onsdag 29 juli 2009

Global Climate Modeling: Fate of Humanity?

Policies to control climate change must be based on global climate models which are mathematical models of the circulation of the atmosphere and the ocean of our planet in the form of the Navier-Stokes equations of fluid flow plus thermodynamics with various energy sources, referred to as General Circulation Models GCM.

Recently G8-leaders have decided to put a limit to global warming of 2C.

The key question is the reliability/accuracy GCM. Since GCM is based on computational solution of the Navier-Stokes equations, also referred to as Computational Fluid Dynamics CFD, it is of interest to investigate what is state-of-the-art of CFD.

The areodynamics of flight is governed by the Navier-Stokes equations, and one may ask if today CFD allows computational modeling of the take-off of a jumbo-jet, for example. The answer is a disappointing NO: We have to wait 50 years before computers are powerful enough! Says the state-of-the-art expert Parviz Moin.

Of course you ask: If we cannot computationally model the flight of an airplane, how can we expect to model the immensely much more complicated system of the the global climate?

Or maybe more is possible than what state-of-the-art says, see Fear of Flying and  Why It Is Possible to Fly.

Maybe it is possible to predict global climate with better Navier-Stokes solvers, which can be used today on existing computers. Waiting 50 years for the predictions may not be very meaningful, like waiting one week to get the prediction of tomorrows weather.

As concerns the state-of-the-art of CFD and fluid mechanics in general, see the Interview with Editors of Journal of Fluid Mechanics JFM.

Since the fate of humanity critically depends on the reliability and accuracy of computational global climate modeling, it would seem that no efforts should be spared to bring forward the best possible competence in this area. It does not seem to be represented by JFM, which poses a serious problem since JFM is supposed to be the leading journal of fluid mechanics. What can be done? But who cares? Do you?

4 kommentarer:

  1. CFD is of course a subtle matter, and weather prediction is consequently somewhat limited.

    Climate projection as opposed to weather prediction depends on CFD but not as tightly as might be suspected. The system is dissipative and the dominant uncertainties are not fluid dynamical ones. Rather coarse CFD cores suffice as the large scale statistical closures of boundary forcings and moist convection dominate the dynamics problem. Radiative transfer problems are equally dependent on statistical closures.

    It is not obvious that improved CFD will improve climate models in the foreseeable future. Performance and computational scaling are of greater interest than computational accuracy for those working on the dynamics of air and ocean. It is for this reason that the CFD community is not engaged. It is not really an interesting CFD problem.

    That is to say that existing CFD seems sufficient to ensure that models with differing cores and similar forcings or parameterizations have trajectories that are better matches than are models with similar cores and different forcings or parameterizations.

    It would be a good thing if more mathematical and computational physicists would take an interest in the climate problem. What usually seems to happen is they enter with too much confidence and start attacking the problem in inopportune ways.

    While the current state of the art is rather underestimated, there are good reasons for concern that little additional progress is immediately foreseeable. This being the case it is necessary to make decisions based on the limited information already in hand.

    Unfortunately in the present politically charged environment it is rather difficult to convey nuances of confidence in various aspects of the situation.

    On the other hand, one hopes that fluid dynamicists will at least be in a position to draw on the rich experimental literature of the 1960s regarding rotating and stratified fluids, and understand that the possibility of abrupt flow regime shifts is a significant and concerning one.

  2. Global climate prediction has to use best possible CFD to deliver best possible results. There is no reason to use use second-rate CFD, unless you believe that all CFD is nonsense. Do think that way?

  3. There is no reason to use "second-rate" CFD, but there is no reason to throw too much computational resources at the CFD, because they don't seem to dominate the uncertainties of the problems addressed by GCMs as matters stand at present.

    Certainly efforts to improve the CFD are ongoing, but in fact the boundary conditions, the forcings, and the parameterized processes are much worse than the CFD as things stand.

    Nevertheless the models are already suitable for many purposes.