1}$ for PCA, implying that nonlinear feature extraction captures fine-scale structure that linear bases miss.
Calibrated prediction intervals are available in both space and time: on WRF data, empirical coverage for nominal 95%, 90%, and 80% intervals lands within roughly 2 percentage points on average.
Because CESAR trains in under 30 minutes on a 65,536-location, 240-hour field using two GPUs, it is computationally feasible as an operational forecasting tool for regions without real-time high-resolution weather model output.
The same CAE-plus-ESN architecture transfers to a 2D Burgers' equation benchmark, showing the method is not specific to wind but applies to other nonlinear spatio-temporal fields with short training records.

Reading between the lines

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

If the method transfers beyond this single simulation, the natural deployment is to retrain CESAR on multi-season WRF runs and validate against surface observations, since the paper's single 24-hour test window cannot establish operational reliability by itself.
The daytime/nighttime skill split noted in the paper implies the headline gain is likely concentrated in stable nighttime hours; a targeted evaluation of convective daytime hours would reveal where the next modeling effort is needed.
The two-block decomposition is modular: replacing the CAE with a graph convolutional encoder would extend CESAR to irregular observation networks, which are common in observational wind data.
The uncertainty-quantification recipe, dropout for space and weight-resampling for time, could be reused with other autoencoder and state-space pairings, offering a template for calibrated deep spatio-temporal forecasting beyond wind energy.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Two things to know about arXiv:2412.10578. First, the architecture—a convolutional autoencoder feeding an echo state network, framed as a nonlinear hierarchical state-space model—is a genuinely new combination in the cited literature, and it looks like a sensible one for short, high-dimensional spatio-temporal series. Second, as written, the paper cannot reproduce its own central numbers: equation (4) uses softmax on a single output channel, which returns 1 everywhere, so the reconstruction MSE in Table 1 is impossible. That has to be a typo—presumably they meant a sigmoid or linear activation—but it is load-bearing and needs fixing before the results can be trusted.

What the paper does well: the CAE+ESN idea is clean and computationally efficient; training takes minutes on a GPU for the Burgers example and under 30 minutes for the WRF field. The simulation study on 2D Burgers with ten independent runs gives a reasonable check of reconstruction and forecasting skill across repeated initial conditions. The authors also ship code and data for the Burgers part, which is a real plus.

The soft spots are in the application. The headline 'up to 17%' improvement over LSTM comes from a single 24-hour test window (hours 218–240) of one 10-day WRF summer simulation. There is no rolling-origin evaluation, no second season, no multiple initialization, and no comparison against observational wind data. The paper itself notes that daytime forecasts are much worse than nighttime ones, so the chosen window may flatter the method. That is the primary weakness and it directly undermines the generality of the headline claim.

There is also a circularity in the uncertainty quantification: the dropout rate is selected after seeing the test data so that empirical coverage matches the nominal level. That makes the coverage tables in Table 3 look better than they should. And the wind-power section uses a constant shear coefficient 1/7, which is standard but adds another layer of assumption.

Overall, the methodological idea is worth taking seriously, and the paper is clearly written. But the evaluation design is too thin to support the abstract's claim, and the activation inconsistency means the exact model is unspecified. I'd send it to peer review—it deserves a serious referee—but I would insist on fixing the activation, adding rolling-origin or multi-period evaluation, and re-deriving the dropout calibration on a validation set.

For your reading group, it's a maybe. If you work on ESN dimension reduction or spatio-temporal deep learning, cite it as related work after a fix; I wouldn't cite the current version's numbers.

Referee Report

4 major / 5 minor

Summary. The paper proposes CESAR, a two-step spatio-temporal forecasting model that combines a convolutional autoencoder (CAE) for spatial dimension reduction with an echo state network (ESN) for temporal dynamics, and frames the construction as a hierarchical state-space model. Inference is two-stage: the CAE is trained by reconstruction error, and the ESN is then fit on the extracted latent features. Uncertainty quantification is attempted via dropout for the spatial component and ensemble ESN draws for the temporal component. The method is tested on a 2D Burgers' equation and on a 10-day, 450m-resolution WRF simulation over Riyadh, with reported reconstruction and forecast improvements over PCA, kriging, ARIMA, persistence, and LSTM, and an abstract claim of wind-speed and wind-power forecast improvements up to 17% over the best alternative method.

