arxiv: 2605.03152 · v1 · submitted 2026-05-04 · 📊 stat.ME

Recognition: unknown

Scalable generative modeling of non-Gaussian spatio-temporal fields via autoregressive Gaussian processes

Carrie J. Lei-Cramer , Jian Cao , Matthias Katzfuss

Authors on Pith no claims yet

Pith reviewed 2026-05-08 17:40 UTC · model grok-4.3

classification 📊 stat.ME

keywords generative modelingspatio-temporal fieldsGaussian processesautoregressive modelstransport mapsnon-Gaussian distributionsclimate modelingscalable methods

0 comments

The pith

An autoregressive construction with Gaussian process conditionals models non-Gaussian spatio-temporal fields scalably.

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

The paper sets out to build generative models for spatio-temporal fields that show complex nonstationary and non-Gaussian behavior due to varying and nonlinear dependencies. It does so by writing the joint density as a product of univariate conditional distributions, each represented by a Gaussian process inside an autoregressive transport-map construction. The Gaussian process prior regularizes the fit, allowing use with small training sets, while data-dependent sparsity in conditioning sets supports scaling to very large fields. A variant enables forward sampling or prediction in time from incomplete trajectories. The method is tested successfully on non-Gaussian climate data containing tens of millions of points.

Core claim

The authors establish that representing the joint density of a spatio-temporal field as a product of univariate conditionals, each modeled by a Gaussian process in an autoregressive transport-map construction, allows accurate generative modeling of non-Gaussian and nonstationary fields. This setup includes regularization from the prior for small samples and uses data-dependent sparsity to ensure scalability to high dimensions. Accuracy is demonstrated on climate-model output with tens of millions of data points, along with a forward-in-time variant.

What carries the argument

The autoregressive transport-map construction with Gaussian process conditionals and data-dependent sparsity in the conditioning sets.

If this is right

Generative modeling becomes feasible for applications such as stochastic weather generators and climate-model surrogates.
The prior regularization makes the method suitable even when only a small number of training samples are available.
Data-dependent sparsity allows handling of high-dimensional distributions with tens of millions of points.
The time-forward variant supports sampling or prediction from incomplete space-time trajectories.
Empirical results confirm accuracy for non-Gaussian climate-model fields at large scales.

Where Pith is reading between the lines

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

This approach could apply to other domains with high-dimensional non-Gaussian spatio-temporal data, such as ocean modeling or air quality monitoring.
Full generative capabilities may improve uncertainty quantification in downstream tasks like risk assessment for climate events.
Integrating physical laws as constraints within the conditionals could produce more realistic and consistent samples.
Controlled experiments on synthetic data with known dependence structures would help quantify how well the sparsity preserves key interactions.

Load-bearing premise

The autoregressive factorization and Gaussian process conditionals can accurately represent the nonstationary and non-Gaussian dependencies, and that the data-dependent sparsity does not omit important interactions.

What would settle it

Drawing many samples from the fitted model on the climate dataset and verifying whether their joint statistics, including non-Gaussian features like tails and nonlinear correlations, match those of the original data; significant mismatch would falsify the claim of accurate representation.

Figures

Figures reproduced from arXiv: 2605.03152 by Carrie J. Lei-Cramer, Jian Cao, Matthias Katzfuss.

**Figure 1.** Figure 1: Maximin (a) and time (b) ordering in a space–time domain. Rows show weaker (top) and stronger (bottom) temporal dependence. The i-th location is red, ordered points are blue, and unordered points are gray. Circles represent the conditioning radius for cm(i) (m = 3). Columns correspond to early (i = 9) and later (i = 39) ordering stages. 7 view at source ↗

**Figure 2.** Figure 2: Decay relationships under maximin ordering (left) and time ordering (right). Results are from a Gaussian field on a regular spatio-temporal grid (N = 103 ) with a Mat´ern covariance (ν = 0.3, ρ = 0.5). Top row: Conditional variance (d 2 i (◦) and e θd,1 ℓ θd,2 i (−)). Bottom row: Average squared regression coefficients (b 2 i,k, over i = 1, . . . , N for a fixed k). Note the periodic jumps in the time-orde… view at source ↗

