GuillaumeGay commited on
Commit
7bbbbf6
·
verified ·
1 Parent(s): d3b697a

small_commune_v2, frozen 9 November 2026

Browse files
Files changed (7) hide show
  1. LICENSE +21 -0
  2. README.md +188 -0
  3. example.py +104 -0
  4. forecasts_2026.csv +0 -0
  5. metadata.json +153 -0
  6. model.joblib +3 -0
  7. usual_rates.csv +154 -0
LICENSE ADDED
@@ -0,0 +1,21 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ MIT License
2
+
3
+ Copyright (c) 2026 Guillaume Gay
4
+
5
+ Permission is hereby granted, free of charge, to any person obtaining a copy
6
+ of this software and associated documentation files (the "Software"), to deal
7
+ in the Software without restriction, including without limitation the rights
8
+ to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
9
+ copies of the Software, and to permit persons to whom the Software is
10
+ furnished to do so, subject to the following conditions:
11
+
12
+ The above copyright notice and this permission notice shall be included in all
13
+ copies or substantial portions of the Software.
14
+
15
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16
+ IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17
+ FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
18
+ AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19
+ LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20
+ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21
+ SOFTWARE.
README.md ADDED
@@ -0,0 +1,188 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: mit
3
+ language:
4
+ - en
5
+ - fr
6
+ library_name: sklearn
7
+ pipeline_tag: tabular-classification
8
+ tags:
9
+ - wildfire
10
+ - fire-weather-index
11
+ - tabular-classification
12
+ - scikit-learn
13
+ - logistic-regression
14
+ - france
15
+ - var
16
+ - open-data
17
+ ---
18
+
19
+ # firepredict: daily wildfire-start risk for the communes of the Var (France)
20
+
21
+ A small logistic regression that gives each of the 153 communes of the Var (83), France, a daily chance that a fire of 0.1 ha or more starts there. It is built only from open data. Code, data pipeline and full report: [github.com/guillaumegay13/firepredict](https://github.com/guillaumegay13/firepredict).
22
+
23
+ > **Research preview.** This is not an official warning tool. Fire danger warnings in France are issued by Météo-France and the prefectures. The every-year results below are exploratory (see [Limitations](#limitations)); the clean test is the 2026 one.
24
+
25
+ ## What it predicts
26
+
27
+ For one commune and one target day: the chance that **at least one fire of 0.1 ha or more starts in the commune on the target day**.
28
+
29
+ - The forecast is issued on the morning of the day before the target day.
30
+ - It uses data up to two days before the target day (the last complete day of station data).
31
+ - No input uses recent fire records. BDIFF publishes a year's fires months later, so a live forecast cannot know them. Automated tests in the GitHub repository check the time rules and that no input changes when fire records change.
32
+
33
+ Version: `small_commune_v2`, frozen on 9 November 2026 (git commit `5fa2bc1`), trained on every labelled year.
34
+
35
+ ## Inputs
36
+
37
+ A scikit-learn pipeline: `StandardScaler` then `LogisticRegression` (default settings, `max_iter=1000`). Five inputs, in this order:
38
+
39
+ | Input | Definition | Unit / transform | Standardised weight |
40
+ |---|---|---|---|
41
+ | `log_usual_rate` | Commune's usual fire rate: average number of days per year with at least one fire start of 0.1 ha or more, 2006–2025 without 2024 (`usual_rates.csv`) | log(rate + 0.01) | +0.589 |
42
+ | `log_local_fwi` | Fire Weather Index two days before the target day, from the commune's nearest Météo-France stations | log(1 + FWI), FWI has no unit | +0.858 (strongest) |
43
+ | `local_wb_30d` | Rain minus potential evaporation (Météo-France ETP) over the 30 days that end two days before the target day, nearest stations | mm, negative = dry | −0.562 |
44
+ | `orange_period` | Target day in February or March (brush burning allowed under conditions) | 0 / 1 | +0.230 |
45
+ | `red_period` | Target day in June to September (fire use banned) | 0 / 1 | −0.261 |
46
+
47
+ A missing input is replaced with its training median (`fill_values` in `metadata.json`).
48
+
49
+ The FWI here is an approximation computed from daily station values (daily maximum temperature, minimum humidity, mean wind, daily rain), not the official noon-based product.
50
+
51
+ ## How to use
52
+
53
+ ```bash
54
+ uv run --with scikit-learn==1.9.1 --with joblib --with pandas python example.py
55
+ ```
56
+
57
+ `example.py` checks the SHA-256 fingerprints, loads the model and computes the chance for example rows. The pipeline was saved with **scikit-learn 1.9.1**; older versions (for example 1.7.2) fail to load it. Joblib files are pickles: load only files whose fingerprint you trust (`model_sha256` in `metadata.json`).
