The upcoming KNMI’23 climate scenarios – what will they bring our drinking water companies?

The climate scenarios produced by the Dutch national meteorological institute KNMI’s are something of a gold standard in The Netherlands, used in many studies and design to ensure climate resilience. Since the publication in 2014 of their latest set of climate scenarios for the Netherlands, there have been many developments in climate science. KNMI is preparing a new set of scenarios, based on the latest IPCC report (6th assessment report, part I, published in 2021) and the underlying CMIP6 model simulations, to be published in October 2023. In order to ensure that these scenarios meet the requirements of their users, a stakeholder workshop was organized on October 10, 2022. The first part consisted of a series of presentations, with about 200 participants, both live and online. This blog post summarizes the information that was shared in the presentations and reflects on their relevance for the drinking water sector. As the KNMI scenarios focus on the Netherlands, so does the summary of the shared information. However, my reflections on these should be applicable to a wider geographical extent.

Summary of the generation and general view on the KNMI’23 climate scenarios

The new set of scenarios, just as the previous one, is intended to cover the range of possible futures. Four scenarios are defined spanning two axes. The first represents emission scenarios, and ICCP’s SSP1-2.6 at the low end and SSP5-8.5 at the high end. The former appeared to be considered almost infeasible by now by the presenters, the latter was presented as very pessimistic. This axis essentially reflects the emissions behavior of our global civilization during the coming decades and beyond.

The second axis represents two classes of results that are observed in the set of models, which either show a projected increase or a decrease in annual precipitation for the Netherlands. The scenarios focus on The Netherlands, but the associated modelling work considers the complete basins of the Rhine and Meuse rivers, that bring large amounts of fresh water into the country.

It is notable that the four scenarios do not differ significantly from the KNMI’14 scenarios in terms of the associated amounts of global warming, even though ICCP AR6’s SSP1-2.6 is slightly warmer (mean 0.2 by degrees) than AR5’s RCP2.6 and AR6’s SSPS-8.5 is also somewhat warmer than AR5’s RCP8.5 (mean by 0.1 degrees, upper bound by 0.3 degrees). But they may remain suitable to bound the likely parameter space. One of the speakers noted later on that the range spanned by the new KNMI scenarios is based on emission scenarios rather than the response of the climate system, for which the models show a significant uncertainty range in themselves. And it was stressed that a lot of discussions both internally and with international experts underly the selection and definition of the scenarios.

In any case, there are considerable differences between the two generations (i.e. 2014 and the upcoming 2023) of KNMI scenarios, the most important of which is the new class of models that show an overall decrease of annual precipitation. Note that all scenarios predict a decrease in summer precipitation and an increase in winter precipitation; the difference is in the amounts and the yearly balance, see Figure 1.

Figure 1: Changes in precipitation for the SSP5-8.5 models for Winter, Spring (Lente), Summer (Zomer), Fall (Herfst) and the whole year (jaar), for the entire group of models (white) and the wet (blue) and dry (red) sets. Source: Frank Selten, KNMI.

Some interesting insights from the presentations and discussions:

  • Land use has a huge impact on the inception of droughts through a feedback mechanism – as the soil dries up, the heat consumption by evaporation is lost and therefore the air above a drier soil heats up much more quickly, reinforcing the evaporation of the remaining moisture; also, less evaporation means fewer clouds and therefore more solar radiation reaching the surface. Forests are thus much better at keeping the soil moist that grassland.
  • The models do include the potential weakening of the “Gulf stream” (part of the Atlantic Meridional Overturning Circulation, AMOC) as an emerging feature. This is responsible for the generation and stability of high pressure zones west of Ireland that promote an easterly wind in the Netherlands in summer and that are responsible for the decrease in precipitation as discussed and shown in Figure 1.
  • No significant changes in wind are expected, apart from an increasing likelihood of tropical cyclones reaching Europe in fall.

An important consequence of the changing climate in the Netherlands is the more frequent occurrence of droughts, which result from the balance of precipitation and evaporation. The observations show 4 droughts in the past 5 years, which are insufficient to statistically speak of a trend. However, a regional analysis does show a significant trend in the precipitation deficit for the entire country in spring and for the southeastern part also in summer. In particular in the high emission scenario based on SSP5-8.5, an increase in droughts can be expected both for the dry and the wet classes of models.

