Development of the basics for rapid damage indication maps after storm events

After a major storm event, spaceborne remote sensing enables an initial overview of the damage in the forest. Methods based on optical satellite data have the advantage that they are very intuitive and the data are very easy to interpret. However, the disadvantage is that no images of the earth's surface can be taken in cloudy conditions. For example, after the winter storm "Burglind" (2 and 3 January 2018), well-illuminated and cloud-free optical Sentinel-2 images of certain areas were not available for up to three months. This developed method is based on Sentinel-1 radar data, which are acquired independently of cloud cover.

Time required for an evaluation after a storm event

The accuracy of the method for generating storm damage indication maps depends on the number of Sentinel-1 acquisitions processed. With 3 acquisitions, a first overview is already possible: Areal damages with a minimum size of 0.5 ha can be detected with an accuracy of 65%. A longer waiting time for 7 acquisitions increases the accuracy to 75%. For now, these accuracies apply to flat areas only, since the method could not be validated in topographically more complex areas due to insufficient reference data. Depending on the region, it takes a different time period to obtain the required number of acquisitions because the acquisition frequency of Sentinel-1 is varying spatially (see, e.g., for Switzerland in Fig. 1).

Method

The method for generating indications for areal forest damage is based on differences in radar backscatter before and after the storm event. After the storm, backscatter is increased from forest areas affected by windthrow. Several factors influencing radar backscatter were investigated. Three factors had a decisive influence on the performance of the method: the vegetation height, the liquid water content within the snowpack and the topography.

• Only from vegetation higher than 15 m, it is possible to detect reliable differences in radar backscatter between intact forest and windthrows. This means that relevant stands can be covered.
• Liquid water inside the snowpack influences radar backscatter to such an extent that it made it impossible to successfully apply the method in areas with wet snow.
• Due to the acquisition geometry, radar sensors do not reliably measure backscatter in areas with steep slopes (> 60°).

The method was developed based on these findings. Thus, indications for storm damages are given in stands with a minimum height of 15 m. In the "Burglind" case, large parts of the foothills of the Alps were affected by wet snow, so these areas were excluded from the evaluation. In the future, this influence should specifically be investigated to make appropriate adjustments to the method to enable the generation of indications of storm damage in these areas as well. Furthermore, we excluded those forest areas from which reliable radar measurements cannot be made.

Visualisation and distribution of the results

The result of the method is an indication map for storm damage. To not imply an exact mapping with the result, the indications are depicted in a generalised way as a grid of hectares or as point indications in the form of coordinate pairs. Fig. 2 shows an example for the depiction of the point indications. An automatic cluster algorithm adjusts the appearance of the points according to the selected region.