Ukraine Fire Perimeters 2023

The data set contains the fire perimeters for Ukraine for 2023 sourced from remote sensing data. The locations and dates of burning were detected using MODIS/VIIRS products on thermal anomalies. This information was used to select pre- and post-fire Copernicus Sentinel 2 L2A imagery. Satellite images were screened from clouds, cloud shadows and combined into median composite mosaics. The image mosaics were created using all available imagery within a 14-day time window before (pre-fire mosaic) and after (post-fire mosaic) each fire. Fire perimeters were visually delineated by comparing post and pre-fire image mosaics using a SWIR2–NIR–Red band combination. The spatial accuracy of fire perimeters corresponds to Sentinel 2 data at 20-m spatial resolution.

Delineated fire perimeters were intersected with the Copernicus Dynamic Land Cover map at 100 m resolution (v.3.0.1) to extract burned areas of five land cover classes according to the following reclassification scheme of the original pixel values.

Land cover class

Original pixel values

Coniferous forest (LC_1)

111

Broadleaved forest (LC_2)

112,113,114,115,116

Other natural landscape (LC_3)

121,124,125,126, 20,30,60,90

Agricultural land (LC_4)

40

Settlement (LC_5)

50

MetaData

Field name

Description

id
Unique identifier of each fire perimeter

longitude
Longitude referring to centroids of fire perimeters

latitude
Latitude referring to centroids of fire perimeters

date
Date of burning in the format DD.MM.YYYY

year
Year

month
Month

day
Day of the month

week
Week number in the year

reg
The capital city name of the Ukrainian regions (oblast) associated with each fire

occupied
A binary indicator variable referring to the locations of each fire within the Russian-occupied territory (1) or within the territory controlled by the Government of Ukraine (0) for a given date

buff_30km
A binary indicator variable referring to the location of each fire within a 30-km buffer zone on both sides of the front line. In contrast to a daily progression front line used to detect "occupied" territory, the buffer was calculated using the farthest position of a frontline towards territories controlled by the Government of Ukraine

emerald
A binary indicator variable referring to locations of each fire within the Emerald network

 Burned areas by land cover (LC) type (in hectares) within the fire perimeter according to the Copernicus Dynamic Land Cover map at 100 m resolution (v.3.0.1)

LC_0
N/A

LC_1
Coniferous forest

LC_2
Broadleaved forest 

LC_3
Other natural landscape

LC_4
Agricultural land 

LC_5
Settlement

本数据集涵盖2023年乌克兰境内的火灾边界数据，数据源自遥感观测资料。研究人员通过热异常的MODIS/VIIRS产品识别出火灾发生的位置与日期，以此选取火灾前后的哥白尼（Copernicus）哨兵2号（Sentinel 2）L2A级影像。随后对卫星影像进行去云、去除云阴影处理，并合成为中值合成镶嵌图。影像镶嵌图的制作基于每场火灾发生前14天时间窗口内的所有可用影像（火灾前镶嵌图），以及火灾发生后14天时间窗口内的所有可用影像（火灾后镶嵌图）。研究人员通过对比火灾前后镶嵌图的SWIR2–NIR–Red波段组合，人工勾绘得到火灾边界。该火灾边界的空间精度与20米空间分辨率的哨兵2号数据一致。

将勾绘完成的火灾边界与分辨率为100米的哥白尼动态土地覆盖地图（v.3.0.1）进行空间相交分析，依据原始像素值的重分类规则，提取5类土地覆盖类型的过火面积。

### 土地覆盖分类与对应原始像素值
1. 针叶林（LC_1）：原始像素值为111
2. 阔叶林（LC_2）：原始像素值为112、113、114、115、116
3. 其他自然景观（LC_3）：原始像素值为121、124、125、126、20、30、60、90
4. 农用地（LC_4）：原始像素值为40
5. 居民地（LC_5）：原始像素值为50

### 元数据字段说明
| 字段名 | 字段描述 |
| --- | --- |
| id | 每个火灾边界的唯一标识符 |
| longitude | 火灾边界质心的经度坐标 |
| latitude | 火灾边界质心的纬度坐标 |
| date | 火灾发生日期，格式为DD.MM.YYYY |
| year | 火灾发生年份 |
| month | 火灾发生月份 |
| day | 火灾发生当月的具体日期 |
| week | 火灾发生所在的年度周数 |
| reg | 对应每场火灾所在乌克兰州（oblast）的首府名称 |
| occupied | 二元指示变量，用于标识指定日期下每场火灾的位置：1代表俄罗斯占领区域，0代表乌克兰政府控制区域 |
| buff_30km | 二元指示变量，用于标识每场火灾是否位于战线两侧30公里缓冲区内。与用于识别“占领区”的每日战线进展不同，该缓冲区基于战线向乌克兰政府控制区推进的最远端位置计算得到 |
| emerald | 二元指示变量，用于标识每场火灾是否位于翡翠网络（Emerald network）范围内 |

### 按土地覆盖类型划分的过火面积
根据分辨率为100米的哥白尼动态土地覆盖地图（v.3.0.1），提取火灾边界内按土地覆盖类型（LC）划分的过火面积，单位为公顷：
| 字段 | 类型说明 |
| --- | --- |
| LC_0 | 无数据（N/A） |
| LC_1 | 针叶林 |
| LC_2 | 阔叶林 |
| LC_3 | 其他自然景观 |
| LC_4 | 农用地 |
| LC_5 | 居民地