arxiv: 2605.04164 · v1 · submitted 2026-05-05 · 💻 cs.LG · physics.ao-ph· physics.comp-ph

Recognition: 3 theorem links

· Lean Theorem

Enabling Real-Time Training of a Wildfire-to-Smoke Map with Multilinear Operators

Dan J. Krofcheck, John D. Jakeman, Joseph Crockett, Zachary Morrow

Pith reviewed 2026-05-08 17:46 UTC · model grok-4.3

classification 💻 cs.LG physics.ao-phphysics.comp-ph

keywords wildfire modelingsmoke predictionmultilinear operatorsprincipal component analysisreal-time predictionaerosol optical depthmachine learning approximation

0 comments

The pith

Multilinear operators on principal components enable fast wildfire smoke mapping from ignition time.

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

This paper develops a method using multilinear operators to predict smoke concentration fields from time since ignition in wildfire scenarios. It first reduces the fields to principal components and learns a map between their coefficient powers. The result is a model that trains in under 30 seconds and runs in milliseconds while matching the accuracy of expensive Monte Carlo sampling and beating other classifiers on smoke detection. A reader would care because it makes long-term smoke impact forecasting practical for health and infrastructure planning by avoiding full simulations for every fuel scenario.

Core claim

The authors claim that principal components of time-since-ignition and smoke fields from training simulations allow learning a multilinear map from powers of input coefficients to output coefficients. When applied to the Upper Rio Grande Watershed, this produces aerosol optical depth predictions equal in accuracy to Monte Carlo sampling but with fewer than half the coupled model calls, and smoke detection with 65% IoU and 0.95 AUC on holdout data, improving on prior methods' 0.15 IoU and 0.61 AUC.

What carries the argument

A multilinear operator learned to map powers of principal component coefficients of the time-since-ignition field to those of the smoke concentration field.

If this is right

Training on collected data takes less than 30 seconds on CPU.
Forward calls complete in less than 1 ms.
Aerosol optical depth accuracy equals Monte Carlo sampling with under half the model calls.
Smoke detection reaches 65% IoU and 0.95 AUC on holdout data, exceeding similar published classifiers.

Where Pith is reading between the lines

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

Such operators could accelerate modeling in other complex environmental systems like air pollution or climate impacts.
The speed allows exploring many more fuel treatment and succession scenarios than direct simulation permits.
Real-time smoke maps might support emergency response by updating as new data arrives.

Load-bearing premise

The learned multilinear map and principal components will generalize beyond the specific fuel distributions and events in the training simulations to new conditions.

What would settle it

A direct comparison of predicted smoke fields to those from full coupled simulations or satellite observations in a wildfire event featuring fuel types or ignition patterns outside the training set.

Figures

Figures reproduced from arXiv: 2605.04164 by Dan J. Krofcheck, John D. Jakeman, Joseph Crockett, Zachary Morrow.

**Figure 1.** Figure 1: Three approaches for computing the quantity of interest ( view at source ↗

**Figure 2.** Figure 2: Classification performance of linear model. The solid black line is the pointwise median view at source ↗

**Figure 3.** Figure 3: Linear model predictions and observations at different IoU percentiles. The rows in a view at source ↗

**Figure 4.** Figure 4: Correlation of IoU with smoke footprint extent. view at source ↗

**Figure 5.** Figure 5: Classification performance of quadratic model. The solid black line is the pointwise view at source ↗

**Figure 6.** Figure 6: Quadratic model predictions and observations at different accuracy percentiles. The rows view at source ↗

**Figure 7.** Figure 7: Left: Changing observational threshold α (linear operator). Center: Changing PCA accuracy (linear operator). Right: Changing regularization (quadratic operator). All quantities are computed on the validation set, not the holdout set. 17 view at source ↗

**Figure 8.** Figure 8: Classification accuracy on different distributions than were seen in training. view at source ↗

**Figure 9.** Figure 9: Classification performance using a GP to map between input and output images of fixed view at source ↗

read the original abstract

Wildfires are a major producer of fine particulate matter, impacting human health and the electrical grid. Accurately forecasting smoke impacts over long time scales incorporates fuel treatment strategies, natural fuel succession, and stochastic events like lightning strikes. However, predicting smoke for each fuel distribution with a forward simulation of a coupled fire-atmosphere model is computationally infeasible. Moreover, relatively simple fire models are tractable to run in many long-time scenarios but do not capture smoke transport. We use data-driven multilinear operators to predict a smoke concentration field from knowledge of the time since ignition for two quantities of interest: aerosol optical depth and smoke detection. Our method first computes the principal components of time-since-ignition and smoke concentration fields and then learns a map from powers of the input coefficients to the output coefficients. We apply our learned operator to smoke prediction in the Upper Rio Grande Watershed. After collecting training data, learning the approximation weights on a CPU takes less than 30 seconds, and each forward call takes less than 1 ms. On a proxy for aerosol optical depth, we obtain equal accuracy to Monte Carlo sampling with fewer than half as many coupled model calls. For smoke detection, we obtain an intersection-over-union (IoU) of 65% and an area under the receiver operating characteristic curve (AUC) of 0.95 on holdout data. Our method is significantly more accurate than the most similar published smoke classifier, which obtains an IoU and AUC of 0.15 and 0.61, respectively, on a 2015 bushfire in Australia.