**Figure 3.** Figure 3: Visualization of one ensemble member from the Central America regional precipitation dataset. Each panel view at source ↗

**Figure 4.** Figure 4: Log-score comparison between our ARGP framework (using global maximin ordering) and a parametric view at source ↗

**Figure 5.** Figure 5: Unconditional sample generation. Top row: true unobserved (test) global surface temperature fields. Middle view at source ↗

**Figure 6.** Figure 6: Conditional temporal forecasts given the first 10 time frames. Top row: true unobserved test data trajectory. view at source ↗

read the original abstract

Generative modeling of spatio-temporal fields is crucial for a variety of applications, including stochastic weather generators and climate-model surrogates. However, many such fields exhibit complex dependence structures that vary across space and time and are nonlinear, resulting in nonstationary and non-Gaussian joint distributions. Our approach represents the joint density of a spatio-temporal field as a product of univariate conditional distributions and models these conditionals using Gaussian processes within an autoregressive transport-map construction. This prior distribution provides regularization, making our method suitable for a small number of training samples. Data-dependent sparsity in the conditioning sets ensures scalability to high-dimensional distributions. We also propose a variant of the method designed to sample or predict forward in time from a given incomplete space-time trajectory. We demonstrate the accuracy and scalability of our approach on non-Gaussian climate-model output with tens of millions of data points.

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

A practical autoregressive GP transport map method for large non-Gaussian spatio-temporal data, with the sparsity choice as the main thing to verify.

read the letter

The paper's main contribution is a scalable autoregressive construction for generative modeling of non-Gaussian spatio-temporal fields. It represents the joint as a product of conditionals, each handled by a Gaussian process within a transport map, with data-dependent sparsity for high dimensions and a forward-in-time variant. This combination is new in its specifics, and the approach does well by incorporating a GP prior for regularization when training data is limited. The claim of handling tens of millions of points on climate data suggests practical scalability if the experiments back it up. The soft spot is the sparsity rule. Data-dependent selection of conditioning sets is essential for computation, but the abstract gives no indication of theoretical bounds or extensive ablations to confirm it captures all relevant non-local dependencies in nonstationary fields. This could lead to biased generative distributions, as the stress test points out. Without those details, it's hard to gauge how well the method preserves the structure. I would like to see more on how the sparsity thresholds are chosen and whether they introduce any systematic error in the generated fields. This work is for spatial statisticians and climate researchers building surrogates or stochastic generators. It deserves a serious referee to examine the full derivations, experimental setup, and validation against baselines.

Referee Report

2 major / 1 minor

Summary. The paper proposes a generative modeling framework for non-Gaussian spatio-temporal fields that factorizes the joint density as a product of univariate conditional distributions. Each conditional is modeled via a Gaussian-process-parameterized transport map in an autoregressive construction. A GP prior supplies regularization for small training sets, while data-dependent sparsity in the conditioning sets is used to achieve scalability. A forward-in-time sampling/prediction variant is also introduced, and the method is demonstrated on non-Gaussian climate-model output containing tens of millions of points.

Significance. If the central claims hold, the work would supply a flexible, regularized approach to sampling from complex nonstationary and non-Gaussian spatio-temporal distributions at scales relevant to climate applications. The combination of autoregressive factorization, transport maps, and GP priors addresses a recognized gap between expressive density estimation and practical scalability with limited data.

major comments (2)

[Construction of the autoregressive transport map and sparsity rule] The scalability claim rests on data-dependent sparsity of conditioning sets, yet no theoretical bound or consistency result is supplied showing that the chosen sparsity rule preserves all statistically relevant long-range dependencies typical of climate fields; without such safeguards the generative distribution can be biased even if the GP prior regularizes the small-sample regime.
[Numerical experiments] The experimental demonstration on climate data reports no quantitative error metrics, baseline comparisons, or ablation studies on the sparsity thresholds; this leaves the accuracy and scalability assertions without the concrete verification needed to support the central claim.

minor comments (1)

[Method overview] The abstract and early sections would benefit from an explicit equation for the transport-map parameterization of each conditional to make the construction reproducible.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and detailed review. The comments highlight important aspects of our work that we will address to strengthen the manuscript. We respond to each major comment below.

read point-by-point responses

Referee: [Construction of the autoregressive transport map and sparsity rule] The scalability claim rests on data-dependent sparsity of conditioning sets, yet no theoretical bound or consistency result is supplied showing that the chosen sparsity rule preserves all statistically relevant long-range dependencies typical of climate fields; without such safeguards the generative distribution can be biased even if the GP prior regularizes the small-sample regime.

