1. PUBLICATION CORRESPONDING TO THESE DATA

 

Sloan, Sean*; Locatelli, Bruno; Andela, Niels; Cattau, Megan E.; Gaveau, David; Tacconi, Luca. 2022 ‘Declining Severe Fire Activity on Managed Lands in Equatorial Asia’. Communications Earth & Environment. DOI: 10.1038/s43247-022-00522-6.

 

*Corresponding author email: sean.sloan@viu.ca

 

 

2. ABSTRACT OF THE DATA

 

The GIS data and corresponding attribute data described here pertain to discrete fire events, their severity, and their ignitions, as derived on the basis of daily MODIS Collection 6 MCD14ML active-fire detections (AFDs).  Data on fire events and their ignitions are provided separately, as two data files.  These data files on fire events and ignitions may however be linked to each other by the data user.  Fire-event severity is quantified per fire event and reported in the data file for fire events.

 

A fire event is a cluster of MODIS Collection 6 MCD14ML active-fire detections (AFDs) wherein each AFD has a spatial (<=1-km) and temporal (<=4-day) proximity to another AFD in the same fire event, inferring thus a relational co-occurrence amongst AFDs in time and space.  In other words, a fire event is considered a likely occasion of burning wherein all constituent AFDs are related to each other in time and space, either directly (as for proximate AFDs) or indirectly (as in the case of a large area of fire activity that spread progressively over time and space from an initial source).

 

Each fire event has a designated ignition AFD, being the AFD of the fire event with the earliest detection date.  A given fire event can have more than one ignition AFD if the ignitions all share same earliest detection date.  The ignition AFD(s) is the nominal initial source of the burning described by the corresponding fire event.  All other, non-ignition AFDs of a fire event are deemed its ‘propagation’ AFDs, since these AFDs follow from the ignitions, temporally and spatially.

 

See Figure 4 in the publication by Sloan et al. for an illustration of the geography of fire events and their ignition AFDs. 

 

Fire events and their ignitions were derived from standard science-quality MODIS Collection 6 MCD 14ML AFD data, commonly referred to as fire ‘hotspot’ data.  Data were detected by both the Terra and Aqua satellite sensors daily for Indonesia between July 2002 and December 2019.  Information on these input data are provide by the two citations below.  The publication of Sloan et al. provides methodological details on how the MODIS Collection 6 MCD 14ML AFD data were processed into discrete fire events and ignitions.

 

EarthData. MODIS Collection 6 Active-Fire Detections standard scientific data (MCD14ML), NASA EarthData, https://earthdata.nasa.gov/firms (2019).

 

Giglio, L., Schroeder, W. & Justice, C. O. The Collection 6 MODIS active fire detection algorithm and fire products. Remote Sensing of Environment 178, 31-41, (2016).

 

 

3. DATA FILES

 

Two data files are distributed here – one for discrete fire events, and another for the ignition AFDs of each fire event.  The data files are provided in a GIS-compatible format, and also as a generic text format, as described below.

 

3.1 GIS VERSION

 

Data files in GIS-compatible format are provided as ‘feature classes’ within an ArcGIS file geodatabase ‘Sloan_MODIS_FireEvents_Ignitions_2002_2019.gdb’.  These data files can be viewed and manipulated using either ArcGIS Desktop or ArcGIS Pro software.  There is one feature class for fire events, and another file for ignitions.

 

Sloan_MODIS_FireEvents_Ignitions_2002_2019.gdb\nfire4_all_spatial_fire_2002_2019_joins_sp_LC

This file pertains to fire events.  All AFDs of a given fire event are included, without differentiation as to whether the AFDs are ignition AFDs or other (propagation) AFDs.  Fire events are assigned unique ID values and basic attribute data.

 

Sloan_MODIS_FireEvents_Ignitions_2002_2019.gdb\nfire4_all_spatial_fire_2002_2019_igs_sp_LC

This file pertains to ignitions. Only ignition AFDs are included for a given fire event.  Fire events corresponding to the ignitions are assigned unique ID values and basic attribute data.

 

3.2 CSV TEXT VERSION

 

Both data files are also supplied as comma-separated value (CSV) text files for viewing and manipulation in non-GIS software, such as Excel, text editors, or any statistical software.  The text files can also be read into various GIS software.  CSV-formatted files have the same file name and attribute fields as the corresponding GIS-formatted data files.  These CSV-formatted data files (as well as the GIS-formatted data files) include attribute data on the latitude and the longitude of each AFD.  

 

Attribute field names are included as the first row of values in a CSV file. 

 

No ‘text qualifiers’ like quotations (“ ”) or inverted commas (‘’) are used to designate text/string values within the CSV file.  Text values appear directly between commas in the CSV data file, e.g.,  …,Kalimantan_Southern,…  .

