ECMWF Newsletter #177

Global seasonal fire danger forecasts available in the Climate Data Store

Francesca Di Giuseppe
Christopher Barnard
Eduardo Damasio-Da-Costa


On behalf of the EU's Copernicus Emergency Management Service (CEMS), ECMWF has recently widened the fire danger data offering in the Climate Data Store (CDS) to include a set of fire danger forecasts with lead times up to seven months. The CDS is run by the EU’s Copernicus Climate Change Service (C3S) implemented by ECMWF. The new data are provided by the Global ECMWF Fire Forecast (GEFF) model, which uses the Centre's seasonal forecasts to drive three different fire danger models developed in Canada, the United States and Australia. The dataset is made openly available for the period 1981 to 2022 and will be updated regularly, providing a resource to assess the predictability of fire weather at the seasonal timescale. The dataset complements the availability of real-time seasonal forecasts provided by the European Forest Fire Information System (EFFIS).

Availability of seasonal fire danger forecast

Global fire danger seasonal hindcasts have been generated using ECMWF’s SEAS5 seasonal forecasts for the period of 1981 to 2022 as input to GEFF. The GEFF model is open source and available from a public repository under an APACHE2 license. The current version is 4.1. Data are archived in the CDS with several advantages: open access via a user-friendly web interface; bulk access via a convenient API; integration with the CDS toolbox for performing server-side operations; and shared visualisation and data analysis tools.

The fire danger seasonal forecast dataset has global coverage and a spatial resolution of about 0.25 degrees (about 35 km). Natively, data are laid out over an octahedral reduced Gaussian grid (O320) and archived as GRIB2, a standard format published by the World Meteorological Organization. Users can also request data in NetCDF format, which implies an internal remapping data transformation. Forecasts are issued monthly, on the first day of each month, with a lead time of 216 days (about seven months). The dataset is updated regularly with a delay of a few months compared to real time, which is available through CEMS.

Skill of seasonal fire danger forecast

Forecasting fire danger is key in preventing fires and taking protection measures as it improves the readiness of fire professionals and enables the timely and efficient allocation of resources. A limited number of studies show that, besides well-established fire danger forecasts with lead times of a few days, skillful predictions of fire danger are possible up to the seasonal timescale for Mediterranean Europe. Local soil moisture anomalies and heatwaves have been identified as an important source of this predictability. Seasonal forecasting of fire weather conditions throughout the world have also been found to correlate with large-scale climate patterns, such as the El Niño–Southern Oscillation (ENSO) and the Indian Ocean Dipole. This shows that fire weather conditions could potentially be predicted months in advance for various seasons and regions.

Prediction of monthly mean fire danger anomalous conditions between 2013 and 2022 over Indonesia.
Prediction of monthly mean fire danger anomalous conditions between 2013 and 2022 over Indonesia. Months are classified as above or below the 1981–2022 climate mean using percentiles. Anomalies from ERA5 fire danger are compared to SEAS5 fire danger forecasts for increasingly longer lead times to highlight the predictability of anomalous conditions. Months outside the traditional fire season are masked out. They are months in which the mean Fire Weather Index is lower than a third of the year’s maximum. The ENSO index helps to identify years of strong positive or negative anomalies, in which El Niño or La Niña conditions are established. These years correspond to periods of high predictability, for which anomalous conditions could be predicted up to seven months before.

Looking at the Anomaly Correlation Coefficient (ACC) of the released dataset, verified against the fire danger calculated using ECMWF’s ERA5 reanalysis, we found that globally anomalous conditions for fire weather can be predicted with confidence one month ahead. In some regions the prediction can be extended to two months ahead. In most situations beyond this horizon, forecasts do not improve on climatology.

However, an extended predictability window, up to 6–7 months ahead, is possible when anomalous fire weather is the result of large-scale phenomena, such as ENSO. This is a climate pattern characterised by the warming of the surface waters in the central and eastern tropical Pacific Ocean during El Niño conditions. Those conditions often lead to a shift in rainfall patterns, resulting in reduced precipitation in Southeast Asia, including Indonesia. This can create drier-than-normal conditions, especially in peatland areas, making them more susceptible to fires. The conditions established by a strong El Niño exacerbate landscape flammability, but it is human activities that play a significant role in igniting fires. In Indonesia, particularly in the regions of Sumatra and Kalimantan, land clearing practices such as slash-and-burn agriculture, illegal logging, and peatland drainage for agriculture have been responsible for extensive burning in the past. The release of large amounts of smoke and pollutants into the atmosphere has affected air quality not only in Indonesia but also in neighbouring countries, such as Malaysia and Singapore, generating international health emergencies.

The establishment of a positive or negative ENSO is usually monitored using a multivariate index obtained by extracting the leading combined Empirical Orthogonal Function (EOF) of five different variables over the tropical Pacific basin (30°S–30°N and 100°E–70°W). During a strong positive or negative ENSO, seasonal prediction of fire weather is enhanced up to seven months ahead (see the figure). Efforts to mitigate the impact of fires during ENSO events in Indonesia could therefore benefit from an early warning system at this timescale. This could help to guide land management practices and implement fire prevention and suppression measures before the burning takes place.

The dataset can be accessed at