Evidence of human activity near a fire site does not provide conclusive proof that the fire itself was anthropogenic. However, the presence of objects such as roads or logging sites close to the starting point of a fire does indicate a high probability that the fire had a human cause, particularly in regions with such a low population density.

No, the total could be greater. The map only includes the largest fires from 2018 (those with an area of more than 1,000 hectares). Also, not all traces of human activity can be seen through satellite images. By the same token, we cannot be certain that every fire that began close to a site of human activity was anthropogenic (see above).

Probably not. The map was produced by our experts manually, using visual decoding of satellite images, so it is possible that some objects have been missed. If you have noticed such a case, please let us know, and we will supplement our data.

At the moment there is no accurate data on lightning strikes, and it is not possible - using the available data - to develop a uniform methodology to account for them in this research. It is possible some wildfires are caused by dry thunderstorms, but it is not possible to identify which ones.

These are the regions of Russia that have been most exposed to forest fires (on average over the last decades).

We intentionally did not combine the results of the study in four regions into one with the average conclusions, because this would not allow us to explore the specific issues facing each region. However, there is no reason to believe the situation is radically different in other Siberian regions. The regions analysed in this study have a particularly low population density - in principle, the proportion of wildfires caused by humans should be higher in other, more densely populated areas.

The map combines two types of satellite image. The first is a mosaic of pictures from the satellite Sentinel-2, for the period from August 1, 2018 to October 1, 2018, with the least cloudiness (for readability) in the synthesis of True Color and SWIR channels. In most cases, these pictures were taken before the snow fell but after the end of the fires, therefore active burning on these pictures is not visible, but the final edges of fires are shown. Sometimes there were no suitable low-cloud shots for the desired date, in the case we took earlier shots. As a result some of the images may not show some fires or those fires don’t have final edges (for the analysis itself, the cloudiness threshold was set significantly lower). These images clearly show logging sites and other large man-made objects, the resolution of this data is 10 meters for True Color synthesis, 20 meters for SWIR synthesis.

The second type of satellite images used is a mosaic of ESRI high resolution satellite imagery. These are images with a resolution of about one meter, and smaller objects are visible on them (small roads, detached buildings). However, these images were taken long before the period covered by the analysis (the exact date can be found out only for each specific fragment, it may be from several months to more than a year ago) and there may be some objects missing because they appeared after the image was taken (for example, new logging sites). Also these images may show fires active at the time of the shot or fires from previous years. For the analysis we used all available high and ultra-high resolution images and took into account their temporal resolution.