 

Note two points of caution for working with these CSV data:

 

i)                    Microsoft Excel may be used for a partial view of the data file nfire4_all_spatial_fire_2002_2019_joins_sp_LC.csv, but it is not recommended for working with this data file.  This is because the number of records/rows in this csv file slightly exceeds that maximum that may be read by Excel, which is just over 1 million.  This limitation does not apply to the other csv file, however.

 

ii)                  The GIS-formatted data files employ ‘null values’ in their attribute tables, and so the corresponding ‘values’ in the CSV-formatted data files are similarly null.  For null values, no value whatsoever is ascribed, not even 0.  In the syntax of a CSV file (apparent upon opening the file in any text editor like Microsoft Notepad), a null value is denoted by two consecutive commas without any value, text, or space between them.  If a CSV file were opened in Excel, a cell assigned a null value would be blank, not 0 or otherwise.  This denotes the correct transcription of the GIS-formatted data.  This feature will not impede the correct reading of these CSV data by whatever software.  Users are made aware of this feature merely to ensure the proper input of these data into whatever software.

 

 

4. DATA STRUCTURE / GEOGRAPHY

 

The GIS-formatted data files are ‘point data’, i.e., they map the geography of AFDs as individual ‘points’, in keeping with how these MODIS MCD14ML AFD data were originally structured.  For the GIS-formatted data files, each record/row in its corresponding attribute tables corresponds geographically to single AFD ‘point’, regardless of whether that AFD belongs to a fire event comprised of many AFDs.  In the parlance of GIS files, the files depict ‘single-part’ point features.  The unique ID field [nfireID2] serves to denote the fire event to which a given AFD belongs.

 

Similarly, for the CSV-formatted data files, each record/row of values corresponds to a single AFD.

 

There are 1,232,377 records for the data file ‘nfire4_all_spatial_fire_2002_2019_joins_sp_LC’.

There are 720795 records for the data file ‘nfire4_all_spatial_fire_2002_2019_igs_sp_LC’.

 

 

5. ATTRIBUTE FIELDS

 

In the data files, while some attribute fields pertain to the individual AFD as the unit of observation (e.g., the land-cover class coincident with the AFD), other attribute fields correspond to the larger ‘fire event’ to which the individual AFD belongs (e.g., the total duration of fire activity for the fire event).  Accordingly, for certain attribute fields pertaining to the fire event as a whole, their values will appear ‘duplicated’ in the data file amongst those individual AFDs (records) that constitute the fire event in question.  Whether a given attribute field pertains to the individual AFD or to its constituent fire event is denoted below for each field. 

 

Each AFD is assigned a unique ID field denoting its constituent fire event, [nfireID2].  This field is consistent between both data files, so that attribute data for a given fire event may be ‘matched’ to attribute data for its corresponding ignition AFD(s), and vice versa, on the basis of the common value of the field [nfireID2].

 

Note that many attributes below are as originally defined/measured by the input MCD 14ML data, or are derived directly thereof.

 

5.1 DATASET nfire4_all_spatial_fire_2002_2019_joins_sp_LC

 

 

 

 

 

 

5.2 DATASET nfire4_all_spatial_fire_2002_2019_igs_sp_LC

 

 

 

5.3. LAND-COVER ATTRIBUTE DATA

 

As noted in Section 5.1 and Section 5.2, the nominal values of the attribute fields ‘CCI_LC####’ correspond to column 1 of the table below.  These hierarchical values, and their corresponding land-cover classes labels in column 2 of the table, pertain to the land-cover classification of the Copernicus Climate Change Initiative Land-Cover Product of the European Space Agency.   This classification has an annual temporal resolution and 300-meter spatial resolution.  Pertinent citations for these land-cover data are below:

 

ESA. Annual land-cover product, 1992 to 2019/present, based on MERIS 300-m and ancillary SPOT, AVHRR, Sentinel-3 and PROB-V satellite data. European Space Agency (ESA) European Centre for Medium-Range Weather Forecasts (ECMFW) Copernicus Climate Change Service (C3S) Climate Change Initiative (CCI), https://cds.climate.copernicus.eu/cdsapp#!/dataset/satellite-land-cover?tab=overview; http://maps.elie.ucl.ac.be/CCI/viewer/download.php; http://www.esa-landcover-cci.org/ (2020).

 

Pérez-Hoyos, A., Rembold, F., Kerdiles, H. & Gallego, J. Comparison of global land cover datasets for cropland monitoring. Remote Sensing 9, (2017).

 

Columns 3 and 4 in the table below illustrate how the original land-cover classes of the Copernicus Climate Change Initiative Land-Cover Product were reclassified for analysis in Sloan et al.  

 

6. METHODOLOGICAL SUMMARY

 

Full methodological details of how daily MODIS Collection 6 MCD 14ML active-fire detections (AFDs) were rendered as discrete fire events and ignitions are provided in the current publication by Sloan et al., as well as by citations therein to the following earlier sources:

 

Sloan, S., et al. (2021). "Fire prevention in managed landscapes: Recent successes and challenges in Indonesia." Mitigation and Adaptation Strategies for Global Change 26: Article 32.

