Modeling wildfire dynamics using FLAM coupled with deep learning methods
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
We improve the accuracy of modeling burned areas using the FLAM model by identifying the hidden relationships between human and natural impacts on wildfire suppression efficiency using the deep learning-based methods. \n \nThe wildfire climate impacts and adaptation model (FLAM) is able to capture impacts of climate, population, and fuel availability on burned areas. FLAM uses a process-based fire parameterization algorithm with a daily time step. The model uses daily temperature, precipitation, relative humidity and wind speed to assess climate impacts on ignition probability and fire spread. The key features implemented in FLAM include fuel moisture computation based on the Fine Fuel Moisture Code (FFMC) of the Canadian Forest Fire Weather Index (FWI), and a procedure to calibrate spatial fire suppression efficiency. \n \nThe coupled FLAM and deep learning approach consists in the following steps. First, using FLAM we calibrate the suppression efficiency map by comparing model output with observed burned area (satellite data). Secondly, we use deep learning methods to identify and assess the drivers behind the calibrated map. The features used in the analysis include several socio-economic factors, including accessibility, GPP, land use maps, as well as burned areas and other parameters modeled by FLAM. Our approach allows classifying those features by their importance and find correlations between them. Finally, we implement the output of deep learning network to estimate the spatial suppression efficiency within FLAM (instead of calibrating it), and validate the approach using observed burned area. \n \nThe proposed approach is implemented using the Google Earth Engine platform that provides flexibility in terms of input data sets and visualization tools. We will present the case study for Indonesia at 0.083 arc degree spatial resolution. It is planned to consider climate change impacts in more detail. \n \nModeling burned areas and suppression efficiency can help the implementation of fire prevention policies for decision maker and provide important information for building adequate and cost-efficient fire response infrastructure.
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.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
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 it