Description of database variables

 

Censored data

 

Censored data refer to frequency distributions reporting the number or relative frequency of lightning-ignited wildfires for which the exact holdover times are not known but the lower and upper bounds of the time intervals surrounding the holdover times. Holdover time refers to the time elapsed from lightning-induced ignition to wildfire detection.

 

Variable

Format

Description

Study_id

Character

ID code referring to the original study and dataset from which the data were collected. Each frequency distribution of holdover time has a unique ID code.

Reference

Character

In-text citation of the original study from which the data were collected.

Time_interval

Character

Duration of the time interval. Note that not all the time intervals from the same frequency distribution must have the same duration. Thus, the unit might be different for each time interval and is reported after the value.

Time_interval_d

Numeric

Duration of the time interval in days. For example, a time interval of 12 hours = 0.5 days.

Lower_limit_d

Numeric

Lower bound of the time interval in days.

Upper_limit_d

Numeric

Upper bound of the time interval in days. The upper bound of the last interval in some frequency distributions is unknown.

N_fires

Integer

Number of lightning-ignited wildfires with an estimated holdover time within the time interval. Some studies did not report the specific number of wildfires per time interval.

RF

Numeric

Relative frequency of lightning-ignited wildfires in the time interval.

CRF

Numeric

Cumulative relative frequency of lightning-ignited wildfires for each frequency distribution of holdover time.

Original_data

Character

How the frequency distribution was reported in the original study. N = number of wildfires; P = relative frequency.

Data_location

Character

Where the data were reported within the original study. For example, a specific figure or table.

Collection_method

Character

Method used to collect the data. Copied from original = the data were copied from texts, tables, figures or appendices; WebPlotDigitizer = the data were extracted from figures using the WebPlotDigitizer tool (https://automeris.io/WebPlotDigitizer/); Personal communication = the data were provided by one of the authors of the original study.

 

Non-censored data

 

Non-censored data refer to estimated values of continuous holdover time for each single lightning-ignited wildfire. Nine datasets of non-censored data are provided (ID codes of the studies are included). All the values of holdover time are reported in hours. Note that each dataset has a different number of lightning-ignited wildfires for which the holdover times were estimated.

 

Ancillary data

 

Ancillary data refer to additional information describing and summarizing the studies from which the holdover times were estimated.

 

Variable

Format

Description

Data identification-related variables

Study_id

Character

ID code referring to the original study and dataset from which the data were collected. Each frequency distribution of holdover time has a unique ID code.

Reference

Character

In-text citation of the original study from which the data were collected.

Publication_type

Character

Type of original study from which the data were collected. Paper = peer-reviewed article published in a scientific journal; Proceeding = article published in the context of a scientific conference; Report = scientific or technical report; Thesis = Bachelor, Master or Doctoral thesis; Unpublished = unpublished data.

Spatial variables

Study_area

Character

Territory in which the lightning-ignited wildfires were studied.

Country

Character

Country of the study area.

ISO_code

Character

Country code or region code of the study area according to ISO 3166 (https://www.iso.org/iso-3166-country-codes.html). Some study areas do not fit a unique particular subdivision of a country but a part of it or several regions.

Spatial_scale

Character

Spatial extent of the study area. Local < 10,000 km2; Regional = 10,000-1,000,000 km2; Continental = 1,000,000-100,000,000 km2; Global > 100,000,000 km2.

Biome

Character

Most common biome of the study area according to Olson et al. 2001 (https://doi.org/10.1641/0006-3568(2001)051[0933:TEOTWA]2.0.CO;2).

Ecozone

Character

Biogeographic realm of the study area according to Olson et al. 2001 (https://doi.org/10.1641/0006-3568(2001)051[0933:TEOTWA]2.0.CO;2).

Climate_class

Character

Most common climate class of the study area according to the Köppen-Geiger climate classification by Beck et al. 2018 (https://doi.org/10.1038/sdata.2018.214). The climate class must be considered as indicative since the study areas have different climate classes, and wildfire coordinates are not available to attribute properly a climate class to each study area.

Temporal variables

Start_year

Integer

Starting year of the study period.

End_year

Integer

Ending year of the study period.

Length_year

Integer

Length of the study period in years.

Min_time_h

Numeric

Minimum time interval in hours of the censored data.

Max_time_h

Integer

Maximum time interval in hours of the censored data.

Fire-related variables

Number_fires

Integer

Total number of lightning-ignited wildfires for which the holdover times were estimated. Often the holdover time was not estimated for every lightning-ignited wildfire studied.

