Fire Module

Using this document, you will get a general understanding of the functions and datasets on the GWIS map-viewer. Other resources/platforms that might serve as an alternative to the GWIS are briefly mentioned in the document.
Introduction
The Global Wildfire Information System (GWIS) is a joint initiative of the Group on Earth Observations (GEO) and Copernicus that provides fire danger forecast and near-real time information on active fires. The portal contains five applications: (1) Current situation viewer, (2) current statistics, (3) country profiles, (4) long-term fire weather forecast, (5) data & services.

Current Situation Viewer
The situation viewer (Figure 1) is the primary application of the GWIS that delivers information on forecasts and damage assessments. Additional layers such as country boundaries, human settlement, protected areas, and land cover are also available.

Forecasts
The Forecasts module is divided into two sections: Fire danger forecast and lightning forecast.

Fire Danger Forecast
This section has three data sources as shown in Figure 4: ECMWF (European Centre for Medium-Range Weather Forecasts) with 8km spatial resolution, NASA’s GEOS-5 with 28km spatial resolution, and Météo-France with 10km spatial resolution.

ECMWF is both a research institute and an operational service, delivering weather predictions and other weather-related data globally. It has a significant role in the Copernicus Program and is the primary data source for forecasts on GWIS.

Fire Weather Index
FWI stands for Fire Weather Index and consists of a series of numerical subindices that are indicators of potential fire behavior in common boreal fuel types [1] . The structure of the FWI is visualized in Figure 5. This system accounts for the effects of fuel moisture on fire behavior and comprehends three levels of information: (1) fire weather observations, (2) fuel moisture codes, and (3) fire behavior indexes [2] . These parameters are unitless, and as follows [3] :

• Fine Fuel Moisture Code (FFMC): It represents the moisture content of litter and other cured fine fuels. It is an indicator of sustained flaming ignition and fire spread. Generally, fires begin to ignite at FFMC values near 70, and the maximum probable value that will ever be achieved is 96.
• Duff Moisture Code (DMC): It represents the average moisture content of loosely compacted organic layers of moderate depth. It relates to the probability of lightning ignition and fuel consumption.
• Drought Code (DC): It represents the average moisture content of deep, compact organic layers. It relates to the consumption of heavier fuels and the effort required to extinguish a fire.
• Initial Spread Index (ISI): It is a combination of wind speed and FFMC representing rate of spread without the variable influence of fuel.
• Build Up Index (BUI): It is a combination of DMC and DC representing total fuel available to the spreading fire. It is correlated with fuel consumption.
• Fire Weather Index (FWI): It is based on the ISI and the BUI, representing intensity of the spreading fire as energy rate per unit length of fire front. It is often used as a single integration of fire weather.

National Fire Danger Rating System (NFDRS)
The U.S. index called National Fire-Danger Rating System (NFDRS; visualized in Figure 7) consists of functions of fuel type, topography, and weather. There are four components in this system [5] :

• Spread Component (SC): It is the rating of the forward rate of spread of a head fire. It integrates the effect of wind, slope, and fuel bed and fuel particle properties.

• Energy Release Component (ERC): It is based upon the estimated potential available energy released per unit area in the flaming zone of a fire. It is dependent upon the same fuel characteristics as the spread component (SC). The day-to-day variations of the ERC are caused by changes in the moisture contents of the various fuel classes, including the 1000 hour time lag class. ERC is derived from predictions of the rate of heat release per unit area during flaming combustion and the duration of the burning. Expressed in BTUs per square foot.

• Burning Index (BI): A measure of fire intensity. BI combines the Spread Component (SC) and Energy Release Component (ERC) to relate to the contribution of fire behavior to the effort of containing a fire. BI has no units, but in general it is 10 times the flame length of a fire.

• Ignition Component (IC): IC is related to the probability of a firebrand producing a fire that will require suppression action. It is mainly a function of the 1-hour time lag (fine fuels) fuel moisture content and the temperature of the receptive fine fuels. IC has no units but represents a probability from 1 to 100 percent.

