Climate Science: Roger Pielke Sr. Research Group News

Professor Chris Castro alerted us to the following paper;

Barnett, T. P, and D. W. Pierce, 2008: When will Lake Mead go dry? Water Resour. Res., 44, W03201, doi:10.1029/2007WR006704.

The abstract reads

“A water budget analysis shows that under current conditions there is a 10% chance that live storage in Lakes Mead and Powell will be gone by about 2013 and a 50% chance that it will be gone by 2021 if no changes in water allocation from the Colorado River system are made. This startling result is driven by climate change associated with global warming, the effects of natural climate variability, and the current operating status of the reservoir system. Minimum power pool levels in both Lake Mead and Lake Powell will be reached under current conditions by 2017 with 50% probability. While these dates are subject to some uncertainty, they all point to a major and immediate water supply problem on the Colorado system. The solutions to this water shortage problem must be time-dependent to match the time-varying, human-induced decreases in future river flow.”

The text includes the statements

“We consider human-induced reductions in runoff of 10 to 30%, in accordance with estimates from global climate models and statistical analysis, and take these reductions to be linear in time over the next 50 years (i.e., runoff slowly decreases until it reaches a total reduction of, say, 10% below current levels in 2057)”;

“….we begin with deterministic estimates of when the live storage will be depleted by global warming-driven runoff reductions alone, without the outside impacts of evaporation and natural variability in the river flow”;

“The climate models which have produced estimates of decreasing runoff have a host of problems of their own in handling the water budget from coarse resolution (little in the way of Rocky Mountains) to the variety of ways they handle soil processes and vegetation representations. However, a recent study of changes in hydrology of the western U.S. over that last 50 years shows several of the models, when run with observed anthropogenic forcings, reproduce extremely well the observed changes in river flow timing, snow pack decline and increasing air temperatures in the western United States [Barnett et al., 2008]. So these models, while not perfect, have a message to tell; a message supported by their ability to reproduce well the last 50 years of multivariate hydrological observations“;

and 

“…..the Colorado River will continue to lose water in the future, if the global climate models are correct.”

This paper correctly identifies that there is risk associated with the limited water available from the Colorado River. Indeed their statement that

“Tree ring data suggest the long term flow of the Colorado experiences more variability than has been observed over the last century [NAS, 2007]. These data also suggest prolonged droughts far worse and more extensive than seen in the last 100 years of flow record on the River are possible”

shows that the water resource is at risk regardless of how humans have altered the system. This is a conclusion we also reached in our paper

 Pielke Sr., R.A., N. Doesken, O. Bliss, T. Green, C. Chaffin, J.D. Salas, C. Woodhouse, J.L. Lukas, and K. Wolter, 2005: Drought 2002 in Colorado - An unprecedented drought or a routine drought?Pure Appl. Geophys., Special Issue in honor of Prof. Singh, 162, 1455-1479, doi:10.1007/200024-005-2679-6.

However, the paper suffers from their reliance on the multi-decadal global models as quantitative predictions of what will happen in terms of climate in the coming years. They even recognize this in their text “…..the Colorado River will continue to lose water in the future, if the global climate models are correct.” 

Thus while Climate Science agrees that there is a significant concern on water available from the Colorado River, and planning should be a major priority with respect to long-term drought, the multi-decadal global model predictions are just hypotheses and their use as part of the computation as definitive, skillful predictions to present quantitative probabilities of Lake Mead drying out is misleading to the policymakers. This is yet another example of overselling the skill that exists in using these models as predictions. The large amounts of precipitation this past winter (2007-2008) in large areas of the West should be a wake-up call on the serious limitations of the IPCC models.

The discussion with Real Climate continues. The updated comments as of Saturday April 26 are at

Comment On Real Climate’s Post On The Relevance Of The Sensitivity Of Initial Conditions In The IPCC Models

Follow Up (April 27 2008)

Ray - In searching for what Professor Lorenz has said on this issue, please  see Chaos Avant-Garde: Memories of the Early Days of Chaos Theory

 In this essay he writes,

“Returning now to the question as originally posed, we notice some additional points not yet considered. First of all, the influence of a single butterfly is not only a fine detal - it is confined to a small volume. Some of the numerical methods which seem to be well adapted for examining the intensification of errors are not suitable for studying the dispersion of errors from restricted to unrestricted regions. One hypothesis, unconfirmed, is that the influence of a butterfly’s wings will spread in turbulent air, but not in calm air”

This certainly would rule out the butterfly in the jar! More importantly, he recognized that there remain questions about the “butterfly effect”, one of which is when small pertubations result in altering larger scale atmospheric flow, and when they do not.

