pith. sign in

arxiv: 2606.24076 · v1 · pith:GKS5Y23Lnew · submitted 2026-06-23 · 📊 stat.ME

A Non-Stationary Spatio-Temporal Covariance Model with Dynamic Advection Effects for Rainfall Data

Pedro Nasevicius Ramos , Guilherme Ludwig This is my paper

Pith reviewed 2026-06-25 23:13 UTC · model grok-4.3

classification 📊 stat.ME
keywords spatio-temporal covarianceadvection effectsmixture modelnon-stationaryrainfall datawind directionBayesian estimationMCMC
0
0 comments X

The pith

A mixture of advection covariance models allows rainfall data to exhibit wind direction changes over estimated time intervals.

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

The paper proposes a non-stationary spatio-temporal covariance model formed as a mixture of components, each with advection effects that increase covariance along a chosen orientation vector to represent wind-driven cloud movement. Standard single-component advection models assume one fixed direction for the whole period, but the mixture lets the effective direction switch by assigning different components to different time intervals. The authors build a Markov chain Monte Carlo algorithm to carry out Bayesian estimation of all parameters, including the interval boundaries. They apply the procedure to observations from a severe rainfall event in southeastern Brazil. This construction directly targets the rigidity of traditional advection models when wind patterns evolve during an event.

Core claim

We propose a non-stationary model constructed using a mixture of spatio-temporal covariance models with advection effects; namely, models that have larger covariance values along an orientation vector in the spatio-temporal index set, that simulate wind direction and cloud movement. We show that a mixture of such models can allow for wind direction change in data during (estimated) time intervals, unlike traditional models that use rigid advection effects. We construct a MCMC procedure for Bayesian estimation, and illustrate the method with the analysis of a severe rainfall event from the southeastern region of Brazil.

What carries the argument

Mixture of advection covariance models, each component carrying its own orientation vector that produces larger covariance along a direction simulating wind, with the mixture allowing the active direction to change across estimated time intervals.

If this is right

  • The model can represent non-stationary rainfall covariance by letting advection direction adapt at data-driven time points.
  • Bayesian MCMC supplies joint inference on covariance parameters, orientation vectors, and the timing of direction shifts.
  • Application to real Brazilian rainfall data illustrates recovery of dynamic wind effects during a severe event.
  • The mixture removes the single fixed-direction assumption that limits traditional advection-based spatio-temporal models.

Where Pith is reading between the lines

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

  • The same mixture structure could apply to other directional transport processes such as pollutant plumes or ocean surface currents that exhibit regime shifts.
  • If interval boundaries prove stable, the fitted components could support regime-aware short-term prediction of rainfall movement.
  • Physical wind observations could be added as covariates to anchor the estimated orientation vectors and reduce label-switching among components.

Load-bearing premise

The rainfall process exhibits distinct intervals of relatively constant advection direction that can be recovered by the mixture components without severe identifiability problems or overfitting.

What would settle it

Rainfall records in which wind direction changes continuously without clear stationary intervals, and where the mixture model shows no improvement in fit over a single rigid advection model or produces unstable component assignments.

Figures

Figures reproduced from arXiv: 2606.24076 by Guilherme Ludwig, Pedro Nasevicius Ramos.

Figure 1
Figure 1. Figure 1: Trace plots across 10 independent MCMC simulations (columns) versus posterior samples of the parameters [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: South American air masses during summer (left) and winter (middle); region of interest (right). [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Hourly aggregated data from rainfall stations, where log-rainfall intensity (color) and wind direction (segments, [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Traceplot of the posterior distribution of [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Posterior distribution of λ0 | others, λ1 | others and λ2 | others (spatial variances, top row), θ1 | others, θ2 | others (spatial dependence, mid row) as well as V1 | others and V2 | others (advection direction; spatial histograms in R 2 , bottom row). The posterior mode of τ | others is 13:00. Salvaña, M. L. O., Lenzi, A., and Genton, M. G. (2023). Spatio-temporal cross-covariance functions under the lag… view at source ↗
Figure 6
Figure 6. Figure 6: Predicted maps with posterior average of [PITH_FULL_IMAGE:figures/full_fig_p010_6.png] view at source ↗
read the original abstract

We propose a non-stationary model constructed using a mixture of spatio-temporal covariance models with advection effects; namely, models that have larger covariance values along an orientation vector in the spatio-temporal index set, that simulate wind direction and cloud movement. We show that a mixture of such models can allow for wind direction change in data during (estimated) time intervals, unlike traditional models that use rigid advection effects. We construct a MCMC procedure for Bayesian estimation, and illustrate the method with the analysis of a severe rainfall event from the southeastern region of Brazil.

