I was fortunate enough to attend the CMIP 2026 community workshop in Kyoto, Japan from 09-13 March 2026. The CMIP community as a whole works around the word “uncertainty”and in this context it is a technical challenge like noise of the climate system, model configuration uncertainties and it mostly is a measurable variable. If you were to ask the same question to a stakeholder, the answers would come out completely different. To them- port authorities, local governments- uncertainty is closely linked to the language of budget, risk and safety. In the long run, we as modellers need to play the game with stakeholders in converting this raw data into a shared strategy.
The modellers approach
As a case study, in the Indian Ocean (IO), a highly dynamic basin in terms of ocean surface waves, there isn’t information on its projected changes within the CMIP framework. Despite being identified as one the climate drivers and the role wind-waves play in coastal stability, there is persistent lack of information in CMIP6- and the latest CMIP7- to be used for regional climate studies. To address this, wave modellers often use wind data, sea ice concentration, currents etc to force third generation wave models to produce wave projection datasets.
To a modeller, the validation of these data sets with ground truth (observation data) is the first and foremost step. This step is crucial in understanding and acknowledging that even the best models carry uncertainties. The use of this word here means a deviation of the model results from the observed truth of the sea state. In this framework we use tools like Taylor diagram, M-score, Empirical Orthogonal Function to compare our results and quantify the biases and uncertaitnties within the model in representing the wave fields of the IO. For a modeller, a correlation of 0.84 between the model and insitu dataset is an example of “robust forcing”.
The stakeholder’s approach
Imagine conveying these m-scores and correlations to a port authority (a port that is vulnerable to high waves and has high volatility towards flash flooding events). To this stakeholder, an uncertainty for example, the the presence of Southern Ocean swells in North IO not being modelled well is more of a reliability issue than a statistic. In this case, the uncertainty we speak of is not a measurable variable but the difference between a coastal or offshore project that succeeded or failed. The stakeholders are more concerned about the return period of these swell events more than the average or trend value of wave height. This communication breakdown between a modeller and a stakeholder often arises due to our focus on one word that means entirely different things in different contexts and different communities.
I am unable to conclude this blog with anything more insightful as this entre thing has just been a small thought in my brain and I wanted to put it out. I will be collecting more data on this particular topic and forming a robust foundation of knowledge on how these gaps can be bridged. While we as a community putforth petabytes of data within different frameworks, it would be interesting to know the number of sea walls, number of protected coasts and number of coasts reclaimed from the verge of complete erosion.