58
+
59
+ ```python
60
+ import joblib, json, numpy as np, pandas as pd
61
+
62
+ model = joblib.load("model.joblib")
63
+ meta = json.load(open("metadata.json"))
64
+ rates = pd.read_csv("usual_rates.csv", dtype={"insee": str}).set_index("insee").usual_rate
65
+
66
+ X = pd.DataFrame([{
67
+ "log_usual_rate": np.log(rates["83137"] + 0.01), # Toulon
68
+ "log_local_fwi": np.log1p(35.0), # FWI two days before the target day
69
+ "local_wb_30d": -140.0, # mm over 30 days
70
+ "orange_period": 0,
71
+ "red_period": 1, # target day in July
72
+ }])[meta["inputs"]].fillna(pd.Series(meta["fill_values"]))
73
+ print(model.predict_proba(X)[0, 1]) # about 0.0034
74
+ ```
75
+
76
+ To build the inputs from raw data (weather download, station matching, FWI), use the GitHub repository: `firepredict.model_communes.build_panel` and `firepredict.freeze.predict`, or `uv run python -m firepredict.pipeline forecast` for today's forecast.
77
+
78
+ ## Files
79
+
80
+ | File | Content |
81
+ |---|---|
82
+ | `model.joblib` | The fitted pipeline (SHA-256 `a1e64b40ed3aa7c274d4b2ec01f24e93a80beb976c3e03ce6d2a3e837056b243`) |
83
+ | `usual_rates.csv` | Usual fire rate per commune (INSEE code), used for forecasts |
84
+ | `metadata.json` | Inputs, weights, fill values, training years and rows, raw-data fingerprints, file fingerprints |
85
+ | `forecasts_2026.csv` | Retrospective 2026 test forecasts: `target_day`, `insee`, `chance` (47,889 rows) |
86
+ | `example.py` | Load the model and compute a chance |
87
+ | `LICENSE` | MIT licence of the code and model |
88
+
89
+ ## Training data
90
+
91
+ - **Fire records:** BDIFF (Base de Données sur les Incendies de Forêts en France), Var, 2006–2025 without 2024. Only fires of 0.1 ha or more are counted.
92
+ - **Weather:** Météo-France daily climatological data, all Var stations.
93
+ - **Burning calendar:** Var prefectoral orders on fire use of 16 May 2013 and 2 November 2025.
94
+
95
+ 1,052,640 commune-days, of which 902 positive commune-days (from 910 fires). A positive commune-day is a day on which at least one fire of 0.1 ha or more started in the commune.
96
+
97
+ ## Evaluation
98
+
99
+ Leave-one-year-out over 19 years (2006–2025 without 2024): each year is hidden in turn, the model learns on the other years, then forecasts the hidden year. The unit is a positive commune-day (902 in total). Ranges are 95% ranges from resampling years.
100
+
101
+ **Against two references:**
102
+
103
+ | Reference | ROC AUC (reference → model) | Years better | 15 communes/day: caught | 10% of commune-days: caught |
104
+ |---|---|---|---|---|
105
+ | FWI alone (Var) | 0.718 → 0.782 | 19 / 19 | 85 → 213, +151% (+104% to +221%) | 230 → 354, +54% (+33% to +74%) |
106
+ | Local FWI alone | 0.728 → 0.782 | 18 / 19 | 123 → 213, +73% (+46% to +111%) | 256 → 354, +38% (+20% to +58%) |
107
+ | FWI × commune history | 0.757 → 0.782 | 16 / 19 (p < 0.001) | 222 → 213, −4% (−10% to +2%) | 307 → 354, +15% (+2% to +30%) |
108
+
109
+ - **Against the FWI alone**, the model is clearly better.
110
+ - **Against the FWI multiplied by each commune's usual fire rate**, the model ranks commune-days better, but the gain in fires caught is about zero at practical warning levels (15 communes a day: −4%, −10% to +2%).
111
+ - **On its very top warnings the model does worse than FWI × history:** 59 against 76 at the 1% most alarming commune-days. Its pooled PR AUC is also lower (0.0030 against 0.0034).
112
+
113
+ **What each input adds** (nested, ROC AUC):
114
+
115
+ | Inputs | ROC AUC |
116
+ |---|---|
117
+ | Commune history + local FWI | 0.767 |
118
+ | + burning calendar | 0.775 |
119
+ | + 30-day dryness (this model) | 0.782 |
120
+ | + buildings and roads in wildland (IGN BD TOPO) | 0.781 |
121
+
122
+ Buildings and roads near vegetation add nothing once each commune's fire history is known, so they are not in the model.
