In September and October 2023, the Belgian Climate Centre had the pleasure to welcome its first intern ever. During her internship, Frederieke Devriendt, Master student in Bioscience Engineering at the KULeuven, analysed the temperature change projections for Belgium of the latest global models from the CMIP6 project and regional models from the CORDEX project.
The aim of the analysis was to identify the impact of spatial resolution and emission scenarios on the projection of temperature change in Belgium. Therefore, Frederieke compared the outcomes from the CMIP6 and CORDEX project, which differ from each other in their level of spatial detail and the definition of their scenarios. The global climate models (GCMs) from CMIP6 have a spatial resolution around 100 km ranging between 80km and 250km, which means that the models may have 1 to 9 data points covering Belgium, depending on their specific resolution. For the regional models (RCMs) from the EURO-CORDEX project about one fourth of the 50 model projections has a resolution of 50km (EUR-44) while the others have a resolution of 12.5 km (EUR-11), equivalent to almost 200 data points in Belgium.
To compare the two types of models, Frederieke focused on the end of the century (2070-2100) and the most pessimistic scenarios (ssp585 and rcp 85) by looking into different levels on which they could differentiate:
Intensity of warming
Changing of the extremes – how are the extreme temperatures changing compared to the average temperatures?
Temporal differentiation – is the warming changing over the different seasons?
Spatial differentiation – are there differences in warming between different regions in Belgium?
RESULTS
1. Changing of extremes compared to the average & intensity of warming
The below figure shows the change of the probability distribution for temperature over Belgium for the period 2070-2100 under ssp585. The black dots indicated by ”av” show the changes in the average temperature of tas. The values P1, P5, P10, P90, P95 and P99, on the other hand, are different percentiles of the distribution of daily temperature against the different models. P1 and P99 are the most extreme values of tas predicted over the period 2070-2100: for P1, 99% of all values of tas are higher, so they are the 1% coldest values. For P99, 99% of all values are smaller, i.e., they are the 1% warmest values.
The global models show that the heat extremes (red) warm much more than the average
temperatures and that the cold extremes (yellow) warm a little more than the average
temperature. For the cold extremes, the regional models show the same as the global models. For the heat extremes, most of the regional models show a stronger warming as the average temperature, but some models, especially the ones that show a higher warming of the average temperatures, show a smaller warming of the heat extremes than of the average temperature.
In general, the intensity of warming is higher for the global models, especially for the heat
extremes.
2. Temporal differentiation
The below figure shows the annual cycle of the temperature change over Belgium under climate change for the different scenarios (in °C).
Global models show a peak in warming in summer. The warming remains relatively constant during the rest of the year. The regional models also show the strongest warming in summer, but the difference in warming with the other seasons is less extreme than in the global models. There is also a decrease of warming in spring.
3. Spatial differentiation
For both global and regional models small gradients are visible (more warming
in Wallonia and less in Flanders) in the mean temperature change. These gradients are bigger for the global models and increase for the most pessimistic scenarios.
With regards to the total warming of the mean temperature across Belgium, these data also show more warming for the global models than for the regional models.
There is a big variation between warming predicted by the different models used (within
one project) for the different regions in Belgium. Important to note is that the models do not
represent a realistic urban climate for Brussels. CONCLUSIONS
It can be concluded that the difference between CMIP6 and CORDEX is especially in the intensity of warming and the warming of heat extremes being higher for the global models than for the regional models. The regional models have a potential added value due to their higher spatial resolution, which makes it possible to detect regional feedback mechanisms and provides more detailed information on warming in Belgium. In addition, it is not possible to draw conclusions on spatial differentiation due to a full range of warming projected by the different models (within one project).