Sixth Update (April 27 2008)

A Further Reply By Ray Pierrehumbett

 [Response:Roger, I can’t make sense of what you’re trying to say here. For those picokelvins of temperature to be lost to space, first they have to appear in the atmosphere as an increase of temperature, right? So there you have your change of one digit in the initial conditions, just like in Lorenz’s example. And your statement is just flatly inconsistent with thermodynamics. The butterfly dissipates heat locally, and that heat will be gradually diluted over a larger and large area. So just divide by Cp and there’s your answer. Do you think there’s some way to magically teleport the heat away, leaving the fluid to heal back to exactly the same condition it would have had without the flap? That’s really a stretch. Your remarks about simple models and GCM’s don’t make much sense to me either. The GCM doesn’t resolve butterfly-scale motions, but once you have influenced a dynamic variable (e.g. temperature) at a resolved scale, any number of actual twin experiments in GCM’s confirm the divergence. If you are claiming there’s some fundamental difference between sensitive dependence to large scale changes in a GCM and sensitive dependence in the atmosphere, I’d like to see some evidence to back up that claim. The success of GCM’s in short term weather forecasting would be pretty much impossible to reconcile with such a claim. –raypierre]

 My Reply

You are correct in that you and I probably agree on most issues in chaos and nonlinear dynamics. All NWP and climate models show the sensitivity of large scale circulation features to initial conditions when perturbations are inserted in their initial state or in their parameterizations (these are all much larger effects than the energy that a butterfly places in the system). We also agree that the added heat from a butterflies flapping wings results in a slightly different system than if this flapping did not occur. However, the issue is whether the heat (the “information”) from this effect can translate (teleconnect) to larger scale so as to result in alterations in large scale features. 

Even Issac Held seemed to indicate that there is a lower limit to when this upscale effect can occur (i.e. this ability disappears when the flow becomes laminar); he said in this thread

“the scale of the perturbation has to be larger than what is often referred to as the Kolmogorov microscale, the scale below which the flow is effectively laminar, to avoid being damped out immediately. This scale is typically a few millimeters in the atmosphere….”

I agree with this, but maintain that the smallest turbulent scales also are damped out due to the physics of non-motion transfers (i.e. radiative transfers) of energy. I have been in communication with Professor Ekyholt on this question, and he and I agree that you are misinterpreting the butterfly effect for very small scale perturbations. We will be preparing a paper on this to demonstrate that there is  lower limit to which the “butterfly effect” applies.

On a separate note, I see commenters on this thread are somehow skewing this discussion to be on climate change. It is not. This issue of the scale at which the “butterfly effect” occurs is a pure discussion of the science such as we all used to have as graduate students and need more of!

Also, you questioned as to why Roy Spencer posted a guest weblog. The answer is that he has introduced a novel and important new perspective into how variations in atmospheric/ocean circulations can result in alterations in the global average radiative balance. Disagreements with his results and conclusions should be on his science. I invite others (including any interested Real Climate climate scientist) to post unedited guest weblogs on Climate Science.  

 Additional Response From Ray  Pierrehumbert

 [Response:Regarding the butterfly in the room — even in a jar in the room — sure I think it’s likely that it would ultimately affect the large scale weather. Look at it this way: Temperature has a dynamic influence through buoyancy. The heat dissipated by the butterfly might warm the room by a few tens of microkelvins, say. That increased temperature will change the heat flow between the house and the environment, which will ultimately change the temperature of some parcel of air by a few nanokelvins. Then before you know it, some parcel of air the size of the state of Illinois has a temperature different by maybe a few picokelvins. I guarantee that if you take a GCM and change the temperature of the air over Illinois by a few picokelvins (given sufficient arithmetic precision) that that will lead to divergence of the large scale forecast given infinite time. I have seen no indication either in dynamical systems theorems or in numerical experiment to suggest that anything else would be the case. –raypierre]

My Reply

Ray- We certainly disagree with respect to the butterfly in the room in a jar.  :-). Other readers of Real Climate (and Climate Science) can make up their own minds on this.

You are, however, taking the concept of chaos too narrowly and are focusing on idealizations (simple illustrative models and GCMs) of  how the real atmosphere (and climate system) works. You are ignoring the consequences of the dissipation of kinetic energy into heat within a open system. The “picokelvins” of heat, even if they could cause such a temperature perturbation over the state of Illinois (which it would not), would be lost to space long before an “infinite” time were reached.

Fourth Update (April 26 2008)

 Additional Response From Ray  Pierrehumbert

[Response:Have a look at Isaac’s remark above. I think what you probably have in mind is the possibility that if a perturbation is at a scale where you have primarily downscale energy cascade to the dissipation range, it might never project on the large scale quantities whose behavior determines large scale predictability loss. Given the nature of turbulence, it is hard to absolutely exclude this possibility a priori, but for this to happen, there would have to be ZERO leakage to large scales. Not just small but ZERO. That is exceedingly unlikely, and would be contrary to most of what is know about turbulent cascades. As a practical matter, I do agree that if the initial perturbation is at sufficiently small scales, the projection on large scales would be small enough that it could take an exceedingly long time before it affected the evolution of the large scales. –raypierre]

My Reply [posted on Real Climate]

Ray - Thank you for getting involved in this discussion.  The question of the leakage time scale is, of course needed, in order to determine when the exceedingly long time scale becomes infinite (in terms of where the heat goes).  If we both agree that ALL of the turbulence quickly dissipates into heat when the flapping stops, then what is your estimate of the residence time of this heat within the atmosphere before it is lost to space?