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 multilinear PCA surrogate gives fast smoke predictions that beat the prior method on in-distribution holdout but lacks tests on the out-of-distribution cases needed for its intended use.

read the letter

The paper's core contribution is a surrogate model that predicts smoke concentration from time-since-ignition using principal component analysis followed by a multilinear operator on the coefficient powers. On their Upper Rio Grande Watershed data, this runs fast enough for many scenarios and delivers holdout performance that matches Monte Carlo sampling for aerosol optical depth at less than half the cost, while achieving 65% intersection-over-union and 0.95 AUC for smoke detection. They also compare against a prior smoke classifier from a 2015 Australian bushfire and report substantially better numbers. The training of the weights takes less than 30 seconds on a CPU and each prediction is under a millisecond. That combination of speed and accuracy on the proxy task is the practical takeaway. The multilinear map in reduced space is the part that seems fresh. Reducing both input and output fields to PCA coefficients and then learning a map from powers of the inputs to the outputs avoids running the full coupled fire-atmosphere model for each new fuel map. The abstract gives clear metrics and a direct comparison, which helps. The main limitation is that all the reported results come from holdout data drawn from the same ensemble of simulations used for training. The introduction highlights the need to handle fuel treatments, natural succession, and lightning strikes, but those variations are not represented in the training set and no separate out-of-distribution tests are shown. Without that, it's hard to know how well the learned operator will work for the long-term impact assessments the authors have in mind. The method itself is straightforward supervised learning in a reduced space, with no circularity in the predictions. The free parameters are the number of principal components, the degree of the multilinear terms, and the fitted weights. This paper is aimed at people building fast approximations for wildfire smoke transport in environmental modeling. Someone looking for a lightweight way to generate many smoke scenarios from simpler fire models would get concrete value from the reported timings and accuracy numbers. It has enough novelty in the operator construction and enough empirical support to warrant sending it out for peer review, though the referees will likely ask for generalization experiments on new fuel conditions.

Referee Report

2 major / 1 minor

Summary. The manuscript proposes extracting principal components from time-since-ignition and smoke concentration fields generated by coupled fire-atmosphere simulations, then learning a multilinear (polynomial) map from powers of the input coefficients to the output coefficients to predict aerosol optical depth and smoke detection fields. Applied to the Upper Rio Grande Watershed, the approach achieves CPU training of the weights in under 30 seconds and inference in under 1 ms. On holdout data it reports equal accuracy to Monte Carlo sampling for an AOD proxy using fewer than half the coupled model calls, plus smoke detection IoU of 65% and AUC of 0.95, outperforming a prior published classifier (IoU 0.15, AUC 0.61 on a different 2015 Australian event).

Significance. If the in-distribution performance generalizes, the method offers a practical route to real-time smoke mapping that could support long-term forecasting scenarios at far lower cost than repeated full simulations. The sub-30-second training and sub-millisecond inference times, together with the direct Monte Carlo comparison, are concrete strengths. The multilinear operator construction is a reasonable reduced-order modeling choice for this setting.

major comments (2)

[Abstract] Abstract: the reported holdout metrics (AOD parity with Monte Carlo at <50% calls; smoke IoU 65%, AUC 0.95) are given without any description of the data exclusion protocol, the number of retained principal components, the chosen multilinear degree, hyperparameter selection procedure, or error bars / variability across runs or seeds. These omissions make the central performance claims only partially verifiable.
[Abstract] Abstract: the motivating use case explicitly requires accurate predictions under new fuel distributions, natural succession, and stochastic ignitions (e.g., lightning) not present in the training ensemble, yet all quantitative results are confined to in-distribution holdout draws from the same Upper Rio Grande Watershed simulations; no out-of-distribution test cases are described, which directly undercuts the long-term forecasting claim.

minor comments (1)

[Abstract] Abstract: the comparison to the 2015 Australian bushfire classifier is performed on a different geography, fuel type, and event; the manuscript should explicitly note that the datasets are not matched to avoid overstating the relative improvement.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for the constructive comments, which have helped us improve the clarity and scope of the manuscript. We address each major comment below.

read point-by-point responses

Referee: [Abstract] Abstract: the reported holdout metrics (AOD parity with Monte Carlo at <50% calls; smoke IoU 65%, AUC 0.95) are given without any description of the data exclusion protocol, the number of retained principal components, the chosen multilinear degree, hyperparameter selection procedure, or error bars / variability across runs or seeds. These omissions make the central performance claims only partially verifiable.

Authors: We agree that these details are necessary for verifiability. In the revised manuscript we have expanded the abstract (while remaining within length limits) to state: an 80/20 random train/holdout split with no spatial or temporal leakage; retention of the leading 8 principal components for the time-since-ignition fields and 6 for the smoke fields (capturing >95 % of variance); multilinear degree 2; hyperparameters chosen by 5-fold cross-validation on the training set; and all metrics reported as mean ± one standard deviation across 10 independent random seeds. The same information has also been added to the Methods and Results sections for completeness. revision: yes
Referee: [Abstract] Abstract: the motivating use case explicitly requires accurate predictions under new fuel distributions, natural succession, and stochastic ignitions (e.g., lightning) not present in the training ensemble, yet all quantitative results are confined to in-distribution holdout draws from the same Upper Rio Grande Watershed simulations; no out-of-distribution test cases are described, which directly undercuts the long-term forecasting claim.

Authors: We accept this criticism. Our current training ensemble is drawn exclusively from the Upper Rio Grande Watershed with a single fixed fuel map, so true out-of-distribution experiments under altered fuels or different geographic regions are not feasible with the data we possess. We have therefore revised the abstract and added a dedicated Limitations paragraph that (i) explicitly qualifies the long-term forecasting claim to scenarios for which representative training simulations can be obtained, (ii) notes that retraining or domain-adaptation techniques would be required for new fuel distributions, and (iii) lists out-of-distribution generalization as an important direction for future work. revision: partial

standing simulated objections not resolved

We do not have access to additional coupled fire-atmosphere simulations under new fuel distributions, natural succession, or different watersheds that would be required to perform the requested out-of-distribution tests.

Circularity Check

0 steps flagged

No significant circularity; standard supervised learning pipeline with independent holdout evaluation.

full rationale

The paper extracts principal components from an ensemble of coupled fire-atmosphere simulations, fits a multilinear polynomial map from input coefficients to output coefficients on training data, and reports performance metrics on separate holdout draws from the same ensemble. No equations reduce the reported AOD parity, IoU, or AUC values back to the fitted weights by construction. The derivation chain is self-contained against the holdout benchmark and does not rely on self-citation chains or ansatzes that presuppose the target result.

