Extreme Weather #15 - Modeling Tropical Cyclones
Biases in modeled SSTs can lead to large changes in modeled TCs
In #1-#6 of the “Extreme Weather” series we looked at trends in Tropical Cyclones (TCs) from the perspective of chapter 11 of the 6th assessment report of the IPCC (AR6). The six parts were summarized here.
The report breaks up each type of extreme weather, reviews recent trends and then covers attribution and future projections.
Both attribution and future projections rely primarily on climate models. We looked at some of the ideas of attribution in the “Natural Variability, Attribution and Climate Models” series.
AR6 has a section: Model Evaluation, on p. 1587, before it moves into Detection and Attribution, Event Attribution.
How good are models at reproducing Tropical Cyclones?
Accurate projections of future TC activity have two principal requirements: accurate representation of changes in the relevant environmental factors (e.g., sea surface temperatures) that can affect TC activity, and accurate representation of actual TC activity in given environmental conditions.
Suppose in the future we had a model that was amazing at reproducing tropical cyclones when a variety of climate metrics were accurately reproduced. However, if the climate model didn’t reproduce these metrics reliably we still wouldn’t get a reliable answer about future trends in tropical cyclones.
As a result tropical cyclones are a major modeling challenge.
Here’s an example, referenced by this section of the IPCC report: “The impact of climate model sea surface temperature biases on tropical cyclone simulations”, by Wei‐Ching Hsu and colleagues from 2019.
The paper is a bit involved but I’ll try to show a few of their main results.
Their first graphic shows that there are substantial errors (“biases”) in modeled sea surface temperatures (SSTs) in the tropics - the graphic shows the difference between models and observations:
Therefore, reliable climate model simulations of TC activity depend on a realistic representation of modes of climate variability on seasonal to multi-decadal timescales, as well as climatological mean SSTs.
However, severe SST biases exist in the tropical Pacific and Atlantic in most of the Intergovernmental Panel on Climate Change (IPCC) coupled global climate models (GCMs).
And later they add:
Even though the mechanisms that cause the SST biases have been widely studied, the impact of the biases on TC simulations and seasonal predictions has not been fully understood.
Their paper uses a higher resolution climate model, a bit more suited to the dynamics of tropical cyclones. They compare modeled TCs using correct sea surface temperatures with the model results using the warm and cold SST biases (separately and then combined) in the modeled ocean.
This paper doesn’t compare models with observations, a fact that might enrage some commentators. I’ll be covering other papers that do just that, so sit tight.
We’ll focus on the first graph below, top left, which shows the simulated Accumulated Cyclone Energy (ACE) in the Atlantic.
The first set of values, CTRL, is the model using real SSTs (the range is from the different model runs).
The second set of values (AtlWB) uses real SSTs but includes just the warm biases in the Atlantic from models. The third set (AtlCB) includes just the cold biases. The fourth (AtlTB) uses the "full Atlantic biases”:
The simple story is that for this basin, warm biases have only a small effect on total cyclone energy, but cold biases have a major effect - the modeled ACE is about one third of what the model predicts with correct temperatures.
If you move along to the middle top and right top graphs you see the effect on the Pacific TCs from Atlantic temperature errors.
Big differences from actual model errors in SSTs.
Here’s the same set of results, this time from the errors in the Pacific SST. I’ve highlighted the results for the eastern North Pacific:
This time the warm bias roughly doubles the ACE in the eastern Pacific, whereas the cold bias has a negligible effect.
Conclusion
Using climate models to simulate Tropical Cyclones is a major challenge due to their fast and relatively small scale dynamics.
Even with a higher resolution model we see that incorrectly modeled SSTs causes large changes in the model simulations of TC energy.
References
The impact of climate model sea surface temperature biases on tropical cyclone simulations, Wei‐Ching Hsu, Christina M. Patricola & Ping Chang, Climate Dynamics (2019)