Number_records

Integer

Total number of time intervals in which the frequency distribution of holdover time data was divided.

Fire_detection

Character

Source of wildfire discovery data. Fire database = information on cause, location and time of wildfire detection comes from a wildfire database; Remote sensing = information on cause, location and time of wildfire detection comes from remote sensing data, (e.g., satellite images).

Fire_data_source

Character

Dataset used to extract lightning-ignited wildfire data. For remote sensing data, the name of the instrument or the data product may be included between the parenthesis. ABoVE = Ignitions, Burned Area, and Emissions of Fires in Alaska, Yukon and Northwest Territories; FIRMS = Fire Information for Resource Management System; MODIS = Moderate Resolution Imaging Spectroradiometer; VIIRS = Visible Infrared Imaging Radiometer Suite.

Lightning-related variables

LLS

Character

Lightning Location System (LLS) used to obtain lightning data. LLSs are usually composed of a network of ground-based sensors. ALDIS = Austrian Lightning Detection and Information System; ALDN = Alaska Lightning Detection Network; CGLSS = Cloud to Ground Lightning Surveillance System; CLDN = Canadian Lightning Detection Network; ENTLN = Earth Networks Total Lightning Network; EUCLID = European Cooperation for Lightning Detection; GLD360 = Global Lightning Detection Network; GPATS = Global Position and Tracking System ; LLP = Lightning Location and Protection Detection System; MSCLLS = Meteorological Service of Catalonia Lightning Location System; NLDN = National Lightning Detection Network; OMNRLLS = Ontario Ministry of Natural Resources Lightning Location System; STARNET = Sferics Timing and Ranging Network; WWLLN = World Wide Lightning Location Network.

Lightning_level

Character

Level at which lightning data were matched with wildfire data: “Stroke” and “Flash”. A Cloud-to-Ground (CG) stroke involves a downward leader channel and a return stroke of the discharge, while a CG flash is a cluster of subsequent CG strokes.

DE_pct

Character

Detection Efficiency (DE) refers of the expected percentage of lightning discharges reported by the LLS. Often studies distinguish between DE of CG flashes and CG strokes.

LA_km

Character

Location Accuracy (LA) in km or lightning location error usually refers to the expected median error between the reported CG stoke locations by the LLS and the real ground strike points. However, studies may also refer to lightning LA as the median value of the semi-major axis of 50% probability confidence ellipses of CG strokes.

Methodology-related variables

Method

Character

Methodological approach used to estimate holdover times of lightning-ignited wildfires. Storm time = elapsed time between discovery of the wildfire and the most recent lightning storm over the area of the wildfire (analogous to the minimum holdover time). Formerly this time was often reported by personnel from lookouts. Lightning match = at least one lightning is identified as the probable ignition source by searching within the lightning data obtained from a LLS.

Buffer_distance_km

Character

Maximum buffer radius in km around the wildfire ignition point used to select potential igniting lightning.

Temporal_window_d

Character

Maximum temporal window backward in days from the wildfire detection time used to select potential igniting lightning, (i.e., the maximum holdover time allowed).

Max_holdover_d

Character

Maximum estimated holdover time in days. The maximum estimated holdover time cannot be longer than the maximum temporal window applied.

Selection_criteria

Character

Criteria used to select the most likely igniting lightning. These criteria include: “Minimum holdover time”, “Daily minimum distance”, “Maximum proximity index”, and “Decision tree”. See Moris et al. 2020 (https://doi.org/10.1016/j.agrformet.2020.107990) for a brief description of common methods to find igniting lightning.

Data entry-related variables

Dataset

Integer

Reference number used to distinguish the holdover time distribution in case different methods were applied to the same dataset for estimating holdover times within the same original study.

Data_collector

Character

Person who filled the data records.

Date_entry

Character

Date on which the data records were filled. Format dd/mm/yyyy.

Data_check

Character

Whether or not the data records were double-checked by a different person from the one who filled the data records: “Yes” or “No”.

Comments

Character

Additional notes about the data or original study. Several comments may be included.

 

References

 

Complete list of references included in the database.

 

Variable

Format

Description

Study_id

Character

ID code referring to the original study and dataset from which the data were collected. Each frequency distribution of holdover time has a unique ID code.

Citation

Character

Complete citation of the original study from which the data were collected.

DOI

Character

Permanent web address (URL) associated with the Digital Object Identifier (DOI) of the original study from which the data were collected. Some studies do not have any DOI.