Extreme Weather #17 - Modeling Tropical Cyclones
What controls the frequency of tropical cyclones?
Here’s an interesting review paper from 2021 on one aspect of tropical cyclone research, by a cast of luminaries in the field - Tropical Cyclone Frequency by Adam Sobel and co-authors.
Plain language summaries are a great idea and this paper has one:
In this paper, the authors review the state of the science regarding what is known about tropical cyclone frequency. The state of the science is not great. There are around 80 tropical cyclones in a typical year, and we do not know why it is this number and not a much larger or smaller one.
We also do not know much about whether this number should increase or decrease as the planet warms - thus far, it has not done much of either on the global scale, though there are larger changes in some particular regions.
No existing theory predicts tropical cyclone frequency.
Why is this important?
Tropical cyclone frequency strongly controls the total hazard and risk from tropical cyclones, because no other aspect of a tropical cyclone matters if it does not occur. We are particularly interested in how tropical cyclone frequency is related to climate: what factors in the climate system determine the climatological frequency in the present and recent past, its natural variability, and its possible changes as a consequence of anthropogenic global warming.
Here is a graphic of trends in each region of the various categories of TCs (tropical cylcones), with cat 3-5 being the most of interest, as they are more severe. Note that these are out over the ocean, measured by satellites, only a tiny number hit land:
By eye, one sees an upward trend in total frequency in the North Atlantic basin since the start of the satellite era, a downward trend in the Australian region (Chand et al., 2019), and an increase in the number of intense storms in the North Indian Ocean (Balaji et al., 2018). Trends in other basins are either subtle or clearly absent..
They dig into climatic factors that are believed to influence TC formation:
Genesis indices can help us to understand some of the relative frequency differences between one basin and another, and between different times in the same basin, whether due to the seasonal cycle or internal variations such as those due to ENSO. But they do not explain the overall global frequency, nor the absolute frequency in any basin.
Every index, to our knowledge, requires empirical calibration to observations, explicitly or implicitly, such that the absolute frequency is not meaningfully predicted. The empirical basis of existing genesis indices also undercuts their predictions of how global frequency will change with global warming. Different indices can give opposite answers—either an increase in frequency with warming or a decrease—even when they behave very similarly in the present climate.
[Emphasis added].
I found this one illuminating. In a number of papers I’d read in recent weeks, I had been suprised by the correspondence between climate models and observations in total TCs. This explains the reason.
In early thinking on the topic, there was a belief that a warmer climate would produce more tropical cyclones, because the area over which the SST exceeds the threshold for deep convection and tropical cyclogenesis—historically somewhere in the 26–27°C range—would increase. That argument has, by now, long been understood to be wrong because the threshold itself rises with warming as the troposphere warms moist adiabatically..
In very simple terms, the temperature difference between the surface and higher up in the atmosphere is one driver for TCs. Both warm together.
Since the early 2000s, when global atmospheric models with resolutions in the 25–50 km range began to be practical for climate studies, most such models have predicted decreases, or at most no change, in global TC frequency. Assessment reports and review articles have largely based their statements about future frequency change on these results, though generally with low confidence. Recently, however, a few studies have simulated increases..
This illustrates one challenge of numerical modeling of complex problems. You can find lots of simulation results with the same conclusion. Until you start improving the models and find that they reach different conclusions.
They review various modeling approaches and comment:
Increasing resolution in both atmospheric and ocean models holds promise as a route to alleviating biases in simulated TC activity, including those associated with atmosphere-ocean coupling as well as those associated with representing the TCs per se, particularly as the representation of unresolved processes, including convection, remains uncertain in simulating and projecting TC frequency. Although global cloud resolving (kilometer-scale) models have been developed, the duration of the simulations performed with these models are typically shorter than what is needed for TC projections. Overcoming this challenge requires advancements in supercomputing that are unlikely to take place within the next decade.
[Emphasis added]. Suppose we have a model with a 100km x 100km grid - current highest resolutions GCMs that are routinely used to predict climate out to 2100. We want improve it to 1km x 1km to resolve important processes like TCs.
So we need 100 x 100 = 10,000x more processing power. That’s a lot, but it turns out yet more is needed. We need to increase our vertical resolution as well. Let’s say just 10x. So we need 100,000x more processing power.
Did we forget anything? A current GCM typically steps forward in time every 6 hours. But now we have higher resolution we need a shorter time step - maybe every 10 minutes. So we need over 1,000,000x more processing power. Something has to give.
For people interested in understanding a bit more about current theories on tropical cyclone formation this is a great paper to digest. It’s freely available.
References
Tropical cyclone frequency, Adam H. Sobel et al, Earth's Future (2021)