123
+
124
+ **Calibration** holds in the bands used by the map: commune-days predicted at 0.1–0.3% had a mean prediction of 0.167% and an observed rate of 0.169%; those at 0.3–1%, 0.38% predicted against 0.45% observed. Above 1% there are only 17 commune-days, too few to check.
125
+
126
+ **Robustness** (model against FWI × commune history):
127
+
128
+ - Without 2023: ROC AUC 0.787 against 0.761 (better in 16 of 18 years).
129
+ - Fires of 1 ha or more: ROC AUC 0.780 against 0.720 (better in 16 of 19 years).
130
+
131
+ **Location within a day is hard to improve.** None of the fixed rankings tested puts more than about 25% of fire starts in the day's 15 highest-ranked communes. This is what the rankings tested reach, not a proven ceiling.
132
+
133
+ Larger models (CatBoost, the Timestral time-series model) were tested in the project too. They do no better and are less consistent from year to year.
134
+
135
+ These results were corrected after an independent review of the code and claims on 9 November 2026 (inputs built from recent fire records were removed, two small leaks were closed, wording was corrected). See the "Update after an independent review" section of [`reports/phase1_results.md`](https://github.com/guillaumegay13/firepredict/blob/main/reports/phase1_results.md).
136
+
137
+ ## 2026 retrospective test and live archive
138
+
139
+ The inputs were chosen after looking at results across all years, so the scores above are somewhat optimistic. Two cleaner tests are running:
140
+
141
+ - **Retrospective 2026 test, blind to the outcomes.** On 9 November 2026, before any 2026 fire record was published, the model was frozen and its forecasts for 1 January – 9 November 2026 were saved in `forecasts_2026.csv` (SHA-256 `2be6091ab25539dc47bb751822d2a9b2b48f21218f8238bf0be1e1653dc8c6d6`). They were all computed on that day, from the weather available then. This is a retrospective test, not a live trial. When BDIFF publishes the 2026 fires, the forecasts will be scored as they are.
142
+ - **Live archive.** From 9 November 2026, each day's forecast from the frozen model is saved in the [`forecasts/`](https://github.com/guillaumegay13/firepredict/tree/main/forecasts) folder of the GitHub repository and committed the same day, so the git history dates it.
143
+
144
+ ## Limitations
145
+
146
+ - **Exploratory results.** The inputs and the 30-day window were chosen after looking at the same years used for evaluation, so p-values and ranges are optimistic. The ranges are conditional on the chosen model.
147
+ - **Approximate FWI.** Computed from daily station values, not the official noon-based FWI or Météo-France's forecasts.
148
+ - **One department.** The model is trained and tested on the Var only. It has not been checked on other departments.
149
+ - **Label issues.** 2024 is excluded: BDIFF holds only 3 Var fires for 2024, against 50–380 in other years. In 2023 BDIFF started recording many more tiny fires, so only fires of 0.1 ha or more are counted.
150
+ - **Commune-level locations only.** BDIFF gives the commune of each fire, not its position.
151
+ - **Burning calendar.** The 2013 and 2025 rules are applied to all years. The 2006–2012 rules and the prefect's extra red days are not included.
152
+ - **Past features use later knowledge.** Weather stations were chosen by their coverage over 2006–2026, and the leave-one-year-out test trains on years after the one it scores.
153
+ - **Small chances.** Most predicted chances are below 1% per commune-day; they are useful to rank communes and days, not as yes/no forecasts.
154
+
155
+ ## Data and credits
156
+
157
+ The model was trained on the data below. This repository does not contain BDIFF or Météo-France raw data; the code in the GitHub repository downloads it from the official sites. `usual_rates.csv` holds yearly averages per commune derived from BDIFF.
158
+
159
+ | Data | Provider | Licence |
160
+ |---|---|---|
161
+ | Fire records 2006–2025 (BDIFF) | Ministère de l'Agriculture, [bdiff.agriculture.gouv.fr](https://bdiff.agriculture.gouv.fr) | see below |
162
+ | Daily weather, all Var stations | Météo-France, via [data.gouv.fr](https://www.data.gouv.fr/datasets/donnees-climatologiques-de-base-quotidiennes) | Licence Ouverte / Etalab 2.0 |
163
+ | Burning calendar | Var prefectoral orders of 16 May 2013 and 2 November 2025 | public acts |
164
+ | Commune boundaries (to match each commune to its nearest stations) | geo.api.gouv.fr (IGN, INSEE) | Licence Ouverte / Etalab 2.0 |
165
+
166
+ The wider project (GitHub repository) also uses IGN BD TOPO, DDTM du Var burned-area outlines, Corine Land Cover 2018 and school holiday data; none of them is an input of this model.
