the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 26 Aug 2025
CONFEX: A Database for CONUS Fire EXtent
Abstract. This article presents the CONUS Fire EXtent (CONFEX) database. This database, based on the VIIRS S-NPP 375 m data product, provides wildfire perimeter, centroid, ignition location, start and end date for the period 2012–2023, for the CONUS and Alaska regions. The algorithm takes hotspot locations from VIIRS S-NPP, clusters them into actual wildfires based on DBSCAN clustering and calculates the perimeter and centroid of each cluster, attaching a geodata frame to each cluster or fire. When validated for some recent large fires against the CALFIRE database, an F1 score of 85–96 % and a CSI of 74–93 % were found, showing the efficiency of the algorithm in aggregating hotspots spatially and temporally accurately. This is the first publicly available high-resolution wildfire extent dataset developed for the CONUS and Alaska regions using VIIRS S-NPP 375 m data product. The database provides a valuable resource for researchers to understand the complexities of the fire regimes in the CONUS and Alaska regions.
- Preprint
(1460 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed
- RC1: 'Comment on essd-2025-462', Anonymous Referee #1, 18 Sep 2025
-
RC2: 'Comment on essd-2025-462', Anonymous Referee #2, 17 Oct 2025
The manuscript entitled “CONFEX: A Database for CONUS Fire EXtent” by Qadiri & Cerrai applied the DBSCAN clustering method to the 375m SNPP VIIRS fire detections for estimating fire extents at each SNPP overpass across the CONUS and Alaska. The authors produced a fire extent dataset – CONFEX from 2012 to 2023.
Satellite-based sub-daily fire extents can be a useful data resource for some key applications (e.g., fire management and fire safety, especially in wildland and urban interfaces). The authors aim to produce such a fire extent dataset, so-called CONFEX. However, the data quality (i.e., accuracy and uncertainty) of the CONFEX has not been comprehensively and quantitatively assessed using available datasets, failing to justify the usefulness of the dataset and meet the journal ESSD’s ‘high-quality data’ criteria. Furthermore, the manuscript’s presentation quality is very poor due to the lack of (1) introducing the state-of-the-art fire extent/area datasets and fire extent mapping algorithms, (2) highlighting the knowledge gap that this study would contribute to, (3) fully discussing data accuracy and uncertainty, and algorithm limitations. I will elaborate on my above concerns below.
1. Unassessed data quality. First, the CONFEX contains ~1.2 million fire extents (polygons), yet the authors merely compared the CONFEX’s extents with the perimeters from CalFire over three large wildfires and reported a confusion matrix. For such a large volume dataset, its usefulness is very low without knowing the accuracy and uncertainties. Comparisons of fire perimeters over three fires are far from sufficient. To reveal the accuracy of this dataset, comprehensive evaluation is needed by following standard evaluation practices (as also applied in Chen et al. 2022 that is also cited in this manuscript, and Oliva & Schroeder 2015) with available good-quality fire data. It looks like the authors were not aware of such data. One good choice would be the USGS’ 30m Landsat-based MTBS dataset that maps fire perimeter and severity from 1984 to present. Another dataset for assessing sub-daily fire extents would be daily fire extents from the National Interagency Fire Center (NIFC), as also used in Chen et al. (2022).
Second, after checking the DBSCAN algorithm and the involved parameters in “README_CONFEX_v2.md”, it looks like the authors set “min_samples = 1” in the DBSCAN algorithm. A minimal sample threshold of “1” would lead to delineate all single VIIRS fire detections as individual fire extents, which seems do not make sense. This explains why there are 664,784 (~55%) fire extents of the CONFEX containing only one VIIRS fire detection. It is unreliable to draw a polygon around one VIIRS fire pixel as a fire extent because conditions to flag a VIIRS pixel as a fire detection depend on both fire temperature and fire size. For the DBSCAN algorithm itself, it is discouraged to set the minimal points to less than three. For the CONFEX, fire extents containing 1 - 2 VIIRS fire detections account for ~75% (899,871) of all (1.2 million).
