Page images
PDF
EPUB

HEARING OF THE SUBCOMMITTEE ON ENERGY AND ENVIRONMENT
COMMITTEE ON SCIENCE

U.S. HOUSE OF REPRESENTATIVES

on

Countdown to Kyoto-Part 1: The Science of a Global Climate Change Agreement

Tuesday, October 7, 1997

Post-Hearing Questions and Answers

Dr. Alan Robock,

Professor, Department of Meteorology,
University of Maryland

Climate Model Uncertainties

Q1.

Al.

Chapter 5 of the 1995 IPCC scientific notes that, “Clouds, the hydrological cycle and the treatment of the land surface remain the largest areas of uncertainty in the climate models." You have testified regarding convective processes and water vapor. Are there other areas of uncertainty where you think the models need to be improved?

The areas mentioned by IPCC, clouds, the hydrological cycle, and the treatment of the land surface, are the main areas.

Arctic Ice and GCMs

Q2.

A2.

A recent article in the Journal of Climate by D.S. Battisti and others suggests that current climate models do not represent sea ice in the Arctic regions very well, citing the variability in Arctic sea ice the authors say is not captured adequately in General Circulation Models (GCMs). As GCMs predict the greatest warming will occur over the polar latitudes in winter, how much confidence can we place in these predictions in light of the research by Battisti et al.?

I have not studied this paper, but indeed there are many components of climate models that could be improved. If sea ice is wrong, we do not know if it is because of the sea ice parameterization, or ocean currents or clouds. But the greatest warming is simulated over continents, not the sea ice, so I doubt that these inaccuracies are more important than any

Flux Adjustments in GCMs

Q3.

A3.

It appears that one problem with GCMs is the widespread use of flux adjustments— what laymen would call “fudge factors”—to make the models comport with reality. In a recent article in Science, Robert Kerr makes the observation that, “Climate modelers have been ‘cheating' for so long it's almost become respectable." Mr. Kerr goes on to say that this a practice most researchers do not like. Would you explain what these arbitrary adjustments are and why they are necessary?

First, I disagree with this characterization by Richard Kerr. The topic is explained extremely clearly in a recent article by Meehl et al. (EOS, vol. 78, October 14, 1997, pp. 445, 446, 451), so I will just quote from them:

"... Differences between model-simulated and observed quantities indicate systematic errors. Such errors show where and in what ways the models are succeeding or failing to reproduce the behavior of the atmosphere, ocean, sea ice, and land surface under current climate conditions. For example, a typical systematic error is warmer-than-observed sea surface temperatures off the west coasts of the subtropical continents. This error is usually associated with a poor simulation of the low-level stratocumulus clouds. A lack of sufficient cloud cover in these regions allows too much sunlight to the reach the ocean surface. Sea surface temperatures then become warmer than the observed temperatures.

"The atmosphere and the underlying ocean surface interact with each other through fluxes of heat, fresh water, and momentum. These fluxes are determined by net radiation, temperature of the overlying atmospheric surface layer, precipitation, evaporation from the surface, and the force of the wind acting on the ocean surface. The ocean, sea ice, and land surface the influence the atmosphere via surface temperature, soil moisture, snow and sea ice distributions. When the model components are coupled together, errors in the fluxes and corresponding surface conditions result in errors in the coupled climate simulation of temperature, pressure, moisture, winds, ocean currents, and rainfall. A technique call flux adjustment (also referred to as flux correction) is sometimes used to overcome these simulation errors and bring the couple climate simulation into better agreement with observations. About half the models in CMIP1 [Coupled Model Intercomparison Project 1] use this technique.

"Flux adjustments are designed to bring the couple model simulation into closer agreement with observations. As such, there are constant additive terms, not interactive or restorative terms, which modify fluxes between model components. Therefore since terms are simply added and the model is not being restored to some observed state, the model is still free to drift away from present-day climate.

"In the case of a lack of sufficient low-level clouds in the example mentioned above, the flux adjustment would be calculated to reduce the heat flux into the

ocean. Thus the sea surface temperatures would be somewhat cooler and agree better with observations.

"Once calculated, the flux adjustments remain constant in model simulations of present-day and future climate. Flux adjustment ensures that the physical climate feedbacks in the models are operating in the correct climatic range so that perturbations are appropriately modeled. For instance, 'albedo feedback' is unimportant for climate change. Warming of the surface melts snow and ice, thereby reducing the surface albedo. This leads to an enhances absorption of incoming solar radiation that heats the surface, more snow and ice melt, and so on in a feedback loop (cooling drives the loop in the opposite sense). If the control climate simulated in the models has too much or too little snow and ice, the nature of the response to a climate perturbation will be affected.

"Since the flux adjustment makes the coupled model simulation agree better with observations, most coupled models that use flux adjustment simulate present-day climate better than the models that do not. Nevertheless, the various component models used by different modeling groups tend to have similar systematic simulation errors before the flux adjustment technique is applied in the coupled simulations. If the feedbacks in a non-flux adjusted coupled models are affected (the albedo feedback, for instance), the climate simulated by an unflux-adjusted model would be compromised. Conversely, the magnitude of the flux adjustmentas in the case of too few low-level clouds in the simulation-is a measure of the mismatch between component models. Such inconsistencies could perhaps mask the lack of a missing physical feedback mechanism in the coupled model.”

