the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 29 Jul 2025
The Western United States MTBS-Interagency database of large wildfires, 1984–2024 (WUMI2024a)
Abstract. Wildfire regimes of the western United States (US) have changed dramatically since the 1980s but our understanding of the causes and effects of these changes is limited by a lack of a quality-controlled, publicly available database of wildfire events that (1) spans from the 1980s to present, (2) represents wildfires across a wide range of sizes, and (3) includes mapped fire perimeters and the area burned within. Here we present an updated and improved Western US MTBS-Interagency database (WUMI2024a) of wildfire occurrences, perimeters, and burned-area maps, covering the period 1984–2024 and the geographic domain of the 11 westernmost states in the contiguous US. The database represents 22,464 wildfires ≥1 km2 in size, which we compile by merging seven publicly available government databases. For over 46% of wildfires in our database (more than 10,300 wildfires), the maps of fire perimeters and area burned are based on 30-m satellite data provided by the US government’s Monitoring Trends in Burn Severity (MTBS) project, allowing our mapping and assessments of total area burned to account for unburned areas within fire boundaries. For another 24% of fires, our database includes perimeter observations provided by non-MTBS sources, meaning that only 30% of fire occurrences are without perimeter observations. For these fires we tentatively assume perimeters are circular centred on the ignition location. Fires without perimeter observations are relatively small and over 95% of area burned in the database is associated with fires with observed perimeters. The fire perimeters and burned area maps will aid assessment of the landcover types that burn and can be used to improve simulations of how historical fires have affected ecosystems and smoke emissions. The WUMI2024a can be quickly updated as new and improved data become available. The WUMI2024a dataset and the code used to produce the dataset are available at http://datadryad.org/share/Ox4oxdwdrhkmjUTpke7QgkfF--h-RLRbmMzGBhSmOr4 (Williams et al., 2025).
- Preprint
(2424 KB) - Metadata XML
-
Supplement
(3210 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on essd-2025-366', Anonymous Referee #1, 27 Aug 2025
-
AC1: 'Reply on RC1', A. Park Williams, 08 Nov 2025
The comment was uploaded in the form of a supplement: https://essd.copernicus.org/preprints/essd-2025-366/essd-2025-366-AC1-supplement.pdf
-
AC1: 'Reply on RC1', A. Park Williams, 08 Nov 2025
-
RC2: 'Comment on essd-2025-366', Anonymous Referee #2, 29 Sep 2025
Williams et al. introduce WUMI2024a, a harmonized, quality-controlled database of 22,464 large wildfire events across the 11 western U.S. states from 1984 – 2024. The product merges seven public government datasets and provides, for each event, a start date, location, final size, a perimeter (observed where available, circular when missing), and a 1-km grid of fractional area burned. About 46% of events inherit 30 m burned-area maps and perimeters from MTBS; another 24% have perimeters from non-MTBS sources; only 30% lack observed perimeters. Given the importance and lack of good quality data for wildfire regime assessment, the work of Williams et al. is highly relevant and useful for the community.
Major comments:
1.
The paper introduces an alternate event list that replaces MTBS “parent” fires with linked non-MTBS “sub-fires,” and even reassigns sizes so sub-fires sum to the parent (Sec. 2.4). It is not clear what decision logic was used to separate single events vs. multiple events. How are overlaps across days/IDs handled? What are the implications for trend analysis of counts vs. burned areas when switching between parent and sub-fire lists?
2.
In terms of the dataset quality-control, did you quantify how sensitive the final counts and areas are to the chosen thresholds (e.g., L160 “we treat them as duplicates if they are within 25 km, 3 days, and 10% of the larger of the fire sizes”)? How were these thresholds chosen?
3.
Section 2.2.2 would benefit from a table or flowchart to outline the merging logic.
4.
The manuscript carefully notes under-representation of pre-1992 non-forest fires and potential artifacts post-2020. Based on these weaknesses, could you add a paragraph/section that outlines “best practices” and makes recommendations for potential users?
