In temperate Europe, fire is already here: The case of The Netherlands

Landscape fires are usually not associated with temperate Europe, yet not all temperate countries record statistics indicating that actual risks remain unknown. Here we introduce new wildfire statistics for The Netherlands, and summarize significant events and fatalities. The period 2017–2022 saw 611 wildfires and 405 ha burned per year, which Copernicus’ European Forest Fire Information System satellite data vastly underestimate. Fires burned more heathland than forest, were small (mean fire size 1.5 ha), were caused by people, and often burned simultaneously, in Spring and in Summer drought. Suppression, restoration and traffic delays cost 3 M€ year−1. Dozens of significant events illustrate fire has never been away and has major societal impact amidst grave concerns for firefighter safety. Since 1833, 31 fatalities were reported. A legal framework is needed to ensure continuity of recordkeeping, as the core foundation of integrated fire management, to create a baseline for climate change, and to fulfill international reporting requirements. Supplementary Information The online version contains supplementary material available at 10.1007/s13280-023-01960-y.


Appendix S1
Next-to-roadside fires Fires classified as verge fires or next-to-road fires (bermbrand) are not automatically included in the list of wildfires sourced from the Dutch national dispatch system, given that an internal analysis showed that the very large number of next-to-road fires in the dispatch system (>3000 in 2018 alone) often do not concern wildfires but other fires like a burning shrub as a result of a person burning weeds growing between their driveway pavement.Because some of these roadside fires are actual wildfires, we count next-to-road fires in the dispatch system only if the data show that the fire service scaled up the fire and requested more resources upon arrival.
Detailed methodological information regarding fire locations, and classification of fire size, vegetation type and fire cause -Location: location information in the Dutch national emergency dispatch system concerns the name of the village, town or city, as well as the nearest street.For fires that occur away from built up areas the actual location of the burned site may therefore be hund reds of meters or even a few kilometers away from the location registered in the database.
-Fire size: because most fires are very small, fire size is not determined through satellite analysis but estimated from photo and video material, or in the field , although this is not systematically done for all fires.For ~60% of fires, size information is not available.
-Vegetation type: in line with the estimate of fire size, the vegetation type affected is based on field observation by local fire service personnel, ground-based imagery and news reports, mostly for the larger fires (fires that exceed ~100 m 2 ).

Quality checks to ensure completeness of data collection
Regarding the data obtained from the Dutch national dispatch system, three main procedures are used to ensure all relevant fires are included in the database.In the Netherlands, fire incidents are reported nationally in the Gemeenschappelijk Meldkamer Systeem (GMS), that stores raw data for each incident, and in which the operator assigns labels to fire incidents.GMS is therefore the place where fires are classified as landscape fires or not.Incidents in GSM are publicly communicated in real -time through the so-called P2000 network which is accessible via a range of user interface websites.Since 2020, a second quality check is done through a private Telegram group that automatically screens the P2000 records for relevant fires.Additional information about these fires is subsequently obtained from GMS and from any media coverage summarized by OBI4wan.This may include a possible cause, and information about the possible size (described in the text and/or interpreted from accompanying photo or video material).In the case that public information about a fire is contradicting, an internal fire service quality check is done.Finally, since 2019 a final check of relevant fires is done at the end of each calendar year, when the manually sourced data is cross checked with raw data from GMS.The reason this check is done is to capture fires that were erroneously labeled as a wildfire, as GMS will have the corrected label while errors are not corrected in P2000.On the other hand, GMS may inadvertently contain wildfire trainings, which are removed from our database through this final c ross check.Since we did not have access to GMS in 2017 and 2018 the database may slightly underestimate the total number of fires in those years.Based on our cross checks since 2019 we expect this likely concerns < 25 fires.
In addition to the abovementioned internal steps, since 2021 we also ask the local fire services to report information, as a quality check and also to understand more about the fire itself .This includes information on the vegetation type it burned in, the coordinates of the site and the possible origin, fire size, fire behavior including spot fires, whether the fire jumped any roads, waterways or other fire breaks, and any fire cause and origin investigation that was carried out.This questionnaire is currently sent out v ia email or embedded in the fire investigation documentation; response rate is moderate at 30% and plans are made to make the questionnaire more accessible, for instance by making it available on the tablets inside each fire engine.

Comparison of current fire statistics with pre -1994 data
The annual number of fires between 2017 and 2022 was 611 ± 306 (from 212 to 949 fires each year), approximately double the number of wildfires counted between 1945 and 1994 (248 ± 181 fires, ranging from 51 to 904 per year).Because of a higher proportion of fires < 1 ha in the current dataset (89% now, vs. 77% pre-1994), this higher number of fires in the most recent statistics may be explained by the fact that the pre-1994 reporting excluded fires that were stopped when still very small whereas current reporting considers all wildfires regardless of size.Another reason why the current data collection shows larger fire numbers may be that the pre-1994 statistics reportedly underestimated fires outside of forests (IKC , 1995).Regarding burned area, which is often dominated by a small number of very large fires, differences between current data (405 ± 385 ha, from 18 to 1073 ha) and pre-1994 (699 ± 810 ha, from 34 to 3653 ha) were small.Individual fire size of current and past statistics was also similar, with 2% and 2.6% of fires being larger than 10 hectare, respectively (Fig. 2c, 5i).

Table S1
For 40 days between 2017-2022, ≥15 fires occurred; data sorted by date.Three series of (nearly) consecutive days in Aug 2018, Apr 2020 and Sept 2022 are highlighted in bold.