Monthly burned area and forest fire carbon emission estimates for the Russian Federation from SPOT VGT

Abstract

Russian boreal forests contain around 25% of all global terrestrial carbon, some of which is released to the atmosphere when the forests burn. Whilst it is well known that fire is widespread in the boreal environment, there is a lack of good quality quantitative data on the extent of fire activity in Russian forests and on its interannual variation. This study provides one of the first comprehensive monthly satellite-based studies of fires occurring across the entire Russian Federation using a single, standardised methodology designed to map burned areas down to a size of 2 km 2. Using data from SPOT VEGETATION (VGT), we detect newly burned pixels via a series of multi-temporal spectral reflectance differencing criteria. For the year 2001, the method is applied to 21 VGT 10-day syntheses (S10) scenes covering the Russian fire season. We map 2764 fires with a total area of 41,782 km 2, and our methodology successfully detects all fires present in a comparison Landsat ETM+ data set, although it underestimates their size by on average of 18%. Using frequency–size relations, we estimate that 3790 fires of 1–2-km 2 area are likely to have remained unobserved by our method across the entire Russian region. Taking these corrections into account, we calculate the total burned area for the Russian Federation in 2001 as 51,546 km 2, with 38,512 km 2 occurring in forest and 13,034 km 2 in other land use classes. Fire activity is strongest in August in Eastern Siberia and the northern part of the Russian Far East, and in May and October in the southern part of the Russian Far East. Using these data, we estimate direct carbon emissions from these Russian forest fires to be 39.3–55.4 Mt, five to eight times that from the 2001 North American boreal forest fires and around 11–17% of that year's Russian industrial carbon emissions. This methodology will, in the future, be applied to the full VGT archive to quantify burned area and direct carbon emissions over a 5-year period in order to better assess the interannual variation in burned area and emissions and the relation to local climate.