arxiv: 2605.07225 · v1 · submitted 2026-05-08 · 📊 stat.AP

Recognition: 2 theorem links

· Lean Theorem

Spatiotemporal dynamics of wind-speed volatility

Ariane Nidelle Meli Chrisko , Philipp Otto

Authors on Pith no claims yet

Pith reviewed 2026-05-11 01:15 UTC · model grok-4.3

classification 📊 stat.AP

keywords wind speedvolatilityspatiotemporalGARCHspatial dependenceforecastingheight dependence

0 comments

The pith

Properly modeling spatial dependence in the mean is essential for reliable wind-speed volatility inference and forecasting.

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

This paper applies a spatiotemporal volatility model to daily wind-speed data from 141 stations in Northern Italy at two heights. It combines a spatial mean specification with GARCH-type volatility dynamics that incorporate both local past shocks and spatially weighted information from neighboring stations. The results demonstrate that accurate spatial modeling of the mean produces well-behaved residuals and supports trustworthy statistical inference. Forecast accuracy depends heavily on how the mean is specified, with simple distance-based models performing robustly once spatial effects are included in the mean. Volatility shows greater persistence at higher altitudes, and a multivariate version uncovers dependence between the two measurement heights.

Core claim

A parsimonious spatiotemporal volatility framework based on GARCH-type dynamics, in which conditional variance depends on past local shocks and spatially aggregated information from neighbouring stations using distance-based and directionally informed weight matrices, shows that properly modelling spatial dependence in the mean is essential for well-behaved residuals and reliable inference, while forecast performance is strongly driven by the mean specification.

What carries the argument

The spatiotemporal volatility model combining spatial mean specification with GARCH dynamics using distance-based weight matrices for aggregating neighbor information.

If this is right

Properly accounting for spatial dependence in the mean leads to well-behaved residuals and reliable inference.
Flexible mean structures perform better when residual spatial dependence remains, while parsimonious distance-based models yield robust out-of-sample forecasts once spatial interactions are captured.
Volatility persistence increases with height.
A multivariate extension reveals cross-height dependence.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Similar spatiotemporal GARCH models could be tested on wind data from other regions to check generalizability of the spatial mean importance.
Accounting for height-dependent persistence may help in designing better vertical wind profiles for atmospheric applications.
The cross-height dependence suggests potential for joint modeling of multi-level data to improve overall volatility predictions.

Load-bearing premise

The chosen distance-based and directionally informed weight matrices adequately capture the true spatial interactions in wind-speed volatility across the monitoring network.

What would settle it

If residuals from a model without spatial mean modeling show no spatial autocorrelation and yield better or equal out-of-sample forecasts, the claim that spatial mean modeling is essential would be challenged.

Figures

Figures reproduced from arXiv: 2605.07225 by Ariane Nidelle Meli Chrisko, Philipp Otto.

**Figure 2.** Figure 2: Raw wind speed, STL-deseasonalised series, and AR(1) residuals at 10 m and 100 m for [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗

**Figure 3.** Figure 3: Autocorrelation functions (ACF) and squared ACF of deseasonalised wind speed and [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

**Figure 4.** Figure 4: Station-wise GARCH(1,1) and EGARCH(1,1) coefficient estimates for wind speed at 10 m b [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗

**Figure 5.** Figure 5: Annual mean wind speed per station (ws10 vs ws100). [PITH_FULL_IMAGE:figures/full_fig_p012_5.png] view at source ↗