Also, as another thought example, if a butterfly flaps its wings inside a room with the doors shut, would you still maintain that this has an influence on atmospheric circulation at large distances? All of the heat generated would be absorbed by the walls of the room, and subsequent heat conduction is, of course, laminar.  An analogous behavior will occur in a very stable boundary layer (and any region of the atmosphere for such small perturbations), and if we can agree on this “exception” than we have made progress in understanding this issue. My point here is that if there is an part of the process which results in complete loss of the turbulent flow, then it is not communicated over large distances.

Issac’s Held’s answer also actually contains part of the answer on this issue.  If the turbulence dissipates into heat, as  illustrated in the above example,  than its further behavior can be described by non-turbulent behavior. As he explained, he was “was thinking that the scale of the perturbation has to be larger than what is often referred to as the Kolmogorov microscale, the scale below which the flow is effectively laminar, to avoid being damped out immediately. This scale is typically a few millimeters in the atmosphere “.  This is what occurs with the flapping of the wings of a butterfly; all of its energy dissipates into heat and the spatial structure of this heated air is less than a few mm.  To disprove this total transfer downscale, one would have to show that a coherent turbulent structure remains  and becomes progressively larger in scale and/or is monitored propagating away from the location of the flapping wings as a coherent disturbance of the air flow; in both cases,  while still retaining the conservation of total energy.  Since the total energy of the flaps of the butterfly’s wings must be accounted for (as kinetic energy in the turbulence, heat) what is your estimate of the magnitude of this energy that reaches thousands of kilometers away, as well as the path this energy would take to get there?

Third Update (April 25 2008)

Further Response From Gavin Schmidt

 Response: As we said above, this is what you believe. Why you accused us of misrepresenting you is a mystery. However, your claim about Ekykholt’s belief is contradicted by his quote above. He states very specifically that exponential growth saturates at the time the perturbation reaches the size of the attractor. That, for the atmosphere, is very large indeed and is certainly large scale enough to encompass storms thousands of miles away. Isaac can certainly speak for himself, but as far as I know there is no demonstration that there is a minimum scale below which perturbations do not grow. Such a thing may exist, but your certainty on the matter seems a little overconfident. Perhaps you’d care to point out a reference on the subject? - gavin]

 My Reply [posted on Real Climate] Gavin  - I am glad this discussion is continuing. I will be having more to say on this next week in a weblog on Climate Science, however, you are failing to distinguish between an open and closed system, and between the real world and models.  With nonlinear atmospheric models such as analyzed by Professor Lorenz, the results for large scale features are sensitive to the initial conditions regardless of how small they are. This is because the system is closed.  The real world climate system, however, is not closed, such that energy (i.e. in the form of heat) can leak out of the system.  In the case of such a small perturbation as the flap of a butterfly wing, the kinetic energy of the small amount of turbulent air that it generates will quickly dissipate into heat, once the flapping stops. Radiative loss of this heat to space will prevent the flapping to have any effect at large distances.  

This is one of the reasons that you are mistaken in stating that “there is no demonstration that there is a minimum scale below which perturbations do not grow.” If a perturbation in the system (i.e. the atmosphere) dissipates into heat, it can be lost to the system before affecting atmospheric features at large distances. I will have more on this topic on my weblog next week, and will post a comment on Real Climate when it appears.

Second Update (April 24 2008)

 Gavin Schmidt has replied

Response:You misinterpreted this back on the original thread and you are misinterpreting it here again. However, just repeating the same argument is pointless. Since I agree with Dr. Eykholt’s statement, and so do you, let’s just leave it at that. (if other readers are interested in what this is about, please go to the original thread. The clue is that ‘larger scales’ in the Eykholt quote means the attractor itself (i.e. climate), while RP thinks he means the large scale flow (i.e. the specific position on the attractor)). - gavin]

and 

My Response is

 Gavin- I agree readers can go through the thread to see the discussion. However, you are misrepresenting my views. Rich Eykholt and I are in 100% agreement on this subject. The question that was being discussed is whether an atmospheric perturbation as small as a real world butterfly could actually affect large scale weather features thousands of kilometers away. The answer, as given by Professor Eykholt, is NO under any circumstance. The perturbation has to be much larger (Issac Held, as I recall said meters in his NPR interview; I suspect it is a few kilometers or more) for a perturbation to affect an atmospheric feature thousands of kilometers away.