Axiom & Free-Parameter Ledger

3 free parameters · 2 axioms · 0 invented entities

The method rests on standard dimensionality reduction and a learned multilinear regression; the only fitted elements are the operator weights and the choice of component count and polynomial degree, all determined from training data rather than ad-hoc invention.

free parameters (3)

number of principal components
Chosen to represent the dominant modes of time-since-ignition and smoke fields; exact count not stated in abstract.
multilinear degree / powers
The order of powers used in the coefficient map; not specified in abstract.
learned multilinear weights
The approximation coefficients fitted during the <30-second CPU training step.

axioms (2)

standard math Principal component analysis yields an effective low-dimensional basis for both input time-since-ignition and output smoke fields.
Invoked when the method first computes the principal components of the fields.
domain assumption A multilinear function of input coefficients can approximate the mapping to output smoke coefficients.
Core modeling choice that enables the fast forward evaluation.

pith-pipeline@v0.9.0 · 5603 in / 1518 out tokens · 28944 ms · 2026-05-08T17:46:09.775191+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.

Foundation.BranchSelection / Cost.FunctionalEquation branch_selection / washburn_uniqueness_aczel unclear
Our method first computes the principal components of time-since-ignition and smoke concentration fields and then learns a map from powers of the input coefficients to the output coefficients.
Foundation.AlphaCoordinateFixation J_uniquely_calibrated_via_higher_derivative unclear
we limit ourselves to d≤2 ... With practical simplifications ... we can derive a closed-form expression for the approximation weights with a linear operator.
Foundation.RealityFromDistinction (zero-parameter chain) reality_from_one_distinction unclear
We use the validation set to tune hyperparameters ... We use λ=10^5, which we determined with hyperparameter tuning.

Reference graph

Works this paper leans on

94 extracted references · 77 canonical work pages · 1 internal anchor

[1]

& Freitas, R

J. T. Abatzoglou and A. P. Williams. “Impact of Anthropogenic Climate Change on Wildfire across Western US Forests.”Proc. Nat. Acad. Sci. USA113.42 (2016), pp. 11770–11775.doi:10.1073/pnas. 1607171113

work page doi:10.1073/pnas 2016
[2]

Computer-BasedModelsofWildlandFireBehavior:AUser’sManual

F.A.Albini.“Computer-BasedModelsofWildlandFireBehavior:AUser’sManual.” Tech.rep.United States Forest Service, 1976.url:https://www.frames.gov/catalog/8178

1976
[3]

Data-Driven-Dased Analysis and Modeling for the Impact of Wildfire Smoke on PV Systems

A. J. Ali, L. Zhao, and M. H. Kapourchali. “Data-Driven-Dased Analysis and Modeling for the Impact of Wildfire Smoke on PV Systems.”IEEE Trans. Ind. Appl.60.2 (2023), pp. 2076–2084.doi:10.1109/ TIA.2023.3292807

work page arXiv 2023
[4]

Ecological restoration of southwestern ponderosa pine ecosystems: a broad perspective

C. D. Allen, M. Savage, D. A. Falk, K. F. Suckling, T. W. Swetnam, T. Schulke, P. B. Stacey, P. Morgan, M. Hoffman, and J. T. Klingel. “Ecological restoration of southwestern ponderosa pine ecosystems: a broad perspective.”Ecol. Appl.12.5 (2002), pp. 1418–1433.doi:10 . 1890 / 1051 - 0761(2002)012[1418:EROSPP]2.0.CO;2

2002
[5]

Fire Emission Production Simulator (FEPS) User’s Guide, Version 1.0

G. K. Anderson, D. V. Sandberg, and R. A. Norheim. “Fire Emission Production Simulator (FEPS) User’s Guide, Version 1.0.” Tech. rep. United States Forest Service, 2004.url:https://www.frames. gov/catalog/6921

2004
[6]

Aids to Determining Fuel Models for Estimating Fire Behavior

H. E. Anderson. “Aids to Determining Fuel Models for Estimating Fire Behavior.” Tech. rep. INT- GTR-122. United States Forest Service, 1982.doi:10.2737/int-gtr-122

work page doi:10.2737/int-gtr-122 1982
[7]

BEHAVE: Fire Behavior Prediction and Fuel Modeling System—BURN Subsystem, Part 1

P. L. Andrews. “BEHAVE: Fire Behavior Prediction and Fuel Modeling System—BURN Subsystem, Part 1.” Tech. rep. INT-GTR-194. United States Forest Service, 1986.doi:10.2737/INT-GTR-194

work page doi:10.2737/int-gtr-194 1986
[8]

Current Status and Future Needs of the BehavePlus Fire Modeling System

P. L. Andrews. “Current Status and Future Needs of the BehavePlus Fire Modeling System.”Int. J. Wildland Fire23.1 (2013), 21–33.doi:10.1071/wf12167

work page doi:10.1071/wf12167 2013
[9]

A Review of a New Generation of Wildfire–Atmosphere Modeling

A. Bakhshaii and E. A. Johnson. “A Review of a New Generation of Wildfire–Atmosphere Modeling.” Can. J. Forest Res.49.6 (2019), pp. 565–574.doi:10.1139/cjfr-2018-0138

work page doi:10.1139/cjfr-2018-0138 2019
[10]

Proceedings of the National Academy of Sciences114(11), 2946–2951 (2017)

J. K. Balch, B. A. Bradley, J. T. Abatzoglou, R. C. Nagy, E. J. Fusco, and A. L. Mahood. “Human- Started Wildfires Expand the Fire Niche across the United States.”Proc. Nat. Acad. Sci. USA114.11 (2017), pp. 2946–2951.doi:10.1073/pnas.1617394114

work page doi:10.1073/pnas.1617394114 2017
[11]

A North American Hourly Assimilation and Model Forecast Cycle: The Rapid Refresh

S. G. Benjamin, S. S. Weygandt, J. M. Brown, M. Hu, C. R. Alexander, T. G. Smirnova, J. B. Olson, E. P. James, D. C. Dowell, G. A. Grell, H. Lin, S. E. Peckham, T. L. Smith, W. R. Moninger, J. S. Kenyon, and G. S. Manikin. “A North American Hourly Assimilation and Model Forecast Cycle: The Rapid Refresh.”Mon. Weather Rev.144.4 (2016), pp. 1669 –1694.doi:1...

