I was alerted to two publication which provide further examples of the misuse and overselling of what is possible with multi-decadal global climate predictions [the colleague requested remaining anonymous].
The first report was published in 2006 and is titled
Building Climate Change Scenarios of Temperature and Precipitation in Atlantic Canada using the Statistical Downscaling Model (SDSM) October 2006 by Gary S. Lines, Michael Pancura and Chris Lander of the Meteorological Service of Canada, Atlantic Region.
The abstract reads [highlight added]
Atlantic Canada is situated in a very diverse environmental area along the east coast of Canada, spanning almost 20 degrees of latitude and 20 degrees of longitude. The climate of the region is varied, encompassing both marine and continental regimes and influenced by several major ocean currents and mountain ranges. In order to best describe the expected climate change impacts for the region, climate change scenarios and climate variables must be developed on a regional, or even site-specific, scale.
Two methods exist that could potentially provide this information, output from a Regional Climate Model (RCM) and statistical techniques to “downscale” climate variables from global climate models. Since the RCM capability for Canadian territory is presently being developed and output for Atlantic Canada is not readily available, the statistical techniques were explored to generate the downscaled climate variables in that region.
Homogenized daily mean, maximum and minimum temperature, and quality controlled precipitation data for 14 sites across Atlantic Canada over the last 30 years was taken from the Historical Canadian Climate Database and used as the basis for developing the initial statistical relationships. Essentially, a predictor-predictand relationship is defined between global climate model values and the observed values at specific sites. Future climate variables (predictors) are then extracted from various model experiments. Those predictors are used to provide downscaled climate variables (predictand) that are applicable to those specific observed data sites. The resulting values are intended for use by climate change impacts researchers who want to apply climate variables on a regional scale in future climate impact studies. These researchers’ interests span many sectors including agriculture, forestry, biodiversity and natural resources. The statistical techniques are embodied in the Statistical Downscaling Model (SDSM) developed by Rob Wilby et al., King’s College, London. The model results are primarily from the Canadian coupled global climate model version 1 (CGCM1) from the University of Victoria, in British Columbia.
The monthly, seasonal and annual results show that in general downscaled SDSM values differed from, and in most cases were greater than, the raw CGCM1 global grid box projections, due presumably to local climatic forcing, and that the SDSM downscaling skill (as represented by explained variance) was highest for temperature (69-79%), lowest for precipitation (7-18%) and in both cases showed only slight spatial variability.
Users of these projections should be aware of the limitations of the methods used, and that downscaling using other GCM models running the same emission scenarios may produce slightly different but equally plausible results.
My Comment: While Rob Wilby worked on this project in 2006, it is clear from our 2012 EOS Forum article and our AGU chapter, in which Rob is a co-author, he now agrees that type 4 downscaling (from multi-decadal climate predictions) is a flawed approach and is being oversold to the impacts communities. To equate “explained variance” to “skill” with respect to future climate is an example of this misrepresentation of what is possible. Except for the application of the SDSM model to the observed climate (which is very much a worthwhile activity), the creation of multi-decadal predictions of the future climate by Lines et al 2006 is a waste of funds.
Our two publications are
Pielke Sr., R.A., R. Wilby, D. Niyogi, F. Hossain, K. Dairuku, J. Adegoke, G. Kallos, T. Seastedt, and K. Suding, 2012: Dealing with complexity and extreme events using a bottom-up, resource-based vulnerability perspective. AGU Monograph on Complexity and Extreme Events in Geosciences, in press. https://pielkeclimatesci.files.wordpress.com/2011/05/r-365.pdf
Pielke Sr., R.A., and R.L. Wilby, 2012: Regional climate downscaling . what.s the point? Eos Forum, in press [it will appear in EOS next week]. https://pielkeclimatesci.files.wordpress.com/2011/10/r-361.pdf
In the Pielke and Wilby 2012 article we concluded with respect to the statistical regional downscaling of multi-decadal climate predictions (which we refer to as Type 4 downscaling) that
“Type 4 statistical downscaling uses transfer functions developed for the present climate, fed with large scale atmospheric information taken from Earth system models representing future climate conditions. It is assumed that statistical relationships between real-world surface observations and large-scale weather patterns will not change. Type 4 downscaling has practical value but with the very important caveat that it should be used for model sensitivity experiments and not as predictions…….It is, therefore, inappropriate to present type 4 results to the impacts community as reflecting more than a subset of possible future climate risks.”
The second publication is
Bootsma, A., Gameda, S. and McKenney, D. W. 2005. Impacts of potential climate change on selected agroclimatic indices in Atlantic Canada. Can. J. Soil Sci. 85: 329–343.
This paper illsustrates how model predictions, which have shown no skill at predicting changes in climate statistics, are inappropriately used to claim forecast skill in the coming decades.
The abstract reads
“Agroclimatic indices (heat units and water deficits) were determined for the Atlantic region of Canada for a baseline climate (1961 to 1990 period) and for two future time periods (2010 to 2039 and 2040 to 2069). Climate scenarios for the future periods were primarily based on outputs from the Canadian General Circulation Model (GCM) that included the effects of aerosols (CGCMI-A), but variability introduced by multiple GCM experiments was also examined. Climatic data for all three periods were interpolated to a grid of about 10 to 15 km. Agroclimatic indices were computed and mapped based on the gridded data. Based on CGCMI-A scenarios interpolated to the fine grid, average crop heat units (CHU) would increase by 300 to 500 CHU for the 2010 to 2039 period and by 500 to 700 CHU for the 2040 to 2069 period in the main agricultural areas of the Atlantic region. However, increases in CHU for the 2040 to 2069 period typically varied from 450 to 1650 units in these regions when variability among GCM experiments was considered, resulting in a projected range of 2650 to 4000 available CHU. Effective growing degree-days above 5°C (EGDD) typically increased by about 400 units for the 2040 to 2069 period in the main agricultural areas, resulting in available EGDD from 1800 to over 2000 units. Uncertainty introduced by multiple GCMs increased the range from 1700 to 2700 EGDD. A decrease in heat units (cooling) is anticipated along part of the coast of Labrador. Anticipated changes in water deficits (DEFICIT), defined as the amount by which potential evapotranspiration exceeded precipitation over the growing season, typically ranged from +50 to –50 mm for both periods, but this range widened from +50 to –100 mm when variability among GCM experiments was considered. The greatest increases in deficits were expected in the central region of New Brunswick for the 2040 to 2069 period. Our interpolation procedures estimated mean winter and summer temperature changes that were 1.4°C on average lower than a statistical downscaling procedure (SDSM) for four locations. Increases in precipitation during summer and autumn averaged 20% less than SDSM. During periods when SDSM estimated relatively small changes in temperature or precipitation, our interpolation procedure tended to produce changes that were larger than SDSM. Additional investigations would be beneficial that explore the impact of a range of scenarios from other GCM models, other downscaling methods and the potential effects of change in climate variability on these agroclimatic indices. Potential impacts of these changes on crop yields and production in the region also need to be explored.”
My Comment: This study is a waste of funds. The actual “additional investigation” that would be beneficial would be to show in a hindcast mode that the multi-decadal and decadal forecasts have i) any skill in simulating the current climate statistics including major weather features such as ENSO, the PDO and the NAO, and ii) any skill in predicting CHANGES in these climate statistics over decadal and multi-decadal time scales. Until and unless the models can show skill in both of these measures, these studies are providing policymakers and the public (as well as impact scientists) with a false level of confidence in their robustness. As a result, faulty policy decisions are likely to result.