167
+
168
+ **BDIFF.** Source : Base de Données sur les Incendies de Forêts en France, [bdiff.agriculture.gouv.fr](https://bdiff.agriculture.gouv.fr). Droits de reproduction réservés et strictement limités. BDIFF data may be reused for personal, associative or professional purposes, unchanged, with this source cited, and may not be sold or rented.
169
+
170
+ ## Licence
171
+
172
+ Code and model under the MIT licence (see `LICENSE`). Data under the licences above.
173
+
174
+ ## Citation
175
+
176
+ ```bibtex
177
+ @software{gay2026firepredict,
178
+ author = {Gay, Guillaume},
179
+ title = {firepredict: daily wildfire-start risk for the communes of the Var, France, from open data},
180
+ year = {2026},
181
+ url = {https://github.com/guillaumegay13/firepredict},
182
+ note = {Model small\_commune\_v2, frozen 9 November 2026}
183
+ }
184
+ ```
185
+
186
+ ## Contact
187
+
188
+ Guillaume Gay, guillaume.gay@protonmail.com. Feedback from fire services, researchers and local authorities is very welcome.
example.py ADDED
@@ -0,0 +1,104 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """Load the frozen firepredict model and compute a chance for one commune-day.
2
+
3
+ Run from this folder:
4
+ uv run --with scikit-learn==1.9.1 --with joblib --with pandas python example.py
5
+
6
+ The model answers: what is the chance that at least one fire of 0.1 ha or more
7
+ starts in this Var commune on the target day? The forecast is issued on the
8
+ morning of the day before the target day and uses weather up to two days
9
+ before the target day.
10
+
11
+ The five inputs (one row per commune and target day):
12
+
13
+ log_usual_rate log(usual_rate + 0.01). usual_rate is the commune's average
14
+ number of days per year with at least one fire start of
15
+ 0.1 ha or more (BDIFF, 2006-2025 without 2024). Take it from
16
+ usual_rates.csv by INSEE code.
17
+ log_local_fwi log(1 + FWI). FWI is the Fire Weather Index (no unit) of the
18
+ day two days before the target day, computed from the
19
+ commune's nearest Météo-France stations with daily values
20
+ (maximum temperature, minimum humidity, mean wind, daily
21
+ rain). It is an approximation, not the official FWI product.
22
+ local_wb_30d Rain minus potential evaporation (Météo-France ETP), in mm,
23
+ summed over the 30 days that end two days before the target
24
+ day, at the commune's nearest stations. Negative = dry.
25
+ orange_period 1 if the target day is in February or March (brush burning
26
+ allowed under conditions, Var prefectoral orders), else 0.
27
+ red_period 1 if the target day is in June to September (fire use
28
+ banned), else 0.
29
+
30
+ A missing input is replaced with the training median stored in
31
+ metadata.json ("fill_values"), as the frozen pipeline does.
32
+
33
+ To build these inputs from raw data (weather download, station matching,
34
+ FWI), use the code at https://github.com/guillaumegay13/firepredict
35
+ (firepredict.model_communes.build_panel and firepredict.freeze.predict).
36
+
37
+ This is a research preview, not an official warning tool. Fire danger
38
+ warnings in France come from Météo-France and the prefectures.
39
+ """
40
+
41
+ import hashlib
42
+ import json
43
+ from pathlib import Path
44
+
45
+ import joblib
46
+ import numpy as np
47
+ import pandas as pd
48
+ import sklearn
49
+
50
+ HERE = Path(__file__).parent
51
+
52
+
53
+ def sha256(path: Path) -> str:
54
+ return hashlib.sha256(path.read_bytes()).hexdigest()
55
+
56
+
57
+ # The pipeline was saved with scikit-learn 1.9.1; older versions fail to load it.
58
+ if sklearn.__version__ != "1.9.1":
59
+ print(f"warning: scikit-learn {sklearn.__version__} found, the model was saved with 1.9.1")
60
+
61
+ metadata = json.loads((HERE / "metadata.json").read_text())
62
+
63
+ # joblib files are pickles: only load files whose fingerprint you trust.
64
+ assert sha256(HERE / "model.joblib") == metadata["model_sha256"], "model.joblib does not match metadata.json"
65
+ assert sha256(HERE / "forecasts_2026.csv") == metadata["forecasts_2026"]["sha256"], "forecasts_2026.csv changed"
66
+
67
+ model = joblib.load(HERE / "model.joblib")
68
+ rates = pd.read_csv(HERE / "usual_rates.csv", dtype={"insee": str}).set_index("insee").usual_rate
69
+
70
+
71
+ def inputs(insee: str, fwi: float, wb_30d_mm: float, target_month: int) -> pd.DataFrame:
72
+ """One row of the five model inputs, in the order the model expects."""