work page doi:10.1175/mwr-d-15-0242.1 2016
[12]

Envi- ronmental Modelling & Software 97, 201–210

J. P. Bolte, D. W. Hulse, S. V. Gregory, and C. Smith. “Modeling Biocomplexity – Actors, Landscapes and Alternative Futures.”Environ. Model. Softw.22.5 (2007), pp. 570–579.doi:10.1016/j.envsoft. 2005.12.033

work page doi:10.1016/j.envsoft 2007
[13]

The Evolution, Complexity and Diversity of Models of Long-Term Forest Dynamics

H. Bugmann and R. Seidl. “The Evolution, Complexity and Diversity of Models of Long-Term Forest Dynamics.”J. Ecol.110.10 (2022), pp. 2288–2307.doi:10.1111/1365-2745.13989. 22

work page doi:10.1111/1365-2745.13989 2022
[14]

The Contribution of Wildfire to PM2.5 Trends in the USA

M. Burke, M. L. Childs, B. de la Cuesta, M. Qiu, J. Li, C. F. Gould, S. Heft-Neal, and M. Wara. “The Contribution of Wildfire to PM2.5 Trends in the USA.”Nature622.7984 (2023), pp. 761–766.doi: 10.1038/s41586-023-06522-6

work page doi:10.1038/s41586-023-06522-6 2023
[15]

Principal Component Flow Map Learning of PDEs from Incomplete, Limited, and Noisy Data

V. Churchill. “Principal Component Flow Map Learning of PDEs from Incomplete, Limited, and Noisy Data.”J. Comput. Phys.524 (2025), p. 113730.doi:10.1016/j.jcp.2025.113730

work page doi:10.1016/j.jcp.2025.113730 2025
[16]

Flow Map Learning for Unknown Dynamical Systems: Overview, Implemen- tation, and Benchmarks

V. Churchill and D. Xiu. “Flow Map Learning for Unknown Dynamical Systems: Overview, Implemen- tation, and Benchmarks.”J. Mach. Learn. Model. Comput.4.2 (2023), pp. 173–201.doi:10.1615/ JMachLearnModelComput.2023049717

2023
[17]

A Sub-Grid, Mixture–Fraction-Based Thermodynamic Equilibrium Model for Gas Phase Combustion in FIRETEC: Development and Results

M. M. Clark, T. H. Fletcher, and R. R. Linn. “A Sub-Grid, Mixture–Fraction-Based Thermodynamic Equilibrium Model for Gas Phase Combustion in FIRETEC: Development and Results.”Int. J. Wild- land Fire19.2 (2010), pp. 202–212.doi:10.1071/WF07116

work page doi:10.1071/wf07116 2010
[18]

A Small-Scale Dynamic Model Using a Terrain-Following Coordinate Transformation

T. L. Clark. “A Small-Scale Dynamic Model Using a Terrain-Following Coordinate Transformation.” J. Comput. Phys.24.2 (1977), pp. 186–215.doi:10.1016/0021-9991(77)90057-2

work page doi:10.1016/0021-9991(77)90057-2 1977
[19]

Description of a Coupled Atmosphere–Fire Model

T. L. Clark, J. Coen, and D. Latham. “Description of a Coupled Atmosphere–Fire Model.”Int. J. Wildland Fire13.1 (2004), pp. 49–63.doi:10.1071/WF03043

work page doi:10.1071/wf03043 2004
[20]

Multi-Domain Simulations of the Time-Dependent Navier–Stokes Equa- tions: Benchmark Error Analysis of Some Nesting Procedures

T. L. Clark and W. D. Hall. “Multi-Domain Simulations of the Time-Dependent Navier–Stokes Equa- tions: Benchmark Error Analysis of Some Nesting Procedures.”J. Comput. Phys.92.2 (1991), pp. 456– 481.doi:10.1016/0021-9991(91)90218-A

work page doi:10.1016/0021-9991(91)90218-a 1991
[21]

A CoupledAtmosphere–FireModel: Convective Feedback on Fire-Line Dynamics

T. L. Clark,M. A. Jenkins, J.Coen, and D.Packham. “A CoupledAtmosphere–FireModel: Convective Feedback on Fire-Line Dynamics.”J. Appl. Meteorol. Climatol.35.6 (1996), pp. 875 –901.doi:10. 1175/1520-0450(1996)035<0875:ACAMCF>2.0.CO;2

1996
[22]

A Coupled Atmosphere–Fire Model: Role of the Convective Froude Number and Dynamic Fingering at the Fireline

T. Clark, M. Jenkins, J. Coen, and D. Packham. “A Coupled Atmosphere–Fire Model: Role of the Convective Froude Number and Dynamic Fingering at the Fireline.”Int. J. Wildland Fire6.4 (1996), pp. 177–190.doi:10.1071/WF9960177

work page doi:10.1071/wf9960177 1996
[23]

Modeling Wildland Fires: A Description of the Coupled Atmosphere–Wildland Fire Model

J. L. Coen. “Modeling Wildland Fires: A Description of the Coupled Atmosphere–Wildland Fire Model.” Tech. rep. NCAR/TN-500+STR. National Center for Atmospheric Research, 2013.doi:10. 5065/D6K64G2G

2013
[24]

WRF-Fire: Coupled Weather–Wildland Fire Modeling with the Weather Research and Forecasting Model

J. L. Coen, M. Cameron, J. Michalakes, E. G. Patton, P. J. Riggan, and K. M. Yedinak. “WRF-Fire: Coupled Weather–Wildland Fire Modeling with the Weather Research and Forecasting Model.”J. Appl. Meteorol. Climatol.52.1 (2013), pp. 16–38.doi:10.1175/JAMC-D-12-023.1

work page doi:10.1175/jamc-d-12-023.1 2013
[25]

SolarEnergy Resource Availability under Extreme and Historical Wildfire Smoke Conditions