Significance. If the claims hold, CESAR would be a useful, scalable nonlinear alternative for high-resolution spatio-temporal wind forecasting, and the hierarchical framing is a genuine bridge between deep learning constructions and statistical space-time models. The paper's strengths include a clearly described two-step training scheme with modest computational cost, a reproducible simulation study on Burgers' equation, and public code and data for the simulation. The significance is limited, however, by the thin empirical evaluation: the headline WRF gain rests on a single 24-hour test window from one 10-day simulation, and the uncertainty-quantification calibration is tuned on the test period. These issues prevent the paper, as currently written, from supporting the strong operational-forecasting claims in the abstract.

major comments (4)

[Section 5.1, Table 2, Abstract] The headline 16.9% CESAR-over-LSTM improvement is computed from hours 218-240 of a single 10-day WRF summer run, i.e., one test window, one initialization, and one season. The text itself notes in Section 5.1 and in the supplementary material that daytime forecasts are markedly worse than nighttime forecasts, so the diurnal composition of this particular window matters. Without a rolling-origin evaluation, repeated initializations, multiple seasons, or at least a per-hour breakdown of all methods, the reported advantage cannot be distinguished from period-specific behavior. The empirical claim in the abstract should be softened or, preferably, supported by a more robust evaluation protocol.
[Section 3.2, Eq. (4), Table 1] Equation (4) states that the final CAE layer uses a softmax activation. For the WRF application the output has a single channel (p=1), and softmax over one logit is identically 1, which cannot yield the reconstruction errors reported in Table 1. Even for p>1, a softmax over variables would force the reconstructed components to sum to 1, which is not a natural model for wind speed. The manuscript therefore does not fully specify the activation that produced the reported results; a linear (or other) final activation should be stated explicitly, and the equation corrected.
[Section 4.4, Section 5.1, Table 3] The dropout rate is selected to make prediction intervals as calibrated as possible (Section 4.4) and the WRF application then uses the 'optimal' rate phi=0.3 (Section 5.1). If this selection is based on coverage computed on the same test period reported in Table 3, the calibration numbers are circular and should not be presented as evidence of calibrated uncertainty. The authors should either choose the dropout rate with a validation split or report a sensitivity analysis over rates before assessing test coverage.
[Section 3.5, Table 3] Spatial and temporal uncertainties are quantified separately and reported separately, but the paper does not explain how to combine them into a single forecast interval for the wind-speed or wind-power quantities of interest. Since Section 3.5 states that the aim is calibrated prediction intervals, the absence of a combined calibration check leaves the practical UQ claim incomplete. At minimum, the paper should state whether the two sources are meant to be combined, and if so, how.

minor comments (5)

[Table 1, Table 2] The caption units are given as 'x10^-1 ms^-1', but the entries are MSE values and the units should be (m/s)^2 (or x10^-1 (m/s)^2). Please clarify.
[Section 4.2] The sentence 'The CAE drastically outperforms the all other methods' contains a grammatical error ('the all other').
[Section 5.2] The phrase 'one step aheat forecasts' appears in the wind power section; it should read 'one-step ahead forecasts'.
[References] Dowell et al. (2022a) and (2022b) are the same reference, and Crippa et al. (2021a) and (2021b) are identical; these duplicates should be consolidated.
[Section 5.1, Figure 5] The text says Figure 5 shows '1-step ahead forecast (row 1) and average forecasts in time (row 2) across the training set', while the caption says 'average wind speed forecasts across space for tau=24 hours'. This discrepancy should be reconciled.