Keetch-Byram Drought Index (KBDI)
KBDI is based on soil moisture storage capacity and is expressed in hundredths of an inch of soil moisture depletion [10] . The index is developed to function throughout a wide range of climatic and rainfall conditions in forested or wildland areas. The KBDI values range from 0 to 800, with 800 indicating extreme drought and 0 indicating saturated soil [11] .

Lightning Forecast
Lightning strikes are the dominant cause of forest fires [12] . On GWIS, the lightning forecast is provided by ECMWF (European Centre for Medium-Range Weather Forecasts), from their Integrated Forecasting System (IFS) [13] . Lightning prediction is acquired by exploiting its dependence on particular weather conditions and by averaging over sufficiently large spatial and temporal scales [14] . On GWIS, it is displayed in 6 classes from very low to extreme (see Figure 5 for an example).

Fire emissions data is provided by ECMWF. A selection of fire emissions types allows compatibility with the emissions estimates of the EFFIS. The data include black carbon, carbon dioxide, sulfur dioxide, organic carbon, non-methane hydro-carbon, total carbon in aerosols, methane, carbon monoxide, nitrogen oxide, and particulate matter.

The global fuel layer (Figure 10) is also available within the module. Fuel classes were aggregated from “Global Fuelbed Dataset“, by Pettinari., 2016 [15].

Current Statistics
The portal provides statistics at country level, as well as regions of interest such as Brazilian Legal Amazon, and the Arctic Monitoring Assessment Program. It is divided into two parts: (1) GWIS estimates of burned areas and number of fires, (2) seasonal trend of burned areas, emissions, number of fires, and thermal anomalies.

In the first part, for each country, a graph is presented for total burned area in hectares and number of fires covering the last 10 years (currently from 2012 to 2022).

In the second part, five types of data are presented for each country (see Figure 11 for an example), such as:

• Weekly burned areas + cumulative
• Weekly emissions + cumulative
• Weekly number of fires + cumulative
• VIIRS weekly thermal anomalies + cumulative
• MODIS weekly thermal anomalies + cumulative

Country Maps
In this section, four types of data presented: Monthly burned area (for a specific year and month), cumulative burned area (for a specific year), fire frequency (2002 – 2019) and median day of burning (2002 – 2019)

Available data can be displayed both at country scale and regional scale. The year can be changed on the top-left side of the screen; however, it is not possible to change the date range for fire frequency and median of day burning – which is locked at the 2002 – 2019 range.

Every map window has a layer tool, in which the user can change the base layer of a certain map (will not affect others) and add additional layers such as land cover and administrative boundaries.

Country Charts
This section delivers various charts at country and regional level. It is divided into two sections, burned area & number of fires and emissions. The data within these parts can be observed both in multi-year and single-year.

Available charts for burned area & number of fires:

Multi-Year

• Yearly burned area & number of fires
• Yearly burned area seasonality
• Yearly burned area by land cover (area in ha & percentage)
• Average monthly burned area by land cover (area in ha & percentage)
• Average monthly burned area seasonality & number of fires
• Monthly burned area vs. historical
• Fire size distribution and contribution to total burned area
• Average monthly fire size distribution vs historical
• Average monthly fire size per year

Single-Year

•  Yearly burned area by land cover (area in ha & percentage)
• Monthly burned area seasonality & number of fires
• Average monthly fire size & number of fires
• Fire size distribution and contribution to total burned area
• Average monthly fire size distribution

In the emissions part, there are 3 different charts available but 4 in total, as the yearly emissions & burned area chart is also presented using GFED (Global Fire Emissions Database). Others are processed using a FAOSTAT methodology. All the charts below are available both in multi-year and single-year tabs.

