[geeks] More on global warming

[Dr Robert Pasken]

On Mon, 24 Dec 2007, Shannon Hendrix wrote:

> On Dec 22, 2007, at 4:16 PM, Dr Robert Pasken wrote:
>
>>> The models are themselves unreliable, that is the issue being
>>> discussed.
>>>
>>> Cordially
>>>
>>> Patrick Giagnocavo
>>>
>>
>> On what grounds do you conclude that climate models are unreliable,
>> what
>> studies, model output, what statistical analysis of results do you
>> base
>> this claim?
>
> I can't speak for him, but what I've seen is that the models
> frequently do not agree with each other, and often cannot generate
> known or historical conditions.
>
> If they were very accurate, you'd expect several models to come to the
> same general conclusions, or be able, given historical input, to
> predict the present.
>
> When I was being taught to do models in college, those were two of the
> most fundamental tests to perform: can your model show known
> situations, and do they agree with each other.
>
> Do you think we've reached that stage yet?
>
>
> -- 
> Shannon Hendrix
> shannon at widomaker.com
> _______________________________________________
> GEEKS:  http://www.sunhelp.org/mailman/listinfo/geeks
>

Model verification is done at several levels. First thing to do is discard 
the notion that different models MUST produce the same results. The reason 
that different models exist in the first place is to allow different model 
physics to be tested. Consider the instance where the flow lowest 1km is 
considered to be more important than the 6km flow. This model has 
additional vertical layers in the lowest 1km rather than at 6km. By 
definition it will produce different results than model with the most 
layers at 6km. Should the model use spectral methods or finite 
differencing to compute dereivatives. A simple assesment of possible 
combinations of methods produces a lot of different models with possibly 
different results, but this is exactly the point of the different models.

Models are verified in a lot of different ways. A simple method is to 
start with todays data and run the model backwards several hundred years 
to determine if it reproduces history. The biggest flaw with this method 
is verification data doesn't extend back in time far enough nor in detail 
enough to do a good comparison. Example: I am modeling the development of 
hurricane Helene as it changed from an African Easterly wave into a 
tropical storm. The lastest experiment has the tropical storm 20km north 
of it's true position determined from aircraft data. Without the aircraft 
data I could not determine the tropical storms true position due to the 
extensive cloud sheild. In the late 1700's and early 1800's there was very 
limited surface observations in the US. How do I know if I missed a little 
bit in space and time when verifying the model. A second option is to 
start the model with data from a well documented day and time in the past
and run the model forward to compare the forward results against 
observations. This method allows me to determine if a "busted forecast" 
was due to spatial or temporal errors. The dust bowl was forecasted, but 
to far north/south/east/west or earlier/latter than what actually 
happened. Hansen et al (1988) described a model verification that with the 
exception of the timing of temperature declines due to vulcanic eruptions 
forecasted the global average temperature very well. Hansen et al. (1988)
assumed vulcanic eruption would occur, but assigned the timing of these events
at 50 years apart. The model predicts to 2020 and so far the global average
temperatures are within the error bars. That model is over 20 years old.


The biggest problem is that the global warming nay sayers are cherry 
picking results. Micheals of University of Virginia testified before 
congress that Hansen's model runs were wrong because Hansen's model over 
predicted current temperatures.  What Michaels did was edit a figure from 
Hansen's paper to remove all other model runs leaving only the model run 
with what Hansen explicitly stated in the paper as being a totally 
unrealistic increase in the amount of CO2 and was designed to place an 
upper bound on the rate of temperature increases. The Hansen figure also 
included a model run with a decline in atmospheric CO2 and with many runs 
with varying rates of CO2 increases.