Authors: We agree that the paper does not supply formal theoretical bounds or consistency guarantees for the data-dependent sparsity rule. The sparsity mechanism selects conditioning sets using empirical dependence measures (such as lagged correlations or mutual information thresholds computed from the training data), which are intended to retain the dominant long-range structures present in the observed climate fields. The Gaussian-process parameterization of each transport map supplies additional regularization that is particularly useful when the number of training realizations is small. While we do not claim that the rule is universally consistent for arbitrary non-Gaussian spatio-temporal processes, we will add a new discussion subsection that (i) explicitly states the empirical nature of the sparsity choice, (ii) describes the risk of under-representing weak long-range dependencies, and (iii) outlines possible future theoretical directions. This addition will clarify the scope of the current claims without overstating theoretical support. revision: partial
Referee: [Numerical experiments] The experimental demonstration on climate data reports no quantitative error metrics, baseline comparisons, or ablation studies on the sparsity thresholds; this leaves the accuracy and scalability assertions without the concrete verification needed to support the central claim.

Authors: We accept that the original experimental section relies primarily on visual and qualitative assessment of generated fields. In the revised manuscript we will augment the climate-data demonstration with quantitative metrics, including continuous ranked probability score (CRPS) for predictive distributions and estimated log-likelihood on held-out space-time locations. We will also add direct comparisons against two baselines: a standard separable Gaussian process and a simpler autoregressive model without transport maps. Finally, we will include an ablation study that varies the sparsity threshold and reports the resulting trade-off between generative fidelity (via the quantitative metrics) and wall-clock time. These additions will supply the concrete verification requested. revision: yes

Circularity Check

0 steps flagged

No circularity in derivation chain

full rationale

The paper constructs the joint density via the standard chain-rule factorization into univariate conditionals, each modeled by a Gaussian-process transport map. This is an explicit modeling choice, not a reduction of any output to a fitted input or self-defined quantity. Data-dependent sparsity is introduced as a pragmatic scalability device without claiming it preserves all dependencies by construction or via a self-citation uniqueness theorem. No load-bearing step renames a known result, smuggles an ansatz through prior self-work, or equates a prediction to its own fitting procedure. The derivation remains self-contained against external probabilistic and GP foundations.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 0 invented entities

The central claim rests on the validity of the autoregressive conditional factorization and the modeling power of Gaussian processes inside transport maps; sparsity is introduced for scalability. Limited information is available from the abstract alone.

free parameters (1)

Sparsity selection thresholds or rules
Data-dependent sparsity in conditioning sets likely requires choices or thresholds that are not specified in the abstract.

axioms (2)

domain assumption The joint density factors exactly into a product of univariate conditional distributions
Core modeling choice stated in the abstract for the autoregressive construction.
domain assumption Gaussian processes can accurately represent the conditional distributions after transport-map adjustment
Assumed to handle non-Gaussianity and spatio-temporal dependence.

pith-pipeline@v0.9.0 · 5449 in / 1538 out tokens · 73870 ms · 2026-05-08T17:40:48.801477+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

298 extracted references · 180 canonical work pages

[1]

Scalable Bayesian Transport Maps for High-Dimensional Non-Gaussian Spatial Fields , url =

Matthias Katzfuss and Florian Schäfer , doi =. Scalable Bayesian Transport Maps for High-Dimensional Non-Gaussian Spatial Fields , url =. Journal of the American Statistical Association , keywords =
[2]

arXiv preprint arXiv:2412.08820 , year=

Precision and Cholesky Factor Estimation for Gaussian Processes , author=. arXiv preprint arXiv:2412.08820 , year=

work page arXiv
[3]

Journal of the American Statistical Association , volume =

Permutation-based Factorizations of the Gaussian Likelihood , author =. Journal of the American Statistical Association , volume =. 2018 , publisher =

