Guest post by Rud Istvan
Fabius Maximus alerted Charles the Moderator to a new alarming ‘agitprop’ paper in Nature Communications (here) concerning global warming induced sea level rise (SLR). Since I previously contributed to WUWT on SLR and its lack of acceleration, CtM asked if I would provide a guest post reviewing this new SLR paper. I read it and explain below why you don’t have to. The post title is h/t to Fabius Maximus’ alerting email to CtM.
The alarming new paper by Kulp and Strauss is titled (paraphrased) “New DATA triple estimates of SLR vulnerability”. One knows immediately that agitprop follows, since the paper is about a new way to model coastal land elevations. Models produce data only in alarmist climate circles.
The paper has two parts. The first is a new way to remove a known high bias in coastal elevations derived by SRTM (satellite altimetry like that used to estimate SLR). The land problem is that satellite altimetry ‘sees’ the land’s top surface, not the actual land underneath a forest canopy or a cityscape. The alternative much more accurate and expensive method, airplane flown LIDAR transects, has only mapped the US and Australian coastal regions of those two wealthy countries. Comparison shows an SRTM positive US median bias of +3.7 meters!
The new method used for SRTM bias correction is a software neural network (a form of artificial intelligence, AI) incorporating 23 variables. The AI was trained on the US LIDAR/SRTM discrepancy, and then validated using Australia. That is a robust AI software methodology. Good work.
Unfortunately, the AI bias correction was good but not great. For those two countries, AI cut the root mean square error (RMSE) of SRTM compared to LIDAR ‘roughly in half’ (translation, to ‘only’ ~1.8 meters). The discussion in the paper uses South Florida as a physical example of the remaining causal bias. The neural network failed because South Florida’s coasts are densely populated with many high-rise condos (in between is a virtually uninhabitable Everglades), compared to the entire US coastline it was trained on. As an example of new AI induced error, the AI model applied to ‘current’ global coastal populations estimated that 110 million people already live below mean highest high water (MHHW). Which is sort of true for Miami’s South Beach during King tides, but way more people than just Dutchmen on their polders. I give kudos to the authors for being honest and illustrative about the remaining and newly introduced SRTM uncertainties after their AI ‘improvement’.
My only criticism of this first part is that the paper explicitly did NOT make the AI code available to scrutinize for logic and coding mistakes.
With this new AI model of SRTM bias ‘sort of firmly except not’ in hand, the bulk of the paper then uses it to TRIPLE SLR climate alarm.
To do that, the paper starts with a survey of some (ten to be exact, footnotes 3-12) of the many already alarming SLR papers that exist. The paper uses some of the usual suspects WUWT has previously discussed, including several WAIS instability speculations. That survey concludes +20-30cm of SLR by 2050 (so SLR had better hurry up and start accelerating), and +70-100cm by 2100 for RCP4.5. By comparison, most serious observational SLR papers predict less than 20-30cm by 2100 under business as usual, not a meter under RCP4.5. No SLR acceleration is evident in long record differential GPS corrected (for vertical land motion) tide gauges. So ~2.2mm/year with closure equals about 22 cm for the 21st century. The climate alarm survey bias is evident from that fact alone.
It then proceeds to model how many people end up under water. The first scenario illustrated in the body of the paper (most are in the SI) is ‘median K17/RCP8.5/2100’. It is the basis for the climate disasters illustrated by paper Figure 1 for the Pearl River Delta, Bangladesh, Jakarta, and Bangkok. (I checked the paper MSM PR. Figure 1’s Bangladesh and Bangkok are very popular.)
This scenario identifier needs explicit decoding, which no one I could find in the MSM PR about this paper, or using parts of Figure 1, has yet done.
1.Checking footnotes, we find that this new paper’s SLR scenario K17 expressly ‘includes early onset WAIS instability’. The old and thoroughly discredited ‘WAIS slides into the sea’ about Pine Island Glacier and the Amundsen Embayment.
2. RCP8.5 is the physically impossible but real bad one from AR5.
3. 2100 is sea level that year, based on CMIP5 climate models that provably run hot in the tropical troposphere by a factor of over 3x.
Applying that doubly impossible and 3x hot erroneously climate modeled scenario to the ‘current’ populations of global coastal areas, the new AI model ‘data’ shows that over 400 million people will be flooded out. The paper then carefully notes that the actual climate impact will be much worse because of population growth—so the new estimate is conservative.
Still high bias understating disaster! Conservative estimate! Triply alarming!