Minor comments:
1. The abstract cites 22,464 events, while Sec. 2.4 first lists 22,228 parent events and then 22,464 with sub-fire replacement. Could you clarify where the difference comes from?
2. L180 – 185 Where ignition points fall just outside the CalFire perimeter and are “snapped” to the nearest boundary, mention whether snapping is recorded as a flag and provide displacement statistics (median/95th percentile).
Citation: https://doi.org/10.5194/essd-2025-366-RC2 -
AC2: 'Reply on RC2', A. Park Williams, 08 Nov 2025
The comment was uploaded in the form of a supplement: https://essd.copernicus.org/preprints/essd-2025-366/essd-2025-366-AC2-supplement.pdf
-
AC2: 'Reply on RC2', A. Park Williams, 08 Nov 2025
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 1,406 | 68 | 29 | 1,503 | 42 | 32 | 44 |
- HTML: 1,406
- PDF: 68
- XML: 29
- Total: 1,503
- Supplement: 42
- BibTeX: 32
- EndNote: 44
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
general comments
The objective of the paper is very interesting and extremely useful. Although remote sensing increasingly plays and important role in monitoring the earth surface, and in particular wildfire activity, it lacks the temporal extent to bring the needed confidence of results that assess the role and interaction of fire in the earth systems. This paper addresses this gap by merging different datasets to build a comprehensive and consistent geo-database of wildfire activity over western US. The paper reads well and is clear. In terms of presentation, the paper could benefit from a visual diagram, in the methods section, on how the datasets interact and the rules (temporal and spatial) applied to remove duplicate fire records. In addition, for a paper on a new compiled dataset, it would benefit presenting all the attributes in a table, including also the definition/characteristics that are contained in dataset attributes. This not only makes it clear what the measurements and classification are, but also quickly informs potential users on the fit-for-purpose of the dataset is for their application.
specific comments
The purpose of the paper is achieved but the work is not of exceptional quality and compleness. I was expecting that for such an important topic - where one needs to merge old and new and different information types – that the new dataset and its merging methodology would set a standard on how it could be done with the perspective that it would be regularly updated with attributes that are of important to fire managers, ecologist, fire ecologist and climate researchers, to name a few. Attributes like end-date, fire spread rate, intensity, power, landcover, and fragmentation, can be retrieved and added, especially to recent records. The dataset, although very useful, is limited to what is commonly recorded by the different datasets and struggles to show added value in terms of new information useful for different applications
The authors, although focusing on achieving the higher quality records for the dataset, do not supply a quality/confidence indicator associated with each record. Meaning that users that may want to screen the dataset to remove data that could be uncertain. The compiled dataset should offer a confidence layer. It that regard, the paper fails to set a standard for dataset compilation, and it comes across that quality is reduced to removing duplicates. The authors should consider developing a confidence indicator.
The authors used several rules to remove duplicate records when merging the different datasets. These rules are mostly based on chosen thresholds in space and time. The authors should state what is the rationale behind these, and why the chosen values.
Circular areas are never representative of fire scars. Over flat grasslands under constant conditions burn scars can appear ellipsoidal, but over other landscapes they are rarely circular. By representing them as such, the authors risk overlaying a fire scar over bare soil or water, requiring users to treat the data prior to using it when landcover is important. This kind of inconsistencies should be avoided.
technical corrections
Line 37: sentence could finish with a reference to a study supporting the statement.
Line 83-84: this sentence can be confusing, I recommend reminding that the datasets includes fires that are lower than 1km and that proportion of BA is what is recorded in at every 1km2 gridcell.
Line 96: the different class of severity is presented here, it is not clear where these come from and if the classification is retained for further use. If these are all the range of possible classes provided by MTBS, and are no longer used, I recommend removing these as it will confuse the reader expecting such a classification.
Line 157: 200 kms is a large distance for exclusion, what is the rational for it and the impact of choosing a smaller distance.
Line 162-163: I assume by “keep the first fire in the database” it is meant retaining the record with the earliest date. If so, please make it clear.
Line 330: I might have missed but it is not clear what ST/C&L stands for