Multisource Remote Sensing and Geospatial Analysis of Vineyard Wildfire Impacts and Resilience: The 2019 Kincade Fire
Parastoo Farajpoor, Mahla Ardebili Pour, Mohammad Bagher Ghiasi, Mohammadreza Narimani
eess.IV · cs.CV · physics.ao-ph · physics.geo-ph · stat.AP
Quick verdict
A careful, openly reproducible geospatial audit of 4,581 Sonoma vineyard fields in the 2019 Kincade Fire: lower dNBR over vines is real descriptively but dissolves under boundary, donut, placebo and residual-spatial diagnostics, so no firebreak effect is claimed.
Summary
Working agricultural landscapes are usually treated as background in wildfire research. This study builds an open, event-anchored geospatial stack over the 2019 Kincade Fire in Sonoma County to assess vineyard resilience across five dimensions. Inputs include Sentinel-2 SR Harmonized reflectance (71 usable pre-fire scenes, 24 post-fire), OpenET ensemble monthly ET, gridMET fire weather, USDA SSURGO soils, USGS 3DEP terrain, NOAA/NESDIS smoke polygons, an OpenStreetMap network fixed to the 23 October 2019 ignition date, CAL FIRE damage inspections, and an inventory of airborne lidar, GEDI and ICESat-2 holdings. On 4,581 vineyard fields the authors report pre-fire NDMI 0.015 and NDVI 0.439 with ET correlations of r = 0.567 and 0.639, whereas mapped soil water capacity explains almost nothing. Inside the perimeter, vineyard pixels show mean dNBR 0.130 against 0.337 for wildland pixels, a 61.4 percent difference. The paper then deliberately dismantles that headline: the boundary contrast is only -0.0166 at 100 m, reverses sign at 30 m, vanishes in donut specifications, and a placebo offset yields +0.0624; slope is discontinuous across the boundary. A 250 m spatial GAM reverses the sign entirely, giving vineyard fraction +0.138 with residual Moran's I of 0.519. All fields were beneath mapped smoke at least once (mean 7.78 days), 34.2 percent of road nodes are dead ends, and inside-perimeter greenness stayed lower through 2021. Data are archived in Zenodo and code is on GitHub under MIT.
Review complete
Assessment at a glance
5
risk flags identified
● high 0
● medium 3
● low 2
5 strengths · 5 weaknesses · 5 author questions
Levels reflect verification complexity — not a quality score. Evolve never rates your paper.
Key claims under review
- Vineyard pixels show a substantially lower descriptive dNBR than wildland pixels inside the Kincade perimeter (means 0.130 vs 0.337), but this does not establish a universal vineyard firebreak effect.
- The apparent vineyard-wildland edge contrast is small, scale dependent and non-causal: -0.0166 at 100 m, sign-reversed at 30 m, null in donut specifications, and contradicted by a +0.0624 placebo boundary.
- After conditioning on location, terrain and water use, vineyard fraction is positively associated with dNBR (+0.138), showing the unconditional landscape contrast and the conditional comparison answer different questions.
- Lower immediate spectral impact coexisted with pervasive overhead smoke exposure for all 4,581 fields, a road network with 34.2% dead-end nodes, and a persistent inside-perimeter greenness deficit through 2021.
- No dedicated 2019 airborne lidar acquisition exists over Sonoma County, so event-day parcel-level vineyard fuel loads cannot be derived from open data.
Strengths
The headline result is actively falsified rather than defended
Section 3.3 gives the tempting contrast (vineyard dNBR mean 0.130 vs wildland 0.337, 61.4% lower) and Sections 3.4-3.5 immediately test it to destruction: tau reverses to +0.0446 at a 30 m bandwidth, the 100 m donut estimate is null (+0.0068, p = 0.10), an interior placebo returns +0.0624, and the slope continuity check fails at 3.22 vs 4.48 degrees. Very few applied remote-sensing papers publish their own placebo estimate.
Explicit claim-boundary table ties every result to its analytical unit
Table 2 lists nine subsystems with the unit (field, pixel, boundary segment, 250 m cell, field-day, graph node) and a stated limit: descriptive only, not a causal RDD, not a structure-loss model, not dose or smoke taint, and not realized evacuation performance. Section 2.1 also enforces a temporal rule that post-fire variables never explain pre-fire state.