Circularity Check

1 steps flagged · score 6.0 of 10

Uncertainty-quantification coverage is circular by construction: the dropout rate is selected to optimize calibration and the resulting coverage is then reported as evidence of calibration; the forecast-skill comparison itself is a genuine holdout evaluation.

fitted input called prediction [Section 4.4 (Uncertainty Quantification), same tuning reported in Section 5.1]
"The percentage of parameters dropped was chosen such that the prediction intervals was as calibrated as possible. The optimal choice for the dropout rate was 0.79 or φ = 0.21 in Section 3.5. ... we choose to estimate the uncertainty surrounding the spatial reconstruction using dropout which optimal rate of φ = 0.3"

The UQ goal is stated in Section 3.5 as calibrated coverage: a 95% interval 'must cover the true (unobserved) value ∼95% of the time.' The dropout rate is then explicitly selected to make the intervals 'as calibrated as possible.' Therefore the empirical coverages reported in Table 3 are the objective function used to choose the dropout rate, not an independent evaluation of the method's coverage properties. The same rate is reused for the WRF application (φ = 0.3) and the resulting coverage is again presented as evidence that uncertainty is 'properly estimated.' With no separate validation split and no pre-specified rate, the calibration claim reduces to the fitting criterion: the intervals cover at the nominal rate because the rate was chosen to make them do so.

full rationale

The only circularity found is in the uncertainty-quantification component. Section 3.5 defines the goal of the UQ as calibrated coverage, and Section 4.4 states that the dropout rate 'was chosen such that the prediction intervals was as calibrated as possible,' with an 'optimal' rate of φ = 0.21 for the Burgers’ simulation and φ = 0.3 for the WRF application. Because the dropout rate is selected using the coverage objective and the resulting empirical coverage in Table 3 is then reported as evidence that the intervals are correctly calibrated, the coverage numbers are a restatement of the selection criterion rather than an out-of-sample validation. This is a fitted-input-called-prediction loop for the UQ claim. The headline forecasting comparison is not circular in the same way: CESAR’s ESN is trained on the first 217 hours and evaluated on the held-out last 24 hours, and the reported 16.9% improvement over LSTM is a genuine, if statistically fragile, holdout comparison. The fragility of relying on a single 24-hour test window is an evaluation-design and generalizability concern, not circularity. Similarly, the softmax final activation in Eq. (4) is internally inconsistent with the reported reconstruction MSE, but that is a modeling-specification error rather than a reduction of a prediction to its inputs. I found no load-bearing self-citation chain or imported uniqueness theorem: the WRF data come from prior work by Giani et al. (2022), and the ESN/CAE components are standard techniques cited from the literature. The circularity score is set to 6 because one secondary prediction (calibrated uncertainty) reduces by construction, while the central forecast-skill claim retains independent content.

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

The CESAR architecture introduces no new physical entities. Its additional baggage is mostly data assumptions, hand-chosen hyperparameters, and one internally inconsistent activation choice.

free parameters (5)

Dropout rate phi = 0.21 (Burgers), 0.3 (WRF)
Chosen in Sections 4.4 and 5.1 so that prediction intervals are as calibrated as possible, making coverage claims tuned rather than independently validated.
CAE architecture (layers, filters, stride, epochs, batch size) = L=3, k=3, stride=2, filters 16/32/64 (Burgers) and 32/64/128 (WRF), epochs 500/1000, batch 2/10
Hand-selected; sensitivity analysis for these choices is deferred to supplementary material.
ESN depth and hidden nodes = D=1, nh=64 (Burgers), nh=128 (WRF)
Optimal architecture chosen via a sensitivity study in the supplementary material, so it is a per-dataset tuning choice.
LeakyReLU slope rho = 0.3
Set by hand in Section 3.2 with no sensitivity analysis in the main text.
Shear coefficient kappa = 1/7
Assumed constant in space and time in Section 5.2.1; standard in the power law literature but a modeling choice affecting power estimates.

assumptions (5)

domain assumption WRF simulation output is treated as ground truth for wind speed and power assessment.
No observational validation is provided; all forecast errors are relative to the simulation in Section 2.
domain assumption Gaussian observation equation and min-max normalization are assumed in Section 3.1.
Wind speed is non-Gaussian in general, and the normalization affects the CAE reconstruction targets.
standard math The echo state property and spectral radius constraint are inherited from the ESN literature.
Section 3.3 relies on Jaeger (2001) and Lukosevicius (2012) without re-deriving the property.
standard math The deep ESN equations (5a) to (5g) are adopted from McDermott and Wikle (2019b).
The model is taken as a prior modeling framework without derivation in this paper.
ad hoc to paper The softmax activation in Equation (4) is assumed to reconstruct normalized wind data.
As written, softmax on a single output channel always returns 1, so the paper implicitly assumes a different final activation than the one stated.