K.A.Corwin,J.Burkhardt,C.A.Corr,P.W.Stackhouse,A.Munshi,andE.V.Fischer.“SolarEnergy Resource Availability under Extreme and Historical Wildfire Smoke Conditions.”Nat. Commun.16.1 (2025), p. 245.doi:10.1038/s41467-024-54163-8

work page doi:10.1038/s41467-024-54163-8 2025
[26]

A Computationally Efficient Method for Updating Fuel Inputs for Wildfire Behavior Models Using Sentinel Imagery and Random Forest Classification

A. L. DeCastro, T. W. Juliano, B. Kosović, H. Ebrahimian, and J. K. Balch. “A Computationally Efficient Method for Updating Fuel Inputs for Wildfire Behavior Models Using Sentinel Imagery and Random Forest Classification.”Remote Sens.14.6 (2022), p. 1447.doi:10.3390/rs14061447

work page doi:10.3390/rs14061447 2022
[27]

Exploring Three- Dimensional Coupled Fire–Atmosphere Interactions Downwind of Wind-Driven Surface Fires and Their Influence on Backfires Using the HIGRAD-FIRETEC model

J.-L. Dupuy, R. R. Linn, V. Konovalov, F. Pimont, J. A. Vega, and E. Jiménez. “Exploring Three- Dimensional Coupled Fire–Atmosphere Interactions Downwind of Wind-Driven Surface Fires and Their Influence on Backfires Using the HIGRAD-FIRETEC model.”Int. J. Wildland Fire20.6 (2011), pp. 734–750.doi:10.1071/WF10035

work page doi:10.1071/wf10035 2011
[28]

Spectral Neural Operators

V. S. Fanaskov and I. V. Oseledets. “Spectral Neural Operators.”Dokl. Math.108.2 (2023), S226–S232. doi:10.1134/S1064562423701107

work page doi:10.1134/s1064562423701107 2023
[29]

Quantifying the Impact of Wildfire Smoke on Solar Photovoltaic Generation in Australia

E. Ford, I. M. Peters, and B. Hoex. “Quantifying the Impact of Wildfire Smoke on Solar Photovoltaic Generation in Australia.”iScience27.2 (2024), p. 108611.doi:10.1016/j.isci.2023.108611

work page doi:10.1016/j.isci.2023.108611 2024
[30]

Fully Coupled ‘Online’ Chemistry within the WRF Model

G. A. Grell, S. E. Peckham, R. Schmitz, S. A. McKeen, G. Frost, W. C. Skamarock, and B. Eder. “Fully Coupled ‘Online’ Chemistry within the WRF Model.”Atmos. Environ.39.37 (2005), pp. 6957– 6975.doi:10.1016/j.atmosenv.2005.04.027

work page doi:10.1016/j.atmosenv.2005.04.027 2005
[31]

PnET-Successionv5.1:Comprehensive Description of an Ecophysiological Succession Extension within the LANDIS-II Forest Landscape Model

E.J.Gustafson,B.R.Miranda,B.R.Sturtevant,andZ.Zhou.“PnET-Successionv5.1:Comprehensive Description of an Ecophysiological Succession Extension within the LANDIS-II Forest Landscape Model.” Tech. rep. United States Forest Service, 2023.url:https : / / research . fs . usda . gov / treesearch/66927

2023
[32]

Toward real-time optimization through model reduction and model discrepancy sensitivities

J. Hart, S. A. McQuarrie, Z. Morrow, and B. van Bloemen Waanders. “Toward real-time optimization through model reduction and model discrepancy sensitivities.”Struct. Multidiscip. Optim.69.3 (2026), p. 62.doi:10.1007/s00158-025-04241-2. 23

work page doi:10.1007/s00158-025-04241-2 2026
[33]

Neural and Spectral Operator Surrogates: Unified Construction and Expression Rate Bounds

L. Herrmann, C. Schwab, and J. Zech. “Neural and Spectral Operator Surrogates: Unified Construction and Expression Rate Bounds.”Adv. Comput. Math.50.4 (2024), p. 72.doi:10.1007/s10444-024- 10171-2

work page doi:10.1007/s10444-024- 2024
[34]

Hersbach, B

H. Hersbach, B. Bell, P. Berrisford, S. Hirahara, A. Horányi, J. Muñoz-Sabater, J. Nicolas, C. Peubey, R. Radu, D. Schepers, A. Simmons, C. Soci, S. Abdalla, X. Abellan, G. Balsamo, P. Bechtold, G. Biavati, J. Bidlot, M. Bonavita, G. De Chiara, P. Dahlgren, D. Dee, M. Diamantakis, R. Dragani, J. Flemming, R. Forbes, M. Fuentes, A. Geer, L. Haimberger, S. ...

work page doi:10.1002/qj.3803 2020
[35]

Hotelling

H. Hotelling. “Analysis of a Complex of Statistical Variables into Principal Components.”J. Educ. Psychol.24 (6 1933), pp. 417–441.doi:10.1037/h0071325

work page doi:10.1037/h0071325 1933
[36]

Next Day Wildfire Spread: A Machine Learning Dataset to Predict Wildfire Spreading From Remote-Sensing Data

F. Huot, R. L. Hu, N. Goyal, T. Sankar, M. Ihme, and Y.-F. Chen. “Next Day Wildfire Spread: A Machine Learning Dataset to Predict Wildfire Spreading From Remote-Sensing Data.”IEEE Trans. Geosci. Remote Sens.60 (2022), pp. 1–13.doi:10.1109/TGRS.2022.3192974

work page doi:10.1109/tgrs.2022.3192974 2022
[37]

U.S. Fires Became Larger, More Frequent, and More Widespread in the 2000s

V. Iglesias, J. K. Balch, and W. R. Travis. “U.S. Fires Became Larger, More Frequent, and More Widespread in the 2000s.”Sci. Adv.8.11 (2022), eabc0020.doi:10.1126/sciadv.abc0020

work page doi:10.1126/sciadv.abc0020 2022
[38]

PyApprox: A Software Package for Sensitivity Analysis, Bayesian Inference, Optimal Ex- perimental Design, and Multi-Fidelity Uncertainty Quantification and Surrogate Modeling

