ECMWF lightning forecasts evaluation during fire seasons in Portugal

Lightning represents only 5-10% of fire causes in Europe and 1-2% in Portugal and correctly predicting its occurrence can help to mitigate large forest fires. In 2016, the European Centre for Medium-Range Weather Forecasts (ECMWF) introduced a parametrization for lightning forecast. Therefore, aiming to assess the feasibility of using ECMWF forecasts in the context of natural wildfires, this study analyses the ECMWF data for mainland Portugal over 4 fire seasons (2019 - 2022). The dataset used is composed by the total average lightning density in 3 hours from the ECMWF and the observation data from the detection network from IPMA and grouped for the same times. The data was also aggregated into horizontal resolution grids of 1⁰ and 0.5⁰. To carry out the statistical analysis, two contingency tables were constructed, and some statistical indices were calculated (e.g., BIAS, Success rate, False alarm rate, Threat score, Equitable threat score, True skill score). The results reveal a Bias score bigger than 1, indicating that model tends to overestimate the lightning occurrence. Moreover, the success rate was about 57.7% (1⁰) and 49% (0.5⁰). The temporal analysis indicated a time lag between both data, with the model starting and ending the prediction earlier than observed. Furthermore, two meteorological patterns were found producing lightning during the period studied: the development of a thermal low overlapping with an Upper Level Low and the passage of large-scale systems. These results show that the ECMWF lightning forecast can be used as an additional tool to prevent natural fires, with the model being able to indicate the days and locations of lightning occurrence. The research was funded by the European Union through the CILIFO project (0753-CILIFO-5-E) and also by national funds through FCT Foundation for Science and Technology, I.P. under the PyroC.pt project (PCIF/MPG/0175/2019).