how reviews work

0 comments

Cite this review

Pith. "Pith review of CESAR: A Convolutional Echo State AutoencodeR for High-Resolution Wind Forecasting." pith.science (2026). https://pith.science/paper/EHCZTTQC

@misc{pith2026241210578,
  author       = {Pith},
  title        = {Pith review of: CESAR: A Convolutional Echo State AutoencodeR for High-Resolution Wind Forecasting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EHCZTTQC}},
  note         = {Machine review of arXiv:2412.10578}
}

read the original abstract

An accurate and timely assessment of wind speed and energy output allows an efficient planning and management of this resource on the power grid. Wind energy, especially at high resolution, calls for the development of nonlinear statistical models able to capture complex dependencies in space and time. This work introduces a Convolutional Echo State AutoencodeR (CESAR), a spatio-temporal, neural network-based model which first extracts the spatial features with a deep convolutional autoencoder, and then models their dynamics with an echo state network. We also propose a two-step approach to also allow for computationally affordable inference, while also performing uncertainty quantification. We focus on a high-resolution simulation in Riyadh (Saudi Arabia), an area where wind farm planning is currently ongoing, and show how CESAR is able to provide improved forecasting of wind speed and power for proposed building sites by up to 17% against the best alternative methods.

Figures

Figures reproduced from arXiv: 2412.10578 by the authors.

**Figure 1.** Wind speed (ms−1 ) output from the WRF model over Riyadh, Saudi Arabia for (A) nighttime and (B) daytime on 2016-07-26. Panel (C) shows the average hourly wind speed across all 10 days from 2016-07-22 through 2016-08-01 convection that transports surface heat up to the top of the atmospheric boundary layer. Figure 1B shows the convective turbulence structure (at noon local time of July 26, 2016), which can be observ… view at source ↗

**Figure 2.** Schematic for a convolutional autoencoder for two dimensional. Red represents the input [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗

**Figure 3.** Simulation from the two dimensional Burgers’ equation (7) with periodic boundary [PITH_FULL_IMAGE:figures/full_fig_p016_3.png] view at source ↗

Figures from the paper (3 more)

**Figure 4.** Figure 4: Comparison between the true data and the CESAR and ARIMA forecasts for both [PITH_FULL_IMAGE:figures/full_fig_p017_4.png]

**Figure 5.** Figure 5: Forecast of the WRF wind speed data for proposed CESAR method and the ARIMA [PITH_FULL_IMAGE:figures/full_fig_p021_5.png]

**Figure 6.** Figure 6: Wind power forecasts from CESAR for 2016/07/31 01:00:00 UTC for the domain of [PITH_FULL_IMAGE:figures/full_fig_p023_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

68 extracted references · 61 canonical work pages

[1]

write newline

" write newline "" before.all 'output.state := FUNCTION format.url url empty "" url if FUNCTION article output.bibitem format.authors "author" output.check author format.key output output.year.check new.block format.title "title" output.check new.block crossref missing format.jour.vol output format.article.crossref output.nonnull format.pages output if ne...
[2]

Li, and H

Ai, X., S. Li, and H. Xu (2023). Wind speed prediction model using ensemble empirical mode decomposition, least squares support vector machine and long short-term memory. Frontiers in Energy Research\/ 10

work page 2023
[3]

Bateman, H. (1915). Some recent researches on the motion of fluids. Monthly Weather Review\/ 43\/ (4), 163--170

work page 1915
[4]

Bonas, M. and S. Castruccio (2023). Calibration of spatial forecasts from citizen science urban air pollution data with sparse recurrent neural networks. Annals of Applied Statistics\/ 17\/ (3), 1820--1840

work page 2023
[5]

Datta, C

Bonas, M., A. Datta, C. K. Wikle, E. L. Boone, F. S. Alamri, B. V. Hari, I. Kavila, S. J. Simmons, S. M. Jarvis, W. S. Burr, D. E. Pagendam, W. Chang, and S. Castruccio (2024). Assessing predictability of environmental time series with statistical and machine learning models. Environmetrics\/ , e2864

work page 2024
[6]