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.

Referee Report

1 major / 0 minor

Summary. The paper proposes a non-stationary spatio-temporal covariance model for rainfall data constructed as a mixture of covariance models that incorporate advection effects along orientation vectors. The central claim is that the mixture permits wind direction changes during estimated time intervals, in contrast to traditional models with rigid advection. The authors outline an MCMC procedure for Bayesian estimation and illustrate the approach on a severe rainfall event from southeastern Brazil.

Significance. If the mixture components can be shown to recover distinct advection intervals reliably, the model would offer a useful extension for handling non-stationarity in environmental spatio-temporal processes, particularly where wind or flow directions shift over time. The Bayesian MCMC framework provides a natural route to uncertainty quantification, which is a positive feature for applied work in rainfall modeling.

major comments (1)
  1. [Abstract] Abstract: the central claim that mixture components recover time intervals of constant advection direction without severe identifiability problems is load-bearing for the paper's contribution, yet the abstract (and by extension the described construction) provides no indication of ordering constraints, time-localized component probabilities, or simulation-based checks against label switching and posterior multimodality. Standard finite mixtures on covariance parameters are known to be vulnerable to these issues, and a single-event illustration cannot rule out overfitting artifacts.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the detailed report and the opportunity to respond. We address the major comment on the abstract and identifiability below.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim that mixture components recover time intervals of constant advection direction without severe identifiability problems is load-bearing for the paper's contribution, yet the abstract (and by extension the described construction) provides no indication of ordering constraints, time-localized component probabilities, or simulation-based checks against label switching and posterior multimodality. Standard finite mixtures on covariance parameters are known to be vulnerable to these issues, and a single-event illustration cannot rule out overfitting artifacts.

    Authors: We agree that the abstract should more clearly indicate the mechanisms used to address identifiability. The model construction employs time-localized component probabilities to permit changes in advection direction across estimated intervals and imposes ordering constraints on the orientation vectors to limit label switching. We will revise the abstract to include a concise reference to these features. We also acknowledge that the single real-data illustration leaves open the possibility of overfitting or multimodality artifacts; we will add a targeted simulation study in the revision to demonstrate recovery of distinct advection intervals under the MCMC procedure. revision: yes

Circularity Check

0 steps flagged

No circularity: model is a constructive proposal with standard estimation

full rationale

The paper proposes a mixture of advection-based spatio-temporal covariance models as a direct modeling choice to capture changing wind directions, then applies standard MCMC for Bayesian fitting and illustrates on rainfall data. No derivation step reduces to its own inputs by construction, no fitted parameters are relabeled as predictions, and no load-bearing claims rest on self-citations or imported uniqueness theorems. The mixture construction and its claimed flexibility are presented as an ansatz whose validity is assessed via data illustration rather than tautological equivalence. This is the normal case of a self-contained modeling paper.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 0 invented entities

Abstract-only; specific free parameters and assumptions cannot be fully enumerated without the manuscript. The model implicitly relies on standard Gaussian process assumptions for covariance modeling.

free parameters (2)
  • mixture component count
    Number of components in the mixture is a modeling choice likely selected or estimated but unspecified in abstract.
  • advection orientation vectors
    Direction vectors per component are parameters to be fit from data.
axioms (1)
  • domain assumption The underlying process admits a Gaussian process representation with the proposed mixture covariance.
    Standard assumption invoked for spatio-temporal covariance modeling.

pith-pipeline@v0.9.1-grok · 5615 in / 1143 out tokens · 31314 ms · 2026-06-25T23:13:55.149864+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

70 extracted references · 1 canonical work pages

  1. [1]

    001/2011/SEGER/LAIME/CSC/INMET: Rede de EstaC'C5es MeteorolC3gicas AutomC!ticas do INMET , author =

    NOTA TC CNICA No. 001/2011/SEGER/LAIME/CSC/INMET: Rede de EstaC'C5es MeteorolC3gicas AutomC!ticas do INMET , author =. 2011 , url=

    2011

  2. [2]

    Statistics for

    Cressie, Noel and Wikle, Christopher K , publisher =. Statistics for

  3. [3]

