From Climate Etc.
by Frank Bosse
Neither the trend analysis nor the model-observation comparison supports the conclusions of the attribution study that found:
“The combined change, attributable to human-induced climate change, is roughly a doubling in likelihood and a 7% increase in intensity.”
Starting September 11, there was a heavy rainfall event in parts of Austria, Poland and the Czech Republic. First assessments point to a record high value of precipitation in a wide area, as the result of a “Vb weather condition”, named after the historical classification of the tracking directions of low-pressure fields in Europe. In a Vb weather condition, a low pressure area tracks to the Mediterranean Sea, thereafter north-eastward and ends usually in the Baltic region of Europe. A Vb condition is very often associated with much rain in Mid-and Eastern Europa and flood events e.g. 1997 (River Oder) and 2002 (River Elbe).
However, an “Attribution study” appeared only a few days afterwards. It’s Core message about the event (cited in media) was:
“The combined change, attributable to human-induced climate change, is roughly a doubling in likelihood and a 7% increase in intensity.”
To evaluate the robustness of the claim, the full text of the attribution study was downloaded.
The meteorological classification of the event in question involves several atmosphere dynamics features. The triggering event was an “Arctic Outbreak”, an extreme northward displacement of the intertropical convergence zone (ITCZ) was also involved. To make the things worse, there was a stable blocking high pressure field north of the area in question, so the precipitation area was relatively stationary and could not move north towards the Baltic Sea as usual.
The key question is whether the thermodynamic element (related to warming from “climate change”) contributing to the events described can actually be quantified with some robustness, as it was claimed in the attribution study.
The attribution study describes trend analyses of observational data (E-Obs.) and (weather) model-observational reanalysis data (ERA5) for the time 1950-2023 (2024). The data used are available via the “KNMI Climate Explorer” permitting one to evaluate the numbers. The study uses the GMST GISS Dataset to describe the connection of heavy rain in Mid Europe to a warmer world. The attribution study states:
“All datasets show similar trends across the region, with increasing trends…” (see section 3.1)
The same dataset is used here, but averaged over the European area in question rather than globally. The mean temperature anomaly 1950-2023 in the region 20°W-25°E; 35°N-65°N is shown below. This region includes more land (which is faster warming than the ocean) than the roughly 30% global mean land fraction.
Fig.1: The temperature time series (GISS) in the European area. The figure was generated with the KNMI Climate Explorer.
The (not too surprising) observation: From 1950 to about 1981 the temperatures did not show any rise. Anthropogenic warming, manifested in the mean temperatures, started around 1981, not 1950.
To calculate the trends in the “RX4days” precipitation (which is the accumulation of 4 days of precipitation) the data for ERA5 were recalculated for 1950-2024:
Fig.2: The outstanding event in September 2024 is clearly visible. It makes the trendslope 1981-2024 ( green) positive (a “one year trend”), for 1981-2023 (black) it’s zero. The figure was generated with ChatGPT.
The ordinary least square (OLS)-Trend 1950-2024 (blue) is robustly positive (p=0.025) as the attribution study states. However, it did not mention that the trend become insignificant after the late 1960s when calculated to 2024. If the increasing trend from 1950 to 2024 were attributable to the “human induced climate change” after 1981, one would NOT expect that the trend slope to 2024 is completely insignificant (p=0,32) and for 1981-2023 (black) it’s zero. In the light of these findings the OLS-trends to 2024 might be more a result of internal variability. Over the period 1950-1981 with no warming (see Fig. 1), the most positive trend slope (orange) of RX4day was 2 times steeper than in 1981-2024, when the forced warming was observed.
The study evaluates the climate models used for the attribution analysis. Many models belong to the CMIP6 family. It iss well known that those models face substantial difficulties when it comes to atmosphere dynamics owing to their low resolution. The Multi Model Mean shows no skill in the study area (46°N- 52°N; 11°E- 24°E) with respect of the model-observation (E-Obs.) spatial correlation for precipitation.
Fig.3: The spatial correlation between precipitation of the CMIP6 Multi Model Mean with the Observations (E-Obs.) for the warm seasons 1975-2023. The figure was generated with the KNMI Climate Explorer.
A meaningful correlation should be a prerequisite for blaming the anthropogenic warming as simulated in the CMIP6 models, for a distinct extreme precipitation event based on model comparisons with the real world.
In Table 4.1 of the attribution study the models were evaluated, some of them (only a few) were labelled as “good” when it comes to precipitation. The “IPSL-CM6A-LR” model was labelled as “reasonable”. The spatial correlation 1950-2023 to E-OBS observations during months where Vb-events were observed is shown below, also for the “good” model “EC Earth 3”, both with below 20%, indistinguishable from random noise:
Fig. 4: There is no skill (white for the zero- correlation) in selected models of the study. The figure was generated with the KNMI Climate Explorer.
Neither “IPSL-CM6A-LR” (left) nor in “EC Earth3” (right), labelled as “good” in the study, have any skill when it comes to the spatial correlation of precipitation with the real world. Nor does the model MPI-ESM1-2LR (“reasonable” in Table 4.1 of the study), but not shown here.
In the end it seems dubious to attribute an extreme precipitation event to climate change when using the CMIP6 models. The ocean warming is for sure a source for more evaporation and also for more rain, although the proportional rise in precipitation with warming is only a fraction of the rise in evaporation.
However, the influence of atmosphere dynamics is overwhelming and hampers the attribution of single extreme weather events based on thermodynamics arguments.
Conclusion
After a closer look, neither the trend analysis nor the model-observation comparison supports the conclusions of the attribution study.
The issue of unsound extreme weather event attribution studies is not limited to extreme precipitation. As this recent article by Roger Pielke Jr explains, attribution studies for all types of extreme weather event are in general highly dubious, and appear to be undertaken more for “political” than for scientific purposes.
Related