**Figure 6.** Figure 6: Spatial interaction networks implied by the alternative spatial weight matrices for the [PITH_FULL_IMAGE:figures/full_fig_p027_6.png] view at source ↗

read the original abstract

Wind-speed processes exhibit substantial temporal variability and spatial dependence, yet volatility dynamics across monitoring networks remain relatively unexplored. This study investigates the spatiotemporal behaviour of wind-speed volatility using daily observations from 141 stations in Northern Italy over 2016--2021, with measurements at 10 m and 100 m enabling the analysis of spatial and vertical dependence. We adopt a parsimonious spatiotemporal volatility framework based on GARCH-type dynamics, in which conditional variance depends on past local shocks and spatially aggregated information from neighbouring stations. The approach combines a spatial mean specification with structured volatility models using distance-based and directionally informed weight matrices. Results show that properly modelling spatial dependence in the mean is essential for well-behaved residuals and reliable inference. Forecast performance is strongly driven by the mean specification: flexible structures perform better when residual spatial dependence remains, while parsimonious distance-based models yield robust out-of-sample forecasts once spatial interactions are captured. Persistence increases with height, and a multivariate extension reveals cross-height dependence.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper applies a spatiotemporal GARCH setup to Northern Italy wind data and shows that mean-model spatial terms matter more for residuals and forecasts than the volatility layer itself, with persistence rising at higher altitudes.

read the letter

The core finding is that once you put a flexible spatial structure into the conditional mean for wind speeds, the residuals behave better and out-of-sample forecasts improve even with fairly simple distance-based volatility weights. The authors use daily data from 141 stations at 10 m and 100 m, build weight matrices that mix distance and direction, and report that persistence in volatility increases with height while a multivariate version picks up cross-height links. That vertical comparison and the explicit mean-versus-volatility split are the parts that feel like a modest but usable extension of existing GARCH work to this setting. The dataset size and the out-of-sample checks give the results some grounding that pure simulation papers often lack. The main soft spot is the pre-chosen weight matrices. If topography, elevation changes, or non-stationary wind patterns matter more than the distance-plus-direction rules capture, then the apparent gains from the mean specification could partly reflect leftover spatial misspecification rather than clean separation of mean and volatility effects. The abstract is also light on exact fitting details, residual plots, and robustness to alternative weights, so a referee would want to see those diagnostics expanded. This is applied spatial statistics aimed at wind-energy people who already use GARCH-type models and want to add a spatial layer without going full multivariate. It is not reshaping theory, but the empirical patterns are clear enough to be worth checking. I would send it to peer review; the data and the mean-volatility comparison are solid enough to justify referee time even if the weight-matrix choice needs more defense.

Referee Report

3 major / 2 minor

Summary. The paper develops a spatiotemporal GARCH-type volatility model for daily wind-speed observations from 141 Northern Italy stations at 10 m and 100 m heights (2016–2021). It combines flexible or parsimonious spatial mean specifications with volatility dynamics driven by local shocks and spatially aggregated neighbor information via distance-based and directionally informed weight matrices. Central claims are that proper spatial modeling in the mean is required for well-behaved residuals and reliable inference, that forecast performance is driven by the mean specification, that volatility persistence increases with height, and that a multivariate extension reveals cross-height dependence.

Significance. If the empirical separation of mean and volatility effects holds, the work offers a practical framework for spatiotemporal volatility modeling with direct relevance to wind-energy forecasting and risk assessment. The use of real multi-height observational data and out-of-sample forecast comparisons is a positive empirical contribution, though the study relies on pre-specified weight matrices rather than parameter-free derivations or machine-checked proofs.

major comments (3)

[Section 3] Section 3 (model specification): The central claim that modeling spatial dependence in the mean produces well-behaved residuals and reliable inference rests on the pre-specified distance-based and directionally informed weight matrices adequately representing the true spatial interactions. No sensitivity analysis or comparison against topography- or elevation-informed alternatives is reported; if these matrices omit key features, apparent gains from flexible mean models may reflect residual spatial misspecification rather than genuine mean-volatility separation.
[Section 4] Section 4 (estimation and diagnostics): The abstract and methods provide no explicit details on the fitting procedure for GARCH parameters, convergence criteria, or residual diagnostics (e.g., spatial autocorrelation tests on standardized residuals). Without these, the assertion of 'well-behaved residuals' after spatial mean modeling cannot be independently verified and undermines the reliability of the reported persistence and forecast results.
[Section 5] Section 5 (forecast results): The statement that 'parsimonious distance-based models yield robust out-of-sample forecasts once spatial interactions are captured' requires quantitative comparison (e.g., specific RMSE or CRPS values across mean specifications in a table). The current qualitative summary does not allow assessment of effect sizes or whether the improvement is statistically significant after accounting for multiple testing.