    Static and roving sensor data fusion for spatio-temporal hazard mapping with application to occupational exposure assessment , volume =

    Ludwig, Guilherme and Chu, Tingjin and Zhu, Jun and Wang, Haonan and Koehler, Kirsten , journal =. Static and roving sensor data fusion for spatio-temporal hazard mapping with application to occupational exposure assessment , volume =

  4. [4]

    Journal of the American Statistical Association , volume=

    Nonseparable, stationary covariance functions for space--time data , author=. Journal of the American Statistical Association , volume=. 2002 , publisher=

    2002

  5. [5]

    Journal of the Royal Statistical Society: Series C (Applied Statistics) , volume=

    Geostatistical inference under preferential sampling , author=. Journal of the Royal Statistical Society: Series C (Applied Statistics) , volume=. 2010 , publisher=

    2010

  6. [6]

    A statistical approach to some basic mine valuation problems on the Witwatersrand , journal =

    Krige, Danie G , year =. A statistical approach to some basic mine valuation problems on the Witwatersrand , journal =

  7. [7]

    Economic Geology , volume=

    Principles of geostatistics , author=. Economic Geology , volume=. 1963 , publisher=

    1963

  8. [8]

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

    Spatial interaction and the statistical analysis of lattice systems , author=. Journal of the Royal Statistical Society: Series B (Methodological) , volume=. 1974 , publisher=

    1974

  9. [9]

    Advances in applied probability , volume=

    The intrinsic random functions and their applications , author=. Advances in applied probability , volume=. 1973 , publisher=

    1973

  10. [10]

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

    Some statistical methods connected with series of events , author=. Journal of the Royal Statistical Society: Series B (Methodological) , volume=. 1955 , publisher=

    1955

  11. [11]

    2007 , publisher=

    Model-based Geostatistics , author=. 2007 , publisher=

    2007

  12. [12]

    Journal of statistical software , volume=

    Spatstat: an R package for analyzing spatial point patterns , author=. Journal of statistical software , volume=

  13. [13]

    Statistics for

    Cressie, Noel , year=. Statistics for

  14. [14]

    1 , author=

    An Introduction to the Theory of Point Processes, vol. 1 , author=. 2003 , publisher=

    2003

  15. [15]

    2010 , publisher=

    Spatial Statistics and Modeling , author=. 2010 , publisher=

    2010

  16. [16]

    1999 , publisher=

    Interpolation of spatial data: some theory for kriging , author=. 1999 , publisher=

    1999

  17. [17]

    Journal of the American Statistical Association , volume=

    Classes of nonseparable, spatio-temporal stationary covariance functions , author=. Journal of the American Statistical Association , volume=. 1999 , publisher=

    1999

  18. [18]

    Journal of the American Statistical Association , volume=

    Nonparametric estimation of nonstationary spatial covariance structure , author=. Journal of the American Statistical Association , volume=. 1992 , publisher=

    1992

  19. [19]

    Mathematical Geology , volume=

    Nonseparable space-time covariance models: some parametric families , author=. Mathematical Geology , volume=. 2002 , publisher=

    2002

  20. [20]

    Environmetrics , volume=

    A general class of nonseparable space--time covariance models , author=. Environmetrics , volume=. 2011 , publisher=

    2011

  21. [21]

    arXiv preprint arXiv:2004.08724 , year=

    Modeling nonstationary and asymmetric multivariate spatial covariances via deformations , author=. arXiv preprint arXiv:2004.08724 , year=

    arXiv 2004

  22. [22]

    Journal of Computational and Graphical Statistics , volume=

    Covariance tapering for interpolation of large spatial datasets , author=. Journal of Computational and Graphical Statistics , volume=. 2006 , publisher=

    2006

  23. [23]

    Journal of Computational and Graphical Statistics , volume =

    Benjamin Shaby and David Ruppert , title =. Journal of Computational and Graphical Statistics , volume =. 2012 , publisher =

    2012

  24. [24]

    arXiv preprint arXiv:1710.05013 , volume=

    Methods for analyzing large spatial data: A review and comparison , author=. arXiv preprint arXiv:1710.05013 , volume=

    Pith/arXiv arXiv

  25. [25]

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

    Fixed rank kriging for very large spatial data sets , author=. Journal of the Royal Statistical Society: Series B (Statistical Methodology) , volume=. 2008 , publisher=

