By: Benoit Vanniere
I’ll take the opportunity of this blog to present a result which continues to puzzle me and which I still haven’t found a full explanation for: why does tropical cyclone precipitation depend so little on the resolution of climate models?
This result was obtained by analysing the models of HighResMIP, a protocol that compares high-resolution global climate models, introduced in CMIP6. European research centres coordinated their contributions to HighResMIP in the Horizon2020 European project PRIMAVERA and performed simulations in the range of horizontal resolutions going from 1 to 1/4degree. This upper limit of ¼ degree was not the result of a random choice but was rather guided by feasibility and physical considerations. First, it matched the capability of an emerging class of GCMs, given model performance and CPU power, at the time the protocol was designed. Second, at resolutions finer than ¼ degree, atmospheric models enter a grey zone in which the assumptions used to parameterise atmospheric convection can be questioned.
In some cases, regional climate models may prove sufficient to assess the impact of resolution on physical processes and would represent a cheaper alternative to global models. However, to answer some other scientific questions, it is not possible to use regional climate models. For instance, if one wanted to evaluate the upscale effect of mesoscale processes and fine-scale air-sea interactions on the global circulation and teleconnections. To answer all those questions, we need global high-resolution models.
One example of such research topics is the interactions of Tropical cyclones (TC) with their large-scale environment: what environmental factors control the interannual variability of TC frequency? how will it change in a warmer world? When starting this study, I was interested in the effect of the environment on TC precipitation.
When the horizontal resolution of a climate model is increased, TC become more intense, with deeper minimum sea-level pressure of TCs and stronger surface winds (e.g., Roberts et al., 2020). Because the latent heat released by precipitation plays a key role in TC intensification, we would also expect TC precipitation to reach larger values in high-resolution models.
Figure 1: (a) Distribution of TC minimum sea-level pressure, (b) precipitation averaged in a 1-degree cap and (c) precipitation averaged in a 5-degree cap, for five HighResMIP models (HadGEM3-GC31, ECMWF-IFS, EC-Earth3, CNRM-CM6-1 and CMCC-CM2). All quantities are 6-hourly. The low-resolution is represented by the solid curve and the high-resolution by the dashed curve.
We evaluate TC precipitation in five HighResMIP models, in which TCs have been systematically identified with the same tracking algorithm TRACK (Vannière et al., 2020). Precipitation was averaged, respectively, within a radius a 1-degree radial cap around the centre of the TC, to account for the inner core precipitation, and within a 5-degree radial cap, to account for the total TC precipitation. From Fig. 1b, HR models simulate larger precipitation rates in this inner region, as we anticipated.
More surprisingly, however, the climatological distribution of the total TC precipitation is in remarkable agreement between the low- and high-resolution configuration of a given model (Fig 1c), despite high-res models simulating more intense TCs (Fig 1a). In addition, the distribution of 5deg TC precipitation is much more sensitive to models’ formulation than to resolution.
Figure 2: The azimuthally averaged moisture budget of the 200 strongest TCs in the Northern Hemisphere over the period [1950-2015] in (a) CNRM-CM6-1 and (b) EC-Earth3.
To better understand this, we computed the water budget of the composite of the 200 strongest TCs in each model (Fig. 2). At first order the main balance is between precipitation and moisture convergence. The evaporation flux at surface is a small term in comparison. When resolution increases, the bulk of precipitation and moisture convergence gets closer to the TC centre, but their contribution to the budget of the entire TC (i.e., the area below the curve) does not change much. Assuming the cyclone is axisymmetric, we define the radius of closure as the distance at which the radially integrated surface evaporation balances the TC precipitation. For the 5deg TC precipitation, we find a radius of closure of ~1500 km +/- 200km in the HighResMIP models. Those results are in line with those of Trenberth et al. (2007) who showed that the water budget of TCs Katrina and Ivan were closed at a radius of ~1600 km. A radius of closure as large as 1500 km might seem very large considering that the largest surface evaporation occurs really close to the core. Although this large evaporation flux plays a crucial role in the dynamics of the storm by changing the specific humidity of an air parcel converging towards the centre of the storm and increasing the parcel’s convective available potential energy, it represents a negligible fraction of the total water budget!
Interestingly, we find that the radius of closure does not depend on model resolution. Hence, the moisture budget of a tropical cyclones is the result of a large-scale balance that low- and high-resolution models seem to represent equally well!
An implication of this result is that the details of physical processes in the inner core, does not play a major role on the climatological distribution of TC precipitation. Instead, this might be the signature of a mechanism acting at a scale larger than the inner core, which would be well captured by LR models. One potential candidate is the IR feedback, which was shown to participate in the intensification of tropical cyclones in some recent work (Ruppert et al. 2020). We can also speculate that this is the expression of a large-scale radiative constraint of TC convection: Jakob et al. (2019) showed that a significant fraction of the tropical atmosphere was in radiative convective equilibrium at time scales of 1 day and over areas comparable in size to the closure of the TC water budget.
Next steps? The analysis of the models of Horizon Europe project NextGEMS, comparing global storm resolving models, will allow us to test whether our results remain true at resolutions ~4km and when convection is resolved. The preparatory modelling campaign DYAMOND suggested this might be the case.
References:
[1] Roberts, M., and Coauthors, 2020: Impact of model resolution on tropical cyclone simulation using the HighResMIP–PRIMAVERA multimodel ensemble. J. Climate. 33, 2557-2583, https://doi.org/10.1175/JCLI-D-19-0639.1
[2] Vannière, B., and Coauthors, 2020: The moisture budget of tropical cyclones in HighResMIP models: large-scale environmental balance and sensitivity to horizontal resolution. J. Climate 33, 8457-8474, https://doi.org/10.1175/JCLI-D-19-0999.1
[3] Ruppert, James H., and Coauthors, 2020: The critical role of cloud–infrared radiation feedback in tropical cyclone development. Proc. Natl. Acad. Sci. 117, 27884-27892, https://doi.org/10.1073/pnas.2013584117
[4] Trenberth, K., and Coauthors, 2007: Water and energy budgets of hurricanes: Case studies of Ivan and Katrina. J. Geophys. Res.: Atmos. 112, D23107,
https://doi.org/10.1029/2006JD008304
[5] Jakob, C., and Coauthors, 2019: Radiative convective equilibrium and organized convection: An observational perspective. J. Geophys. Res.: Atmos. 124, 5418-543, https://doi.org/10.1029/2018JD030092