Several large Swedish wildfires during recent decades were caused by forestry machinery in operation, fires for which there is still no characterisation. We combined 18 years of data on dispatches, weather and fire danger and interviewed forestry workers to understand the spatial, temporal and weather distributions of these fires, and their underlying mechanisms. We estimate the average annual number of ignitions from forestry machinery in Sweden at 330–480 (2.0±0.4 ignitions per 1000ha clear-felling) of which 34.5 led to firefighter dispatches, constituting 2.2% of all forest fire dispatches and 40% of area burnt. Soil scarification causes the most ignitions and the main mechanism is likely high-inertia contact between discs and large stones, causing sparks igniting dry humus or moss, countering reports suggesting that such metal fragments cannot fulfil ignition requirements. We found a spatial relationship between forestry machine ignitions and abundance of large stones, represented by a Boulder Index generated from a nationwide dataset. Further, 75% of the dispatches occurred on days with relative humidity <45%, Duff Moisture Code (Canadian system) >26 and Fire Weather Index >12. 75% of the area burned when Fire Weather Index was >20. Results suggest machine-caused forest fires can be largely avoided by cancelling operations in stony terrain during high-risk weather.

The highly managed forests of Fennoscandia are dominated by coniferous (Scots pine and Norway spruce) stands with occasional inclusion of deciduous trees (Betula spp, Populus tremula, Quercus robur). Without active pre-commercial thinning, broadleaved deciduous trees would be more common and it is often assumed this would reduce flammability. Generally there is little information on surface fuel structure in Fennoscandia and its relation to the dominant tree species. We therefore evaluated fuel characteristics in a sample of Swedish mature, closed-canopy forest stands dominated by Pinus sylvestris, Picea abies, Quercus robur, Betula pendula, Populus tremula, as well as some mixed-species stands. Stand properties (tree species and basal area of the stand) were measured on site and fuel bed composition was determined by destructive sampling of two 0.25 m2 plots per stand (species, fuel bed depth, live weight, % coverage). Dry weight was evaluated in a laboratory environment. The measured parameters were used as input to the BEHAVE wildfire model, to assess relative fire behaviour for these different fuel assemblies. The results show that the moss/lichen layer was thicker and had a larger dry weight under pure pine stands compared to that under spruce stands, most likely due to better sunlight penetration to the forest floor in pine stands and structural differences of the needle litter. BEHAVE modelling suggested both fireline intensity and rate-of-spread to be substantially higher for pine stands. As for the broadleaved deciduous tree species, even a small inclusion in the coniferous stands severely affected the moss/litter layer. Leaf litter formed horizontal, multi-layered packs that replaced the porous structure and high surface-to-volume ratio of the moss/litter with a more compacted pure litter layer, resulting in reduced flammability. The BEHAVE models presented here are available at https://www.ri.se/sites/default/files/2021-12/SwedishForestFuels_0.zip.

No abstract

Summary This work aims to explore possible ways of improving the precision of ignition measurements in the cone calorimeter. Both inherent repeatability of parts of the testing equipment and operator‐dependent variations are considered. Inherent repeatability is indicated to be slightly improved if the test samples used are circular rather than square. Operator‐dependent variation is discussed in terms of the method used for determining ignition. Four procedures are compared, namely, visual observation, usage of a light sensor, and looking at the peak of the second and first derivatives of the mass loss and heat release curves, respectively. Results indicate that the preferable operator‐independent method depends on the test conditions; the derivative of the heat release rate is an alternative to the mass loss rate derivative when the scale is of standardised quality. A light sensor for ignition time observation is a good option when the surrounding light is not changed during the test.