2018
[4]

Spatial Statistics , volume =

Limitations on Low-Rank Approximations for Covariance Matrices of Spatial Data , author =. Spatial Statistics , volume =. 2014 , publisher =

2014
[5]

and Vertenstein, Mariana and Worley, Patrick H

Dennis, John M. and Vertenstein, Mariana and Worley, Patrick H. and Mirin, Arthur A. and Craig, Anthony P. and Jones, Philip W. and Mickelson, Shawn A. and Jacob, Robert L. , title =. International Journal of High Performance Computing Applications , volume =. 2012 , doi =

2012
[6]

Bulletin of the American Meteorological Society , volume=

The Community Earth System Model: A framework for collaborative research , author=. Bulletin of the American Meteorological Society , volume=. 2013 , doi=

2013
[7]

, number =

Stein, Michael L. , number =. 2011 , journal =. doi:10.1214/11-AOS909 , issn =

work page doi:10.1214/11-aos909 2011
[8]

and Schmidt, H

Elbern, H. and Schmidt, H. and Talagrand, Olivier and Ebel, A. , number =. 2000 , journal =. doi:10.1016/S1364-8152(00)00049-9 , issn =

work page doi:10.1016/s1364-8152(00)00049-9 2000
[9]

and Katzfuss, Matthias and Wikle, Christopher K

Stroud, Jonathan R. and Katzfuss, Matthias and Wikle, Christopher K. , number =. 2018 , journal =. doi:10.1175/MWR-D-16-0427.1 , arxivId =

work page doi:10.1175/mwr-d-16-0427.1 2018
[10]

and Tawn, Jonathan A , number =

Coles, S.G. and Tawn, Jonathan A , number =. 1996 , journal =

1996
[11]

, number =

Handcock, Mark S and Stein, Michael L. , number =. 1993 , journal =

1993
[12]

1999 , booktitle =

Knuth, Kevin H , pages =. 1999 , booktitle =

1999
[13]

2010 , journal =

Stegle, Oliver and Parts, Leopold and Durbin, Richard and Winn, John , number =. 2010 , journal =. doi:10.1371/journal.pcbi.1000770 , issn =

work page doi:10.1371/journal.pcbi.1000770 2010
[14]

Di Narzo, A. F. and Cocchi, D. , number =. 2010 , journal =. doi:10.1111/j.1467-9876.2009.00700.x , issn =

work page doi:10.1111/j.1467-9876.2009.00700.x 2010
[15]

and Smith, Richard L

Katzfuss, Matthias and Hammerling, Dorit M. and Smith, Richard L. , number =. 2017 , journal =. doi:10.1002/2017GL073688 , issn =

work page doi:10.1002/2017gl073688 2017
[16]

and Buxton, Bruce E and Craigmile, Peter F

Paul, Rajib and Cressie, Noel and Calder, Catherine A. and Buxton, Bruce E and Craigmile, Peter F. and Li, Hongfei and Mcmillan, Nancy J and Morara, Michele and Sanford, Jessica and Santner, Thomas J and Zhang, Jian , number =. 2007 , booktitle =

2007
[17]

IEEE Sensor Array and Multichannel Signal Processing Workshop , author =

Proc. IEEE Sensor Array and Multichannel Signal Processing Workshop , author =
[18]

, pages =

Xu, Ganggang and Liang, Faming and Genton, Marc G. , pages =. 2015 , journal =. doi:10.5705/ss.2013.085w , issn =

work page doi:10.5705/ss.2013.085w 2015
[19]

and Berliner, L

Wikle, Christopher K. and Berliner, L. Mark , number =. 2007 , journal =. doi:10.1016/j.physd.2006.09.017 , issn =

work page doi:10.1016/j.physd.2006.09.017 2007
[20]

and Fricker, T.E

Ferro, C.A.T. and Fricker, T.E. , month =. 2012 , journal =. doi:10.1002/qj.1924 , issn =

work page doi:10.1002/qj.1924 2012
[21]

and Gelfand, Alan E

Berrocal, Veronica J. and Gelfand, Alan E. and Holland, David M. , url =. 2011 , journal =

