Scottish Fire and Rescue Service (SFRS) - wildfire: incident reporting system - data analyses

9 Synthesis & recommendations

9.1 Synthesis & overview

The Incident Reporting System (IRS) is a very well-structured dataset and a valuable resource for keeping records of the characteristics of different types of fires occurring in Scotland. However, our analysis found that it is extremely difficult to use information recorded in the IRS to run an automated filtering and identification of wildfires, and hence we had to resort to extensive visual inspection of satellite imagery to complement the filtering process. This is due to uncertainties in the recording of fire incident locations and ambiguities in the description of fuel types, with regards to the type of vegetation communities or habitats they describe. We attempted to consider these issues to a certain degree in our analysis by using different datasets of habitat types and land use classes at different scales and resolutions. However, the results of comparing locations and fuel type information from the identified IRS wildfires with respective information derived from burnt area datasets showed that there are extensive differences and disparities in fuel type composition between actual burnt areas and IRS records.

Nevertheless, the selected wildfires from IRS were identified and selected using a robust filtering process and were verified with extensive visual inspection; hence we are confident that wildfires identified are indeed wildland fires affecting seminatural habitats. Around 7.3% (9,725 fires) of the outdoors IRS fire incidents for the 2009-2020 period were identified as wildfires; we cannot ascertain whether this number of wildfires is realistic for the studied period or an underestimation of wildfire activity in Scotland because we are lacking robust estimates of wildfires occurrence. It has been suggested previously that recorded fire incidents are likely to be biased towards smaller burns in more accessible and/or more urban areas (Davies and Legg, 2016); this is because reporting of fires in rural upland areas is likely to be incomplete for social reasons (e.g., due to reluctance of gamekeepers and land-managers in areas where traditional managed burning is still prevalent to involve the SFRS) or due to sheer remoteness. However, in this study we identified several big wildfires within the IRS that caused extensive damage to important seminatural habitats in remote rural areas, and in several cases we also managed to associate them with their respective burnt areas. These results provide an indication that, despite issues mentioned above, the IRS dataset is a good resource for identifying important wildfire incidents in Scotland, even in the most remote areas.

Our extensive investigation of patterns of wildfire occurrence identified three (3) main types of wildfires occurring in Scotland: a) Wildfires caused in very remote and remote rural areas affecting mainly heathlands/shrublands and bogs and peatlands that seem to be caused mainly by accidental ignitions; b) Wildfires in accessible rural areas and smaller urban centres that affected grasslands and woodlands; and c) Small fires close to settlements and urban centres affecting mainly grasslands that seemed to be caused by deliberate ignitions. Wildfires in remote areas tended to be bigger and were mainly caused by bonfires or other intentional burns that got out of control, which might be associated with traditional land management practises or tourism/recreational activities. Small deliberate fires in or close to urban centres seemed to be driven mainly by anti-social behaviour or relatively trivial cases of negligence with fire, while the causes of fires in the more accessible rural areas could be a mixture of burns getting out of control or careless handling of equipment or other heat sources along with ignitions from anti-social behaviour.

Despite differences in fuel type compositions, burnt area sizes, motives of ignition and accessibility, wildfires in Scotland exhibit a clear seasonal pattern with most fire incidents occurring during the months in spring, with most monthly fires recorded in April. This temporal pattern was common in other countries with similar bioclimatic conditions and vegetation communities (Republic of Ireland, rest of the UK and west coast of Norway). However, we need to consider that predicted warmer summers and increased frequencies of drought in Scotland could cause the occurrence of big and particularly damaging wildfires also in summer months, as is the case in central parts of Sweden. Finally, our Highland LA case study showed that the IRS wildfires can be a useful resource for ignition risk modelling and mapping using statistical methods. Maps of ignition risk classes and/or predicted fire locations could be useful tools when used alongside other datasets related to fire danger and risk for enhancing wildfire suppression and mitigation capability.

9.2 Recommendations

Based on the results of our analysis, we propose a set of recommendations with the aim to improve the utility of the IRS dataset for detecting wildfires and to make IRS fire incidents a more relevant and reliable resource for wildfire planning and decision making in Scotland and for supporting wildfire research and analysis. Recommendations for improvement proposed are related mainly to describing and characterising fuels and recording wildfire location in relation to point of ignitions and distances to actual burnt areas.

Although the IRS provides a comprehensive list of categories for recording fuel types, these are currently not appropriate for easily identifying wildfires as they can be ambiguous as to specific vegetation types or communities impacted by respective fire incidents. For example, we found that the "Grassland, pasture, grazing etc" property type includes fires on seminatural, grazed or improved grasslands and that fires belonging to the "Grassland, pasture, grazing etc" or the "Scrub land" property types can include grassland, heathland, or bog fires. This is problematic because different habitat or vegetation communities have different structure which influences fire behaviour. In addition, the IRS doesn't include a peatlands category, even though bog and peatland fires are quite frequent and very damaging to particularly sensitive and important habitats. Therefore, adopting a new, comprehensive system for reporting and describing fuels could greatly improve both the identification of wildfires in the IRS, but also its value as a resource for wildfire planning and analysis.

In our opinion the National Vegetation Classification (NVC) Phase 1 system or the EUNIS habitat classes of Table 4.2 would be a potentially good option for broadly describing fuels within the IRS, as both systems strike a good balance between ecological specificity and moderate training and natural history skill requirements and could be relatively easily adapted for fuels. An implication of this approach is that there are significant differences in fire behaviour across stages of the Calluna (heather) cycle (relevant to fire behaviour in peatlands and heathlands), so this information would probably need to be incorporated in the system. In this respect, our recommendation is that we need to a) collate all available fuels information collected across multiple projects and conduct analyses to reliably classify fuels based on structure and predicted fire behaviour; and b) compile this information into an illustrated fuels guide complete with predicted fire behaviours that could be used to support training of a select group of SFRS personnel. This approach could be trialled first in regional SFRS offices where the more damaging wildfires are expected to occur to assess its feasibility and identify weaknesses or knowledge gaps.

In addition, getting an accurate read on fire location is quite important, especially for determining the ignition point that can be helpful in trying to understand fire causation. A first step could be to add a new field in the IRS stating whether the recorded location is the ignition point or some other location, e.g., where the fire resources were marshalled. In addition, at least for the bigger wildfires, it would be helpful to record multiple locations at the fireline or burnt area perimeter. This information could be used by SFRS staff to quickly draw rough polygons of burns using free and easy to use web mapping tools (e.g., OpenSteetMap: https://www.openstreetmap.org) and use them to calculate burnt area size to better inform the Outdoor Damage Area fields in the IRS, which currently are either incomplete or partially inaccurate. Moreover, the recorded locations could be used by others working with fires (e.g., NatureScot staff or researchers) to help with detecting and digitising burn scars using satellite imagery to refine calculations of burnt area sizes and conduct further analysis (e.g., for fire severity). During our analysis we encountered difficulties when trying to detect burnt areas that corresponded to IRS wildfires by just using the recorded location information. We have shown that using basic GIS skills it is feasible to conduct fairly accurate delineations of burnt areas. At the same time, NatureScot staff are currently digitising some wildfire extents from analysis of Sentinel imagery. It is suggested that SFRS liaises with NatureScot for flagging IRS wildfire incidents that can be further analysed and characterised with the aim to inform and improve IRS records, as in the case of the burnt/damage area estimates.