This issue, based on our disagreement, would benefit from further quantitative evaluation with both analytic and numerical models. We do have papers on the use of analytic models to examine chaos and nonlinear dynamics which document that we are quite familiar with the subject of sensitivity of the climate system to initial conditions; e.g. see

Pielke, R.A. and X. Zeng, 1994: Long-term variability of climate. J. Atmos. Sci., 51, 155-159.
http://climatesci.colorado.edu/publications/pdf/R-120.pdf

Update (April 24 2008)  : Following is my comment, Gavin Schmidt’s reply, and my response on Real Climate

 Roger A. Pielke Sr. Says:
23 April 2008 at 10:15 AM

Please see http://climatesci.org/2008/04/23/comment-on-real-climates-post-on-the-relevance-of-the-sensitivity-of-initial-conditions-in-the-ipcc-models/

[Response:In the linked piece, you very clearly state that you do not believe that the real world is sensitive to initial condition variations like butterflies. That is all we are discussing here. If you now think that it is, feel free to expound on your viewpoint. We were just trying to make sure that a diversity of points was presented. - gavin]

My Reply: 

Gavin - Thank you for posting my Climate Science link. In terms of actual butterlies, this is clearly explained by an expert in the physics and mathematics of nonlinear dynamics and chaos in geophysical flows, Professor Richard Eykholt (see http://climatesci.org/2005/10/12/more-on-the-butterfly-effect/), where he writes “

Roger: I think that you captured the key features and misconceptions pretty well. The butterfly effect refers to the exponential growth of any small perturbation. However, this exponential growth continues only so long as the disturbance remains very small compared to the size of the attractor. It then folds back onto the attractor. Unfortunately, most people miss this latter part and think that the small perturbation continues to grow until it is huge and has some large effect. The point of the effect is that it prevents us from making very detailed predictions at very small scales, but it does not have a significant effect at larger scales. 

Richard Eykholt”

Original Post

Real Climate has published a well written summary of the seminal accomplishments of Professor Ed Lorenz in the field of deterministic chaos and nonlinear dynamics (see). Professor Lorenz’s contribution to the understanding of the mathematics and physics of geophysical flows (and other dynamic systems) has altered how the science community investigates these processes. I had the opportunity to sit and talk with Professor Lorenz during one of his trips to Colorado State University, and enjoyed and learned from his perspective on the nonlinear aspects of the climate system including its behavior, as with any other nonlinear system with strong feedbacks, as being sensitive to initial conditions.

At the end of the well deserved recognition to Professor Lorenz, Real Climate writes

“So what does this have to do with the IPCC?”

Real Climate then writes

“Even though the model used by Lorenz was very simple (just three variables and three equations), the same sensitivity to initial conditions is seen in all weather and climate models and is a ubiquitous phenomenon in many complex non-linear flows. It is therefore usually assumed that the real atmosphere also has this property. However, as Lorenz himself acknowledged in 1972, this is not directly provable (and indeed, at least one meteorologist doesn’t think it does even though most everyone else does). Its existence in climate models is nonetheless easily demonstratable. “

I am the “one meteorologist”.  Real Climate refers to one of the Climate Science weblogs on this issue that was published (see).

However, Real Climate is wrong in its statement on my research conclusions!  I have written several papers on climate as an initial value problem: e.g. see

Pielke, R.A., 1998: Climate prediction as an initial value problem. Bull. Amer. Meteor. Soc., 79, 2743-2746.

Pielke Sr., R.A., G.E. Liston, J.L. Eastman, L. Lu, and M. Coughenour, 1999: Seasonal weather prediction as an initial value problem. J. Geophys. Res., 104, 19463-19479.

Rial, J., R.A. Pielke Sr., M. Beniston, M. Claussen, J. Canadell, P. Cox, H. Held, N. de Noblet-Ducoudre, R. Prinn, J. Reynolds, and J.D. Salas, 2004: Nonlinearities, feedbacks and critical thresholds within the Earth’s climate system. Climatic Change, 65, 11-38.

Real Climate should report  accurately on the research of others.

What we disagree on is whether the multi-decadal global climate model predictions can be used to accurately quantify the degree of nonlinearity and predictability of the real world climate system (the nonlinearity of the climate system is shown, for example, in the Rial et al paper).

Real Climate, however, reports on the use of a model to investigate this issue. This is a typical mistake they are making; a model is itself a hypothesis and cannot be used to prove anything! The multi-decadal global model simulations only provide insight into processes and interactions, but we must use real world data to test the models. So far, the models have failed, for example,  in their ability to accurately predict the regional weather and climate features we discuss in the Rial et al paper. Lets have more accurate reporting on Real Climate.

There is an interesting exchange of views by Bill Gray, Tom Knutson and Steve Lyons at the Bahamas Weather Conference (Thanks to Bob Ferguson for alerting us to this short video.

Climate Science has just one comment on this video. Tom Knudson claims that the global climate models can be used to test theory (such as his claim on the dominance of CO2 as the driver of climate change).