2011
[22]

and Jones, Andrew S

Kidder, Stanley Q. and Jones, Andrew S. , number =. 2007 , journal =

2007
[23]

2008 , journal =

Manzan, S and Zerom, D , number =. 2008 , journal =. doi:10.1016/j.ijforecast.2007.12.004 , issn =

work page doi:10.1016/j.ijforecast.2007.12.004 2008
[24]

and Datta, Abhirup and Finley, Andrew O

Heaton, Matthew J. and Datta, Abhirup and Finley, Andrew O. and Furrer, Reinhard and Guinness, Joseph and Guhaniyogi, Rajarshi and Gerber, Florian and Gramacy, Robert B. and Hammerling, Dorit M. and Katzfuss, Matthias and Lindgren, Finn and Nychka, Douglas W. and Sun, Furong and Zammit-Mangion, Andrew , number =. 2019 , journal =. doi:10.1007/s13253-018-0...

work page doi:10.1007/s13253-018-00348-w 2019
[25]

and Eidsvik, Jo and Guindani, Michele and Nail, Amy J

Reich, Brian J. and Eidsvik, Jo and Guindani, Michele and Nail, Amy J. and Schmidt, Alexandra M. , number =. 2011 , journal =. doi:10.1214/11-AOAS482 , issn =

work page doi:10.1214/11-aoas482 2011
[26]

2020 , journal =

Katzfuss, Matthias and Gong, Wenlong , number =. 2020 , journal =

2020
[27]

2015 , journal =

Liang, Min and Marcotte, Denis , publisher =. 2015 , journal =. doi:10.1007/s00477-015-1100-y , issn =

work page doi:10.1007/s00477-015-1100-y 2015
[28]

2004 , journal =

Wall, Melanie M , url =. 2004 , journal =

2004
[29]

2013 , journal =

Bolin, David and Lindgren, Finn , month =. 2013 , journal =. doi:10.1016/j.csda.2012.11.011 , issn =

work page doi:10.1016/j.csda.2012.11.011 2013
[30]

, number =

Fasiolo, Matteo and Pya, Natalya and Wood, Simon N. , number =. 2016 , journal =. doi:10.1214/15-STS534 , issn =

work page doi:10.1214/15-sts534 2016
[31]

2014 , journal =

Bradley, Jonathan R and Cressie, Noel and Shi, Tao , arxivId =. 2014 , journal =

2014
[32]

and Zimmerman, M.B

Zimmerman, D.L. and Zimmerman, M.B. , number =. 1991 , journal =

1991
[33]

1999 , journal =

Curriero, FC and Lele, Subhash , number =. 1999 , journal =

1999
[34]

2011 , journal =

Kyung, Minjung , number =. 2011 , journal =. doi:10.1214/11-BA629 , issn =

work page doi:10.1214/11-ba629 2011
[35]

and Dubovik, Oleg and Schechner, Yoav , number =

Xu, Feng and Diner, David J. and Dubovik, Oleg and Schechner, Yoav , number =. 2019 , journal =. doi:10.3390/rs11070746 , issn =

work page doi:10.3390/rs11070746 2019
[36]

, pages =

Horrell, Michael T and Stein, Michael L. , pages =. 2015 , journal =

2015
[37]

2022 , journal =

Zhou, Xingyu and Jiao, Yuling and Liu, Jin and Huang, Jian , volume =. 2022 , journal =. doi:10.1080/01621459.2021.2016424 , issn =

work page doi:10.1080/01621459.2021.2016424 2022
[38]

and Cressie, Noel , number =

Wikle, Christopher K. and Cressie, Noel , number =. 1999 , journal =

1999
[39]

, number =

Storey, John D. , number =. 2002 , journal =. doi:10.1111/1467-9868.00346 , issn =

work page doi:10.1111/1467-9868.00346 2002
[40]

2012 , journal =

Sigrist, Fabio and K. 2012 , journal =

2012
[41]

2005 , journal =

Tzeng, ShengLi and Huang, Hsin-Cheng and Cressie, Noel , number =. 2005 , journal =. doi:10.1198/016214505000000420 , issn =

work page doi:10.1198/016214505000000420 2005
[42]