Third, it is also unclear how the VIIRS fire detections were processed because VIIRS fire detection data also include fire points from other non-vegetation thermal sources. Including all fire points without excluding false alarms surely result in errors.
Fourth, the dataset structure/organization is not user-friendly. Each fire extent is stored separately in an ESRI shapefile, and there are ~1.2 million files in a single zip file. It would be very difficult for users to navigate among files to find the extents of interested fires.
2. Poor presentation quality. First, the Introduction section focuses too much on satellite history, which takes about half of manuscript’s main text, and fails to narrow down to the knowledge gap that this study would target at. The authors failed to introduce the state-of-the-art datasets of fire extents/perimeters (e.g., MTBS, MODIS and VIIRS burned area products, Global Fire Atlas, etc.) and evolving algorithms that map fire extents.
Second, it is not stated how VIIRS fire detections were handled. VIIRS active fire product includes fire detection confidence (i.e., low, medium, and high) and false alarms related to non-vegetations fires. I don’t see any explanation at all.
Third, the DBSCAN clustering algorithm is used but not described and referenced. The authors merely mentioned “DBSCAN” was used and did not explain the main idea of the algorithm and what the potential advantages over other popular algorithms and potential limitations.
Fourth, the authors claim that the main motivation of developing the CONFEX is because available datasets do not provide much fire extent information over northeastern CONUS. This region indeed experienced frequent large wildfires before satellite era, yet fires have been limited during the study period. Thus, I don’t see strong motivation that motivated the authors to develop this dataset (or it is misleading at least.)
Finally, I don’t see any discussion of data accuracy and uncertainty and algorithm limitations at all, which are an essential part for the users to understand the uniqueness and usefulness of the dataset and potential applications.
Others minor comments:
L157-160: the 30 m Landsat-based USGS MTBS fire perimeter and burn severity data (1984 - present) has a much higher spatial resolution than VIIRS. The 375m VIIRS’s spatial resolution is still coarse compared to many meter- and sub-meter- resolution sensors. I would not call it “high resolution”. The statement of SNPP VIIRS overpass is incorrect. SNPP VIIRS only has one daytime overpass and one nighttime overpass at the same location across the CONUS every 24 hours, and fewer more in the Alaska.
L209: in Section 3, burn area is presented in both text and figures. Yet no quantitative assessment results of burned area have been reported. Oliva & Schroeder (2015) found that the accuracy of the VIIRS fire detection-based fire extents and burned areas varies substantially with biome types.
L283: Refences are supposed to be arranged in alphabetical order for journal ESSD.
Citation: https://doi.org/10.5194/essd-2025-462-RC2
Status: closed
-
RC1: 'Comment on essd-2025-462', Anonymous Referee #1, 18 Sep 2025
Qadiri et al.
“CONFEX: 1 A Database for CONUS Fire Extent”
Submitted to Earth Science Datasets
Reviewer comments
Dear Authors,
Thank you for your manuscript describing your new fire-perimeter dataset for CONUS. I think you have correctly identified a useful application of the satellite data to support studies of the environmental and other impacts of wildfires in the United States.
In general terms, I think your revisions should focus on three main areas:
First, I believe the introduction needs to spend less time on the history and diversity of fire remote sensing, and more time describing the many, many previously constructed datasets of North American fire area. Your dataset has specific advantages and disadvantages relative to previous attempts, and that is information you should attempt to compile for the reader.
The domain, spatial and temporal coverage and resolution, and key features of these datasets should be compared with CONFEX to inform the reader. Note that I am not requesting quantitative cross-validation. Here are a few of the datasets I think should be compared to CONFEX in your publication:
- Monitoring trends in burn severity
- https://mtbs.gov/
- https://doi.org/10.1071/wf24137
- Firelytics database: https://firelytics.app/
- FIRED (Fire Events Delineation):
- https://doi.org/10.3390/rs12213498
- https://earthlab.colorado.edu/blog/fired-fire-event-dilineation
- Missoula Fire Lab Emission Inventory
- https://essd.copernicus.org/articles/10/2241/2018/essd-10-2241-2018.html
- https://www.fs.usda.gov/rds/archive/Catalog/RDS-2017-0039
Second, the information that pertains to the actual methods you used is currently interleaved with discussion of older methods—you need to make sure that all the information about your method is in the Methods section, and call back to the introduction only to highlight the parts where you have introduced new innovations.