Bonas, M., D. H. Richter, and S. Castruccio (2024). A physics-informed, deep double reservoir network for forecasting boundary layer velocity. Journal of the American Statistical Association - Applications and Case Studies\/ . in press

work page 2024
[7]

Bonas, M., C. K. Wikle, and S. Castruccio (2024). Calibrated forecasts of quasi-periodic climate processes with deep echo state networks and penalized quantile regression. Environmetrics\/ 35\/ (1), e2833

work page 2024
[8]

Bottou, L. (2010). Large-scale machine learning with stochastic gradient descent. In Y. Lechevallier and G. Saporta (Eds.), Proceedings of COMPSTAT'2010 , pp.\ 177--186. Physica-Verlag HD

work page 2010

Show all 68 references

[9]

Brockwell, P. J. and R. A. Davis (2016). Introduction to Time Series and Forecasting . New York: Springer

2016
[10]

Burgers, J. (1948). A mathematical model illustrating the theory of turbulence. Volume 1 of Advances in Applied Mechanics , pp.\ 171--199. Elsevier

1948
[11]

Castruccio, and M

Chen, W., S. Castruccio, and M. Genton (2021). Assessing the risk of disruption of wind turbine operations in saudi arabia using bayesian spatial extremes. Extreme\/ 24 , 267–292

2021
[12]

Castruccio, M

Chen, W., S. Castruccio, M. G. Genton, and P. Crippa (2018). Current and future estimates of wind energy potential over saudi arabia. Journal of Geophysical Research: Atmospheres\/ 123\/ (12), 6443--6459

2018
[13]

Cressie, N. and G. Johannesson (2008). Fixed rank kriging for very large spatial data sets. Journal of the Royal Statistical Society: Series B (Statistical Methodology)\/ 70\/ (1), 209--226

2008
[14]

Alifa, D

Crippa, P., M. Alifa, D. Bolster, M. G. Genton, and S. Castruccio (2021a). A temporal model for vertical extrapolation of wind speed and wind energy assessment. Applied Energy\/ 301 , 117378

2021
[15]

Alifa, D

Crippa, P., M. Alifa, D. Bolster, M. G. Genton, and S. Castruccio (2021b). A temporal model for vertical extrapolation of wind speed and wind energy assessment. Applied Energy\/ 301 , 117378

2021
[16]

Giani, G

De Moliner , G., P. Giani, G. Lonati, and P. Crippa (2024). Sensitivity of multiscale large eddy simulations for wind power calculations in complex terrain. Applied Energy\/ 364 , 123195

2024
[17]

Dowell, D. C., C. R. Alexander, E. P. James, S. S. Weygandt, S. G. Benjamin, G. S. Manikin, B. T. Blake, J. M. Brown, J. B. Olson, M. Hu, T. G. Smirnova, T. Ladwig, J. S. Kenyon, R. Ahmadov, D. D. Turner, J. D. Duda, and T. I. Alcott (2022a). The high-resolution rapid refresh ...

2022
[18]

Dowell, D. C., C. R. Alexander, E. P. James, S. S. Weygandt, S. G. Benjamin, G. S. Manikin, B. T. Blake, J. M. Brown, J. B. Olson, M. Hu, T. G. Smirnova, T. Ladwig, J. S. Kenyon, R. Ahmadov, D. D. Turner, J. D. Duda, and T. I. Alcott (2022b). The high-resolution rapid refresh ...

2022
[19]

Dubey, S. R., S. K. Singh, and B. B. Chaudhuri (2022). Activation functions in deep learning: A comprehensive survey and benchmark. Neurocomputing\/ 503 , 92--108

2022
[20]

Durbin, J. and S. J. Koopman (2012). Time Series Analysis by State Space Methods . Oxford University Press

2012
[21]

Ecmwf's operational model analysis, starting in 2011

ECMWF (2011). Ecmwf's operational model analysis, starting in 2011. Research data archive at the National Center for Atmospheric Research\/

2011
[22]

Gal, Y. and Z. Ghahramani (2016). Dropout as a bayesian approximation: Representing model uncertainty in deep learning. In M. F. Balcan and K. Q. Weinberger (Eds.), Proceedings of The 33rd International Conference on Machine Learning , Volume 48 of Proceedings of Machine Learn...

