The Evolution of Models - 4

Fig. 1 (click to enlarge)

In previous posts of this series we discussed best practices (The Evolution of Models, 2, 3).

One of those practices was based on the use of historical experience to determine best practices.

Projects that build computer software models which project future events, as well as the software eventually built, work better when a beneficial dataset can be used to drive the projection logic that will eventually be used to project what seems to be the most likely future scenario.

To be sure, Fig. 1 is an example of that, i.e., dataset that were used to very accurately project CO₂ levels from 2009 through 2015 (the model was written before 2009).

Fig. 2 (click to enlarge)

It is an example of accurate datasets being used to build robust projection models.

Taking that up a notch, Fig. 2 shows the model projected out much further, some 85 years, to 2100 (see IPCC CO₂ projections).

For a background and a source for realizing why the datasets are useful in developing projection models, Fig. 3 shows that our current environment has been rendered into quite a different one from the CO₂ environment that has existed for many thousands of years.

Fig. 3 (click to enlarge)

All that having been said, the purpose for paying close attention to the modelling of CO₂ emissions is to use that dataset to enhance a sea level rise (SLR) projection model I am developing for Dredd Blog use.

The injection of green house gases leads to temperature increases, which in turn lead to SLR, which is a catastrophic danger to civilization (New Climate Catastrophe Policy: Triage - 12).

The factor that makes SLR projection more difficult than temperature or CO₂ emission projections is the double lag or delay of impact.

How long does it take, after CO₂ is injected into the atmosphere, for a temperature rise to manifest (first lag), and further, how long after the temperature rises does it take for SLR to manifest (second lag)?

I have used another IPCC projection model that anticipated temperature increase, which has also historically been accurate (The IPCC Record on Global Warming Temperature Projections).

The element of the IPCC modelling that has been inaccurate has been SLR projection, which has consistently underestimated SLR.

So, I am not trying to reinvent the wheel here, I am just trying to make that wheel round, like the other IPCC projection elements that are well rounded.

So, by adding another partner (the CO₂ model) to my SLR projection model (which already uses the 4 deg. C IPCC projection dataset), I hope to improve the SLR projection model.

The average of CO₂ ppm projected by circa 2100, per IPCC, is about 750 ppm, (up from about 400 ppm in the atmosphere now).

The figures I posted recently (before adding the CO₂ dataset), from a slice of the data printout, were:

Year, Greenland, Antarctica, Non-polar, Combined
2015, 0.106398, 0.0444488, 0.58813, 0.738976
2016, 0.143898, 0.0544488, 0.59063, 0.788976
2017, 0.181398, 0.0644488, 0.59313, 0.838976
...
2029, 1.61329, 0.446286, 0.688589, 2.74816
2030, 1.89626, 0.521745, 0.707454, 3.12546
...
2097, 10.1968, 16.2361, -, 27.9129
2098, 10.2911, 16.5003, -, 28.2714
2099, 10.3854, 16.7644, -, 28.6298

(The IPCC Record on Global Warming Temperature Projections - 2). The software model generated similar data with the addition of the CO₂ dataset I just added:

Year, Greenland, Antarctica, Non-polar, Combined
2015, 0.106551, 0.0444897, 0.58814, 0.739181
2016, 0.144205, 0.0545307, 0.59065, 0.789385
2017, 0.181858, 0.0645716, 0.59316, 0.83959
...
2029, 1.61961, 0.447972, 0.68901, 2.75659
2030, 1.90374, 0.52374, 0.707952, 3.13543
...
2097, 10.2382, 16.3025, -, 28.0207
2098, 10.333, 16.5677, -, 28.3806
2099, 10.4277, 16.8329, -, 28.7405

It does not change much, so, if my assumptions and implementation of the dataset has been properly done, it is a reasonable projection.

Fig. 4 (click to enlarge)

Later, I took out the commonly accepted 21 cm. increase in SLR (from 1750 to now) in the database by setting it to zero.

Including it may have caused some confusion concerning SLR "from now on" with that 21 cm. of past history in it.

The graph Fig. 4 shows only SLR beginning at zero SLR, which I think is more clear in the sense that we are now only talking about SLR from now on, rather than including the past.

One impact is that it moves the 3 ft. of SLR somewhat, as shown by comparing the two detail print outs above, with this one:

Year, Greenland, Antarctica, Non-polar, Combined
2015, 0.0376535, 0.0100409, 0.00251024, 0.0502047
2016, 0.0753071, 0.0200819, 0.00502047, 0.100409
2017, 0.112961, 0.0301228, 0.00753071, 0.150614
...
2029, 1.55071, 0.413523, 0.103381, 2.06762
2030, 1.83484, 0.489292, 0.122323, 2.44646
2031, 2.11897, 0.56506, 0.141265, 2.8253
2032, 2.40311, 0.640828, 0.160207, 3.20414
...
2097, 10.1694, 16.268,1.39144, 27.8288
2098, 10.2641, 16.5332,1.41038, 28.2077
2099, 10.3588, 16.7984,1.42933, 28.5865

The main impact is that the 3 ft. SLR shows up circa 2031-32 rather than at 2029-30, but, it also shows how touchy or delicate SLR projections are.

I am going to keep at it until I am satisfied that the assumptions and implementations are correct, then declare how it is done (and share the code with any who would like to check it out).

The next post in this series is here, the previous post in this series is here.