• Yearly/monthly emissions & burned area
• Yearly/monthly emissions & burned area (GFED)
• Yearly/monthly emissions by land cover
• Yearly/monthly CO2 equivalent

Figure 13 shows yearly burned area by land cover in South Africa, from 2002 to 2019. Clearly, “grass/shrubland” is the most affected land cover class by far. The data might differ country by country, considering the fact that whether a particular fire event is a hazard or a controlled fire i.e. agricultural burning – and it also depends on the dominant land cover class in that specific country. Additionally, Figure 14 presents yearly burned area and number of fires for the period of 2002 – 2019 in Central African Republic.

Data Downloads
A global monthly burnt area dataset covering 2002 – 2019, and the GlobFire Fire Perimeters dataset covering 2001 – 2020 period are available to download.

The global monthly burnt area dataset is provided in CSV format, and the area unit is hectares. It is derived from the MCD64A1 product of MODIS Terra & Aqua. The dataset is split by land cover class and includes every country, and their GADM (Database of Global Administrative Areas) Level 0 and Level 1 administrative units.

GlobFire Fire Perimeters dataset is in shapefile format and also derived from the MCD64A1 product. In this dataset, each fire shapefile has a unique fire identification code, the initial date, the final date, the geometry, and a field specifying if it is a daily burned area or a final/last burned area.

Long-Term Fire Weather Forecast
In this application, it is possible to access monthly and seasonal forecasts of temperature and rain anomalies at continental scale. Maps available in this application (see Figure 15 for an example) are processed by EFFIS System based on ECMWF data.

Other Platforms
Besides the GWIS, there are other existing platforms delivering data for fires. In some cases, the depicted platforms below can also be used as an alternative and/or a complement to the GWIS. Especially, the EFFIS services that are not available on the GWIS and the wide-range datasets available on the Climate Engine might serve as additional resources for wildfire monitoring, assessment and management in African countries.

1.    EFFIS
The European Forest Fire Information System (EFFIS) became operational in 2000, for the development and implementation of advanced methods for the evaluation of forest fire danger and mapping of burnt areas at the European scale.

Since the GWIS was built correspondingly to the EFFIS, these two platforms are similar in terms of provided applications – yet there are differences in some applications and the methodology used in some sections.

The current situation viewer of the EFFIS covers Europe, Middle East, North and Central Africa. Some datasets are available globally, such as land cover, protected areas, human settlement, active fires (MODIS & VIIRS), and burnt areas (VIIRS). There are datasets that are specific to the European continent, and some parts of Africa, Middle East, Caucasia, etc. See Figure 17 for a screenshot of EFFIS.

Fuels dataset covers European continent (excluding Belarus and Ukraine, including Iceland and Turkey), fire danger forecast (Météo-France dataset) covers European continent (including Israel and Turkey), and North African countries – whereas the extent of ECMWF data is a bit wider, as the data source covers Central Africa subregion, in addition to North Africa. Therefore, some datasets available on the EFFIS can also be useful for certain countries in Africa.

2.    Climate Engine
The Climate Engine, powered by Google Earth Engine, delivers various earth observation products for different applications. In terms of fires, CEMS (Copernicus Emergency Management Service) FIRE daily dataset with 800m spatial resolution including FWI and NFDRS indices is available (see Figure 15), as well as burnt areas dataset with 500m spatial resolution.

The Climate Engine also delivers Sentinel (2-5P), Landsat (5-7-8-9), and MODIS (Terra-Aqua) with various variables such as vegetation, snow, water indices; land surface temperature, R-G-B-NIR-SWIR-1/2 bands, and their ratios.

Time series and statistical summaries can be generated through the platform, and then downloaded in .csv format. Produced map layers can be downloaded in GeoTIFF format.

3.    Resource Watch
The Resource Watch delivers data on various topics globally. Active fire monitoring data from the VIIRS sensor with 375m spatial resolution is available, as well as burnt areas dataset from MODIS with 500m spatial resolution (See Figure 19).

The Fire Weather Index (FWI), provided by the Global Fire Weather Database (GFWED), air quality and current disasters datasets are also available. The data on the platform are free and open to use for your own analysis and can be downloaded from dataset detail pages.