    2008

  26. [26]

    Journal of the American Statistical Association , volume=

    Hierarchical nearest-neighbor Gaussian process models for large geostatistical datasets , author=. Journal of the American Statistical Association , volume=. 2016 , publisher=

    2016

  27. [27]

    1990 , publisher=

    Spline models for observational data , author=. 1990 , publisher=

    1990

  28. [28]

    Splines minimizing rotation-invariant semi-norms in Sobolev spaces

    Duchon, Jean. Splines minimizing rotation-invariant semi-norms in Sobolev spaces. Constructive Theory of Functions of Several Variables. 1977

    1977

  29. [29]

    2012 , volume =

    Edzer Pebesma , journal =. 2012 , volume =

    2012

  30. [30]

    2006 , author =

    A likelihood ratio test for separability of covariances , journal =. 2006 , author =

    2006

  31. [31]

    Environmetrics , volume=

    A general procedure for selecting a class of fully symmetric space-time covariance functions , author=. Environmetrics , volume=. 2016 , publisher=

    2016

  32. [32]

    Stochastic Environmental Research and Risk Assessment , volume=

    Testing the type of non-separability and some classes of space-time covariance function models , author=. Stochastic Environmental Research and Risk Assessment , volume=. 2018 , publisher=

    2018

  33. [33]

    Journal of Statistical Planning and Inference , volume=

    Testing for separability of spatial--temporal covariance functions , author=. Journal of Statistical Planning and Inference , volume=. 2006 , publisher=

    2006

  34. [34]

    Journal of Statistical Software , year =

    Claudia Cappello and Sandra. Journal of Statistical Software , year =

  35. [35]

    Zidek , title =

    Luke Bornn and Gavin Shaddick and James V. Zidek , title =. Journal of the American Statistical Association , volume =

  36. [36]

    Spatial Statistics , pages=

    Nonstationary cross-covariance functions for multivariate spatio-temporal random fields , author=. Spatial Statistics , pages=. 2020 , publisher=

    2020

  37. [37]

    Advances in Contemporary Statistics and Econometrics , year=

    Lagrangian Spatio-Temporal Nonstationary Covariance Functions , author=. Advances in Contemporary Statistics and Econometrics , year=

  38. [38]

    Gneiting, Tilmann and Kleiber, William and Schlather, Martin , journal=. Mat. 2010 , publisher=

    2010

  39. [39]

    Statistical Science , pages=

    Cross-covariance functions for multivariate geostatistics , author=. Statistical Science , pages=. 2015 , publisher=

    2015

  40. [40]

    Statistical Methods & Applications , volume=

    Covariance tapering for multivariate Gaussian random fields estimation , author=. Statistical Methods & Applications , volume=. 2016 , publisher=

    2016

  41. [41]

    D. R. Cox and Valerie Isham , journal =. A Simple Spatial-Temporal Model of Rainfall , volume =

  42. [42]

    and Waymire, Ed , title =

    Gupta, Vijay K. and Waymire, Ed , title =. Journal of Geophysical Research: Atmospheres , volume =

  43. [43]

    Statistics & Probability Letters , volume=

    Space--time analysis using a general product--sum model , author=. Statistics & Probability Letters , volume=. 2001 , publisher=

    2001

  44. [44]

    Statistical methods for spatio-temporal systems , author=. B. Finkenstaedt, L. Held, and V. Isham (Eds.) , pages=

  45. [45]

    Proceedings of the Royal Society of London: Series A (Mathematical and Physical Sciences) , volume=

    The spectrum of turbulence , author=. Proceedings of the Royal Society of London: Series A (Mathematical and Physical Sciences) , volume=. 1938 , publisher=

    1938

  46. [46]

    Wiley Interdisciplinary Reviews: Computational Statistics , volume=

    30 Years of space--time covariance functions , author=. Wiley Interdisciplinary Reviews: Computational Statistics , volume=. 2021 , publisher=

    2021

  47. [47]

    2005 , publisher=

    Functional Data Analysis , author=. 2005 , publisher=

    2005

  48. [48]

    On semiparametric inference of geostatistical models via local

    Chu, Tingjin and Wang, Haonan and Zhu, Jun , journal=. On semiparametric inference of geostatistical models via local. 2014 , publisher=

    2014

  49. [49]

    Journal of Statistical Software , year =

    Dirk Eddelbuettel and Romain Fran. Journal of Statistical Software , year =

  50. [50]

