Change in ocean subsurface environment to suppress tropical cyclone intensification under global warming

Tropical cyclones (TCs) are hazardous natural disasters. Because TC intensification is significantly controlled by atmosphere and ocean environments, changes in these environments may cause changes in TC intensity. Changes in surface and subsurface ocean conditions can both influence a TC's intensification. Regarding global warming, minimal exploration of the subsurface ocean has been undertaken. Here we investigate future subsurface ocean environment changes projected by 22 state-of-the-art climate models and suggest a suppressive effect of subsurface oceans on the intensification of future TCs. Under global warming, the subsurface vertical temperature profile can be sharpened in important TC regions, which may contribute to a stronger ocean coupling (cooling) effect during the intensification of future TCs. Regarding a TC, future subsurface ocean environments may be more suppressive than the existing subsurface ocean environments. This suppressive effect is not spatially uniform and may be weak in certain local areas.

The percentage of the OCE changes (i.e., ΔOCE (%)) is with respect to (wrt) the result in Supplementary   The grid size used is 2-degree in Lon/Lat. The two study regions are the western North Pacific (WNP) and the North Atlantic (NA). The initial, pre-TC ocean profile at each grid is the boreal TC-season (July-October) averaged profile from the CMIP5 ocean field in each year. As we have 22 CMIP5 ocean fields, the above method is applied to each of the CMIP5 members and at each grid. After obtaining the OCE results for each grid and from each individual member, a multi-model ensemble OCE is obtained based on averaging the OCE results from each member at each grid. Therefore, we are able to obtain not only the multi-model ensemble mean (MME) but also the spread in OCE due the differences in the 22 different ocean fields at each grid.
We have run the 3DPWP model at each grid, for each year (2006 to 2100), over each of the 22 CMIP5 ocean fields, to calculate the TC-induced ocean cooling effect. This is done for both WNP and NA, and for each of the 15 scenarios. Because of the large member of such calculation and the use of CMIP5 data, we believe this is one of the most comprehensive OCE assessments so far obtained for TC global warming research.
The objective of this research is to examine the change in OCE due to the change in future initial ocean conditions under the CMIP5 RCP8.5 scenario. Therefore, we do not add-on the possible change in future TC attributes under global warming. Certainly, it is possible that future change in TC attributes may further modify the OCE change. However, at present, we do not even have a clean, baseline OCE change due to change in the future ocean environment alone. Therefore, we are hesitant to introduce further TC changes into the analyses, especially given the large uncertainty in future TC activity projection. In short, the objective of this research is to examine the OCE change due to change in the initial future ocean environmental condition, but without co-varying the TC attribute changes. With the ongoing improvement in TC attribute projection, the co-varying aspect in TC attributes change can then be added to this baseline assessment in subsequent research.
The advantage of our idealized approach (i.e. the 15 hypothetical TC scenarios, Supplementary The disadvantage of our approach is that there is no track information and uniform TC parameters are applied throughout. Therefore for each scenario, the OCE change under global warming is only from change in ocean condition, but without contribution from possible change in TC attributes, i.e. without the co-varying TC attribute change. As discussed in above, given the issue with TC attribute projection in CMIP5, this work aims to obtain first a well-defined baseline, given future ocean condition change only. With the ongoing improvement in TC attribute projection, the co-varying aspect in TC attributes change can be added to this baseline assessment in subsequent research. As compared to the dynamical-downscaling approach 4-7 , our approach has the advantage of efficiency, and can be applied across 22 CMIP5 ocean fields to assess the spread among ensemble members under TC-ocean coupling condition. Because dynamical-downscaling is a much more expensive and computational-consuming approach, it is more difficult to apply to so many ocean fields individually to compare the performance across the 22 ocean fields (Fig.   4 in the main text and Supplementary Figs. 1-3, 9,10,19).
As from above, our potential intensity approach is not to replace dynamical-downscaling Please also kindly note that the dynamical-downscaling approach for the Atlantic in Bender et al. 5 and Knutson et al. 7 are based on CMIP3 MME ocean subsurface temperature gradient fields. In Bender et al. 5 , CMIP 3 oceanic and atmospheric fields were used. In Knutson et al. 6 , CMIP5 atmospheric field is used but the ocean subsurface temperature gradient field is based on the CMIP3 MME oceanic field (based on MME of 18 CMIP3 models). We thus want to conduct a systematic PI approach to use both ocean and atmospheric fields from a large representation (22) of the latest CMIP5 models. Please note that in Knutson et al. 7 , the across model spread (10 models) are examined for the CMIP3 atmospheric models, but not in the ocean subsurface fields, since the same MME ocean subsurface temperature gradient field is used.
Another issue to note is that we are not sure whether the under-sampling of TC track and frequency issue in CMIP5 may also affect the OCE sampling in the dynamical-downscaling approach 2 , since dynamical-downscaling approach is embedded in the CMIP atmospheric environment. It is uncertain to us that whether there will be an issue on the under-sampling of OCE in these approaches.