However, Tom Knudson makes the very serious mistake of stating that models can test his claim.  The models are hypotheses and cannot test anything! They can be used to improve our understanding of how a system works, but their results must be tested against real-world observational data.

His failure to properly communicate what models really should be used for, unfortunately, permeates popular and media preceptions of the climate change issue, and is resulting in very poor policy decisions (e.g., see the April 15th, 2008 post on Prometheus entitled “Biofuels and Mitigation/Adaption” ).

In the peer reviewed literature, I have emphasized that the IPCC multi-decadal global climate runs, while they refer them as “projections” and also “scenarios” are actually model sensitivity studies since all of the important climate forcings and feedbacks are not included; e.g. see

Pielke, R.A., 1998: Climate prediction as an initial value problem. Bull. Amer. Meteor. Soc., 79, 2743-2746.

Pielke Sr., R.A., 2002: Overlooked issues in the U.S. National Climate and IPCC assessments. Climatic Change, 52, 1-11.

A summary of the types of climate models is given in the weblog

What Are Climate Models? What Do They Do?

Now, from an unlikely source (Real Climate) have come the statements

“A scenario only illustrates the climatic effect of the specified forcing - this is why it is called a scenario, not a forecast. To be sure, the first IPCC report did talk about “prediction” - in many respects the first report was not nearly as sophisticated as the more recent ones, including in its terminology. “

“One should not mix up a scenario with a forecast - I cannot easily compare a scenario for the effects of greenhouse gases alone with observed data, because I cannot easily isolate the effect of the greenhouse gases in these data, given that other forcings are also at play in the real world.”

Real Climate states that the scenarios can

“….. become obsolete, and….. cannot be verified or falsified by observed data, because the observed data have become dominated by other effects not included in the scenario.”

This is the definition of a sensitivity experiment! In other words, policymakers are being given global and regional multi-decadal model results by the IPCC which are not predictions but sensitivity model runs since a variety of important first order climate forcings and feedbacks are not included in the models! [e.g. as reported in Radiative forcing of climate change: Expanding the concept and addressing uncertainties. ]. Real Climate now has finally reported to us this serious limitiation to  the interpretation of the results from climate models.

There is an excellent proposal entitled  A Proposed Test Suite for Atmospheric Model Dynamical Cores by Christaine Jablonowski of the University of Michigan which Climate Science was alerted to from a February 3 2008 weblog on Climate Audit.

The idea for this test is described as

   “Tests of atmospheric General Circulation Models (GCMs) and, in particular, tests of their dynamical cores are important steps towards future model improvements. They reveal the influence of an individual model design on climate and weather simulations and indicate whether the circulation is described representatively by the numerical approach.
      Testing a global 3D atmospheric model is not straightforward. In the absence of non-trivial analytic solutions, the model evaluations most commonly rely on intuition, experience and model intercomparisons. In addition, GCM simulation statistics are typically compared to global reanalysis data while numerical weather forecasts are compared to local observations. Such approaches are not applicable to pure dynamical core assessments that isolate the dynamics package from the physical parameterizations. In general, three different sets of equations are most commonly used in dynamical cores. These include the hydrostatic primitive equations as well as the non-hydrostatic shallow-atmosphere and non-hydrostatic deep-atmosphere equation sets. As modeling groups now move towards the next generation of dynamical cores a standard test suite for hydrostatic and non-hydrostatic dynamics packages on the sphere if highly desirable. This webpage contributes to this effort. It suggests a collection of dynamical core test cases with varying complexity.”

Such evaluations of the global models used to make multi-decadal climate predictions are long overdue. In the paper

Pielke, R.A., 1991: Overlooked scientific issues in assessing hypothesized greenhouse gas warming. Environ. Software, 6, 100-107, I wrote

“The horizontal grid spacing of general circulation models is around 400 km. As shown by Pielke [the first edition of Pielke, 2002], as least four grid increments are required to reasonably represent an atmospheric feature, thus this grid resolution would only permit features 1600 km or larger to be reasonably represented in the models, Since extratropical cyclones often are observed to have horizontal wavelengths as small as 500 km or so, they are poorly represented in these models, Since these features provide the major physical mechanism for the exchange of heat, moisture, and momentum between the subtropics and the polar regions, the inability of GCM representations to adequately represent these exchanges is a serious shortcoming. Tropical cyclones, which also provide an important mechanism for exchanges between the tropics and higher latitude is even more poorly represented since its scales of important physical processes includes the eye wall which can be tens of kilometers in radial size.”

and

“Upwelling of deep, cold ocean waters occurs at a number of locations around the world including the equatorial Pacific, around Antarctica, and off the west coast of North America, northern South America, northwest Africa, southwest Africa, and elsewhere. Caused by the direction and speed of the wind at the ocean surface, these upwelled regions of cold surface waters usually have an extent in one spatial direction of 50km or so. Since atmosphere-ocean GCMs have spatial resolutions on the order of 400km, these important sinks for carbon dioxide are ignored.”

