Tropical cyclones act to intensify El Niño

Tropical cyclones (TCs), some of the most influential weather events across the globe, are modulated by the El Niño–Southern Oscillation (ENSO). However, little is known about the feedback of TCs on ENSO. Here, observational and modelling evidence shows that TC activity in the southeastern western North Pacific can affect the Niño-3.4 index 3 months later. Increased TC activity in July–September can significantly contribute to the intensity of ENSO in October–December by weakening the Walker circulation and enhancing eastward-propagating oceanic Kelvin waves in the tropical Pacific. Thus, the greater the accumulated cyclone energy, the stronger (weaker) the El Niño (La Niña). A new physics-based empirical model for ENSO is constructed that significantly outperforms current models in predicting ENSO intensity from July to December and addressing the problem about the target period slippage of ENSO. Results suggest that TCs may provide significant cross-scale feedback to ENSO.

This is a potentially very strong paper and good fit for Nature Communications with revisions. I suggest the authors dig a bit deeper in the analysis (e.g. mjo) to extend on the earlier model-based studies, and improve the communications part of the paper (why is this important in the broader context of climate change?)

Response:
Firstly, thank the reviewer's support for our study and the constructive comments, which give us a big help to improve the quality of our manuscript. Secondly, in view of the reviewer's comments and suggestions, we gave point-by-point response and have added precisely and revised carefully in this manuscript as follows.

Response:
Thanks reviewers' reminder. We must admit that the studies the reviewer mentioned give us a big help to strengthen the paper and enlighten us to show further robust evidence and physical mechanism which provide a solid basis for our new findings. However, it may cause some misunderstanding to the originality of our work, hence, detailed comparisons between our and previous studies have been shown in Tables A1 and A2.

Fedorov et al.'s work (2010)
Our study

Timescales
They described "a positive feedback between hurricanes and the upperocean circulation in the tropical Pacific Ocean that may have been essential for maintaining warm, El Niño-like conditions during the early Pliocene" We study the modulation of TCs over western Pacific to El Niño on interannual timescale in modern times, which is quite another thing.

Physical mechanism
They thought "In the present climate, very few hurricane tracks intersect the parcel trajectories", which implies TCs will not play a main role in El Niño event in modern times.
We think TCs play an important role in intensity of El Niño event in modern times and propose an entirely different mechanism.

TCs' distribution
For the early Pliocene, TCs cover two bands of the entire zonal extent of the Pacific In modern time, there is no TCs in the central Pacific.

Table A2 Comparison between our and Sriver et al.'s work (2013 & 2016)
Hence, the originality of our work is strong enough, and these results mentioned above can give some support to our results and boost our confidence in our results. And these study have been mentioned in the revised manuscript (lines:47, 113,160, 215).

Key finding
In their 2013 work, they study was based on a special case of 3 successive tropical cyclones occurring in the western North Pacific (during May 2003) in an ocean general circulation model. And by blending into global atmospheric forcing from 1997, and the control simulation reproduced the main characteristics of the observed zonal redistribution of ocean heat in the tropical Pacific during 1997. Their work was a resolution-dependent special case analysis. Also in their 2016 work, they said "A key finding of this study is that tropical cyclones could significantly contribute to global ocean heat and energy budgets, and the magnitude of which depends on ocean grid resolution." As the reviewer said "The authors provide substantial new evidence of these links using new observational data over several decades", we reveal the universal and substantial effect of TCs on El Niño events.

Relationship between ACE and El Niño on interannual timescale
They didn't mention it. We firstly report that the TC activity in the southeastern western North Pacific (WNP) leads El Niño by about 3 months from the perspective of the accumulated cyclone energy (ACE) index.

Mechanism
They just mentioned excitation of equatorial Kelvin and Yanai waves by TCs can contribute to global ocean heat and energy budgets.
We show atmospheric and oceanic process to analyze the modulation of TCs to El Niño. It's undeniable that their result can help us to better understand the oceanic process in our study, but it's not entire route that TCs effect El Niño in oceanic process.

