Some Jobs for Models - Bumps and Wiggles (2):
Introduction: This is the second in a series on understanding climate variability, global warming, and what we might do about it. The series focuses on the past 30 years and the next 30 years.
Back in October I wrote an entry
about a paper by Judith Lean and David Rind
. They take a position on predictability of a measure of “global warming” on a decadal time scale. This is based on an analysis of past natural variability and the assumption that that variability extends into the future. Another recent paper by Keenlyside et al.,
2008 in Nature
, examines the impact of the variability in the Atlantic Ocean on regional and global climate. Keenlyside et al. project that based only on the projection of the observed Atlantic variability into the future, natural cooling will act counter to the projected human-made warming. Lean and Rind assert that their analysis suggests warming even in the presence of this projected cooling. These are both statements that will be subject to “validation” with observations.
Here is Figure 1 from Lean and Rind:
Figure 1 from Lean and Rind
(2009), Geophysical Research Letters
. Figure taken from tinypic.com
. This figure shows the temperature record and the model representation from 1980 to 2030, the subject of this series of articles.
As we think about the development of climate services and the development of policies to manage, essentially, the average temperature of the planet, it is imperative that we start to understand all of the bumps and wiggles in this curve. It will no longer be adequate to say, simply, that the differences between observed warming and predicted warming are “well within” the normal observed variability.
How do we do this? First the problem needs to be broken down.
In the past few weeks I have seen some outstanding presentations by Professor V. Ramanathan
from the University of California San Diego
. (I recommend specifically this part of Ram’s web page
.) In his talk he started by posing the question of why the warming of the Earth’s surface has not occurred as rapidly as predicted. This question requires following the heat. The answer lies in the ocean, where the heat is not only increasing, but it is increasing at different rates in the different oceans. This difference, due to how circulation varies from one ocean to the next, is predicted by model simulations. (See here
) (Does it make sense that if the ocean can take up heat then it can give it back?)
So if we look at the differences between a climate projection and the subsequent observations, then there are a variety of possibilities of why the prediction might be in error. In the case of the previous paragraph, not all of the heat went into the surface air temperature. Another possibility is to look at the details of the ultimate source of the Earth’s energy, the Sun. As many of this blog’s readers know, the Sun had a very long sunspot minimum, suggesting a decline in the energy coming from the Sun (see here
- It’s time for a solar update.) If we are going to thoroughly explain the differences between predictions and observations, then we will need to quantify, better, our knowledge of the output of the Sun.
The effort to quantify the difference between the predictions and observations reveals errors and inadequacies in the models (and the observations) that need to be addressed. So far in this article, we see the need to better represent the coupling of heat transfer between the atmosphere and the ocean as well as our observations and ability to model the Sun. There are those who would say that the presence of such errors means that the models are not up to the task. There are others who see the identification of errors as the opportunity to improve the models and the quality of predictions.
Explicit identification and correction of errors has been one of the best strategies for improving weather forecasts. During the 1990s, the scientists at the European Center for Medium Range Weather Forecasts (ECMWF
, the leading forecast center), started to focus on forecast busts and trying to identify the cause of busts. (A dense presentation on the subject
) This has been so successful that methods to identify, automatically, sources of errors have been developed. (see learned article from a different Jim Hansen and Emanuel
). For errors to be useful to guide incremental improvements to models and observing systems, the forecast has to have enough skill to provide a worthy estimate to start with. We are at that stage.
Climate modeling, prediction, and validation are moving into a new era. The projections and the validation of those projections are good enough to say definitively that the Earth will warm, sea level will rise, and the weather will change. This is actionable information; we need to prepare for this. We need to try to manage the warming to keep it from getting too large; we have stated in the Copenhagen Accord
that we will keep warming to two degrees. Our models are not up to that task. Striving to understand the bumps and wiggles as a forecast problem will identify errors that will be corrected, improve the quality of the models, define the need for new observations, and set the foundation for meaningful predictions on decadal scales.
r Bumps and Wiggles (1): Predictions and Projections And here is Faceted Search of Blogs at climateknowledge.org