SKILL.md

---
name: geoskill-wildfire-spread-modeling
description: '元胞自动机叠加坡度风速燃料湿度修正的过火面积模拟'
---

# 野火蔓延模拟 | Wildfire Spread Modeling

Cellular automaton (CA) based wildfire spread simulation: at each time step, burning pixels ignite unburned pixels in their 8-neighborhood with a given probability; the ignition probability integrates fuel flammability, moisture (wetter fuel is harder to ignite), slope (faster uphill), and wind speed (faster downwind, including the angle relative to wind direction). Once ignited, a pixel never goes out, so the burned area grows monotonically with time; under a fixed random sequence, stronger wind burns more and higher moisture burns less.

## Dependencies / 依赖

```bash
pip install numpy rasterio scipy
```

## Usage / 使用方法

### Basic Usage (Synthetic Data, Offline)

```bash
python geoskill-wildfire-spread-modeling.py --bbox 116.0 39.0 117.0 40.0 --synthetic --output-dir ./out
```

### More Examples

```bash
python geoskill-wildfire-spread-modeling.py --bbox 116 39 117 40 --synthetic --output-dir ./out
python geoskill-wildfire-spread-modeling.py --input fuel.tif --steps 25 --output-dir ./out
python geoskill-wildfire-spread-modeling.py --bbox 116 39 117 40 --wind-speed 4 --wind-dir 90 --steps 20 --synthetic --output-dir ./out
python geoskill-wildfire-spread-modeling.py --bbox 110 35 111 36 --synthetic --quiet --output-dir ./out
```

## Output / 输出

| File | Format | Description |
|---|---|---|
| `burned_area.tif` | GeoTIFF | Final burned extent |
| `arrival_time.tif` | GeoTIFF | Arrival time step (unburned=-1) |
| `fire_params.json` | JSON | Per-time-step burned area series and parameters |

Each run also produces `output-manifest.json` (run manifest, including inputs/outputs/QA summary).

## Data Source / 数据源 / Source

Real mode reads a multi-band GeoTIFF (band1=fuel 0-1, band2=moisture 0-1, band3=slope 0-1); synthetic mode generates fire scenarios offline.

## Privacy / 隐私声明 / Privacy

- Runs offline by default; `--synthetic` mode requires no network at all.
- All processing is done locally; no user data is uploaded.

## License / License

MIT

---

<!-- ===== 中文原文 (Chinese Original) ===== -->

---
name: geoskill-wildfire-spread-modeling
description: '元胞自动机叠加坡度风速燃料湿度修正的过火面积模拟'
---

# 野火蔓延模拟 | Wildfire Spread Modeling

基于元胞自动机（CA）的野火蔓延模拟：每个时间步，燃烧像元向 8 邻域未燃像元以概率点火，点火概率综合燃料可燃性、湿度（越湿越难点燃）、坡度（上坡更快）与风速（顺风更快，含风向夹角）。一旦点燃永不熄灭，过火面积随时间单调不减；固定随机序列下，风速越大过火越多、湿度越大过火越少。

## 依赖

```bash
pip install numpy rasterio scipy
```

## 使用方法

### 基本用法（合成数据，离线）

```bash
python geoskill-wildfire-spread-modeling.py --bbox 116.0 39.0 117.0 40.0 --synthetic --output-dir ./out
```

### 更多示例

```bash
python geoskill-wildfire-spread-modeling.py --bbox 116 39 117 40 --synthetic --output-dir ./out
python geoskill-wildfire-spread-modeling.py --input fuel.tif --steps 25 --output-dir ./out
python geoskill-wildfire-spread-modeling.py --bbox 116 39 117 40 --wind-speed 4 --wind-dir 90 --steps 20 --synthetic --output-dir ./out
python geoskill-wildfire-spread-modeling.py --bbox 110 35 111 36 --synthetic --quiet --output-dir ./out
```

## 输出

| 文件 | 格式 | 说明 |
|---|---|---|
| `burned_area.tif` | GeoTIFF | 最终过火范围 |
| `arrival_time.tif` | GeoTIFF | 到达时间步（未燃=-1） |
| `fire_params.json` | JSON | 逐时间步过火面积序列与参数 |

每次运行还会产出 `output-manifest.json`（运行清单，含输入/产物/QA 摘要）。

## 数据源 / Source

真实模式读取多波段 GeoTIFF（band1=燃料0-1、band2=湿度0-1、band3=坡度0-1）；合成模式离线生成火场。

## 隐私声明 / Privacy

- 默认离线运行，`--synthetic` 模式完全无网络。
- 所有处理在本地完成，不上传任何用户数据。

## License

MIT