J. Jakeman. “PyApprox: A Software Package for Sensitivity Analysis, Bayesian Inference, Optimal Ex- perimental Design, and Multi-Fidelity Uncertainty Quantification and Surrogate Modeling.”Environ. Model. Softw.170 (2023), p. 105825.doi:10.1016/j.envsoft.2023.105825

work page doi:10.1016/j.envsoft.2023.105825 2023
[39]

Verification and Vali- dation for Trustworthy Scientific Machine Learning

J. D. Jakeman, L. A. Barba, J. R. R. A. Martins, and T. O’Leary-Roseberry. “Verification and Vali- dation for Trustworthy Scientific Machine Learning.” 2025. arXiv:2502.15496

work page arXiv 2025
[40]

Review of Wildfire Management Techniques–Part I: Causes, Prevention, Detection, Suppression, and Data Analytics

S. Jazebi, F. de León, and A. Nelson. “Review of Wildfire Management Techniques–Part I: Causes, Prevention, Detection, Suppression, and Data Analytics.”IEEE Trans. Power. Deliv.35.1 (2020), pp. 430–439.doi:10.1109/TPWRD.2019.2930055

work page doi:10.1109/tpwrd.2019.2930055 2020
[41]

Explicit Numerical Diffusion in the WRF Model

J. C. Knievel, G. H. Bryan, and J. P. Hacker. “Explicit Numerical Diffusion in the WRF Model.”Mon. Weather Rev.135.11 (2007), pp. 3808 –3824.doi:10.1175/2007MWR2100.1

work page doi:10.1175/2007mwr2100.1 2007
[42]

Spatiotemporal Bag-of-Features for Early Wildfire Smoke Detection

B. Ko, J. Park, and J.-Y. Nam. “Spatiotemporal Bag-of-Features for Early Wildfire Smoke Detection.” Image Vis. Comput.31.10 (2013), pp. 786–795.doi:10.1016/j.imavis.2013.08.001

work page doi:10.1016/j.imavis.2013.08.001 2013
[43]

WRF-Fire Simulation Coupled with a Fuel Moisture Model and Smoke Transport by WRF-Chem

A. K. Kochanski, J. D. Beezley, J. Mandel, and M. Kim. “WRF-Fire Simulation Coupled with a Fuel Moisture Model and Smoke Transport by WRF-Chem.” 2012. arXiv:1208.1059

work page arXiv 2012
[44]

Toward an Integrated System for Fire, Smoke and Air Quality Simulations

A. K. Kochanski, M. A. Jenkins, K. Yedinak, J. Mandel, J. Beezley, and B. Lamb. “Toward an Integrated System for Fire, Smoke and Air Quality Simulations.”Int. J. Wildland Fire25.5 (2015), pp. 534–546.doi:10.1071/WF14074

work page doi:10.1071/wf14074 2015
[45]

ModelingWildfire Smoke Feedback Mechanisms Using a Coupled Fire–Atmosphere Model with a Radiatively Active AerosolScheme

A.K.Kochanski,D.V.Mallia,M.G.Fearon,J.Mandel,A.H.Souri,andT.Brown.“ModelingWildfire Smoke Feedback Mechanisms Using a Coupled Fire–Atmosphere Model with a Radiatively Active AerosolScheme.”J. Geophys. Res. Atmos.124.16(2019), pp.9099–9116.doi:10.1029/2019JD030558

work page doi:10.1029/2019jd030558 2019
[46]

Multi-Grid Tensorized Fourier Neural Operator for High-Resolution PDEs

J. Kossaifi, N. B. Kovachki, K. Azizzadenesheli, and A. Anandkumar. “Multi-Grid Tensorized Fourier Neural Operator for High-Resolution PDEs.”Trans. Mach. Learn. Res.(2024).url:https : / / openreview.net/forum?id=AWiDlO63bH

2024
[47]

Improved Wildfire Spread Prediction with Time-Series Data and the WSTS+ Benchmark

S. Lahrichi, J. Bova, J. Johnson, and J. Malof. “Improved Wildfire Spread Prediction with Time-Series Data and the WSTS+ Benchmark.” 2025. arXiv:2502.12003

work page arXiv 2025
[48]

LANDFIRE: 40 Fire Behavior Fuel Models – Scott/Burgan

LANDFIRE. “LANDFIRE: 40 Fire Behavior Fuel Models – Scott/Burgan.” 2022.url:http://www. landfire/viewer

2022
[49]

The BlueSky Smoke Modeling Framework

N. K. Larkin, S. M. O’Neill, R. Solomon, S. Raffuse, T. Strand, D. C. Sullivan, C. Krull, M. Rorig, J. Peterson, and S. A. Ferguson. “The BlueSky Smoke Modeling Framework.”Int. J. Wildland Fire 18.8 (2009), pp. 906–920.doi:10.1071/WF07086

work page doi:10.1071/wf07086 2009
[50]

A Deep Learning Approach to Identify Smoke Plumes in Satellite Imagery in Near-Real Time for Health Risk Communication

A. Larsen, I. Hanigan, B. J. Reich, Y. Qin, M. Cope, G. Morgan, and A. G. Rappold. “A Deep Learning Approach to Identify Smoke Plumes in Satellite Imagery in Near-Real Time for Health Risk Communication.”J. Expo. Sci. Environ. Epidemiol.31.1 (2021), pp. 170–176.url:https : //pmc.ncbi.nlm.nih.gov/articles/PMC7796988/

2021
[51]

A Spatial Causal Analysis of Wildland Fire- Contributed PM2.5 Using Numerical Model Output

A. Larsen, S. Yang, B. J. Reich, and A. G. Rappold. “A Spatial Causal Analysis of Wildland Fire- Contributed PM2.5 Using Numerical Model Output.”Ann. Appl. Stat.16.4 (2022), pp. 2714–2731. doi:10.1214/22-AOAS1610. 24

work page doi:10.1214/22-aoas1610 2022
[52]

Neural operator: Graph kernel network for partial differential equations.arXiv preprint arXiv:2003.03485, 2020