73
+ row = {
74
+ "log_usual_rate": np.log(rates.get(insee, np.nan) + 0.01),
75
+ "log_local_fwi": np.log1p(fwi),
76
+ "local_wb_30d": wb_30d_mm,
77
+ "orange_period": int(target_month in (2, 3)),
78
+ "red_period": int(target_month in (6, 7, 8, 9)),
79
+ }
80
+ X = pd.DataFrame([row])[metadata["inputs"]]
81
+ return X.fillna(pd.Series(metadata["fill_values"]))
82
+
83
+
84
+ examples = [
85
+ # (description, INSEE code, FWI two days before, 30-day rain - evaporation in mm, month of target day)
86
+ ("Toulon, dry July day", "83137", 35.0, -140.0, 7),
87
+ ("Toulon, February day", "83137", 8.0, -20.0, 2),
88
+ ("Toulon, wet November day", "83137", 2.0, 60.0, 11),
89
+ ]
90
+
91
+ print(f"model {metadata['version']}, frozen {metadata['frozen_at']}, inputs {metadata['inputs']}")
92
+ print(f"usual rate of Toulon (83137): {rates['83137']:.3f} days with a fire start per year\n")
93
+ for label, insee, fwi, wb, month in examples:
94
+ X = inputs(insee, fwi, wb, month)
95
+ chance = model.predict_proba(X)[0, 1]
96
+ # Same result by hand: standardise, weighted sum, logistic function.
97
+ scaler, logit = model[0], model[-1]
98
+ z = ((X.to_numpy()[0] - scaler.mean_) / scaler.scale_) @ logit.coef_[0] + logit.intercept_[0]
99
+ assert np.isclose(chance, 1 / (1 + np.exp(-z)))
100
+ print(f"{label:28s} {X.round(3).to_dict('records')[0]}")
101
+ print(f"{'':28s} chance of a fire start: {chance:.4%}\n")
102
+
103
+ forecasts = pd.read_csv(HERE / "forecasts_2026.csv", dtype={"insee": str})
104
+ print(f"forecasts_2026.csv: {len(forecasts)} rows, {forecasts.target_day.min()} to {forecasts.target_day.max()}")
forecasts_2026.csv ADDED
The diff for this file is too large to render. See raw diff
 
metadata.json ADDED
@@ -0,0 +1,153 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "version": "small_commune_v2",
3
+ "frozen_at": "2026-11-09T12:35:04+00:00",
4
+ "git_commit": "5fa2bc109a9eccab69a1f26349baec04b89ad6d9",
5
+ "question": "chance that at least one fire >= 0.1 ha starts in the commune on the target day; issued the morning before, from data up to two days before the target day",
6
+ "inputs": [
7
+ "log_usual_rate",
8
+ "log_local_fwi",
9
+ "local_wb_30d",
10
+ "orange_period",
11
+ "red_period"
12
+ ],
13
+ "weights_standardised": {
14
+ "log_usual_rate": 0.5893,
15
+ "log_local_fwi": 0.8582,
16
+ "local_wb_30d": -0.5619,
17
+ "orange_period": 0.2303,
18
+ "red_period": -0.2605
19
+ },
20
+ "fill_values": {
21
+ "log_usual_rate": -1.460061,
22
+ "log_local_fwi": 2.703165,
23
+ "local_wb_30d": -34.599998,
24
+ "orange_period": 0.0,
25
+ "red_period": 0.0
26
+ },
27
+ "training_years": [
28
+ 2006,
29
+ 2007,
30
+ 2008,
31
+ 2009,
32
+ 2010,
33
+ 2011,
34
+ 2012,
35
+ 2013,
36
+ 2014,
37
+ 2015,
38
+ 2016,
39
+ 2017,
40
+ 2018,
41
+ 2019,
42
+ 2020,
43
+ 2021,
44
+ 2022,
45
+ 2023,
46
+ 2025
47
+ ],
48
+ "training_rows": 1052640,
49
+ "training_positive_commune_days": 902,
50
+ "bdtopo_edition": "BDTOPO_3-5_TOUSTHEMES_GPKG_LAMB93_D083_2026-09-15",
51
+ "raw_data": {
52
+ "bdiff/bdiff_83_2006-2025.zip": {
53
+ "sha256": "a215fc8bb6c2a2eeca482bc705de2fc5cc3c87b2634a6e86934ffcd0e7ee4103",
54