So you see that flux correction is a technique to make climate models perform more realistically, and using it gives us greater confidence in their results. As models become better, the amount of flux correction used goes down.

Parameterization in GCMs

Q4.

A4.

Dr. Spencer testified that current GCMs parameterize certain processes, like cloud formation, rather that incorporate the physics in them. Please explain the difference between these two approaches. Which is better?

This is really not a valid distinction. Because clouds have a smaller size than the typical grid size used in climate models (say 200x200 miles), an equation is developed to calculate the cumulative effect of all the clouds in that grid box, rather than model each individual cloud separately. This is called a parameterization. It is not feasible to model each individual cloud, as the computer time required would be too large. But even if one had the computers, the individual clouds are not predictable beyond a few hours, so you would be modeling a group of clouds that would be different from the actual ones that would occur. Furthermore, models of individual clouds are full of parameterizations of processes they cannot consider in detail and individually, such as the growth of cloud droplets, or the

The question is just to what accuracy and detail is it necessary to consider different climate processes in order to simulate correctly the behavior of the entire climate system. The range of different cloud parameterizations now in use in climate models represents one of the areas which potentially can be most improved. But this will require continuing and expanding existing programs in observations and modeling of the details of cloud behavior.

There is nothing in the current parameterizations of clouds that would make one believe that they are biased in one way or another in their effects on climate. Improved, more detailed consideration of clouds will change model sensitivity, but there is no way to know ahead of time how large or in what direction these changes will be, as clouds interact with the rest of the climate system, with very complicated feedbacks.

Moist Convective Processes in GCMs

Q5.

A5.

Concerning GCMs, you testified that, “If you put more energy in than is being taken out, the climate warms." Dr. Spencer's testimony questions whether moist convective processes in GCMs are realistic enough to capture the ways in which the tropospheric temperature profile fluctuates naturally in response to the transport of heat from the surface to the upper troposphere, thus keeping the system in some sort of radiative balance. Do you agree that this is an issue requiring further investigation?

I agree that this area of climate models, and many others, could be improved with more research. But as I stated above, I do not think that this particular aspect of climate models is now biased in a way to make the models too sensitive. Unknowns by definition are unknown, and when we improve models their behavior will be different, but there are many possible changes that can occur as different components become more realistic.

Models and Regional-Scale Predictions

Q6.

A6.

Q7.

The 1995 IPCC stated that, “The reliability of regional-scale predictions is still low, and the degree to which climate variability may change is uncertain." Do you agree with this statement, and if not, why not?

I agree.

Just recently, the IPCC released a report, IPCC Regional Impacts Special Report, that assessed the “vulnerability of natural and social systems of major regions of the world to climate change.” Despite the caveats the report contains that the analysis does not contain “predictions that the climate will change by specific magnitudes in particular regions or countries,” the impression with the general public is that these are likely results of increasing CO2 concentrations. Are predictions of regional

A7.

Q8.

A8.

I think the best way to do such an analysis at the present time is to consider different scenarios of future climate change, and to assess the vulnerabilities of different human activities in different parts of the world to a variety of climate changes. The analysis should particularly look for threshold effects, where large impacts could result from small incremental changes beyond a particular value.

The 1995 IPCC notes the prospect of “more severe droughts and/or floods in some places" but also suggests the possibility of “less severe droughts and/or floods in other places." Your testimony states that “The threat of mid-latitude drought, and resulting crop failures in the breadbaskets of the world, is a significant potential danger." How likely is this to happen and do we really know?

The suggestion of increased variability of climate, with more extremes like floods and droughts, is not very strongly supported by existing research. Some results suggest this and others do not, so I would say that we do not now know whether this will happen.

My statement, however addressed another issue. Virtually all models suggest that soil moisture will decrease in the summer in the midlatitudes as climate warms. This is because in a warmer world, the atmosphere can hold more moisture, and evaporation from the surface increases. Precipitation also increases (linearly), but evaporation increases more rapidly (exponentially), so the net effect is for the soil to dry. This I see as a significant potential danger.

Theoretical Limits of the Predictability of Global Climate

Q9.

A9.

Are there theoretical limits on the predictability of global climate? Please discuss.

Weather can only potentially be predicted for a few weeks into the future. Climate prediction is based not on knowing the initial state of the atmosphere, such as the locations of storms, but on predicting or assuming changes to the boundary conditions of the climate system, and then predicting the envelope for the range of weather variations in the new climate. The boundary conditions include those influenced by humans, such as changing greenhouse gases, aerosols, ozone, contrails, and land surface, and natural ones, such as solar variations and volcanic eruptions. Predicting climate involves two separate steps. One is to improve our theoretical knowledge, which we incorporate into climate models, so that we can calculate the response to changed boundary conditions. (This work also involves a large observational component to learn how the climate system interacts and to monitor changes.) The other is to predict how the boundary conditions will change, and this involves predicting human behavior, an area beyond my expertise, as well as predicting solar and volcanic changes, something not done well now.

« PreviousContinue »