Evaluating foehn development in a changing climate using machine learning and investigating the impact of foehn on forest fire occurrence and severity
OPEN ACCESS
Author / Producer
Date
2020-12
Publication Type
Master Thesis
ETH Bibliography
yes
Citations
Altmetric
OPEN ACCESS
Data
Rights / License
Abstract
Foehn winds have a substantial impact on forest fires all over the world. However, two questions have remained unanswered to the foehn community over the last years. First, while accurate forecasts are viable at the timescale of numerical weather prediction, the long-term evolution of monthly foehn occurrence is still open to debate. Second, while many studies cited the importance of foehn winds on forest fires in single case studies, a definite quantitative link between foehn winds and forest fires prevailed missing. Hence, the scientific contribution of this work is divided into two parts.
In the first part, we explored the possibilities of employing machine learning algorithms to predict foehn within Switzerland from its synoptic fingerprint in climate simulations. Here, we used variables from the ERA-Interim reanalysis and the CESM simulation as inputs for our models. We trained on ERA-Interim data to recognize foehn, then verified the results on a CESM simulation of present-day climate, and finally, predicted foehn on a future warming climate CESM simulation. The best generalization between ERAI and CESM was obtained by including the present-day simulation in the training procedure and simultaneously optimizing two objective functions, namely the negative log loss and squared mean loss, on both datasets, respectively. The model verification showed validity of our approach for most of the months. Finally, we found that south foehn in Altdorf is expected to become more common during spring, while north foehn in Lugano is expected to become more common during two summer months.
In the second part, we analyzed forest fires of the past 40 years and linked them with foehn occurrence from a climatological perspective. We found that foehn duration and foehn strength substantially increase both the number and the severity of forest fires. In detail, we observed that if a day showed foehn presence, it was associated with a 3.4-fold increase in numbers of fires in contrast to a day without foehn presence. Furthermore, if a foehn wind set in during the six hours after fire ignition, it increased the median burned area of such fires by a significant factor of three compared to fires without foehn occurrence.
While we developed and tested both methodologies within Switzerland due to the vast availability of foehn and forest fire data, we extensively documented our approaches. Therefore, we encourage other researchers to apply these frameworks also to foehn winds in other regions of the world.
Permanent link
Publication status
published
External links
Editor
Contributors
Examiner : Sprenger, Michael
Examiner : Jansing, Lukas
Examiner : Jochum, Josef
Book title
Journal / series
Volume
Pages / Article No.
Publisher
ETH Zurich
Event
Edition / version
Methods
Software
Geographic location
Date collected
Date created
Subject
Downslope winds; Foehn; Climate prediction; Machine learning; Forest fires
Organisational unit
02717 - Institut für Atmosphäre und Klima / Inst. Atmospheric and Climate Science