+ "downloaded_at": "2026-10-27T08:10:12+00:00"
55
+ },
56
+ "meteofrance/Q_83_previous-1950-2024_RR-T-Vent.csv.gz": {
57
+ "sha256": "3c0fe43a70278c6a72499802499f15c0e6bfd0cc74db8674eb8bd3961c48c97a",
58
+ "downloaded_at": "2026-10-27T08:10:14+00:00"
59
+ },
60
+ "meteofrance/Q_83_previous-1950-2024_autres-parametres.csv.gz": {
61
+ "sha256": "3d3d672ae7594a321dec071b3dc3567ec397a4c7f1312f880359666dfd59ba6f",
62
+ "downloaded_at": "2026-10-27T08:10:15+00:00"
63
+ },
64
+ "meteofrance/Q_83_latest-2025-2026_RR-T-Vent.csv.gz": {
65
+ "sha256": "1910c95009245f0c79ab288db0141c3f9f187f16c250ef3fb601e03e98e02194",
66
+ "downloaded_at": "2026-11-09T12:29:13+00:00"
67
+ },
68
+ "meteofrance/Q_83_latest-2025-2026_autres-parametres.csv.gz": {
69
+ "sha256": "2f9bdfc107f5984239a08f331ddd5b760ff4858246fbe43de40fc2e33f74c061",
70
+ "downloaded_at": "2026-11-09T12:29:14+00:00"
71
+ },
72
+ "ddtm83/contours_feux_var_1958-2022.zip": {
73
+ "sha256": "c4c10e0abaaa763807104d5bb9690702115a01842e9589881b7630901daa91c2",
74
+ "downloaded_at": "2026-10-27T08:10:33+00:00"
75
+ },
76
+ "ddtm83/contours_feux_var_2008.zip": {
77
+ "sha256": "2ec3c7903ebd3347c0afe2ccda2248b8895c013f166d15aba426719b88690bfd",
78
+ "downloaded_at": "2026-10-27T08:10:35+00:00"
79
+ },
80
+ "ddtm83/contours_feux_var_2010.zip": {
81
+ "sha256": "cc2d8f32dc504617fc44c8ad1bf58c0c5d6fc63d2c32d01641cd4bb3e3afffff",
82
+ "downloaded_at": "2026-10-27T08:10:36+00:00"
83
+ },
84
+ "ddtm83/contours_feux_var_2006.zip": {
85
+ "sha256": "f785f1080380f472bd2eaa5d558052bc22328ff7b2a857d50a959cfc0e4793ae",
86
+ "downloaded_at": "2026-10-27T08:10:37+00:00"
87
+ },
88
+ "ddtm83/contours_feux_var_2009.zip": {
89
+ "sha256": "7fe4864ef1d3577e78824fe23658f8bbda00f92d73efe1b1323b53f0a9f8a24a",
90
+ "downloaded_at": "2026-10-27T08:10:38+00:00"
91
+ },
92
+ "ddtm83/contours_feux_var_2007.zip": {
93
+ "sha256": "e3e9a37bc88e905fe9c80387c93e5c19b7081fd671b771082b3abe838f71144f",
94
+ "downloaded_at": "2026-10-27T08:10:40+00:00"
95
+ },
96
+ "ddtm83/contours_feux_var_2016.zip": {
97
+ "sha256": "8381568f681d081ae1f0f6153c3e6bbab41be7c3799155d50376e6aeb6a3c0e9",
98
+ "downloaded_at": "2026-10-27T08:10:41+00:00"
99
+ },
100
+ "ddtm83/contours_feux_var_2019.zip": {
101
+ "sha256": "fe1bcbb3fe2c282ce6a8a57d2b7f3ae6a16336340a2f507d37a8f26373073066",
102
+ "downloaded_at": "2026-10-27T08:10:42+00:00"
103
+ },
104
+ "ddtm83/contours_feux_var_2017.zip": {
105
+ "sha256": "4bde103598e51621637c8c69f5f3fa04f2e32e19d81e164f302cd8a40823143e",
106
+ "downloaded_at": "2026-10-27T08:10:43+00:00"
107
+ },
108
+ "ddtm83/contours_feux_var_2005.zip": {
109
+ "sha256": "e1b0edf97218a218ed083adb3d144470cf0fa7a21eabec48c6923c7b7e5c1cf2",
110
+ "downloaded_at": "2026-10-27T08:10:43+00:00"
111
+ },
112
+ "ddtm83/contours_feux_var_2004.zip": {
113
+ "sha256": "e558f8508f0d64eb4ffa27af6ba3b808ee5c22d0430ca6cb3fe1e21ee73a9689",
114
+ "downloaded_at": "2026-10-27T08:10:44+00:00"
115
+ },
116
+ "ddtm83/contours_feux_var_2003.zip": {
117
+ "sha256": "95c258c00fce3e11cdbfec8c6e9d3dabceb1014b580ef7424560f887ee6a4224",
118
+ "downloaded_at": "2026-10-27T08:10:45+00:00"
119
+ },
120
+ "ddtm83/contours_feux_var_2015.zip": {
