By Andy May
Nicola Scafetta has just published a new paper in Climate Dynamics examining evidence of the Urban Heat Island (UHI) effect (Scafetta, 2021). The paper is not chargeable and can be downloaded here. In summary, Scafetta shows that some of the recent warming in the HadCRUT 4 global temperature record may be due to the UHI effect. It uses an analysis of daily maximum (Tmax) and minimum (Tmin) temperatures, climate model performance, and a comparison of sea surface temperatures (SST) with land temperatures to estimate the possible impact on the HadCRUT 4 dataset.
The various land temperature records are not specifically corrected for the UHI effect; instead, NOAA and the Hadley Climatic Research Center rely on homogeneity algorithms to smooth out anomalies. NOAA calls their homogenization process “PHA” and the Hadley Center’s algorithm is similar. See this post or the 2009 article by Menne and Williams (Menne & Williams, 2009a) for a discussion of temperature homogenization. While these algorithms lower the temperature in the cities, they also raise the temperatures in the rural areas around the cities. To make matters worse, the past 70 years have been a period of rapid population growth and increasing urbanization. The world population has grown from 2.5 billion people in 1950 to 7.5 billion in 2020. The UHI in London was estimated at up to 2.8 ° C in the summer between 1990 and 2006.
UHI causes Tmin to rise faster than Tmax and the daily temperature range (DTR) to decrease over time. Scafetta is using the Hadley Climatic Research Unit (HadCRUT) temperature records and the Coupled Model Intercomparison Project 5 (CMIP5) data to investigate this issue. He compares an ensemble mean of the CMIP5 Tmax and Tmin data with the HadCRUT data and examines the differences between them.
The CMIP5 models were matched to the global and regional HadCRUT anomalies, so local anomalies such as UHI may be visible in maps of the difference between the two datasets. The CMIP5 models do not parameterize cities, so differences between cities and surrounding areas may indicate that the UHI influence remains in the HadCRUT data set.
Figure 1. These are the global Tmax (red) and Tmin (blue) anomalies of HadCRUT (A) and CMIP5 (B). The differences are shown in C and D. Source: (Scafetta, 2021).
Figure 1 compares the global Tmin (blue) and Tmax (red) anomaly records from HadCRUT4 to the CMIP5 ensemble means that Scafetta used in its study. The records are anomalies from 1945-1954. When comparing the decades 1945-1954 and 2005-2014, the differences between Tmin and Tmax (DTR) are different. The HadCRUT4 DTR is 0.25 and the CMIP5 DTR is 0.1. In both cases, Tmin warmed up more than Tmax.
Figure 2 shows how the Tmin-Tmax anomaly differences are distributed in the HadCRUT4 data set.
Figure 2. Global distribution of HadCRUT-Tmin-Tmax anomalies (DTR). Orange, purple and red mean that Tmin is heating up faster than Tmax. White areas, including the oceans, do not have Tmin and Tmax data. (Scafetta, 2021).
As you can see in Figure 2, most of the land area has a positive value, which means that Tmin is increasing faster than Tmax. The HadCRUT data shown in Figure 2 shows that in large areas of North America and Asia, Tmin increases much faster than Tmax. This can be seen most clearly in rapidly urbanizing China and in the growth regions of the USA and Canada.
In Figure 3 we see how the anomalies of the CMIP5 ensemble Tmin -Tmax (DTR) are distributed. The modeled Tmin-Tmax anomalies are much more dampened and closer to zero than the measured and homogenized values.
Figure 3. Global distribution of CMIP5 Tmin-Tmax anomalies (DTR). At the poles in the far north and in parts of Asia and Central Africa, Tmin warms up a little faster than Tmax. Most of the rest of the world, including the oceans, is near zero. (Scafetta, 2021).
Only in the vicinity of the poles do the models show a sharp increase in Tmin-Tmax as well as scattered areas in Asia and Africa. Greenland is a large island with a very small population of ~ 56,000 and shows hardly any differences between the modeled and measured values. The actual values vary from -0.2 to 0.2 and the modeled values are between 0 and 0.2.