Response:
 The effect of MJO on modulation of TCs to ENSO intensity.
Previous studies (McPhaden 1999, Puy et al. 2016 indicate MJO may play an important role on ENSO, especially onset of El Niño and also effecting TC activities over the north Western Pacific in intraseasonal timescale. Actually this work doesn't involve the onset of El Niño, and we also appreciate the reviewer gives us a valuable reminder to further verify our result.

1) Data and methods
The interpolated outgoing longwave radiation ( Table A3, time series of 91-day running mean PC1 2 + PC2 2 defines the MJO index, in this work, the √ 1 2 + 2 2 greater than or equal to 1 defines MJO event, and others are non-MJO events. MJO is divided into eight phases. Generally, TCs over WNP are easily to generate in phase 47 while the convective center propagates into the Pacific. On the contrary, TC genesis is suppressed in phased 8 and 13 while the convective center is located in the Indian Ocean (Zhao et al. 2015). As a result, MJO events are divide into active (phase 47) and inactive (phase 8 and 13) MJO events. After removing MJO signal from the time series of daily SST anomalies, and converting daily events criterion non-MJO √ 1 2 + 2 2 < 1 MJO_active √ 1 2 + 2 2 ≥ 1 and Phase 4-7 MJO_inactive √ 1 2 + 2 2 ≥ 1 and Phase 8,1-3 data to monthly data, we can obtain the new series of monthly SST anomalies without MJO. Then a 3-month running mean is taken to new series. Monthly accumulated cyclone energy index (ACE) anomalies after removing MJO signal can also be obtain by same method.

Q3
[REDACTED] [REDACTED] Q4 "I also suggest providing more discussion about the importance of these results We're sorry that the inaccurate description in the manuscript may make the reviewer's misunderstanding. Table B1 shows detailed comparison between our and Camargo and Sobel's study (2005).

Camargo and Sobel's work (2005)
Our study

ACE calculation
Traditional calculated method of ACE was employed. "ACE is defined as the sum of the squares of the estimated 6-hourly maximum sustained surface wind speed for all TCs in the basin summed over all 6-h periods".

ACE depends on each TC genesis position.
ACE in each 2 latitude  2 longitude grid cell is defined as the sum of the squares of the estimated 6hourly maximum sustained surface wind speed for all TCs occurring in each grid cell over all 6-h periods. ACE is independent of the TC genesis position, it can better show the actual distribution of each grid cell than that using traditional calculated method.

Entirely different observation
As shown in Fig.3 in their paper, it was obviously that the maximum correlation between ACE and Niño 3.4 occurred when Niño 3.4 leaded ACE.
As shown in Fig. B1a, during the same period as that in Camargo and Sobel's work, it is obviously that the maximum correlation between ACE and Niño 3.4 occur when ACE leads N3.4. During our study's period, this lead relationship is more significant (Fig. B1b). This entirely different observation is owing to the ACE calculation. In the whole basin, the sum value of ACE is same as that using traditional method, but to one part of basin, it's quite different. And in our study, ACE in the southeastern western North Pacific is employed.
Hence, our study differs entirely from Camargo and Sobel's work. And this part has been mentioned in the revised manuscript (lines:57-67). Q2 "This is plausible since TCs are associated with strong westerlies on the equator.

are actually "caused" by the TCs…there is a literature suggesting that there are westerly wind bursts on the equator that precede TC formation cause the TCs.)"
Response:  Daily zonal wind near equator has not a main influence on interannual climatic effect of tropical cyclone (TC) activities over WNP on N3.4 indeed The mean values of maximum zonal wind anomalies with and without TCs occurrence are shown in Fig. B2. We can find that the westerlies wind anomalies with TCs is stronger than that without TC occurrence regardless of whether it is an El Niño developing year. And the ratio between situation with TCs and that without TCs holds approximately equal. This result implies this enhancement of TCs to westerlies wind is independent on the El Niño occurrence. The relationship between WNP ACE index after removing daily zonal wind near the equator and N3.4 is also checked. Area-averaged zonal wind anomaly near the equator (5S5N, 120E180) defines a zonal wind index. Using partial correlation, we obtain daily WNP ACE after removing daily zonal wind near the equator, then converting daily data to monthly data. As shown in Fig. B3a