The next presentation about extreme precipitation focused on convective systems (summer showers). An increase in more intense showers is expected (robust prediction), with likely a stronger increase towards the more extreme precipitation evens. The heavy rainfall that caused extensive flooding, many lives lost and billions of damages in Germany, Belgium and The Netherlands in the summer of 2021 had characteristics of both summer showers and a weather front. It did not become clear in the discussion to what extent the presented predictions also apply to these kinds of extreme weather.

Finally, in a presentation about sea level rise, it was shown that the sea level signal at the Dutch coast is strongly influenced by wind effects, but that nevertheless a significant signal of sea level rise is emerging, in line with global trends. In addition to the scenarios for sea level rise based on climate models that have been shown before, a low-likelihood-high-impact scenario of rapid sea level rise (up to 2 meters by 2100 and many more beyond) due to collapse of ice shelfs was discussed (Figure 2). Recent observations on Antarctica suggest that this extreme scenario may be less unlikely than previously thought.

Figure 2: Predicted sea level changes for the SSP1-2.6 and SSP5-8.5 scenarios (green and pink, respectively) and the low-likelihood-high-impact scenario that includes ice shelf boundary collapse (purple).

Implications for the drinking water sector

Droughts have recently become an almost yearly recurring phenomenon, and water utilities are running into the boundaries of what they can or are allowed to abstract. Surely the most relevant of the findings described above are those on the occurrence of droughts. The occurrence of a new class of models that show an overall decrease in annual precipitation and the decrease of summer precipitation in all model results should make all utilities look forward to the publication of the KNMI’23 scenarios so that they can better plan their water reserves in our warming climate. The inclusion of the complete Rhine and Meuse river basins will be very helpful in modelling the availability of river water for drinking water production, and the higher resolution modelling underlying the new scenarios may be expected to better represent what may be ahead for us. They should give us a better understanding of how we may expect our rivers to behave than earlier studies based on the AR5 and KNMI’14 scenarios, in particular with respect to the low discharge extremes.

The sea level rise scenarios give us something to think about for beyond 2050 in the context of salinization of sources and intrusion of salty water inland in situations of low river discharge. Monitoring the situation in West Antarctica for signs of activation of the ‘low-likelihood-high-impact” scenario remains vital in this context. “Early warning signals of accelerated sea level rise from Antarctica could possibly be observed within the next few decades” (IPCC AR6).

But what about Climate Tipping Points?

One aspect that struck me in the morning program was the limited attention to climate tipping points. These are thresholds in Earth’s climate system that, if crossed, cause natural feedbacks that result in further shifts in Earth’s climate. For example, a little bit of heating beyond such a threshold may result in a significant amount of additional, non-anthropogenic heating, or the crossing of such a threshold may lead to a sudden collapse or reorganization of a large system (such as the extreme sea level rise scenario discussed above by the collapse of the West Antarctic Ice Sheet). Lenton et al. argued in Nature in 2019 that climate tipping points are “too risky to bet against”. And just last month, the same group published a new study in which it was shown that several tipping points are likely to be activated already below 2 degrees global warming relative to pre-industrial temperatures (Armstrong McKay et al., 2022). IPCC’s AR6 report seems to be somewhat less worried about global tipping points, but states that “at the regional scale, abrupt responses, tipping points and even reversals in the direction of change cannot be excluded (high confidence). Some regional abrupt changes and tipping points could have severe local impacts, such as unprecedented weather, extreme temperatures and increased frequency of droughts and forest fires.”

These studies leave me wondering whether it would be wise to consider the possibility of activating other tipping points than that of the West Antarctic Ice Sheet collapse, and describe more similar “low-likelihood-high-impact” cases associated with these in KNMI’s scenarios. A question was asked about the permafrost thaw/methane release tipping point and its positive climate feedback (i.e. the thawing of boreal permafrost is associated with the release of massive amounts of methane, which cause further warming, which cause further permafrost thawing…). The answer was that the models underlying the KNMI scenarios do not include this process. A more generic question about tipping points referring to the Armstrong McKay et al. paper posed in the chat was unfortunately not picked up by the moderator and remained unanswered.

For the WAIS collapse tipping point, it seems that “low-likelihood” should be translated as “we poorly understand the process and can therefore not estimate its likelihood of occurring”. For the permafrost thawing tipping point, similarly, it appears that we do not master the physics of the process sufficiently yet to be able to model and estimate its likelihood. Models can produce likelihood estimates only for those aspects that are well represented in the model and if all relevant aspects are included in the model. In any case, even if we assume these activations of climate tipping points to be unlikely, we (the water sector and society) should understand and include these scenarios in our infrastructure planning, because a low likelihood with a high impact still constitutes a significant risk from a risk management perspective.