Scafetta shows with numerous examples “that the land climate record is influenced by significant non-climatic distortions”. Tmin and Tmax are not present in the SST (Sea Surface Temperature) datasets, but we can compare the SST datasets to the HadCRUT land datasets via the CMIP5 model ensemble. When that happened, Scafetta found that the CMIP5 simulations, after taking into account the thermodynamic differences between land and ocean, were consistent with the warmer land records but significantly overestimated the SST. A land simulation of the temperature difference between the 1940 to 1960 average and the 2000 to 2020 average showed a HadCRUT difference of only 0.06 ° C. The comparison of CMIP5 (+ 0.69 ° C) with HadSST (+ 0.41 ° C) over the oceans resulted in a warming of 0.28 ° C, which is five times higher.
The warming of land temperature according to HadCRUT is about one degree between 1940-1960 and 2000-2020. If the CMIP5 models and HadSST records are accurate, then the land registry entries will have a deviation of + 0.36 ° C. This is almost a 60% error. We have already discussed the big effects of corrections on the temperature record. More information on the subject can be found in this post.
discussion
Scafetta’s study reveals a possible systemic bias in the country’s HadCRUT records. As shown in Figure 2, most of the bias lies in areas of rapid urban development during the study period 1940-2020. There are other anomalies in Bolivia that may be due to rapid deforestation in this area. The anomalies in arid parts of North Africa may be due to an inverse urban effect, as urbanization in this area can create a cooler area compared to the surrounding rural areas.
All data used in the study contain errors. Definitive conclusions cannot be drawn. However, it appears that the land portion of the HadCRUT 4 dataset is warmer than it should be compared to SST. It is also likely that this warm bias got into the CMIP5 models. The most recent DTR values (Tmax-Tmin) have fallen more than predicted by the CMIP5 models. This could be a problem with the models in urban areas, or it could be due to the homogenization algorithms used by the Hadley Climatic Research Center to smear the urban heat island warming over large areas. In both cases, Scafetta has shown that these datasets are inconsistent and that one or more of them may contain significant systemic biases.
One last point. When the data is corrected for the apparent bias described above and compared to the UAH independent global mean lower troposphere temperature (Spencer et al., 2017), we see that the corrected HadCRUT dataset is closer to it than the original in black, in 4B. This comparison shows that the apparent bias found in Scafetta’s study is empirically supported.
In Figure 4A, the original HadCRUT 4.6 data set in black is compared with the data set corrected in Scafetta in red. The mean of the CMIP5 model ensemble in yellow is shown in green along with 106 independent model runs. The 14A and 14B use the same colors. Figure 14B adds the UAH global mean lower troposphere temperature in blue. All curves are anomalies from 1940 to 1960.
Compared to 1940 through 1960, the original HadCRUT curve shows a warming of 0.59 ° C and 0.48 ° C using the Scafetta corrections. The UAH record shows 0.44 ° C. The CMIP5 climate models show a warming of 0.78 ° C.
After Scafetta’s correction, it is possible that non-climatic prejudices have caused a fifth of the reported global warming by HadCRUT since 1940-1960. It is also possible that the CMIP5 climate models overestimate the warming by a third. These are significant problems.
Figure 4. Diagram A shows the individual model runs in green, the CMIP 5 mean in yellow, the uncorrected HadCRUT data set in black and the corrected HadCRUT data set in red. B shows the same HadCRUT-corrected data set in red, the UAH – Lower troposphere dataset in blue and the original HadCRUT dataset in black. The red and black lines in A & B are the same.
Menne, M. & Williams, C. (2009a). Homogenization of temperature series through pairwise comparisons. Journal of Climate, 22 (7), 1700-1717. Retrieved from https://journals.ametsoc.org/jcli/article/22/7/1700/32422
Scafetta, N. (2021 Jan 17). Climate dynamics. Retrieved from https://doi.org/10.1007/s00382-021-05626-x
Spencer, R., Christy, J., Braswell, W. (2017), UAH Version 6, Global Satellite Temperature Products: Methodology and Results, Asia-Pac J Atmos Sci 53: 121-130.
4.3
3
be right
Item rating
Like this:
Loading…
Comments are closed.