121
+ "sha256": "8510807ac795ffdf36900d8df3b5127b724dc0a2d4af9b3d5d7528ea67909419",
122
+ "downloaded_at": "2026-10-27T08:10:46+00:00"
123
+ },
124
+ "calendar/vacances_scolaires.csv": {
125
+ "sha256": "cce6ac7d78721c424122b7eac868f6f68199efeda0dcaf9a85687d7c83bc08be",
126
+ "downloaded_at": "2026-10-27T08:34:33+00:00"
127
+ },
128
+ "geography/communes_83.geojson": {
129
+ "sha256": "82270e1304eb96b5204da728804b3fe7f8c27e8d6fd00264ed40fc36ab12b4da",
130
+ "downloaded_at": "2026-10-27T09:56:17+00:00"
131
+ },
132
+ "geography/corine_2018_83.geojson": {
133
+ "sha256": "d43ef795cee37921767104da118077d90ebbef2ac41654dc7dba3f8f300cd4e0",
134
+ "downloaded_at": "2026-10-27T09:56:30+00:00"
135
+ },
136
+ "bdtopo/BDTOPO_3-5_TOUSTHEMES_GPKG_LAMB93_D083_2026-09-15.7z": {
137
+ "sha256": "2617340d9131f0bbcc94a2fc609d8e0242d6983d1e40388fb6e06a370c955336",
138
+ "downloaded_at": "2026-10-28T14:20:01+00:00"
139
+ },
140
+ "bdtopo/BDTOPO_3-5_TOUSTHEMES_GPKG_LAMB93_D083_2026-09-15.gpkg": {
141
+ "sha256": "1b8f72705af10be7f60bf01853da66c6ef22ee43ebe755ac1204e1c1f9d1381b",
142
+ "downloaded_at": "2026-10-28T14:20:25+00:00"
143
+ }
144
+ },
145
+ "model_sha256": "a1e64b40ed3aa7c274d4b2ec01f24e93a80beb976c3e03ce6d2a3e837056b243",
146
+ "forecasts_2026": {
147
+ "kind": "outcome-blinded retrospective test",
148
+ "first_day": "2026-01-01",
149
+ "last_day": "2026-11-09",
150
+ "rows": 47889,
151
+ "sha256": "2be6091ab25539dc47bb751822d2a9b2b48f21218f8238bf0be1e1653dc8c6d6"
152
+ }
153
+ }
model.joblib ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a1e64b40ed3aa7c274d4b2ec01f24e93a80beb976c3e03ce6d2a3e837056b243
3
+ size 1777
usual_rates.csv ADDED
@@ -0,0 +1,154 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ insee,usual_rate
2
+ 83001,0.21052631578947367
3
+ 83002,0.3684210526315789
4
+ 83003,0.5263157894736842
5
+ 83004,0.7368421052631579
6
+ 83005,0.05263157894736842
7
+ 83006,0.15789473684210525
8
+ 83007,0.5789473684210527
9
+ 83008,0.0
10
+ 83009,0.2631578947368421
11
+ 83010,0.21052631578947367
12
+ 83011,0.3157894736842105
13
+ 83012,0.631578947368421
14
+ 83013,0.0
15
+ 83014,0.0
16
+ 83015,0.2631578947368421
17
+ 83016,0.2631578947368421
18
+ 83017,0.2631578947368421
19
+ 83018,0.21052631578947367
20
+ 83019,0.47368421052631576
21
+ 83020,0.10526315789473684
22
+ 83021,0.21052631578947367
23
+ 83022,0.10526315789473684
24
+ 83023,0.5263157894736842
25
+ 83025,0.10526315789473684
26
+ 83026,0.47368421052631576
27
+ 83027,0.3157894736842105
28
+ 83028,0.42105263157894735
29
+ 83029,0.2631578947368421
30
+ 83030,0.21052631578947367
31
+ 83031,0.5263157894736842
32
+ 83032,0.21052631578947367
33
+ 83033,0.47368421052631576
34
+ 83034,0.47368421052631576
35
+ 83035,0.3157894736842105
36
+ 83036,0.15789473684210525
37
+ 83037,0.10526315789473684
38
+ 83038,0.05263157894736842
39
+ 83039,0.05263157894736842
40
+ 83040,0.10526315789473684
41
+ 83041,0.42105263157894735
42
+ 83042,0.10526315789473684
43
+ 83043,0.5263157894736842
44
+ 83044,0.15789473684210525
45
+ 83045,0.3684210526315789
46
+ 83046,0.42105263157894735
47
+ 83047,0.21052631578947367
48
+ 83048,0.21052631578947367
49