Spatial autocorrelation is diagnosed rather than assumed away
Reproducibility artifacts are concrete and versioned
Multidimensional coverage beyond the burn scar
Weaknesses
The recovery comparison is unadjusted yet reported with a very small p-value
The GAM interval is reported alongside a diagnosis that invalidates it
No field calibration of the central proxy
Smoke exposure is binary and does not distinguish heavy from light plumes
Within-vineyard regression silently drops most fields
Full Review Report
$1.99 · one-timeUnlock the complete preliminary analysis Evolve produced for this paper.
Report preview
Scope & Claims
This paper asks a deliberately bounded question: what can openly available geospatial evidence establish about vineyard impacts during the 2019 Kincade Fire? The scope discipline is the paper's defining feature. Section 1 sets up working agricultural landscapes as a distinct interface rather than a variant of the wildland-urban interface, and Table 2 then binds each of nine subsystems to a unit, a statistic and an explicit claim boundary. The refusal to aggregate anything into a composite resilience score is stated twice (Sections 1 and 4.5).
The claims actually made are: vineyards show a lower descriptive dNBR than wildland; that contrast does not survive as a causal firebreak effect; pre-fire canopy condition coheres with OpenET while mapped soil water capacity does not; potential smoke exposure reached every field; road access is broad but topologically fragile; and inside-perimeter greenness lagged through 2021. Every claim matches what the data can carry. The only place the paper edges past its own fence is Table 3, discussed below.
The remaining sections cover Methodology Assessment, Statistical Evidence, Results & Reproducibility, Limitations & Risks, and Verdict & Recommendations.
- All 5 strengths & 5 weaknesses, analyzed in detail
- 5 risk flags with severity, section & reviewer questions
- Methods & statistical evidence assessment
- 5 questions a peer reviewer would ask
- Full six-dimension evaluation report & recommendations
One-time purchase · this paper only
One-time purchase, this paper only. AI-generated preliminary analysis — not peer review.
Frequently asked questions
Do vineyards act as firebreaks during California wildfires?
This study says the evidence does not support that claim. Vineyard pixels did show lower burn severity than wildland pixels (mean dNBR 0.130 versus 0.337), but once the authors tested the vineyard-wildland edge more carefully the effect largely disappeared: the local contrast was tiny (-0.0166), changed sign at narrow distances, vanished when the immediate edge was excluded, and a deliberately misplaced boundary produced an even larger effect in the opposite direction.
What happened to the vineyards in the 2019 Kincade Fire?
Every one of the 4,581 mapped vineyard fields was beneath a mapped smoke plume on at least one day between 23 October and 10 November 2019, averaging 7.78 days out of a possible 9. Fields inside the fire perimeter also stayed greener-poor than comparable fields outside it through 2021 (recovery NDVI ratios 0.815 versus 0.854), although that difference is unadjusted and coincided with a regional drought.
How many structures burned in the Kincade Fire?
CAL FIRE damage inspections covered 1,568 structures in the dataset used here: 374 destroyed, 6 with major damage, 14 with minor damage, 40 affected and 1,134 with no recorded damage. The fire burned 77,758 acres and was contained on 6 November 2019, after prompting the largest evacuation in Sonoma County history.
Can other researchers repeat this vineyard wildfire analysis?
Yes, unusually well for a geospatial study. All inputs are public federal and state datasets listed in Table 1, the analysis-ready products are archived on Zenodo (doi 10.5281/zenodo.22713181), and the replication code including Google Earth Engine collection identifiers is on GitHub under an MIT license. Most of the workflow is CPU-scale tabular and raster analysis.
About this review
This is an AI-generated preliminary review by Evolve.Pub, produced from the openly available text of the paper. It is not peer review and does not represent the authors' views. The canonical record of the paper is at the publisher link above.
- Figures 2-11 carry substantial quantitative content (cumulative distance distributions, daily smoke area, binned boundary plots, NDVI trajectories) that is not recoverable from the extracted text, so those panels are discussed through the authors' captions and prose.
- Tables are reconstructed from text and some rows, notably Table 3 and Table 4, are interleaved across page breaks; individual digits were read from the text flow rather than from a fixed table layout.
- Reference list contains several arXiv preprints co-authored by the same research group (refs 37, 50, 51); their content was not available and is not relied upon here.
Go deeper on this paper
Wrote a paper? Stress-test it before reviewers do.
Upload your manuscript for the same structured scrutiny — reasoning, statistics, evidence, figures, and reproducibility.