minor comments (2)

[Abstract] Abstract: The phrase 'structured volatility models' is undefined; a brief parenthetical reference to the specific GARCH lag structure or weight-matrix form would improve clarity.
[Section 2] Notation: The distinction between local shocks and spatially aggregated shocks should be denoted consistently with subscripts or superscripts throughout the equations to avoid ambiguity when reading the multivariate extension.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their constructive and detailed comments, which have helped us identify areas where the manuscript can be strengthened for clarity, reproducibility, and robustness. We address each major comment point by point below and indicate the revisions planned for the next version of the manuscript.

read point-by-point responses

Referee: [Section 3] Section 3 (model specification): The central claim that modeling spatial dependence in the mean produces well-behaved residuals and reliable inference rests on the pre-specified distance-based and directionally informed weight matrices adequately representing the true spatial interactions. No sensitivity analysis or comparison against topography- or elevation-informed alternatives is reported; if these matrices omit key features, apparent gains from flexible mean models may reflect residual spatial misspecification rather than genuine mean-volatility separation.

Authors: We acknowledge the importance of verifying the robustness of the chosen weight matrices. Our distance-based and directionally informed specifications follow standard approaches in the spatiotemporal literature for wind-speed data, where geographic distance and prevailing wind directions capture key dependencies in Northern Italy. To directly address the concern that apparent mean-volatility separation might stem from unmodeled topography effects, we will add a sensitivity analysis in the revised Section 3. This will compare results using the original matrices against alternatives that incorporate elevation differences and topographic features derived from the station metadata, thereby providing quantitative evidence on whether the core findings hold under alternative spatial representations. revision: yes
Referee: [Section 4] Section 4 (estimation and diagnostics): The abstract and methods provide no explicit details on the fitting procedure for GARCH parameters, convergence criteria, or residual diagnostics (e.g., spatial autocorrelation tests on standardized residuals). Without these, the assertion of 'well-behaved residuals' after spatial mean modeling cannot be independently verified and undermines the reliability of the reported persistence and forecast results.

Authors: We agree that greater transparency in the estimation and diagnostic procedures is essential for reproducibility. In the revised manuscript, we will expand Section 4 to include a dedicated subsection detailing the maximum-likelihood estimation of the GARCH parameters (using a quasi-Newton optimizer), convergence criteria (e.g., gradient norm tolerance of 1e-6 and maximum iterations), and a full suite of residual diagnostics. These will encompass Ljung-Box tests for serial correlation, Engle's ARCH test for remaining heteroskedasticity, and Moran's I statistic applied to the standardized residuals to formally confirm the absence of spatial autocorrelation after the spatial mean modeling step. revision: yes
Referee: [Section 5] Section 5 (forecast results): The statement that 'parsimonious distance-based models yield robust out-of-sample forecasts once spatial interactions are captured' requires quantitative comparison (e.g., specific RMSE or CRPS values across mean specifications in a table). The current qualitative summary does not allow assessment of effect sizes or whether the improvement is statistically significant after accounting for multiple testing.