Lastly, you should consider including some discussion about the challenges specific to Alaska and the results of including Alaska in your processing.
Good luck with your revisions, and thank you for your hard work.
Citation: https://doi.org/10.5194/essd-2025-462-RC1 - Monitoring trends in burn severity
-
RC2: 'Comment on essd-2025-462', Anonymous Referee #2, 17 Oct 2025
The manuscript entitled “CONFEX: A Database for CONUS Fire EXtent” by Qadiri & Cerrai applied the DBSCAN clustering method to the 375m SNPP VIIRS fire detections for estimating fire extents at each SNPP overpass across the CONUS and Alaska. The authors produced a fire extent dataset – CONFEX from 2012 to 2023.
Satellite-based sub-daily fire extents can be a useful data resource for some key applications (e.g., fire management and fire safety, especially in wildland and urban interfaces). The authors aim to produce such a fire extent dataset, so-called CONFEX. However, the data quality (i.e., accuracy and uncertainty) of the CONFEX has not been comprehensively and quantitatively assessed using available datasets, failing to justify the usefulness of the dataset and meet the journal ESSD’s ‘high-quality data’ criteria. Furthermore, the manuscript’s presentation quality is very poor due to the lack of (1) introducing the state-of-the-art fire extent/area datasets and fire extent mapping algorithms, (2) highlighting the knowledge gap that this study would contribute to, (3) fully discussing data accuracy and uncertainty, and algorithm limitations. I will elaborate on my above concerns below.
1. Unassessed data quality. First, the CONFEX contains ~1.2 million fire extents (polygons), yet the authors merely compared the CONFEX’s extents with the perimeters from CalFire over three large wildfires and reported a confusion matrix. For such a large volume dataset, its usefulness is very low without knowing the accuracy and uncertainties. Comparisons of fire perimeters over three fires are far from sufficient. To reveal the accuracy of this dataset, comprehensive evaluation is needed by following standard evaluation practices (as also applied in Chen et al. 2022 that is also cited in this manuscript, and Oliva & Schroeder 2015) with available good-quality fire data. It looks like the authors were not aware of such data. One good choice would be the USGS’ 30m Landsat-based MTBS dataset that maps fire perimeter and severity from 1984 to present. Another dataset for assessing sub-daily fire extents would be daily fire extents from the National Interagency Fire Center (NIFC), as also used in Chen et al. (2022).
Second, after checking the DBSCAN algorithm and the involved parameters in “README_CONFEX_v2.md”, it looks like the authors set “min_samples = 1” in the DBSCAN algorithm. A minimal sample threshold of “1” would lead to delineate all single VIIRS fire detections as individual fire extents, which seems do not make sense. This explains why there are 664,784 (~55%) fire extents of the CONFEX containing only one VIIRS fire detection. It is unreliable to draw a polygon around one VIIRS fire pixel as a fire extent because conditions to flag a VIIRS pixel as a fire detection depend on both fire temperature and fire size. For the DBSCAN algorithm itself, it is discouraged to set the minimal points to less than three. For the CONFEX, fire extents containing 1 - 2 VIIRS fire detections account for ~75% (899,871) of all (1.2 million).
Third, it is also unclear how the VIIRS fire detections were processed because VIIRS fire detection data also include fire points from other non-vegetation thermal sources. Including all fire points without excluding false alarms surely result in errors.
Fourth, the dataset structure/organization is not user-friendly. Each fire extent is stored separately in an ESRI shapefile, and there are ~1.2 million files in a single zip file. It would be very difficult for users to navigate among files to find the extents of interested fires.
