Periodic Reporting for period 1 - SPFireSD (Seasonal Prediction of Fire danger using Statistical and Dynamical models)
Periodo di rendicontazione: 2017-09-06 al 2019-09-05
Wildfires are the largest source of biomass burning and a great source of pollutants and atmospheric CO2. In addition to having a great impact on the environment, wildfires can also pose a threat to property and human lives and health. The occurrence of fire in natural vegetation is dependent on several factors: human activities, accumulation of fine dead fuels (grass, leaves and twigs) and climatic variability. The Earth's climate undergoes natural variability at seasonal-to-decadal timescales. Informing public sectors that are vulnerable to its variations is a key societal and economical challenge. Operational seasonal climate predictions are now routinely performed around the world and multi-model ensemble forecasts systems provide more realistic forecasts than those provided by a single model. This climate information is used for many applications in fields such as agriculture, health, water management and energy. In light of this, seasonal prediction of wildfire danger appears as a priority for health, safety and economic welfare. Climate is partially predictable on seasonal timescales and operational seasonal climate forecasts show significant skill. Opportunities therefore exist of relying on this climate skill to develop potentially skillful seasonal wildfire forecasting systems. This project proposes to develop and assess seasonal fire prediction capability through a variety of complementary and innovative methods using statistical and dynamical models.
ECMWF’s Global ECMWF Fire Forecast (GEFF) system is used operationally to produce 10-day fire danger forecasts (using the Canadian Fire Weather index as a proxy) at the European and global scales. The GEFFSP (GEFF for Seasonal Prediction) is an adaptation of GEFF for fire danger, using seasonal predictions from the ECMWF’s long-range forecasting system SEAS5 obtained from the C3S Seasonal Forecasts archive. The results of this methodology applied to California and the Iberian peninsula (regions identified as critical for this work) were mixed, as it was shown that the predictability of the Fire Weather Index is poor beyond 1 month leadtime (due to a poor predictability of precipitation beyond 1 month in these areas) during the fire seasons of both areas. However, literature suggests that predictability should be better in regions affected by ENSO (El Niño/Southern Oscillation) such as the Amazon and Indonesia, and a follow-up of this project will be to use the same technique over these regions. A paper detailing the implementation and assessment of the system is under preparation, and further results are expected shortly. This project has also contributed to the understanding of the predictability of seasonal prediction systems of indices such as the Fire Weather Index for predicting fire danger.
The EC-Earth-Fire seasonal prediction system, using fire models from the LPJ-Guess dynamic vegetation model coupled to the EC-Earth climate model, was developed in offline configuration known as the EC-Earth Land Surface Model. Atmospheric output obtained previously from the decadal (and historical) simulations performed by the Climate Prediction Group of the BSC for DCPP and CMIP6 using the Autosubmit workflow manager were archived in a format suitable to force offline LPJ-Guess simulations. This output was used to generate decadal (and historical) simulations with LPJ-Guess in offline mode. Preliminary results show that the model produces slightly less biomass and more fire emissions than reconstructions. Follow-up projects will evaluate the performance of the vegetation and fire model when forced with seasonal-to-decadal predictions of the Global Climate Model version of EC-Earth, as well as comparing the added value of initialization, by comparing the predictions to free-running historical simulations. This work will be used in further projects in which the Climate Prediction Group is involved, in particular the H2020 project CCiCC (Climate-Carbon Interactions in the Coming Century).
The EC-Earth-Fire seasonal prediction system, using fire models from the LPJ-Guess dynamic vegetation model coupled to the EC-Earth climate model, was developed in offline configuration known as the EC-Earth Land Surface Model. Atmospheric output obtained previously from the decadal (and historical) simulations performed by the Climate Prediction Group of the BSC for DCPP and CMIP6 using the Autosubmit workflow manager were archived in a format suitable to force offline LPJ-Guess simulations. This output was used to generate decadal (and historical) simulations with LPJ-Guess in offline mode. Preliminary results show that the model produces slightly less biomass and more fire emissions than reconstructions. Follow-up projects will evaluate the performance of the vegetation and fire model when forced with seasonal-to-decadal predictions of the Global Climate Model version of EC-Earth, as well as comparing the added value of initialization, by comparing the predictions to free-running historical simulations. This work will be used in further projects in which the Climate Prediction Group is involved, in particular the H2020 project CCiCC (Climate-Carbon Interactions in the Coming Century).
This was the first attempt at using the C3S Seasonal Predictions for seasonal predictions of fire danger. The approach shows potential for the implementation of short-term (up to 1 month) fire danger warning systems for the Iberian Peninsula and California, and further work is planned to evaluate the potential for skillful forecasts up to a few months in tropical regions.
The implementation of the EC-Earth-Fire seasonal system has facilitated a new research line in seasonal-to-prediction of the land carbon cycle (including fire danger forecast) and follow-up projects have a potential for broad societal impact in the prediction of wildfires and the global carbon cycle.
The implementation of the EC-Earth-Fire seasonal system has facilitated a new research line in seasonal-to-prediction of the land carbon cycle (including fire danger forecast) and follow-up projects have a potential for broad societal impact in the prediction of wildfires and the global carbon cycle.