2016
[23]

Gao, Q. and M. Zou (2017). An analytical solution for two and three dimensional nonlinear burgers' equation. Applied Mathematical Modelling\/ 45 , 255--270

2017
[24]

Huang, C

Ge, R., F. Huang, C. Jin, and Y. Yuan (2015). Escaping from saddle points - online stochastic gradient for tensor decomposition. In P. Gr \" u nwald, E. Hazan, and S. Kale (Eds.), Proceedings of The 28th Conference on Learning Theory, COLT 2015, Paris, France, July 3-6, 2015 ,...

2015
[25]

Geneva, N. and N. Zabaras (2020). Modeling the dynamics of pde systems with physics-constrained deep auto-regressive networks. Journal of Computational Physics\/ 403 , 109056

2020
[26]

Giani, P. and P. Crippa (2024). On the sensitivity of large-eddy simulations of the atmospheric boundary layer coupled with realistic large-scale dynamics. Monthly Weather Review\/ 152\/ (4), 1057 -- 1075

2024
[27]

Giani, P., M. G. Genton, and P. Crippa (2022). Modeling the convective boundary layer in the Terra Incognita : Evaluation of different strategies with real-case simulations. Monthly Weather Review\/ 150\/ (5), 981--1001

2022
[28]

Tagle, M

Giani, P., F. Tagle, M. G. Genton, S. Castruccio, and P. Crippa (2020). Closing the gap between wind energy targets and implementation for emerging countries. Applied Energy\/ 269 , 115085

2020
[29]

Brownsword, G

Giebel, G., R. Brownsword, G. Kariniotakis, M. Denhard, and C. Draxl (2011). The state-of-the-art in short-term prediction of wind power: A literature overview. Technical report, ANEMOS. plus

2011
[30]

Gwec global wind report

Global Wind Energy Council (2023). Gwec global wind report

2023
[31]

Bengio, and A

Goodfellow, I., Y. Bengio, and A. Courville (2016). Deep Learning . MIT Press. http://www.deeplearningbook.org

2016
[32]

Gualtieri, G. (2019). A comprehensive review on wind resource extrapolation models applied in wind energy. Renewable and Sustainable Energy Reviews\/ 102 , 215--233

2019
[33]

Hannachi, A., I. T. Jolliffe, and D. B. Stephenson (2007). Empirical orthogonal functions and related techniques in atmospheric science: A review. International Journal of Climatology\/ 27\/ (9), 1119--1152

2007
[34]

Hochreiter, S. and J. Schmidhuber (1997). Long short-term memory. Neural computation\/ 9\/ (8), 1735--1780

1997
[35]

Castruccio, and M

Huang, H., S. Castruccio, and M. G. Genton (2021). Forecasting high-frequency spatio-temporal wind power with dimensionally reduced echo state networks. Journal of the Royal Statistical Society - Series C\/ 71\/ (2), 449--466

2021
[36]

Jaeger, H. (2001). The ``echo state'' approach to analysing and training recurrent neural networks-with an erratum note. Bonn, Germany: German National Research Center for Information Technology GMD Technical Report\/ 148

2001
[37]

Jaeger, H. (2007). Echo state network. Scholarpedia\/ 2\/ (9), 2330

2007
[38]

Jeong, J., Y. Yan, S. Castruccio, and M. Genton (2019). A stochastic generator of global monthly wind energy with tukey g-and-h autoregressive processes. Statistica Sinica\/ 29 , 1105--1126

2019
[39]

Jolliffe, I. (2014). Principal Component Analysis . John Wiley & Sons, Ltd

2014
[40]

Kingma, D. P. and J. Ba (2014). Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980\/

2014 arXiv
[41]

Kramer, M. A. (1991). Nonlinear principal component analysis using autoassociative neural networks. AIChE Journal\/ 37\/ (2), 233--243

1991
[42]

Kramer, M. A. (1992). Autoassociative neural networks. Computers & Chemical Engineering\/ 16\/ (4), 313--328

