Modelling the Increasing Risk of Damage from Derechos to European Forests

Derechos are a collection of downbursts produced by a group of thunderstorms that lead to widespread straight-line winds. They are most common in the great plains area of the USA but are increasingly being found in the North European Plains and especially in Poland and Belarus.

On 11/12 August 2017 a very strong derecho moved from south to north starting in the Czech Republic and across central Poland and on out into the Baltic Sea. The storm caused six deaths and several dozen injuries and extensive damage to utilities and buildings and to 80,000 ha of forest and 67,000 hectares of agricultural crops. The concern is that such events are likely to become more frequent in the future in the changing climate and forests in regions affected by derechos require adapted management to make them more resilient.

We tested whether a mechanistic wind damage model for forests (ForestGALES) that was originally developed for predicting damage from winter extra-tropical storms could predict the areas of damage caused by the 2017 derecho when combined with high resolution (2 x 2 km) wind field predictions from the AROME mesoscale atmospheric model. Detailed tree inventory data from the Polish National Forest Inventory (NFI) was used together with soil data as inputs to the ForestGALES model to calculate the wind speed at which damage was expected to occur for each NFI plot measured in 2016. These critical wind speeds (CWS) were then compared with the predicted wind speeds at 10 m elevation from the AROME model to give a probability of damage based on a sigmoid damage function.

The predictions of the combined models were tested using Receiver Operator Characteristics (ROC) by adjusting the damge threshold in the sigmoid function and calculating the Area Under the Curve (AUC). An AUC of 0.5 suggests no model discrimination, more than 0.7 is considered as acceptable discrimination, and more than 0.8 as excellent discrimination. For the derecho of 11/12 August using the CWS values predicted by ForestGALES and the gust speeds predicted by the AROME model an AUC of 0.858 and a model accuracy (percentage of correctly identified damaged and undamaged NFI plots) of 77.5% was achieved.

The results suggest the ForestGALES model when used in conjunction with the AROME high-resolution mesoscale model does an excellent job of identifying the forest stands most likely to be damaged. This information can be used to identify which forest stands are most resistant to the extremely strong winds found in derechos, and what characteristics of these stands made them more resistant. Such knowledge can help forest managers create more resilient forests. In addition, such a system could be used to identify the trees and forest stands most at risk of damage before the arrival of a derecho and allow emergency services to anticipate where damage is most likely to be a problem and to organise their response ahead of the storm.