Causation Discovery of Weather and Vegetation Condition on Global Wildfire Using the PCMCI Approach

Wildfire is an important process that affects nature environment and living organisms. In this study, we applied a causal network discovery method called PCMCI which performs in a PC condition selection stage to select relevant conditions and a MCI conditional independent test stage to control false positive rate, to detect casual relationships and time lags between wildfire burned area and weather/drought and vegetation conditions. The results show that for grassland, weather and aridity conditions are dominant indicators to BA (burned area). For shrub land, FWI (fire weather index) is dominant. For sparsely vegetated land cover, which is water or fuel limited region, vegetation growth and health conditions are dominant. For broad leaf forests, radiation is the most important indicator. While for needle leaf forests, temperature is dominant.