Diagnosing imminent flood scenarios with a snow monitoring system
Signs of disasters caused by climate change can be found everywhere around us: We see droughts, fires, hurricanes, and floods in different parts of the world at the same time (thanks, 2020!). But there are some hazards caused by climate change that take less spotlight in the media and are less on our minds. Flash floods caused by snow melt are one of them.
With my article today, I want to focus on the impact climate change has on snow, how snow melt can cause disastrous and deadly flash floods within hours, and why we need to monitor snow properties to increase preparedness, receive early warnings, and potentially save lives.
The impact of climate change
Climate studies show that until 2050, we could witness an increase in average temperature of up to 5°C. This is a fairly high projection, but within the realm of the possible. A comparably low increase of average temperature of “only” 2°C is calculated to increase precipitation intensity by around 10% and run-off by 15%. This would cause a retraction of the snow line to higher altitudes, and increase the potential for more frequent floods and landslides in wintertime.
Even nowadays, small increases in temperature frequently cause massive and rapid snow melt. It can take hours or days until the flooding reaches entire regions, valleys, and villages, where it can bring death and damage to property. These floods are particularly prominent in regions with frozen soil and rainfall following the snow melt, both of which facilitate the severity of the potential flood event.
How can we not only assess the risk of flash floods caused by snow melt more accurately, but also receive an early warning if a flooding is imminent?
Techniques and models for rainfall forecasting have greatly improved the last years and enabled what we know today as highly accurate services. However, rainfall is not the only variable to consider when modelling or predicting a flood event or its potential impact. Other variables such as temperature and snow can be a crucial, regional factor.
However, temperature (and therefore snow) forecasts introduce a comparably higher amount of uncertainty into hydrology models than precipitation. Just 1°C than predicted can cause large amounts of additional runoff in snow-covered areas. An underestimation of snow volume/cover can have the same devastating effect. In addition, ground truth to calibrate temperature and snow models is often not available for remote or not so easily accessible areas.
So that begs the question: What if we could monitor snow changes and integrate it as part of an early warning system for flash floods?
Snow data analysis previous to a flood event — 2019 Gurk river flood
Using DeFROST from the Swiss startup WeGaw, we can monitor snow near real-time on a daily basis from Earth Observation data. This allows us to identify and address rapid changes in snow cover, a factor indicating an imminent flooding event.
On November 15 to 18 2019, the increased water levels of the river Gurk caused a devastating flood event in St. Veit an der Glan and several surrounding towns in Carinthia, Austria. When we analyzed our DeFROST products of the basin of the Gurk from the 14th to the 15th, we were able to detect an undeniable retraction of the snowline:
We can clearly see that a massive amount of snow was lost the day before the flood.
This observation indicates that observing snow near real-time with products such as DeFROST could be a contributing factor to the early diagnostics of flash flood events. However, there is a clear and definite uncertainty in our analysis of the Gurk river flood: Part of our algorithms are assimilating historical weather and temperature data to fill gaps, and during the days in question observable snow data was scarce. While our models perform well on average and give high accuracy outputs, I think this disclaimer is fundamental to maintain transparency and a scientific discourse.
Despite this uncertainty, we see a correlation between the snow cover changes observed through DeFROST data and measurement of the ground station at the Gurk river during November 2019 in the image below.
We checked the water height from the 16th of November onward and were able to clearly trace the spike which represents the actual flooding event.
Conclusions
Flash flood forecasting and diagnosis are extremely tricky and require precise data on and understanding of lots of variables. With the temperature increases caused by climate change, flash flood events caused or worsened by snow melt will become more and more common and harder to predict.
Snow plays a crucial role for flash floods in snow-prone regions and is a variable we need to monitor closely to prevent deaths and property damages. Monitoring snow through Earth Observation and thus without the need of ground sensors or more sophisticated (and expensive) methods can help to understand how a rapid snow melt events happen and hopefully build better early warning systems for flooding on a global scale.
At WeGaw, we want to implement a flash flood warning system for snow-prone areas. Would you like to explore this with us? Connect with us through info@wegaw.com