References

[1]    B. J. Stocks et al., “The Canadian Forest Fire Danger Rating System: An Overview,” For. Chron., vol. 65, no. 6, pp. 450–457, Dec. 1989, doi: 10.5558/tfc65450-6.

[2]    S. Chelli et al., “Adaptation of the Canadian Fire Weather Index to Mediterranean forests,” Nat. Hazards, vol. 75, no. 2, pp. 1795–1810, Jan. 2015, doi: 10.1007/s11069-014-1397-8.

[3]    B. D. Amiro et al., “Fire weather index system components for large fires in the Canadian boreal forest,” Int. J. Wildland Fire, vol. 13, no. 4, pp. 391–400, Dec. 2004, doi: 10.1071/WF03066.

[4]    Y. Safi and A. Bouroumi, “Prediction of forest fires using artificial neural networks,” Appl. Math. Sci., vol. 7, no. 6, pp. 271–286, 2013.

[5]    NWCG Fire Danger Working Team, “Gaining an understanding of the National Fire Danger Rating System. National Wildfire Coordinating Group. United States Department of Agriculture, United States Department of the Interior, National Association of State Foresters.” 2002.

[6]    NWCG, “Fire Danger Background. PMS 437. United States National Wildfire Coordinating Group.,” 2021. https://www.nwcg.gov/publications/pms437/fire-danger/background (accessed Mar. 05, 2023).

[7]    A. J. Dowdy, G. A. Mills, K. Finkele, and W. de Groot, “Index sensitivity analysis applied to the Canadian Forest Fire Weather Index and the McArthur Forest Fire Danger Index,” Meteorol. Appl., vol. 17, no. 3, pp. 298–312, 2010, doi: 10.1002/met.170.

[8]    D. Griffiths, “Improved Formula for the Drought Factor in McArthur’s Forest Fire Danger Meter,” Aust. For., vol. 62, no. 2, pp. 202–206, Jan. 1999, doi: 10.1080/00049158.1999.10674783.

[9]    U. Kc, J. Hilton, S. Garg, and J. Aryal, “A probability-based risk metric for operational wildfire risk management,” Environ. Model. Softw., vol. 148, p. 105286, Feb. 2022, doi: 10.1016/j.envsoft.2021.105286.

[10]  R. R. Heim, “A Review of Twentieth-Century Drought Indices Used in the United States,” Bull. Am. Meteorol. Soc., vol. 83, no. 8, pp. 1149–1166, Aug. 2002, doi: 10.1175/1520-0477-83.8.1149.

[11]  K. Dolling, P.-S. Chu, F. Fujioka, K. Dolling, P.-S. Chu, and F. Fujioka, “Natural variability of the Keetch–Byram Drought Index in the Hawaiian Islands,” Int. J. Wildland Fire, vol. 18, no. 4, pp. 459–475, Jun. 2009, doi: 10.1071/WF06146.

[12]  R. Sarkar, P. Mukhopadhyay, P. Bechtold, P. Lopez, S. D. Pawar, and K. Chakravarty, “Evaluation of ECMWF Lightning Flash Forecast over Indian Subcontinent during MAM 2020,” Atmosphere, vol. 13, no. 9, Art. no. 9, Sep. 2022, doi: 10.3390/atmos13091520.

[13]  P. Lopez, “A Lightning Parameterization for the ECMWF Integrated Forecasting System,” Mon. Weather Rev., vol. 144, no. 9, pp. 3057–3075, Sep. 2016, doi: 10.1175/MWR-D-16-0026.1.

[14]  G. Lentze, “How to predict lightning,” ECMWF, Mar. 05, 2018. https://www.ecmwf.int/en/about/media-centre/news/2018/how-predict-lightning (accessed Mar. 05, 2023).

[15]  M. L. Pettinari, “Global Fuelbed Dataset,” Department of Geology, Geography and Environment, University of Alcala, Spain. PANGAEA, Sep. 23, 2015. doi: 10.1594/PANGAEA.849808.