Wildfire is a complex Earth system process influencing the global carbon cycle, the biosphere and threatening the safety of human life and property. There are three prerequisites of wildfire: fuel availability, an ignition source and special atmospheric conditions to spread the wildfire. Vegetation, hydrologic and atmospheric conditions are considered as influential conditions by providing fuels, fire preconditions and intensifying fires. It is necessary and urgent to improve our understanding of wildfire in order to predict their occurrence. Many studies focused on wildfire prediction or mapping through regression or machine learning methods. Typically, these studies were limited to regional scales, considered an insufficient number of wildfire conditions, and neglected information about the time lag between wildfire and related conditions and therefore provided only inaccurate predictions. In this study we applied the PCMCI approach, a causal network discovery method which in a first stage identifies relevant conditions based on PC (Peter and Clark) conditions, in a second stage a MCI (Momentary Conditional Independence) conditional independent test is used to control false positive rates, to detect casual relationships and reveal time lags between wildfire burned area and atmospheric, hydrologic as well as vegetation conditions. We built the causal networks for each subregions (28 climate zones and 8 vegetation types) globally. The results show that at global scale atmospheric and hydrologic conditions are usually dominant for wildfires, while vegetation conditions show importance in several special regions, e.g. Africa near the equator and middle-high latitudes regions. The time lags between wildfires and vegetation conditions are larger than those of atmospheric and hydrologic conditions which could be related to vegetation growth and fuels accumulation. Our study emphasizes the importance of taking vegetation monitoring into account when predicting wildfires especially for longer lead time forecasts, while for atmospheric and hydrological conditions shorter time lags should be focused on.