Authors: We thank the referee for this suggestion to improve the quantitative presentation of the forecast results. In the revised Section 5, we will replace the qualitative summary with a new table that reports specific out-of-sample RMSE and CRPS values for each combination of mean specification and volatility model. We will also include Diebold-Mariano tests for pairwise forecast accuracy comparisons, with p-values adjusted via the Bonferroni correction to account for multiple testing. This will allow readers to evaluate both the magnitude of improvements and their statistical significance. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical model fitted to external observational data

full rationale

The paper is a standard empirical spatiotemporal analysis that fits GARCH-type volatility models (with spatial mean specifications and pre-specified distance/directional weight matrices) to independent wind-speed observations from 141 Northern Italy stations. All reported results—residual diagnostics, forecast comparisons, persistence by height, and cross-height dependence—derive from out-of-sample performance against held-out data rather than from any equation or parameter that is defined in terms of the fitted outputs themselves. No self-citation chain, ansatz smuggling, or fitted-input-renamed-as-prediction is present; the weight-matrix choice is an explicit modeling assumption, not a self-referential definition.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The paper relies on standard GARCH assumptions and spatial statistics without introducing new entities or deriving results from first principles.

free parameters (1)

GARCH parameters and spatial weight matrices
Conditional variance coefficients and distance/direction-based weights are fitted or selected to match the Italian station data.

axioms (1)

domain assumption Wind-speed volatility follows GARCH-type dynamics where conditional variance depends on past local shocks and spatially aggregated neighbor information
This is the core modeling framework stated in the abstract.

pith-pipeline@v0.9.0 · 5462 in / 1364 out tokens · 51664 ms · 2026-05-11T01:15:26.845359+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

STARMAGARCH(1,1,1,1) with distance-band, k-NN and directional weight matrices; SDPD mean equation; persistence β increasing with height
IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

Moran’s I on residuals and squared residuals; forecasting RMSFE/MAFE on log-variance

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

57 extracted references · 57 canonical work pages

[1]

Econometrica , volume=

Autoregressive conditional heteroscedasticity with estimates of the variance of United Kingdom inflation , author=. Econometrica , volume=

work page
[2]

Environmental Modelling & Software , volume=

A bivariate GARCH model for Sydney Harbour wind components , author=. Environmental Modelling & Software , volume=

work page
[3]

Journal of Econometrics , volume=

Generalized autoregressive conditional heteroskedasticity , author=. Journal of Econometrics , volume=

work page
[4]

Econometrica , volume=

Conditional heteroskedasticity in asset returns: A new approach , author=. Econometrica , volume=

work page
[5]

Energy Economics , volume=

Wind power volatility forecasting using GARCH-type models , author=. Energy Economics , volume=

work page
[6]

Applied Energy , volume=

Asymmetric GARCH-type models for volatility characteristics analysis and wind power forecasting , author=. Applied Energy , volume=

work page
[7]

Journal of Finance , volume=

On the relation between the expected value and the volatility of the nominal excess return on stocks , author=. Journal of Finance , volume=

work page
[8]

Journal of Empirical Finance , volume=

A long memory property of stock market returns and a new model , author=. Journal of Empirical Finance , volume=

work page
[9]

2010 , publisher=

Analysis of Financial Time Series , author=. 2010 , publisher=

work page 2010
[10]

2010 , publisher=

GARCH Models: Structure, Statistical Inference and Financial Applications , author=. 2010 , publisher=

work page 2010
[11]

Energy Conversion and Management , volume=

Empirical investigation on using wind speed volatility to estimate the operation probability and power output of wind turbines , author=. Energy Conversion and Management , volume=. 2013 , publisher=

work page 2013
[12]

International journal of green energy , volume=

Short-term wind speed forecasting: Application of linear and non-linear time series models , author=. International journal of green energy , volume=. 2016 , publisher=

work page 2016
[13]

Comprehensive evaluation of

Liu, Heping and Erdem, Erdal and Shi, Jing , journal =. Comprehensive evaluation of. 2011 , publisher =

work page 2011
[14]

Applied Energy , volume=

Forecasting volatility of wind power production , author=. Applied Energy , volume=. 2016 , publisher=

work page 2016
[15]

Asymmetric

Chen, Hao and Zhang, Jianzhong and Tao, Yubo and Tan, Fenglei , journal=. Asymmetric. 2019 , publisher=

work page 2019
[16]

Journal of Time Series Analysis , volume=

Parametric and semiparametric models for volatility in meteorological time series , author=. Journal of Time Series Analysis , volume=

work page
[17]