The test of the dynamical core fits into these evaluations and assessment of the global climate models as prediction tools. As a necessary condition, when configured to run in a multi-decadal predictive mode they should still be used to make short-term global weather predictions in order to asses their skill at simulating the development and movement of major high and low pressure systems, including tropical cyclones. Moreover, they should be run as seasonal weather predictions using inserted sea surface temperatures at the initial time in order to see if they can skillfully predict the development of El Nino and La Nina events, as well as other circulation patterns such as the North Atlantic Oscillation. If they cannot accurately predict these short term and seasonal weather patterns, they should not believed as valid and societally useful prediction tools on the regional (and even the global average) scale decades into the future.

With the recent evidence of significant disagreement between the IPCC model projections and reality, as diagnosed by surface air and tropospheric temperatures (e.g., see, see and see) and upper ocean heat content (i.e., see), Climate Science is reposting a weblog from 2006 titled “Big Time Gambling With Multi-Decadal Global Climate Model Predictions” by Roger A. Pielke Sr. and Roger A. Pielke Jr.

The weblog reads,

 ”Many advocates for action on climate change, including the IPCC assessments and recent documentaries have promoted a view that global warming will continue through the 21st century, with global warming defined as a steady increase in global average temperatures. This prediction of warming is based on the output of multi-decadal general circulation models and is primarily due to the radiative forcing effect of anthropogenic emissions of CO2. In such models only relatively minor year-to-year variations in global average temperatures are forecast in the upward trend, except when major volcanic eruptions cause short-term (up to a few years) of global cooling. For example, see these projections of the most recent IPCC — none of the models has an obvious multi-year (i.e., >2) decrease in global average temperatures over the next century.

Such predictions represent a huge gamble with public and policymaker opinion. If more-or-less steady global warming does not occur as forecast by these models, not only will professional reputations be at risk, but the need to reduce threats to the wide spectrum of serious and legitimate environmental concerns (including the human release of greenhouse gases) will be questioned by some as having been oversold. For better or worse, a failure to accurately predict the changes in the global average surface temperature, global average tropospheric temperature, ocean average heat content change, or Arctic sea ice coverage would raise questions on the reliance of global climate models for accurate prediction on multi-decadal time scales. Surprises or experience that evolve outside the bounds of model output would likely raise questions even among some of those who have so far accepted the IPCC reports as a balanced presentation of climate science. (For a perspective different than the IPCC on applications of climate models see this.)

The National Research Council published a report in 2002 entitled “Abrupt Climate Change: Inevitable Surprises” (of which RP2 was a committee member). The report raised the issues of surprises in the climate system. One of the surprises (to many) may be that the global climate models are simply unable to accurately predict the variability and trends in the climate metrics that have been adopted to communicate human-caused climate change to policymakers. Among the climate metrics with the most public visibility are the long term trends in global average surface temperature, global average tropospheric temperature, summer arctic sea ice areal coverage, and ocean heat content.

There is some emerging empirical evidence to suggest, however, that the concerns expressed here are worth consideration. The recent dramatic cooling of the average heat content of the upper oceans, and thus a significant negative radiative imbalance of the climate system for at least a two year period, that was mentioned in the Climate Science weblog posting of July 27, 2006, should be a wake-up call to the climate community that the focus on predictive modeling as the framework to communicate to policymakers on climate policy has serious issues as to its ability to accurately predict the behavior of the climate system. No climate model that we are aware of has anticipated such a significant cooling, nor is able to reproduce such a significant negative radiative imbalance.  Meaningless distinctions between “projections” and “predictions” will be unlikely to convince consumers of climate models to overlook experience that does not jibe with modeled output.

[Note added for the March 21 2008 weblog - while the cooling reported above was shown to be an error in the analysis of the upper ocean data by the authors of the Lyman et al at study, their corrected data still shows no warming in the upper oceans for the last 4 years; thus the comment about the failure of the models still applies. There has been no global warming, at least above the 700m level in the ocean, since at least 2004].

There is no greater danger to support for action on important issues of human impacts on the environment than an overselling of what climate science can provide. If the climate behaves in ways that are unexpected or surprising it will be more than just credibility that is lost. Advocates for action should think carefully when gambling with the unknown predictive abilities of climate models. The human influence on the climate system is real, but the climate may not always cooperate.”

Lo, J., Z. Yang, and R. A. Pielke Sr. (2008): Assessment of dynamical climate downscaling methods using the Weather Research and Forecasting (WRF) Model. J. Geophys. Res.,  J. Geophys. Res., doi:10.1029/2007JD009216, in press.

