Wildfire Spread Dataset for Prediction Model
Bibliographic record
Abstract
Wildfire Spread Prediction Dataset and Export Tool This repository contains a publicly available dataset and tools for wildfire spread prediction, focusing on incremental wildfire region segmentation using multi-source remote sensing data. The dataset and associated scripts are developed by researchers from Wuhan University, China, to support research in wildfire monitoring and prediction. Dataset Overview The dataset includes wildfire data from Canada and Alaska (2015–2019), integrating multi-source inputs for training and evaluating models like the RCDA-Net. It comprises: AffineParams: Contains affine transformation parameters for all 256×256 wildfire mask samples (2015–2019), stored in .pkl format (e.g., AffineParams/2015/UID_FIRE_2_transform.pkl). Each file is a tuple (x_origin, pixel_width, 0, y_origin, 0, pixel_height) in meters. SpatialRef: Includes a Test.shp file for retrieving the ESRI:102001 (North America Albers Equal Area Conic) coordinate reference system. test: Contains 1,630 samples with two subfolders: inputs: 12-channel input data (e.g., fire mask, DEM, BGR, NDVI, meteorological factors) as .npy files, each 256×256 pixels.The specific information of each channel is shown in the following table: Channel Data Type Spatial Resolution Temporal Resolution Source Input.1 Wildfire Mask Constant Daily ABoVE set Input.2 DEM 30m Constant ASTER GDEM V003 Input.3-5 Blue/Green/Red Bands 30m Constant Landsat 8/9 Input.6 NDVI 30m Constant Landsat 8/9 Input.7 Wind Speed 0.5°×0.625° Hourly MERRA-2 Input.8 Wind Direction 0.5°×0.625° Hourly MERRA-2 Input.9 Temperature 0.5°×0.625° Hourly MERRA-2 Input.10 Precipitation 0.5°×0.625° Hourly MERRA-2 Input.11 Humidity 0.5°×0.625° Hourly MERRA-2 Input.12 Air Density 0.5°×0.625° Hourly MERRA-2 Output.1 Wildfire Increment Mask Constant Daily ABoVE set labels: Single-channel incremental fire masks as .npy files, each 256×256 pixels. train: Contains 6,501 samples with two subfolders: inputs: 12-channel input data as .npy files, each 256×256 pixels. labels: Single-channel incremental fire masks as .npy files, each 256×256 pixels. Export Tool The Export2Tif.py script and its compiled executable Export2Tif.exe (generated via PyInstaller) enable users to convert .npy samples from the test or train datasets into GeoTIFF files with geographic information. The tool requires: Dependencies: Python 3.x with numpy, osgeo (GDAL/OGR), and pickle (included in standard library). For Export2Tif.exe, no Python installation is needed. Input File: ExportParams.txt with three lines: test or train: Specifies the dataset source. : The unique identifier of the sample (e.g., 8). : The fire occurrence date (e.g., 2018-06-06). Usage Prepare ExportParams.txt: Example content: test 8 2018-06-06 Place this file in the same directory as Export2Tif.py or Export2Tif.exe. Run the Script: Python Version: Execute python Export2Tif.py in a terminal with the required libraries installed. Executable Version: Double-click Export2Tif.exe or run it via command line. Output: GeoTIFF files are saved in the ExportResults/UID_FIRE_ _ folder, including 12 input channel files (Input.1.tif to Input.12.tif) and one output label file (Output.tif). Each file retains the ESRI:102001 projection and affine transformation from the corresponding .pkl file. Notes Ensure the AffineParams, SpatialRef, test, and train folders are in the working directory. The script pauses on errors (e.g., missing files) to allow debugging; press any key to continue. For large-scale batch processing, modify Export2Tif.py to loop through multiple ExportParams.txt files. Citation If you use this dataset or tool, please cite the following: Huang, X., Meng, Q., Fu, J., & Zou, Q. (2026). RCDA-Net: a residual contextual dual attention network for wildfire spread region prediction. International Journal of Remote Sensing, 1–34. https://doi.org/10.1080/01431161.2026.2619148. Contact Xiaoxuan Huang: huangxiaoxuan@whu.edu.cn Jianhong Fu: fu_jianhong@whu.edu.cn Qin Zou: qz@whu.edu.cn Qingxiang Meng (Corresponding Author): mqx@whu.edu.cn; Affiliation: Wuhan University, Wuhan 430072, Hubei, China License This dataset and tool are released under the MIT License. See LICENSE for details.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.003 |
| Meta-epidemiology (narrow) | 0.003 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.002 | 0.002 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.003 | 0.001 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.030 | 0.029 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".