Journal of Time Series Analysis , author =

A distance measure for classifying. Journal of Time Series Analysis , author =. 1990 , publisher=. doi:10.1111/j.1467-9892.1990.tb00048.x , abstract =

work page doi:10.1111/j.1467-9892.1990.tb00048.x 1990
[18]

2025 , eprint=

Exponential Spatiotemporal GARCH Model with Asymmetric Volatility Spillovers , author=. 2025 , eprint=

work page 2025
[19]

Applied Energy , volume=

ARMA-based approaches for forecasting the tuple of wind speed and direction , author=. Applied Energy , volume=

work page
[20]

Journal of the Royal Statistical Society: Series C , volume=

Using conditional heteroskedasticity models for density forecasting of wind speed and direction , author=. Journal of the Royal Statistical Society: Series C , volume=

work page
[21]

and Lunde, Asger , journal=

Hansen, Peter R. and Lunde, Asger , journal=. A forecast comparison of volatility models: Does anything beat a

work page
[22]

Statistics for Spatial Data , author =

work page
[23]

Journal of the American Statistical Association , volume =

Geostatistical Space–Time Models, Stationarity, Separability and Full Covariance Structures , author =. Journal of the American Statistical Association , volume =

work page
[24]

Journal of Applied Energy , volume =

A hybrid wavelet transform and neural network approach for short-term wind speed forecasting , author =. Journal of Applied Energy , volume =. 2014 , note =

work page 2014
[25]

International Journal of Numerical Modelling: Electronic Networks, Devices and Fields , volume=

Wind speed forecasting based on wavelet transformation and recurrent neural network , author=. International Journal of Numerical Modelling: Electronic Networks, Devices and Fields , volume=. 2020 , publisher=

work page 2020
[26]

Journal of Econometrics , volume =

Long memory processes and fractional integration in econometrics , author =. Journal of Econometrics , volume =

work page
[27]

Journal of Official Statistics , volume=

STL: A seasonal-trend decomposition procedure based on loess , author=. Journal of Official Statistics , volume=. 1990 , publisher=

work page 1990
[28]

IEEE Transactions on Energy Conversion , volume =

Short-term prediction of wind farm power: A data mining approach , author =. IEEE Transactions on Energy Conversion , volume =. 2010 , publisher =

work page 2010
[29]

Applied Energy , volume =

Prediction of wind power ramp rates: A data-mining approach , author =. Applied Energy , volume =. 2009 , publisher =

work page 2009
[30]

Decision Making Under Uncertainty in Electricity Markets , author =

work page
[31]

Authorea Preprints , year=

Statistical modeling of the space-time relation between wind and significant wave height , author=. Authorea Preprints , year=

work page
[32]

European Journal of Operational Research , volume=

Forecasting wind speed with recurrent neural networks , author=. European Journal of Operational Research , volume=. 2012 , publisher=

work page 2012
[33]

Applied Energy , volume=

Forecasting wind power--modeling periodic and non-linear effects under conditional heteroscedasticity , author=. Applied Energy , volume=. 2016 , publisher=

work page 2016
[34]

IEEE Transactions on Energy conversion , volume=

Wind power density forecasting using ensemble predictions and time series models , author=. IEEE Transactions on Energy conversion , volume=. 2009 , publisher=

work page 2009
[35]

Long-term wind power and global warming prediction using

Y. Long-term wind power and global warming prediction using. Journal of Industrial and Management Optimization , volume=. 2024 , publisher=

work page 2024
[36]

2010 IEEE 11th International Conference on Probabilistic Methods Applied to Power Systems , pages=

Advantages of ARMA-GARCH wind speed time series modeling , author=. 2010 IEEE 11th International Conference on Probabilistic Methods Applied to Power Systems , pages=. 2010 , organization=

work page 2010
[37]