+ 83049,0.21052631578947367
50
+ 83050,1.5263157894736843
51
+ 83051,0.8421052631578947
52
+ 83052,0.15789473684210525
53
+ 83053,0.21052631578947367
54
+ 83054,0.10526315789473684
55
+ 83055,0.3157894736842105
56
+ 83056,0.5789473684210527
57
+ 83057,0.3157894736842105
58
+ 83058,0.6842105263157895
59
+ 83059,0.10526315789473684
60
+ 83060,0.15789473684210525
61
+ 83061,1.0
62
+ 83062,0.21052631578947367
63
+ 83063,0.5263157894736842
64
+ 83064,0.15789473684210525
65
+ 83065,0.42105263157894735
66
+ 83066,0.42105263157894735
67
+ 83067,0.2631578947368421
68
+ 83068,0.5789473684210527
69
+ 83069,0.9473684210526315
70
+ 83070,0.3157894736842105
71
+ 83071,0.3684210526315789
72
+ 83072,0.9473684210526315
73
+ 83073,0.6842105263157895
74
+ 83074,0.05263157894736842
75
+ 83075,0.21052631578947367
76
+ 83076,0.15789473684210525
77
+ 83077,0.15789473684210525
78
+ 83078,0.05263157894736842
79
+ 83079,0.10526315789473684
80
+ 83080,0.2631578947368421
81
+ 83081,0.2631578947368421
82
+ 83082,0.2631578947368421
83
+ 83083,0.3157894736842105
84
+ 83084,0.05263157894736842
85
+ 83085,0.42105263157894735
86
+ 83086,0.6842105263157895
87
+ 83087,0.42105263157894735
88
+ 83088,0.21052631578947367
89
+ 83089,0.10526315789473684
90
+ 83090,0.47368421052631576
91
+ 83091,0.2631578947368421
92
+ 83092,0.3157894736842105
93
+ 83093,0.0
94
+ 83094,0.15789473684210525
95
+ 83095,0.3684210526315789
96
+ 83096,0.3157894736842105
97
+ 83097,0.21052631578947367
98
+ 83098,0.05263157894736842
99
+ 83099,0.42105263157894735
100
+ 83100,0.3157894736842105
101
+ 83101,0.631578947368421
102
+ 83102,0.10526315789473684
103
+ 83103,0.15789473684210525
104
+ 83104,0.21052631578947367
105
+ 83105,0.0
106
+ 83106,0.15789473684210525
107
+ 83107,0.7368421052631579
108
+ 83108,0.10526315789473684
109
+ 83109,0.15789473684210525
110
+ 83110,0.05263157894736842
111
+ 83111,0.10526315789473684
112
+ 83112,0.15789473684210525
113
+ 83113,0.15789473684210525
114
+ 83114,0.21052631578947367
115
+ 83115,0.15789473684210525
116
+ 83116,0.2631578947368421
117
+ 83117,0.10526315789473684
118
+ 83118,0.3157894736842105
119
+ 83119,0.10526315789473684
120
+ 83120,0.3684210526315789
121
+ 83121,0.47368421052631576
122
+ 83122,0.0
123
+ 83123,0.2631578947368421
124
+ 83124,1.0526315789473684
125
+ 83125,0.10526315789473684
126
+ 83126,0.42105263157894735
127
+ 83127,0.42105263157894735
128
+ 83128,0.0
129
+ 83129,0.21052631578947367
130
+ 83130,0.0
131
+ 83131,0.05263157894736842
132
+ 83132,0.05263157894736842
133
+ 83133,0.5263157894736842
134
+ 83134,0.3684210526315789
135
+ 83135,0.0
136
+ 83136,0.15789473684210525
137
+ 83137,1.631578947368421
138
+ 83138,0.2631578947368421
139
+ 83139,0.3684210526315789
140
+ 83140,0.8421052631578947
141
+ 83141,0.21052631578947367
142
+ 83142,0.15789473684210525
143
+ 83143,0.5263157894736842
144
+ 83144,0.15789473684210525
145
+ 83145,0.47368421052631576
146
+ 83146,0.3684210526315789
147
+ 83147,0.10526315789473684
148
+ 83148,0.42105263157894735
149
+ 83149,0.3157894736842105
150
+ 83150,0.3684210526315789
151
+ 83151,0.0
152
+ 83152,0.10526315789473684
153
+ 83153,0.47368421052631576
154
+ 83154,0.47368421052631576