The big picture and the details – impressions from EGU’23

Climate, ice and tipping points

So many talks and posters deal with different aspects and details of our changing climate, ice caps, and their interactions. We have known the general direction of relevant trends in our environment (temperature, droughts, superfluous precipitation) for quite some time, so this is, to some degree, filling in the details. A lot of work is also being done on the stability of (climate or ice cap) (meta)stable states, switching between them, and tipping points, the activation of which may lead to such a state change, potentially moving our climate beyond scenarios that are currently modelled in global circulation models. For example, think of the possible switching off of the Atlantic Meridional Overturning Circulation, comprising ocean currents which transport large amounts of heat from the tropics to more northern regions. These state changes may be rapid and result in significant changes in the weather we experience. The latest IPCC report mentions tipping points as events with a low likelihood and high impact. In his presentation “Tipping points – a challenge for climate projections”, prof. Thomas Stocker described that we see abundant evidence for the activation of climate tipping points in paleoclimate proxies and that the models we have right now for making climate change projections barely resolve (i.e. do not have sufficient resolution) the processes that are relevant for activation of tipping points. This means as he confirmed in response to my question on this, that we currently cannot estimate the likelihood of the activation of these climate tipping points, and therefore they constitute events of high impact and unknown rather than low likelihood.

Compounding risks and impacts, multi-hazards, permacrises

The conference also paid significant attention to natural hazards, the occurrence of which is exacerbated by climate change, population growth, and changes in land use. These events are starting to appear consecutively or even coincide more and more often. For example, the recent earthquakes in Turkey and Syria were followed shortly by torrential rains that caused further landslides and hindered aid and recovery. Also, longer-term crises like the covid pandemic and the invasion of Ukraine can be part of such a mix. Terms like multicrisis or even permacrisis are being applied to these situations. The coincidence of multiple crises makes dealing with them more difficult, as attention and resources are stretched thin. A special symposium was dedicated to this topic, in which four experts shed light on the issues from their own respective fields. The first speaker, Dr Philip Ward, gave an overview of particular challenges that arise from multi crises. These include a diversity in language/terminology on compounding risks, which lead to confusion, the lack of a clear framework and guidelines for practitioners, an underdeveloped understanding of the potential feedback between hazards and vulnerabilities, and a lack of in-depth case studies. But science is advancing rapidly! The second speaker, Dr Tina Comes, stressed the importance of a strategic perspective in dealing with multi crises. Decisions that are taken in the heat of the moment in one crisis may have long-term consequences that were not considered. Dr Elena Rovenskaya reflected on how science should evolve to help society prepare for and deal with compound events. Her recommendations included strengthening transdisciplinary research, definition of protocols for quickly putting together teams of researchers in response to emerging crises, improving knowledge diffusion within sciences, increasing trust in science and science literacy of the public, and improving the science-policy interface. The final speaker, Dr Arthur Malantowicz, described which measures and structures have been put in place in the European Union to better deal with multi crises, changing from a response-driven to a forward-looking paradigm, bringing together all stakeholders.

Trends in data analysis and machine learning

The past half-decade has shown an explosion of machine learning applications to water-related applications (and more so outside the water domain). This approach continues to be the dominant method in the presentations that I witnessed, people are starting to marry it to more traditional process-based models in so-called scientific or physics-based machine learning (interesting examples in the presentations of Löwe and Daniel&Cominola). Indeed, a whole session was devoted to “Hydroinformatics: data analytics, machine learning, hybrid modelling, optimisation”.

Implications for the water sector

To start with the last item, machine learning is becoming more and more common in the water sector, and hybrid modelling is gaining attention. Indeed, we will be discussing this topic in next Tuesday’s Hydroinformatics knowledge exchange meeting with the water utilities participating in the joint research program BTO. There are opportunities to improve machine learning predictions by including some physics. A major challenge remains the operation of ML models outside the range of their training data. In our changing climate and hydrosphere, that is a situation we will continue to encounter and a domain in which traditional mechanistic models tend to perform better. We need to explore how the marriage of both approaches may bring us the best results.

Multicrises and permacrises are relevant scenarios for the water industry. The diminishing availability of and growing demand for water may already have put some water industry members in a more or less permanent crisis mode. Also, in geographies where this is not (yet) the case, the possibility of multiple hazards compounding risks needs to be on the radar. Many of the insights and recommendations from the dedicated session are also applicable to our sector.

And to conclude, it is important that we, both the water sector and society as a whole, are aware of the existence of high-impact climate tipping points, of the fact that we have evidence of their activation in the geological record, and of the fact that we do not know the likelihood of their being activated in our time. As argued by Timothy Lenton et al. in Nature (2019), they are “too risky to bet against.”

PS

For more impressions, read the blogs of my colleagues Ina Vertommen, Karel van Laarhoven, and Janine de Wit.