2. Poor presentation quality. First, the Introduction section focuses too much on satellite history, which takes about half of manuscript’s main text, and fails to narrow down to the knowledge gap that this study would target at. The authors failed to introduce the state-of-the-art datasets of fire extents/perimeters (e.g., MTBS, MODIS and VIIRS burned area products, Global Fire Atlas, etc.) and evolving algorithms that map fire extents.
Second, it is not stated how VIIRS fire detections were handled. VIIRS active fire product includes fire detection confidence (i.e., low, medium, and high) and false alarms related to non-vegetations fires. I don’t see any explanation at all.
Third, the DBSCAN clustering algorithm is used but not described and referenced. The authors merely mentioned “DBSCAN” was used and did not explain the main idea of the algorithm and what the potential advantages over other popular algorithms and potential limitations.
Fourth, the authors claim that the main motivation of developing the CONFEX is because available datasets do not provide much fire extent information over northeastern CONUS. This region indeed experienced frequent large wildfires before satellite era, yet fires have been limited during the study period. Thus, I don’t see strong motivation that motivated the authors to develop this dataset (or it is misleading at least.)
Finally, I don’t see any discussion of data accuracy and uncertainty and algorithm limitations at all, which are an essential part for the users to understand the uniqueness and usefulness of the dataset and potential applications.
Others minor comments:
L157-160: the 30 m Landsat-based USGS MTBS fire perimeter and burn severity data (1984 - present) has a much higher spatial resolution than VIIRS. The 375m VIIRS’s spatial resolution is still coarse compared to many meter- and sub-meter- resolution sensors. I would not call it “high resolution”. The statement of SNPP VIIRS overpass is incorrect. SNPP VIIRS only has one daytime overpass and one nighttime overpass at the same location across the CONUS every 24 hours, and fewer more in the Alaska.
L209: in Section 3, burn area is presented in both text and figures. Yet no quantitative assessment results of burned area have been reported. Oliva & Schroeder (2015) found that the accuracy of the VIIRS fire detection-based fire extents and burned areas varies substantially with biome types.
L283: Refences are supposed to be arranged in alphabetical order for journal ESSD.
Citation: https://doi.org/10.5194/essd-2025-462-RC2
Data sets
CONFEX: CONUS and Alaska Fire EXtent database Raja Zubair Zahoor Qadiri and Diego Cerrai https://data.mendeley.com/datasets/sk6jwy7xmg/2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 1,153 | 52 | 18 | 1,223 | 28 | 21 |
- HTML: 1,153
- PDF: 52
- XML: 18
- Total: 1,223
- BibTeX: 28
- EndNote: 21
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Qadiri et al.
“CONFEX: 1 A Database for CONUS Fire Extent”
Submitted to Earth Science Datasets
Reviewer comments
Dear Authors,
Thank you for your manuscript describing your new fire-perimeter dataset for CONUS. I think you have correctly identified a useful application of the satellite data to support studies of the environmental and other impacts of wildfires in the United States.
In general terms, I think your revisions should focus on three main areas:
First, I believe the introduction needs to spend less time on the history and diversity of fire remote sensing, and more time describing the many, many previously constructed datasets of North American fire area. Your dataset has specific advantages and disadvantages relative to previous attempts, and that is information you should attempt to compile for the reader.
The domain, spatial and temporal coverage and resolution, and key features of these datasets should be compared with CONFEX to inform the reader. Note that I am not requesting quantitative cross-validation. Here are a few of the datasets I think should be compared to CONFEX in your publication:
Second, the information that pertains to the actual methods you used is currently interleaved with discussion of older methods—you need to make sure that all the information about your method is in the Methods section, and call back to the introduction only to highlight the parts where you have introduced new innovations.
Lastly, you should consider including some discussion about the challenges specific to Alaska and the results of including Alaska in your processing.
Good luck with your revisions, and thank you for your hard work.