Z. Li, N. Kovachki, K. Azizzadenesheli, B. Liu, K. Bhattacharya, A. Stuart, and A. Anandkumar. “NeuralOperator:GraphKernelNetworkforPartialDifferentialEquations.” 2020.arXiv:2003.03485

work page arXiv 2020
[53]

Fourier Neural Operator for Parametric Partial Differential Equations

Z. Li, N. B. Kovachki, K. Azizzadenesheli, B. Liu, K. Bhattacharya, A. Stuart, and A. Anandkumar. “Fourier Neural Operator for Parametric Partial Differential Equations.”International Conference on Learning Representations. 2021.url:https://openreview.net/forum?id=c8P9NQVtmnO

2021
[54]

A Transport Model for Prediction of Wildfire Behavior

R. R. Linn. “A Transport Model for Prediction of Wildfire Behavior.” Tech. rep. LA-13334-T. Los Alamos National Laboratory, 1997.doi:10.2172/505313

work page doi:10.2172/505313 1997
[55]

Studying Wildfire Behavior Using FIRETEC

R. Linn, J. Reisner, J. J. Colman, and J. Winterkamp. “Studying Wildfire Behavior Using FIRETEC.” Int. J. Wildland Fire11.4 (2002), pp. 233–246.doi:10.1071/WF02007

work page doi:10.1071/wf02007 2002
[56]

QUIC-Fire: A Fast-Running Simulation Tool for Prescribed Fire Planning

R. Linn, S. Goodrick, S. Brambilla, M. Brown, R. Middleton, J. O’Brien, and J. Hiers. “QUIC-Fire: A Fast-Running Simulation Tool for Prescribed Fire Planning.”Environ. Model. Softw.125 (2020), p. 104616.doi:10.1016/j.envsoft.2019.104616

work page doi:10.1016/j.envsoft.2019.104616 2020
[57]

Comparing Smoke Emissions and Impacts under Alternative Forest Management Regimes

J. W. Long, S. A. Drury, S. G. Evans, C. J. Maxwell, and R. M. Scheller. “Comparing Smoke Emissions and Impacts under Alternative Forest Management Regimes.”Ecol. Soc.27.4 (2022).doi:10.5751/ ES-13553-270426

2022
[58]

Kernel Neural Operators (KNOs) for Scalable, Memory-efficient, Geometrically-flexible Operator Learning

M.Lowery,J.Turnage,Z.Morrow,J.D.Jakeman,A.Narayan,S.Zhe,andV.Shankar.“KernelNeural Operators (KNOs) for Scalable, Memory-Efficient, Geometrically-Flexible Operator Learning.” 2025. arXiv:2407.00809

work page internal anchor Pith review Pith/arXiv arXiv 2025
[59]

Learning nonlinear operators via DeepONet based on the universal approximation theorem of operators.Nature Machine Intelligence, 3(3), 218–229 (2021)

L. Lu, P. Jin, G. Pang, Z. Zhang, and G. E. Karniadakis. “Learning Nonlinear Operators via DeepONet Based on the Universal Approximation Theorem of Operators.”Nat. Mach. Intell.3.3 (2021), pp. 218– 229.doi:10.1038/s42256-021-00302-5

work page doi:10.1038/s42256-021-00302-5 2021
[60]

A Comprehensive and Fair Comparison of Two Neural Operators (with Practical Extensions) Based on FAIR Data

L. Lu, X. Meng, S. Cai, Z. Mao, S. Goswami, Z. Zhang, and G. E. Karniadakis. “A Comprehensive and Fair Comparison of Two Neural Operators (with Practical Extensions) Based on FAIR Data.” Comput. Meth. Appl. Mech. Eng.393 (2022), p. 114778.doi:10.1016/j.cma.2022.114778

work page doi:10.1016/j.cma.2022.114778 2022
[61]

Long-Term Exposure to Wildland Fire Smoke PM2.5 and Mortality in the Contiguous United States

Y. Ma, E. Zang, Y. Liu, J. Wei, Y. Lu, H. M. Krumholz, M. L. Bell, and K. Chen. “Long-Term Exposure to Wildland Fire Smoke PM2.5 and Mortality in the Contiguous United States.”Proc. Nat. Acad. Sci. USA121.40 (2024), e2403960121.doi:10.1073/pnas.2403960121

work page doi:10.1073/pnas.2403960121 2024
[62]

Modeling Wildland Fire Propagation with Level Set Methods

V. Mallet, D. Keyes, and F. Fendell. “Modeling Wildland Fire Propagation with Level Set Methods.” Comput. Math. Appl.57.7 (2009), pp. 1089–1101.doi:10.1016/j.camwa.2008.10.089

work page doi:10.1016/j.camwa.2008.10.089 2009
[63]

A Review of Modeling Approaches Used to Simulate Smoke Transport and Dispersion

D. V. Mallia and A. K. Kochanski. “A Review of Modeling Approaches Used to Simulate Smoke Transport and Dispersion.”Landscape Fire, Smoke, and Health. American Geophysical Union, 2023. Chap. 8, pp. 127–148.doi:10.1002/9781119757030.ch8

work page doi:10.1002/9781119757030.ch8 2023
[64]

Coupled Atmosphere–Wildland Fire Modeling with WRF 3.3 and SFIRE 2011

J. Mandel, J. D. Beezley, and A. K. Kochanski. “Coupled Atmosphere–Wildland Fire Modeling with WRF 3.3 and SFIRE 2011.”Geosci. Model Dev.4.3 (2011), pp. 591–610.doi:10.5194/gmd-4-591- 2011

work page doi:10.5194/gmd-4-591- 2011
[65]

An Accurate Fire-Spread Algorithm in the Weather Research and Forecasting Model Using the Level-Set Method

D. Muñoz-Esparza, B. Kosović, P. A. Jiménez, and J. L. Coen. “An Accurate Fire-Spread Algorithm in the Weather Research and Forecasting Model Using the Level-Set Method.”J. Adv. Model. Earth Syst.10.4 (2018), pp. 908–926.doi:10.1002/2017MS001108

work page doi:10.1002/2017ms001108 2018
[66]