Voutilainen, A. and Pyh. 2007 , journal =

2007
[43]

2005 , journal =

Fuentes, Montserrat , number =. 2005 , journal =. doi:10.1016/j.jmva.2004.09.003 , issn =

work page doi:10.1016/j.jmva.2004.09.003 2005
[44]

2012 , journal =

Sang, Huiyan and Huang, Jianhua Z , number =. 2012 , journal =

2012
[45]

and Durrant-Whyte, H.F

Rao, B.S.Y. and Durrant-Whyte, H.F. and Sheen, J.a. , number =. 1993 , journal =. doi:10.1177/027836499301200102 , issn =

work page doi:10.1177/027836499301200102 1993
[46]

2008 , journal =

Kiiveri, Harri T , number =. 2008 , journal =. doi:10.1186/1471-2105-9-195 , issn =

work page doi:10.1186/1471-2105-9-195 2008
[47]

2021 , journal =

Katzfuss, Matthias and Guinness, Joseph , number =. 2021 , journal =. doi:10.1214/19-STS755 , arxivId =

work page doi:10.1214/19-sts755 2021
[48]

and Hooten, Mevin B , number =

Wikle, Christopher K. and Hooten, Mevin B , number =. 2010 , journal =

2010
[49]

2007 , journal =

Gotway, Carol A and Young, Linda J , number =. 2007 , journal =. doi:10.1198/106186007X179257 , issn =

work page doi:10.1198/106186007x179257 2007
[50]

2011 , journal =

Hsu, Nan-jung and Chang, Ya-mei and Huang, Hsin-cheng , number =. 2011 , journal =. doi:10.1002/env.1130 , keywords =

work page doi:10.1002/env.1130 2011
[51]

and Guttorp, Peter , number =

Bellone, E and Hughes, James P. and Guttorp, Peter , number =. 2000 , journal =. doi:10.3354/cr015001 , issn =

work page doi:10.3354/cr015001 2000
[52]

Andrew and Berliner, L

Royle, J. Andrew and Berliner, L. Mark , number =. 1999 , journal =. doi:10.2307/1400420 , issn =

work page doi:10.2307/1400420 1999
[53]

2013 , journal =

Wang, Yueqing and Jiang, Xin and Yu, Bin and Jiang, Ming , number =. 2013 , journal =. doi:10.1080/01621459.2013.796834 , issn =

work page doi:10.1080/01621459.2013.796834 2013
[54]

, number =

Hoff, Peter D. , number =. 2009 , journal =

2009
[55]

Andrew and Berliner, L

Royle, J. Andrew and Berliner, L. Mark and Wikle, Christopher K. and Milliff, Ralph F , editor =. 1999 , booktitle =

1999
[56]

1999 , journal =

Rajagopalan, Balaji and Lall, Upmanu , number =. 1999 , journal =. doi:10.1029/1999WR900028 , issn =

work page doi:10.1029/1999wr900028 1999
[57]

and Wikle, Christopher K

Xu, B. and Wikle, Christopher K. and Fox, N.I. , number =. 2005 , journal =

2005
[58]

, number =

Wikle, Christopher K. , number =. 2002 , journal =. doi:10.1191/1471082x02st036oa , issn =

work page doi:10.1191/1471082x02st036oa 2002
[59]

and Szunyogh, Istvan and Zimin, A

Ott, Edward and Hunt, Brian R. and Szunyogh, Istvan and Zimin, A. V. and Kostelich, Eric J. and Corazza, M. and Kalnay, Eugenia and Patil, D. J. and Yorke, J. A. , pages =. 2004 , journal =. doi:10.1111/j.1600-0870.2004.00076.x , issn =

work page doi:10.1111/j.1600-0870.2004.00076.x 2004
[60]

and Gyarmati, Gyorgyi and Kalnay, Eugenia and Hunt, Brian R

Szunyogh, Istvan and Kostelich, Eric J. and Gyarmati, Gyorgyi and Kalnay, Eugenia and Hunt, Brian R. and Ott, Edward and Satterfield, Elizabeth and Yorke, James A. , number =. 2008 , journal =. doi:10.1111/j.1600-0870.2007.00274.x , issn =

work page doi:10.1111/j.1600-0870.2007.00274.x 2008
[61]