1992
[43]

Rue, and J

Lindgren, F., H. Rue, and J. Lindström (2011). An explicit link between gaussian fields and gaussian markov random fields: the stochastic partial differential equation approach. Journal of the Royal Statistical Society: Series B (Statistical Methodology)\/ 73\/ (4), 423--498

2011
[44]

Liu, D., D. Niu, H. Wang, and L. Fan (2014). Short-term wind speed forecasting using wavelet transform and support vector machines optimized by genetic algorithm. Renewable Energy\/ 62 , 592--597

2014
[45]

Lukosevicius, M. (2012). A practical guide to applying echo state networks. In Neural Networks: Tricks of the Trade , pp.\ 659--686. Springer

2012
[46]

Meier, D

Masci, J., U. Meier, D. Cire s an, and J. Schmidhuber (2011). Stacked convolutional auto-encoders for hierarchical feature extraction. In T. Honkela, W. Duch, M. Girolami, and S. Kaski (Eds.), Artificial Neural Networks and Machine Learning -- ICANN 2011 , Berlin, Heidelberg, ...

2011
[47]

Masters, D. and C. Luschi (2018). Revisiting small batch training for deep neural networks. arXiv:1804.07612\/

2018 arXiv
[48]

McDermott, P. L. and C. K. Wikle (2017). An ensemble quadratic echo state network for non-linear spatio-temporal forecasting. Stat\/ 6\/ (1), 315--330

2017
[49]

McDermott, P. L. and C. K. Wikle (2019a). B ayesian recurrent neural network models for forecasting and quantifying uncertainty in spatial-temporal data. Entropy\/ 21\/ (2), 184

2019
[50]

McDermott, P. L. and C. K. Wikle (2019b). Deep echo state networks with uncertainty quantification for spatio‐temporal forecasting. Environmetrics\/ 30\/ (3), e2553

2019
[51]

Mohseni-Cheraghlou, A. (2021). Fossil fuel subsidies and renewable energies in mena: An oxymoron? https://www.mei.edu/publications/fossil-fuel-subsidies-and-renewable-energies-mena-oxymoron

2021
[52]

Moncada Morales, L. N., M. Bonas, S. Castruccio, and P. Crippa (2024). Forecasting High Resolution Precipitation Events with Logistic Echo State Networks . Journal of Geophysical Research: Machine Learning and Computation\/ 1 , e2024JH000291

2024
[53]

Nurunnabi, M. (2017). Transformation from an oil-based economy to a knowledge-based economy in saudi arabia: the direction of saudi vision 2030. Journal of the Knowledge Economy\/ 8\/ (2), 536--64

2017
[54]

Peterson, E. W. and J. P. Hennessey (1978). On the use of power laws for estimates of wind power potential. Journal of Applied Meteorology (1962-1982)\/ 17\/ (3), 390--394

1978
[55]

Renewables 2024 - global status report

REN21 Secretariat (2024). Renewables 2024 - global status report. Paris, France

2024
[56]

Skamarock, W. C. and J. B. Klemp (2008). A time-split nonhydrostatic atmospheric model for weather research and forecasting applications. Journal of Computational Physics\/ 227 , 3465--3485

2008
[57]

Castruccio, P

Tagle, F., S. Castruccio, P. Crippa, and M. G. Genton (2019). A non-gaussian spatio-temporal model for daily wind speeds based on a multi-variate skew-t distribution. Journal of Time Series Analysis\/ 40\/ (3), 312--326

2019
[58]

Tagle, F., M. G. Genton, A. Yip, S. Mostamandi, G. Stenchikov, and S. Castruccio (2020). A high-resolution bilevel skew-t stochastic generator for assessing saudi arabia's wind energy resources (with discussion). Environmetrics\/ 31\/ (7), e2628

2020
[59]

Larochelle, I

Vincent, P., H. Larochelle, I. Lajoie, Y. Bengio, and P.-A. Manzagol (2010). Stacked denoising autoencoders: Learning useful representations in a deep network with a local denoising criterion. Journal of Machine Learning Research\/ 11\/ (110), 3371--3408

2010
[60]

