Since there are no currently active contests, we have switched Climate CoLab to read-only mode.
Learn more at https://climatecolab.org/page/readonly.
Skip navigation
4comments
Share conversation: Share via:

Patrick Ray

May 23, 2016
08:36

Fellow


1 |
Share via:

Management via Measurement team,

Last day to make refinements to your proposal. Looking forward to seeing the final version.

I have a personal interest in your proposal as it intersects with my work substantially. Here's a question - where I apply them, the GCMs have very poor performance in precipitation extremes, both because they have difficulty at the daily and sub-daily timescale (but are more useful at the longer timescales), and because they often struggle with the underlying climate drivers (whether sea surface temperature oscillations affecting regional climate or local convective behavior contributing directly to extremes). If possible, could you comment on the performance of the GCMs you have used to give meaningful information for Bayesian updating of historical observations of precipitation extremes?

Thank you. Best of luck.


Colin Sullivan

May 23, 2016
10:44

Member


2 |
Share via:
Proposal
contributor

Thanks, patrickray.

Indeed your point is correct: GCMs gave a hard time portraying precipitation (including extremes) for a variety of reasons. This is one of the longstanding, crucial gaps in climate science.

Precipitation extremes are in many cases expected to increase in intensity, duration, and/or frequency as a function of climate change given theory, evidence from observations, and GCM projections.

The Clausius Clapeyron (CC) relation shows that, under ideal conditions, saturation vapor pressure (roughly speaking, the capacity of the local atmosphere to hold moisture) increases exponentially with increasing temperature. Thus under global warming, all else held equal, one could expect precipitation to change in terms of distribution, specifically with longer duration between precipitation events but where those events are more intense. This has been referred to as CC scaling. Of course the rate of scaling (% increase in extreme precipitation per degree increase in Kelvins, say) varies significantly by region. Additionally, the intensity of precipitation extremes depends not only on local mean air temperature but also on moisture dynamics including horizontal and vertical wind velocity. However, generally it would be impossible to assess a GCM's ability to simulate those dynamics, since observations for those processes are not even available. In contrast, in many regions of the world like the US, high quality observations for both air temperature and precipitation do exist. Furthermore, the CC relation is a universal factor although not always the dominant determinant of precipitation extremes intensity.

In this model, we use this knowledge with the following hypothesis: a skillful GCM should show a relationship between same day, local air temperature and extreme precipitation that is the same as the relationship shown by observational data, for a given location.

In other words, is the degree of adherence to approximate CC scaling, as shown by an GCM, appropriate given the relationship between temperature and precipitation extremes exhibited in historical observational data? If observations show a relationship that suggests temperature is not the main driver of extreme precipitation, a GCM should do the same, lest it may be simulating physics improperly in that region. Even if that GCM shows a similar marginal distribution of extreme precipitation as compared to historical observations, if it does not also appropriately capture the precipitation extremes' dependence on temperature as shown by precipitation, it is more likely that GCM only appeared to be skillful and for the wrong physical reasons. Hence, it should be treated as less reliable. This is done formally through a Bayesian architecture.

From a business perspective, the most important feature of this model is that it seeks to quantify uncertainty in any extreme (say, a 5-year return level) -- if all GCMs are doing a poor job, the uncertainty will be wider. 

Hope this is helpful.

Thanks!


Patrick Ray

May 23, 2016
01:25

Fellow


3 |
Share via:

Very thoughtful response.

Thank you. I look forward to discussing this with you in the future.

Good luck.


Colin Sullivan

May 24, 2016
03:13

Member


4 |
Share via:
Proposal
contributor

Thanks for your question!

If any others come up feel to ask here or get in contact directly.