Tree Cover Loss Due to Fires in Argentina: Provincial-Level Analysis and Control Variables (2001–2023)

This analysis uses data from Global Forest Watch (GFW), a platform that monitors forest changes worldwide using satellite imagery and geospatial data. Tree cover is defined as all vegetation (natural forests or plantations) at least five meters tall, with varying levels of canopy density. It represents the presence of trees within a geographically defined area at a spatial resolution of 30 meters and may include both natural forests and tree plantations. According to GFW, dry forests and savannas are represented using canopy thresholds below 10 percent, while dense tropical forests such as the Amazon are mapped using thresholds above 50 or 75 percent. However, for cross-country comparative studies, the platform recommends using thresholds above 30 percent. We adopted these thresholds for Argentina and Brazil. Tree cover loss refers to any activity that results in the complete removal of the tree canopy. These activities may include commodity-driven deforestation (large-scale loss linked to commercial agricultural expansion), shifting agriculture (temporary or permanent loss due to small- and medium-scale farming), forestry (temporary loss due to plantations and harvesting of natural forests, including partial deforestation of primary forests), forest fires (temporary loss in the initial forest disturbance), and urbanization (deforestation for urban expansion). The specific variable used in this analysis, tree cover loss due to fires, does not include forest fires that result in canopy loss at a pixel scale smaller than 30 meters, which are classified as low-intensity fires by GFW. The data represent the initial forest disturbance that replaces the stand during the 2001 to 2023 period. This excludes subsequent fires, such as the burning of trunks after forest clearing for agriculture, or fires that occur following forest regrowth after the initial disturbance within the study period. The dataset is an updated version of the tree cover loss data produced by Global Forest Watch, which incorporates several methodological improvements, including the integration of Landsat 8 imagery in 2013 and algorithmic updates introduced in 2011 and 2015. These changes may affect comparability with earlier datasets. In addition to the main variable, the database includes several control variables that help contextualize environmental change across provinces. These include: the existence of tree cover with canopy density greater than 30 percent in the years 2000 and 2010 (measured in hectares), based on data from Global Forest Watch; average annual provincial daily precipitation (in millimeters), from the National Meteorological Service (Servicio Meteorológico Nacional), based on requested data; average daily provincial temperature, calculated as the mean between maximum and minimum daily temperatures (in degrees Celsius), also from the National Meteorological Service; year-to-year variation in provincial temperature (in degrees Celsius); and year-to-year variation in provincial precipitation (in millimeters), both obtained from the National Meteorological Service. Additionally, the analysis includes fire hotspots as a complementary variable. These refer to satellite-detected points of high thermal activity that indicate active fire events. Fire hotspots help capture the intensity and distribution of fire activity over time and serve as a relevant indicator of forest fire pressure in combination with tree cover loss data. Together, these variables provide a robust framework to examine spatial and temporal patterns of forest disturbance, enabling a deeper understanding of the role of climatic and anthropogenic factors in shaping deforestation dynamics

本分析采用来自全球森林观察（Global Forest Watch，GFW）平台的数据，该平台借助卫星影像与地理空间数据监测全球森林动态变化。

树冠覆盖指所有高度至少5米的植被（天然林或人工林），冠层密度存在差异。该数据以30米空间分辨率表征地理区域内的树木存在情况，涵盖天然林与人工林。据GFW说明，干旱林与稀树草原采用低于10%的冠层阈值进行表征，而亚马孙等稠密热带森林则采用50%或75%以上的阈值进行制图。不过针对跨国比较研究，该平台推荐使用30%以上的阈值。本研究针对阿根廷与巴西采用了上述阈值。

树木冠层损失指所有导致冠层完全移除的活动，此类活动包括：商品驱动型森林砍伐（与商业农业扩张相关的大规模森林损失）、轮作农业（因中小型农业活动导致的临时或永久森林损失）、林业活动（因人工林经营与天然林采伐导致的临时损失，包括原始林的部分砍伐）、森林火灾（初始森林扰动阶段的临时损失）以及城市化（城市扩张引发的森林砍伐）。

本分析采用的特定变量为火灾引发的树木冠层损失，该变量不包含像素尺度小于30米的冠层损失型森林火灾，此类火灾被GFW归类为低强度火灾。本数据集表征的是2001年至2023年间，替代原有林分的初始森林扰动事件，排除后续发生的火灾，例如为农业开垦而清理森林后对树干的焚烧，或是研究期内初始扰动后森林再生长阶段发生的火灾。

本数据集是全球森林观察发布的树木冠层损失数据的更新版本，纳入了多项方法学改进，包括2013年Landsat 8影像的集成，以及2011年与2015年推出的算法更新。此类改进可能会影响与早期数据集的可比性。

除核心变量外，该数据库还包含多项用于厘清各省环境变化背景的控制变量，具体包括：2000年与2010年冠层密度大于30%的树冠覆盖面积（单位：公顷），数据源自全球森林观察；各省年平均每日降水量（单位：毫米），数据源自国家气象局（Servicio Meteorológico Nacional），为申请获取的公开数据；各省每日平均气温，以每日最高与最低气温的均值计算（单位：摄氏度），同样源自国家气象局；各省气温的年际变化量（单位：摄氏度）；各省降水量的年际变化量（单位：毫米）。后两项变量均取自国家气象局。

此外，本分析还纳入火点作为补充变量。火点指卫星探测到的指示活跃火灾事件的高热活动点位。火点有助于捕捉火灾活动的强度与时空分布特征，并可与树木冠层损失数据结合，作为森林火灾压力的有效指示因子。

上述变量共同构建了一套严谨的分析框架，用于探究森林扰动的时空格局，从而更深入地理解气候与人为因素在塑造森林砍伐动态中的作用。