           

Cattau, M. E., et al. (2016). "Sources of anthropogenic fire ignitions on the peat-swamp landscape in Kalimantan, Indonesia." Global Environmental Change 39: 205-219.

 

The publication by Cattau et al. is the foundational methodological paper of Sloan et al. (2021) and the current publication of Sloan et al.

 

In summary, the method of defining a discrete fire event was to cluster MODIS Collection 6 MCD 14ML active-fire detections (AFDs) according to the spatial and temporal proximity of AFDs.  Here such a ‘cluster’ of AFDs constitutes a discrete fire event.  AFDs were clustered progressively and exhaustively such that fire events were defined ‘organically’.  An AFD was incorporated into a fire event whenever (i) the AFD occurred within a 1-km² grid cell that was the same as or adjacent to a grid cell already hosting at least one AFD already in the fire event, and (ii) the AFD occurred within four days of at least one AFD already incorporated by the fire event and in the same or an adjacent grid cell.  These protocols are those adopted by Sloan et al.(2021) and Cattau et al. (2016). 

 

All AFDs in the MODIS Collection 6 MCD 14ML AFD archive for July 2002-December 2019 were used.  AFD Data were processed into fire events across Indonesia on a per-annum, per island basis.  Hence, no fire events span two calendar years, nor of course do any span more than one island.

 

Active fires were detectable as MODIS Collection 6 MCD 14ML AFDs at a temporal resolution of up to four time per day, given the MODIS Terra and Aqua satellite revisit rates (i.e., the frequency with which the satellite sensor returns to observe a given locale).  In practice, AFD rates may be less than this, due to cloud or smoke haze, which obscures the detection of active fires.  Of all AFDs clustered to compose a given fire event, those AFD(s) with the earliest detection date were designated as the nominal ignition AFD(s) of the fire event.  A given fire event can have more than one ignition AFD if such ignition AFDs all share same earliest detection date.   

 

3. DATA FILES

 

Two data files are distributed here – one for discrete fire events, and another for the ignition AFDs of each fire event.  The data files are provided in a GIS-compatible format, and also as a generic text format, as described below.

 

3.1 GIS VERSION

 

Data files in GIS-compatible format are provided as ‘feature classes’ within an ArcGIS file geodatabase ‘Sloan_MODIS_FireEvents_Ignitions_2002_2019.gdb’.  These data files can be viewed and manipulated using either ArcGIS Desktop or ArcGIS Pro software.  There is one feature class for fire events, and another file for ignitions.

 

Sloan_MODIS_FireEvents_Ignitions_2002_2019.gdb\nfire4_all_spatial_fire_2002_2019_joins_sp_LC

This file pertains to fire events.  All AFDs of a given fire event are included, without differentiation as to whether the AFDs are ignition AFDs or other (propagation) AFDs.  Fire events are assigned unique ID values and basic attribute data.

 

Sloan_MODIS_FireEvents_Ignitions_2002_2019.gdb\nfire4_all_spatial_fire_2002_2019_igs_sp_LC

This file pertains to ignitions. Only ignition AFDs are included for a given fire event.  Fire events corresponding to the ignitions are assigned unique ID values and basic attribute data.

 

3.2 CSV TEXT VERSION

 

Both data files are also supplied as comma-separated value (CSV) text files for viewing and manipulation in non-GIS software, such as Excel, text editors, or any statistical software.  The text files can also be read into various GIS software.  CSV-formatted files have the same file name and attribute fields as the corresponding GIS-formatted data files.  These CSV-formatted data files (as well as the GIS-formatted data files) include attribute data on the latitude and the longitude of each AFD.  

 

Attribute field names are included as the first row of values in a CSV file. 

 

No ‘text qualifiers’ like quotations (“ ”) or inverted commas (‘’) are used to designate text/string values within the CSV file.  Text values appear directly between commas in the CSV data file, e.g.,  …,Kalimantan_Southern,…  .

 

Note two points of caution for working with these CSV data:

 

i)                    Microsoft Excel may be used for a partial view of the data file nfire4_all_spatial_fire_2002_2019_joins_sp_LC.csv, but it is not recommended for working with this data file.  This is because the number of records/rows in this csv file slightly exceeds that maximum that may be read by Excel, which is just over 1 million.  This limitation does not apply to the other csv file, however.

 

ii)                  The GIS-formatted data files employ ‘null values’ in their attribute tables, and so the corresponding ‘values’ in the CSV-formatted data files are similarly null.  For null values, no value whatsoever is ascribed, not even 0.  In the syntax of a CSV file (apparent upon opening the file in any text editor like Microsoft Notepad), a null value is denoted by two consecutive commas without any value, text, or space between them.  If a CSV file were opened in Excel, a cell assigned a null value would be blank, not 0 or otherwise.  This denotes the correct transcription of the GIS-formatted data.  This feature will not impede the correct reading of these CSV data by whatever software.  Users are made aware of this feature merely to ensure the proper input of these data into whatever software.