    Fast and elegant numerical linear algebra using the

    Douglas Bates and Dirk Eddelbuettel , journal =. Fast and elegant numerical linear algebra using the. 2013 , volume =

    2013

  51. [51]

    2022 , note =

    Martin Schlather and Alexander Malinowski and Marco Oesting and Daphne Boecker and Kirstin Strokorb and Sebastian Engelke and Johannes Martini and Felix Ballani and Olga Moreva and Jonas Auel and Peter J Menck and Sebastian Gross and Ulrike Ober and Paulo Ribeiro and Brian D Ripley and Richard Singleton and Ben Pfaff and. 2022 , note =

    2022

  52. [52]

    Menck and Marco Oesting and Kirstin Strokorb , journal =

    Martin Schlather and Alexander Malinowski and Peter J. Menck and Marco Oesting and Kirstin Strokorb , journal =. Analysis, Simulation and Prediction of Multivariate Random Fields with Package. 2015 , volume =

    2015

  53. [53]

    2018 , journal =

    Edzer Pebesma , title =. 2018 , journal =. doi:10.32614/RJ-2018-009 , url =

    work page doi:10.32614/rj-2018-009 2018

  54. [54]

    2023 , publisher=

    Spatial Data Science: With Applications in R , author=. 2023 , publisher=

    2023

  55. [55]

    2019 , note =

    geosphere: Spherical Trigonometry , author =. 2019 , note =

    2019

  56. [56]

    Annual Review of Statistics and Its Application , volume=

    Stochastic Models of Rainfall , author=. Annual Review of Statistics and Its Application , volume=. 2024 , publisher=

    2024

  57. [57]

    2024 , note =

    geoR: Analysis of Geostatistical Data , author =. 2024 , note =

    2024

  58. [58]

    2024 , journal =

    Bayesian Analysis and Variable Selection for Spatial Count Data with an Application to Rio de Janeiro Gun Violence , author =. 2024 , journal =

    2024

  59. [59]

    Journal of the American Statistical Association , volume=

    Inconsistent estimation and asymptotically equal interpolations in model-based geostatistics , author=. Journal of the American Statistical Association , volume=. 2004 , publisher=

    2004

  60. [60]

    A general framework for

    Carrizo Vergara, Ricardo and Allard, Denis and Desassis, Nicolas , journal=. A general framework for. 2022 , publisher=

    2022

  61. [61]

    O’Hagan, Anthony , journal=. A

  62. [62]

    Monographs On Statistics and Applied Probability , volume=

    Geostatistical space-time models, stationarity, separability, and full symmetry , author=. Monographs On Statistics and Applied Probability , volume=. 2006 , publisher=

    2006

  63. [63]

    A flexible class of non-separable cross-covariance functions for multivariate space-time data

    Marc Bourotte and Denis Allard and Emilio Porcu. A flexible class of non-separable cross-covariance functions for multivariate space-time data. Spatial Statistics. 2016

    2016

  64. [64]

    Spatial Statistics , volume=

    Isotropy, symmetry, separability and strict positive definiteness for covariance functions: a critical review , author=. Spatial Statistics , volume=. 2019 , publisher=

    2019

  65. [65]

    Poveda, Germ. Testing. Advances in Water Resources , volume=. 2005 , publisher=

    2005

  66. [66]

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

    Approximate Bayesian inference for latent Gaussian models by using integrated nested Laplace approximations , author=. Journal of the Royal Statistical Society: Series B (Statistical Methodology) , volume=. 2009 , publisher=

    2009

  67. [67]

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

    Stationary nonseparable space-time covariance functions on networks , author=. Journal of the Royal Statistical Society: Series B (Statistical Methodology) , volume=. 2023 , publisher=

    2023

  68. [68]

    , author=

    Asymptotic equivalence of Bayes cross validation and widely applicable information criterion in singular learning theory. , author=. Journal of Machine Learning Research , volume=

  69. [69]

    Journal of the American Statistical Association , volume=

    Spatio-temporal cross-covariance functions under the Lagrangian framework with multiple advections , author=. Journal of the American Statistical Association , volume=. 2023 , publisher=

    2023

  70. [70]

    Statistics and Computing , volume=

    Understanding predictive information criteria for Bayesian models , author=. Statistics and Computing , volume=. 2014 , publisher=

    2014