Tippett, Jeffrey L

Anderson, Jeffrey L. , number =. 2003 , journal =. doi:10.1175/1520-0493(2003)131<0634:ALLSFF>2.0.CO;2 , issn =

work page doi:10.1175/1520-0493(2003)131 2003
[62]

2016 , journal =

Poterjoy, Jonathan , pages =. 2016 , journal =. doi:10.1175/MWR-D-15-0163.1 , issn =

work page doi:10.1175/mwr-d-15-0163.1 2016
[63]

2012 , journal =

Anitescu, Mihai and Chen, Jie and Wang, Lei , number =. 2012 , journal =

2012
[64]

, number =

Stein, Michael L. , number =. 2008 , journal =. doi:10.1016/j.jkss.2007.09.001 , issn =

work page doi:10.1016/j.jkss.2007.09.001 2008
[65]

2011 , journal =

Lei, Jing and Bickel, Peter J , pages =. 2011 , journal =. doi:10.1175/2011MWR3553.1 , issn =

work page doi:10.1175/2011mwr3553.1 2011
[66]

1992 , journal =

Carlin, Bradley P and Polson, Nicholas G and Stoffer, David S , number =. 1992 , journal =. doi:10.1080/01621459.1992.10475231 , issn =

work page doi:10.1080/01621459.1992.10475231 1992
[67]

\ Thorncroft, C

Anderson, Jeffrey L. and Anderson, Stephen L. , number =. 1999 , journal =. doi:10.1175/1520-0493(1999)127<2741:AMCIOT>2.0.CO;2 , issn =

work page doi:10.1175/1520-0493(1999)127 1999
[68]

and Gelfand, Alan E

Katzfuss, Matthias , number =. 2017 , journal =. doi:10.1080/01621459.2015.1123632 , issn =

work page doi:10.1080/01621459.2015.1123632 2017
[69]

and Bandyopadhyay, Soutir and Hammerling, Dorit M

Nychka, Douglas W. and Bandyopadhyay, Soutir and Hammerling, Dorit M. and Lindgren, Finn and Sain, Stephan R. , number =. 2015 , journal =

2015
[70]

Mike and Vazquez, Jorge and Armstrong, Edward M

Chin, T. Mike and Vazquez, Jorge and Armstrong, Edward M. , url =. 2013 , journal =

2013
[71]

2006 , journal =

Liu, Xuefeng and Daniels, Michael J , number =. 2006 , journal =. doi:10.1198/106186006X160681 , issn =

work page doi:10.1198/106186006x160681 2006
[72]

, number =

Kalman, R.E. , number =. 1960 , journal =

1960
[73]

and Triantafyllou, G and Korres, G , pages =

Hoteit, Ibrahim and Pham, D -T. and Triantafyllou, G and Korres, G , pages =. 2008 , journal =

2008
[74]

1992 , journal =

Vecchia, AV , number =. 1992 , journal =

1992
[75]

2013 , journal =

Wu, Hao and Wang, Chi and Wu, Zhijin , number =. 2013 , journal =. doi:10.1093/biostatistics/kxs033 , issn =

work page doi:10.1093/biostatistics/kxs033 2013
[76]

, number =

Anderson, Jeffrey L. , number =. 2010 , journal =. doi:10.1175/2010MWR3253.1 , issn =

work page doi:10.1175/2010mwr3253.1 2010
[77]

and Guttorp, Peter and Charles, Stephen P

Hughes, James P. and Guttorp, Peter and Charles, Stephen P. , number =. 1999 , journal =

1999
[78]

2013 , journal =

Reich, Sebastian , number =. 2013 , journal =

2013
[79]

and Katzfuss, Matthias , number =

Kathuria, Dhruva and Mohanty, Binayak P. and Katzfuss, Matthias , number =. 2019 , journal =. doi:10.1029/2018WR023505 , issn =

work page doi:10.1029/2018wr023505 2019
[80]

2010 , journal =

Alonso, Ariel , number =. 2010 , journal =. doi:10.1198/tast.2010.09244 , issn =

work page doi:10.1198/tast.2010.09244 2010

Showing first 80 references.