UPDATE #2 September 6 2011 I have corrected the journal where the Feser et al paper is in press. It will appear in the Bulletin of the American Meteorological Society (BAMS).
UPDATE September 2 2011
I sent the e-mail below to the authors of the paper, and the reply follows
From: Roger Pielke Date: Sat, Aug 27, 2011 at 5:47 PM
Subject: Your in press J of Climate paper Feser et al 2011
To: Hans von Storch , Burkhardt Rockel, Frauke.Feser@hzxx
Hi Dr. Feser, Hans and Burkhardt
I plan to post the information below later this coming week. Your
paper is an important contribution but I am concerned that readers
will interpret that your conclusions apply to the value of dynamic
downscaling from multi-decadal global climate predictions. I would be
glad if you could respond and I will post as a Reply on my weblog.
With Best Regards
From: Frauke Feser Frauke.Feser@xxx to pielkesr@xxx cc Hans von Storch , Burkhardt Rockel
date: Fri, Sep 2, 2011 at 2:21 AM
subject : Your in press J of Climate paper Feser et al 2011
Dear Prof. Pielke,
thanks a lot for your comments on our article. So far it has to our knowledge not been shown that the RCM added value results also apply for future scenario simulations of Type 4. We will take this as a motivation to look into this more systematically soon, until then our results are valid only for Type 2 RCM hindcast simulations.
With best regards,
There is a new paper which adds to our understanding of value-added from Type 2 dynamic downscaling. Type 2 dynamic downscaling is defined in Castro et al (2005) and has recently been defined further in a new submitted paper with Rob Wilby [which I will post on in the future] as
Type 2 dynamic downscaling refers to regional weather (or climate) simulations. In this case, the regional model’s initial atmospheric conditions are forgotten, but results are still depend on the lateral boundary conditions from a numerical global model weather prediction (in which initial observed atmospheric conditions are not yet be forgotten), or a global reanalysis, along with the land surface boundary conditions. Reanalyses such as ERA-40, NCEP, and JRA-55 assimilate spatially discontinuous weather observations in order to estimate temperature, humidity, wind speeds and so forth at grid points covering the entire globe. Downscaling from reanalysis products defines the maximum forecast skill that is achievable with Types 3 and 4 downscaling.
Type 2 dynamic downscaling is quite distinct from Type 4 downscaling [Type 1 is with respect to numerical weather prediction and Type 3 is with respect to seasonal weather prediction, for example, where certain parts of the climate system, such as sea surface temperatures are prescribed]. Type 4 is defined below.
Type 4 dynamic downscaling takes lateral boundary conditions from an earth system model in which coupled interactions between the atmosphere, ocean, biosphere and cryosphere are predicted. Other than terrain, all other components of the climate system are predicted except for human forcings, including greenhouse gas emissions scenarios, which are prescribed. Type 4 downscaling is widely used to provide policymakers with impacts from climate decades into the future.
Unfortunately, the new paper does not highlight that their findings do not apply to Type 4 dynamic downscaling (which is the IPCC multi-decadal climate predictions) except as an upper bound as to what value-added is possible.
Type 2 dynamic downscaling from reanalyses, in particular, has the advantage that the lateral boundary conditions and interior nudging is based on sampling from the continuous real world atmospheric conditions. In stark contrast, Type 4 dynamic downscaling knows nothing about spatial scale smaller than resolved by the global model.
The new paper is
Frauke Feser, Burkhardt Rockel, Hans von Storch, Jörg Winterfeldt, and Matthias Zahn, 2011: Regional Climate Models add Value to Global Model Data – A Review and selected Examples.
J of Climate. BAMS, in press.
The abstract reads [highlight added]
An important challenge in current climate modeling is to realistically describe small scale weather statistics such as topographic precipitation, coastal wind patterns or regional phenomena like polar lows. Global climate models simulate atmospheric processes with increasingly higher resolutions, but still regional climate models have a lot of advantages. They consume less computation time due to their limited simulation area and thereby allow for higher resolution both in time and space as well as for longer integration times. Regional climate models can be used for dynamical downscaling purposes, as their output data can be processed to produce higher resolved atmospheric fields, allowing the representation of small-scale processes and a more detailed description of physiographic details (such as mountain ranges, coastal zones, and details of soil properties).
But does higher resolution add value when compared to global model results? Most studies implicitly assume that dynamical downscaling leads to output fields superior to the driving global data, but little work has been carried out to substantiate these expectations. Here, we review a series of articles that evaluate the benefit of dynamical downscaling by explicitly comparing results of global and regional climate model data to observations. These studies show that the regional climate model generally performs better for the medium spatial scales, but not always for the larger spatial scales.
We conclude that regional models can add value, but only for certain variables and locations; particularly those influenced by regional specifics such as coasts or mesoscale dynamics such as polar lows. Therefore, the decision of whether a regional climate model simulation is required depends crucially on the scientific question being addressed.
The following text documents that this study is about Type 2 downscaling and not Type 4 downscaling
“In this article, efforts to determine such added value in case studies as well as in multi-decadal simulations with different RCMs are summarized and evaluated. The simulations presented here comprise mostly ‘reconstructions’, e. g. simulations of the weather dynamics since 1948 until today of Western Europe or the Northwestern Pacific. Most of these simulations use a grid distance of about 50 km, have been constrained with spectral nudging (von Storch et al., 2000) and use global NCEP/NCAR reanalysis (Kalnay et al. 1996; hereafter referred to as the NCEP reanalysis) as forcing data.”
Among their findings is
The regional model does not add value over the open ocean, due to the lack of orographic details and infrequent meso-scale phenomena here. It may even be worse than the reanalyses, which is reflected by the negative BSSs [Brier Skill Score].
Their statement that
“We conclude that RCMs do indeed add value to global models for a number of
applications, variables, and areas. If examined only at the regional scale, added value emerges very distinctly for many model variables, justifying the additional computational effort of RCM simulations.”
is correct but ONLY for Type 2 dynamic downscaling.