The abstract reads

“The common methodology in dynamical regional climate downscaling employs a continuous integration of a limited-area model with a single initialization of the atmospheric fields and frequent updates of lateral boundary conditions based on general circulation model outputs or reanalysis datasets. This study suggests alternative methods that can be more skillful than the traditional one in obtaining high-resolution climate information. We use the Weather Research and Forecasting (WRF) model with a grid spacing at 36 km over the conterminous U.S. to dynamically downscale the 1-degree NCEP Global Final Analysis (FNL). We perform three types of experiments for the entire year of 2000: 1) continuous integrations with a single initialization as usually done, 2) consecutive integrations with frequent re-initializations, and 3) as 1) but with a 3-D nudging being applied. The simulations are evaluated in a high temporal scale (6-hourly) by comparison with the 32-km NCEP North American Regional Reanalysis (NARR). Compared to NARR, the downscaling simulation using the 3-D nudging shows the highest skill, and the continuous run produces the lowest skill. While the re-initialization runs give an intermediate skill, a run with a more frequent (e.g. weekly) re-initialization outperforms that with the less frequent re-initialization (e.g. monthly). Dynamical downscaling outperforms bi-linear interpolation, especially for meteorological fields near the surface over the mountainous regions. The 3-D nudging generates realistic regional  scale patterns that are not resolved by simply updating the lateral boundary conditions as done traditionally, therefore significantly improving the accuracy of generating regional climate information.”

This paper has very important implications in terms of providing regional and local climate prediction information to policymakers and others. It further confirms our conclusions in the paper

Castro, C.L., R.A. Pielke Sr., and G. Leoncini, 2005: Dynamical downscaling: Assessment of value retained and added using the Regional Atmospheric Modeling System (RAMS). J. Geophys. Res. - Atmospheres, 110, No. D5, D05108, doi:10.1029/2004JD004721.

Since the results deteriorate when the driving large scale atmospheric information is applied only at the lateral boundaries, this means that the regional model is a slave to the parent global scale information. If the global model has errors, it is not possible for the regional model to correct for these errors.  Using  regional and local scale predictions based on dynamic downscaling from multi-decadal global climate model projections to make policy decisions decades into the future, therefore, is  erroneous and misleading.  Those who disagree with this conclusion need to provide quantitative tests that should be used to assess whether dynamic downscaling can predict regional weather patterns in the coming years (such as drought events) that provides skillful and useful information to policymakers. Certainly the forecasts for the winter 2007/2008 in the western USA have been a major bust.

Thanks to Chris Castro for alerting us to this paper (see also his guest weblog “Monsoon on Track to be a Wet One”) .

The article is

Richard Seager, Mingfang Ting, Isaac Held, Yochanan Kushnir, Jian Lu, Gabriel Vecchi,Huei-Ping Huang, Nili Harnik, Ants Leetmaa,2 Ngar-Cheung Lau, Cuihua Li, Jennifer Velez, Naomi Naik Model Projections of an Imminent Transition to a More Arid Climate in Southwestern North America Richard Seager, et al., Science 316, 1181 (2007); DOI: 10.1126/science.1139601

This paper is an clear example of publishing a hypothesis as a scinetific paper. Hypothesis testing is certainly appropriate but this is a blantant example of publishing a paper but its forecasts are not tested. Hindcasting, where the answer is known is an appopriate part of the assessment of a hypothesis but predictions in which the answer is not known are a requirement for a robust evaluation.

The abstract reads

“How anthropogenic climate change will affect hydroclimate in the arid regions of southwestern North America has implications for the allocation of water resources and the course of regional development. Here we show that there is a broad consensus among climate models that this region will dry in the 21st century and that the transition to a more arid climate should already be under way. If these models are correct, the levels of aridity of the recent multiyear drought or the Dust Bowl and the 1950s droughts will become the new climatology of the American Southwest within a
time frame of years to decades.”

The introduction of the article states,

“The Third Assessment Report of the Intergovernmental Panel on Climate Change (IPCC) reported that the average of all the participating models showed a general decrease in rainfall in the subtropics during the 21st century, although there was also considerable disagreement among the models (1). Subtropical drying accompanying rising CO2 was also found in the models participating in the second Coupled Model Intercomparison Project (2). We examined future subtropical drying by analyzing the time history of precipitation in 19 climate models participating in the Fourth Assessment Report (AR4) of the IPCC (3). The future climate projections followed the A1B emissions scenario(4), in which CO2 emissions increase until about 2050 and decrease modestly thereafter, leading to a CO2 concentration of 720 parts per million in 2100. We also analyzed the simulations by these models for the 1860–2000 period, in which the models were forced by the known history of trace gases and estimated changes in solar irradiance, volcanic and anthropogenic aerosols, and land use (with some variation among the models). These simulations provided initial conditions for the 21st-century climate projections. For each model, climatologies were computed for the 1950–2000 period by averaging over all the simulations available for each model. All climate changes shown here are departures from this climatology.”