Huang, Y

Wang, W., Y. Huang, Y. Wang, and L. Wang (2014). Generalized autoencoder: A neural network framework for dimensionality reduction. In Proceedings of the IEEE conference on computer vision and pattern recognition workshops , pp.\ 490--497

2014
[61]

Yao, and S

Wang, Y., H. Yao, and S. Zhao (2016). Auto-encoder based dimensionality reduction. Neurocomputing\/ 184 , 232--242

2016
[62]

Wang, Z., S. H. Holan, and C. K. Wikle (2024). Echo State Networks for Spatio-Temporal Area-Level Data . arXiv:2410.10641\/

2024
[63]

Zammit-Mangion, and N

Wikle, C., A. Zammit-Mangion, and N. Cressie (2019). Spatio-temporal Statistics with R . Boca Raton, FL: CRC press, Taylor & Francis Group

2019
[64]

Wikle, C. K. and M. B. Hooten (2006). Hierarchical Bayesian spatio–temporal models for population spread . In Hierarchical Modelling for the Environmental Sciences: Statistical Methods and Applications . Oxford University Press

2006
[65]

Yoo, M. and C. K. Wikle (2023). Using echo state networks to inform physical models for fire front propagation. Spatial Statistics\/ 54 , 100732

2023
[66]

Zhang, G. (2003). Time series forecasting using a hybrid arima and neural network model. Neurocomputing\/ 50 , 159--175

2003
[67]

Crippa, M

Zhang, J., P. Crippa, M. G. Genton, and S. Castruccio (2021). Assessing the reliability of wind power operations under a changing climate with a non-Gaussian bias correction . The Annals of Applied Statistics\/ 15\/ (4), 1831 -- 1849

2021
[68]

Crippa, M

Zhang, J., P. Crippa, M. G. Genton, and S. Castruccio (2024). Sensitivity analysis of wind energy resources with bayesian non-gaussian and nonstationary functional anova. The Annals of Applied Statistics\/ 18\/ (1), 23--41

2024

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.

[1] [1]

write newline

" write newline "" before.all 'output.state := FUNCTION format.url url empty "" url if FUNCTION article output.bibitem format.authors "author" output.check author format.key output output.year.check new.block format.title "title" output.check new.block crossref missing format.jour.vol output format.article.crossref output.nonnull format.pages output if ne...

[2] [2]

Li, and H

Ai, X., S. Li, and H. Xu (2023). Wind speed prediction model using ensemble empirical mode decomposition, least squares support vector machine and long short-term memory. Frontiers in Energy Research\/ 10

work page 2023

[3] [3]

Bateman, H. (1915). Some recent researches on the motion of fluids. Monthly Weather Review\/ 43\/ (4), 163--170

work page 1915

[4] [4]

Bonas, M. and S. Castruccio (2023). Calibration of spatial forecasts from citizen science urban air pollution data with sparse recurrent neural networks. Annals of Applied Statistics\/ 17\/ (3), 1820--1840

work page 2023

[5] [5]

Datta, C

Bonas, M., A. Datta, C. K. Wikle, E. L. Boone, F. S. Alamri, B. V. Hari, I. Kavila, S. J. Simmons, S. M. Jarvis, W. S. Burr, D. E. Pagendam, W. Chang, and S. Castruccio (2024). Assessing predictability of environmental time series with statistical and machine learning models. Environmetrics\/ , e2864

work page 2024

[6] [6]

Bonas, M., D. H. Richter, and S. Castruccio (2024). A physics-informed, deep double reservoir network for forecasting boundary layer velocity. Journal of the American Statistical Association - Applications and Case Studies\/ . in press

work page 2024

[7] [7]

Bonas, M., C. K. Wikle, and S. Castruccio (2024). Calibrated forecasts of quasi-periodic climate processes with deep echo state networks and penalized quantile regression. Environmetrics\/ 35\/ (1), e2833

work page 2024

[8] [8]

Bottou, L. (2010). Large-scale machine learning with stochastic gradient descent. In Y. Lechevallier and G. Saporta (Eds.), Proceedings of COMPSTAT'2010 , pp.\ 177--186. Physica-Verlag HD

work page 2010