WRF Preprocessing System

National Center for Atmospheric Research. “WRF Preprocessing System.” Version 4.6. 2024.url: https://github.com/wrf-model/WPS

2024
[67]

V Derjaguin, V

S. Osher and J. A. Sethian. “Fronts Propagating with Curvature-Dependent Speed: Algorithms Based on Hamilton–Jacobi Formulations.”J. Comput. Phys.79.1 (1988), pp. 12–49.doi:10.1016/0021- 9991(88)90002-2

work page doi:10.1016/0021- 1988
[68]

Nonlinear Integro-Differential Operator Regression with Neural Net- works

R. G. Patel and O. Desjardins. “Nonlinear Integro-Differential Operator Regression with Neural Net- works.” 2018. arXiv:1810.08552

work page arXiv 2018
[69]

Spence, T

B. Peherstorfer and K. Willcox. “Data-Driven Operator Inference for Nonintrusive Projection-Based Model Reduction.”Comput. Meth. Appl. Mech. Eng.306 (2016), pp. 196–215.doi:10.1016/j.cma. 2016.03.025

work page doi:10.1016/j.cma 2016
[70]

System-Level Feedbacks of Active Fire Regimes in Large Landscapes

N. A. Povak, P. F. Hessburg, R. B. Salter, R. W. Gray, and S. J. Prichard. “System-Level Feedbacks of Active Fire Regimes in Large Landscapes.”Fire Ecol.19.1 (2023), p. 45.doi:10.1186/s42408- 023-00197-0

work page doi:10.1186/s42408- 2023
[71]

The Fire and Smoke Model Evaluation Experiment—A Plan for 25 Integrated, Large Fire–Atmosphere Field Campaigns

S. Prichard, N. S. Larkin, R. Ottmar, N. H. French, K. Baker, T. Brown, C. Clements, M. Dickinson, A. Hudak, A. Kochanski, et al. “The Fire and Smoke Model Evaluation Experiment—A Plan for 25 Integrated, Large Fire–Atmosphere Field Campaigns.”Atmos.10.2 (2019), p. 66.doi:10 . 3390 / atmos10020066

2019
[72]

The REBURN Model: Simulating System-Level Forest Succession and Wildfire Dynamics

S. J. Prichard, R. B. Salter, P. F. Hessburg, N. A. Povak, and R. W. Gray. “The REBURN Model: Simulating System-Level Forest Succession and Wildfire Dynamics.”Fire Ecol.19.1 (2023), p. 38.doi: 10.1186/s42408-023-00190-7

work page doi:10.1186/s42408-023-00190-7 2023
[73]

Lift & Learn: Physics-Informed Machine Learn- ing for Large-Scale Nonlinear Dynamical Systems

E. Qian, B. Kramer, B. Peherstorfer, and K. Willcox. “Lift & Learn: Physics-Informed Machine Learn- ing for Large-Scale Nonlinear Dynamical Systems.”Phys. D Nonlin. Phenom.406 (2020), p. 132401. doi:10.1016/j.physd.2020.132401

work page doi:10.1016/j.physd.2020.132401 2020
[74]

C. E. Rasmussen and C. K. I. Williams.Gaussian Processes for Machine Learning. MIT Press, 2006

2006
[75]

L., Boyle, J

J. Reisner, S. Wynne, L. Margolin, and R. Linn. “Coupled Atmospheric–Fire Modeling Employing the Method of Averages.”Mon. Weather Rev.128.10 (2000), pp. 3683 –3691.doi:10.1175/1520- 0493(2001)129<3683:CAFMET>2.0.CO;2

work page doi:10.1175/1520- 2000
[76]

Restoring Frequent Fire to Dry ConiferForestsDelaystheDeclineofSubalpineForestsintheSouthwestUnitedStatesunderProjected Climate

C. C. Remy, D. J. Krofcheck, A. R. Keyser, and M. D. Hurteau. “Restoring Frequent Fire to Dry ConiferForestsDelaystheDeclineofSubalpineForestsintheSouthwestUnitedStatesunderProjected Climate.”J. Appl. Ecol.61.7 (2024), pp. 1508–1519.doi:10.1111/1365-2664.14689

work page doi:10.1111/1365-2664.14689 2024
[77]

A Mathematical Model for Predicting Fire Spread in Wildland Fuels

R. C. Rothermel. “A Mathematical Model for Predicting Fire Spread in Wildland Fuels.” Tech. rep. INT-115. United States Forest Service, 1972.url:https://research.fs.usda.gov/treesearch/ 32533

1972
[78]

A Landscape Model of Variable Social– Ecological Fire Regimes

R. Scheller, A. Kretchun, T. J. Hawbaker, and P. D. Henne. “A Landscape Model of Variable Social– Ecological Fire Regimes.”Ecol. Model.401 (2019), pp. 85–93.doi:10.1016/j.ecolmodel.2019.03. 022

work page doi:10.1016/j.ecolmodel.2019.03 2019
[79]

Design, Development, and Application of LANDIS-II, a Spatial Landscape Simulation Model with Flexible Temporal and Spatial Resolution

R. M. Scheller, J. B. Domingo, B. R. Sturtevant, J. S. Williams, A. Rudy, E. J. Gustafson, and D. J. Mladenoff. “Design, Development, and Application of LANDIS-II, a Spatial Landscape Simulation Model with Flexible Temporal and Spatial Resolution.”Ecol. Model.201.3 (2007), pp. 409–419.doi: 10.1016/j.ecolmodel.2006.10.009

work page doi:10.1016/j.ecolmodel.2006.10.009 2007
[80]

Standard Fire Behavior Fuel Models: A Comprehensive Set for Use with Rothermel’s Surface Fire Spread Model

J. H. Scott and R. E. Burgan. “Standard Fire Behavior Fuel Models: A Comprehensive Set for Use with Rothermel’s Surface Fire Spread Model.” Tech. rep. RMRS-GTR-153. United States Forest Service, 2005.doi:10.2737/rmrs-gtr-153

work page doi:10.2737/rmrs-gtr-153 2005

Showing first 80 references.