The conclusion reads,

“The six severe multiyear droughts that have struck western North America in the instrumental record have all been attributed (by the use of climate models) to variations in SSTs in the tropics, particularly persistent La Niña–like SSTs in the tropical Pacific Ocean (15–19). The projected future climate of intensified aridity in the Southwest is caused by different processes, because the models vary in their tropical SST response to anthropogenic forcing. Instead, it is caused by risinghumidity that causes increased moisture divergence and changes in atmospheric circulation cells that include a poleward expansion of the subtropical dry zones. The drying of subtropical land areas that, according to the models, is imminent or already under way is unlike any climate state we have seen in the instrumental record. It is also distinct from the multidecadal megadroughts that afflicted the American Southwest during Medieval times (20–22), which have also been attributed to changes in tropical SSTs (18, 23). The most severe future droughts will still occur during persistent La Niña events, but they will be worse than any since the Medieval period, because the La Niña conditions will be perturbing a base state that drier than any state experienced recently.”

Policymakers who adopt the claims of this paper as the current level of scientific understanding of the climate system are being misled. Even an examination of the precipitation (see) since the publication of the paper (where we are in a La Niña!) shows that the article by Seager et all significantly oversimplify our understanding of climate science. This is hardly the universe of all the information needed on drought in the Western USA to plan for the future.

In the coming months, we will continue to follow the weather patterns from the Climate Diagnostic Center and the very informative Drought Monitor website to further assess their hypothesis.

Two papers appeared recently that evaluate in depth the ability of regional models to skillfully simulate and predict weather patterns on a seasonal timescale. These two very important papers are:

Castro, C.L., R.A. Pielke Sr., and J. Adegoke, 2007: Investigation of the summer climate of the contiguous U.S. and Mexico using the Regional Atmospheric Modeling System (RAMS). Part I: Model climatology (1950-2002). J. Climate, 20, 3844-3865.

Castro, C.L., R.A. Pielke Sr., J. Adegoke, S.D. Schubert, and P.J. Pegion, 2007: Investigation of the summer climate of the contiguous U.S. and Mexico using the Regional Atmospheric Modeling System (RAMS). Part II: Model climate variability. J. Climate, 20, 3866-3887.

Their abstracts are

“Fifty-three years of the NCEP–NCAR Reanalysis I are dynamically downscaled using the Regional Atmospheric Modeling System (RAMS) to generate a regional climate model (RCM) climatology of the contiguous United States and Mexico. Data from the RAMS simulations are compared to the recently released North American Regional Reanalysis (NARR), as well as observed precipitation and temperature data. The RAMS simulations show the value added by using a RCM in a process study framework to represent North American summer climate beyond the driving global atmospheric reanalysis. Because of its enhanced representation of the land surface topography, the diurnal cycle of convective rainfall is present. This diurnal cycle largely governs the transitions associated with the evolution of the North American monsoon with regards to rainfall, the surface energy budget, and surface temperature. The lower frequency modes of convective rainfall, though weaker, account for rainfall variability at a remote distance from elevated terrain. As in previous studies with other RCMs, RAMS precipitation is overestimated compared to observations. The Great Plains low-level jet (LLJ) is also well represented in both RAMS and NARR, but the Baja LLJ and associated gulf surges are not.”

and

“Summer simulations over the contiguous United States and Mexico with the Regional Atmospheric Modeling System (RAMS) dynamically downscaling the NCEP–NCAR Reanalysis I for the period 1950– 2002 (described in Part I of the study) are evaluated with respect to the three dominant modes of global SST. Two of these modes are associated with the statistically significant, naturally occurring interannual and interdecadal variability in the Pacific. The remaining mode corresponds to the recent warming of tropical sea surface temperatures. Time-evolving teleconnections associated with Pacific SSTs delay or accelerate the evolution of the North American monsoon. At the period of maximum teleconnectivity in late June and early July, there is an opposite relationship between precipitation in the core monsoon region and the central United States. Use of a regional climate model (RCM) is essential to capture this variability because of its representation of the diurnal cycle of convective rainfall. The RCM also captures the observed long-term changes in Mexican summer rainfall and suggests that these changes are due in part to the recent increase in eastern Pacific SST off the Mexican coast. To establish the physical linkage to remote SST forcing, additional RAMS seasonal weather prediction mode simulations were performed and these results are briefly discussed. In order for RCMs to be successful in a seasonal weather prediction mode for the summer season, it is required that the GCM provide a reasonable representation of the teleconnections and have a climatology that is comparable to a global atmospheric reanalysis.”

These two papers, under the leadership of Dr. Castro, demonstrate the value of using reanalyses as the obeservational framework to test the skill of the regional model simulations. The second study shows, for example, that for regional climate simulations

“to be successful in a seasonal weather prediction mode for the summer season, it is required that the GCM provide a reasonable representation of the teleconnections and have a climatology that is comparable to a global atmospheric reanalysis”,

where the “GCM” is the parent global model.

This means that unless the global model itself skillfully replicates large scale weather patterns, the regional climate model is doomed to produce inaccurate results.

This raises serious issues on the value of dynamically downscaling multidecadal global climate model simulations using regional models, and giving this information to policymakers.