Environmetrics , volume=

Time series analysis of wind speed with time-varying turbulence , author=. Environmetrics , volume=. 2006 , publisher=

work page 2006
[38]

arXiv preprint arXiv:2603.02195 , year=

Comparative Analysis of Spatiotemporal Volatility Models: An Empirical Study on Financial Network Series , author=. arXiv preprint arXiv:2603.02195 , year=

work page arXiv
[39]

Geographical Analysis , volume=

Estimation of Anisotropic, Time-Varying Spatial Spillovers of Fine Particulate Matter Due to Wind Direction , author=. Geographical Analysis , volume=. 2020 , publisher=

work page 2020
[40]

Spatial Statistics , volume=

A multivariate spatial and spatiotemporal ARCH model , author=. Spatial Statistics , volume=. 2024 , publisher=

work page 2024
[41]

Journal of the Royal Statistical Society: Series B (Statistical Methodology) , volume=

Econometric analysis of realized volatility and its use in estimating stochastic volatility models , author=. Journal of the Royal Statistical Society: Series B (Statistical Methodology) , volume=. 2002 , publisher=

work page 2002
[42]

2011 , publisher=

Asset price dynamics, volatility, and prediction , author=. 2011 , publisher=

work page 2011
[43]

arXiv preprint arXiv:2507.14389 , year=

Regional compositional trajectories and structural change: A spatiotemporal multivariate autoregressive framework , author=. arXiv preprint arXiv:2507.14389 , year=

work page arXiv
[44]

Basic laws of turbulent mixing in the surface layer of the atmosphere , author=. Contrib. Geophys. Inst. Acad. Sci. USSR , volume=

work page
[45]

2012 , publisher=

An introduction to boundary layer meteorology , author=. 2012 , publisher=

work page 2012
[46]

2012 , publisher=

Lee, Lung-fei and Yu, Jihai , journal=. 2012 , publisher=

work page 2012
[47]

Scientific Data , volume=

Agrimonia: a dataset on livestock, meteorology and air quality in the Lombardy region, Italy , author=. Scientific Data , volume=. 2023 , publisher=

work page 2023
[48]

Energy conversion and management , volume=

A new hybrid model optimized by an intelligent optimization algorithm for wind speed forecasting , author=. Energy conversion and management , volume=. 2014 , publisher=

work page 2014
[49]

Energy Science & Engineering , year=

Economic and Technical Assessment of Wind Potential Using SARIMAX Time Series Models: Wind Speed Forecasting and Analysis , author=. Energy Science & Engineering , year=

work page
[50]

Heliyon , volume=

A two-stage deep learning-based hybrid model for daily wind speed forecasting , author=. Heliyon , volume=. 2025 , publisher=

work page 2025
[51]

ISOPE International Ocean and Polar Engineering Conference , pages=

Spatio-Temporal Modelling of Wind Speed Variations , author=. ISOPE International Ocean and Polar Engineering Conference , pages=. 2018 , organization=

work page 2018
[52]

Theoretical and Applied Climatology , volume=

Spatio-temporal modelling of wind speed variations and extremes in the Caribbean and the Gulf of Mexico , author=. Theoretical and Applied Climatology , volume=. 2019 , publisher=

work page 2019
[53]

Non-homogeneous hidden

Ailliot, Pierre and Bessac, Julie and Monbet, Val. Non-homogeneous hidden. Journal of Statistical Planning and Inference , volume=. 2015 , publisher=

work page 2015
[54]

1988 , publisher=

Spatial econometrics: methods and models , author=. 1988 , publisher=

work page 1988
[55]

Journal of Time Series Analysis , volume=

A Stationary Spatio-Temporal GARCH Model , author=. Journal of Time Series Analysis , volume=. 2020 , publisher=

work page 2020
[56]

Biometrika , volume=

A dimension-reduced approach to space-time Kalman filtering , author=. Biometrika , volume=. 1999 , publisher=

work page 1999
[57]

2008 , publisher=

Applied spatial data analysis with R , author=. 2008 , publisher=

work page 2008