Response:
Thanks for reviewer's question, firstly, the main comparisons between Fedorov et al.'s work (2010) and ours in following Table B2. We study the modulation of TCs over western Pacific to El Niño in modern times, which is quite another thing.

Physical mechanism
They thought "In the present climate, very few hurricane tracks intersect the parcel trajectories", which implies TCs will not play a main role in El Niño event in modern times.
We think TCs play an important role in intensity of El Niño event in modern times, and propose an entirely different mechanism.

TCs' distribution
For the early Pliocene, TCs cover two bands of the entire zonal extent of the Pacific.
In modern time, there is hardly any TCs in the central Pacific.

 Oceanic process for TC influences on El Niño
In order to verify TCs can change the oceanic situation and further affect the El Niño on interannual timescale, the monthly mean equatorial potential temperature anomalies are examined.

Response:
Thanks for the reviewer's good suggestion. Previous studies (

[REDACTED]
In manuscript, we show the result from an intermediate complexity coupled ocean-atmosphere model under giving atmospheric forcing. Due to our carelessness, we didn't give detailed explanation to this part in last version of manuscript. Although it's not enough complex rather than current fully couple models, previous studies (Cane et al. 1986, Zebiak and Cane 1987, Chen et al. 2004) have proved that it can capture well the features of El Niño. We check its simulated ability for El Niño before we use

Reviewers' comments:
Reviewer #1 (Remarks to the Author): The manuscript is improved. The ideas are interesting and fit with recent literature (though they go into much more depth here). Conclusions are still a bit speculative, and I suggest adding in some text on caveats/limitations based on the reviewers' points. I also suggest a careful edit for English. I recommend acceptance after these minor recommendations.
Reviewer #2 (Remarks to the Author): The authors have done a commendable amount of work to address my earlier review. At the same time, my concerns about the causal chain being argued for have not been much alleviated. For example, Figure 3b clearly shows westerly winds on the equator before the TCs develop, at least as indicated by ACE, and I see no evidence that external influences such as the MJO have been ruled out. Beyond that, my earlier complaint that there are many rather speculative statements stands. For example, on lines 128-130, is is stated that 'The main reason for this is a Hadley-like circulation (20°N−20°S, 135°−170°E) caused by the TCs'. How do we know that it was caused by the TCs?
It would be far better to try to publish this paper in a journal that specializes in ENSO-related research so that experts in the field can begin a discussion of the ideas presented here, before attempting to present it to a broader audience.

Reviewer #1
The manuscript is improved. The ideas are interesting and fit with recent literature (though they go into much more depth here). Conclusions are still a bit speculative, and I suggest adding in some text on caveats/limitations based on the reviewers' points. I also suggest a careful edit for English. I recommend acceptance after these minor recommendations. Response:

Caveats/limitations
Thanks for the reviewer's warm reminder, as the reviewer's saying, more text on caveats/limitations should be added in the closing discussion paragraph indeed.
We have divided the original closing paragraph to two paragraphs: the key results in the first part and a detailed discussion including limitations in the second part. These two processes lead to intensified El Niño. TCs can also affect La Niña (not shown   It would be far better to try to publish this paper in a journal that specializes in ENSO-related research so that experts in the field can begin a discussion of the ideas presented here, before attempting to present it to a broader audience.

Response:
Thanks for the reviewer's helpful comments and suggestions. The manuscript has been revised carefully and more supporting materials have been added. More details and point-to-point responses to the reviewer's comments are listed as follows.

Q1. "Figure 3b clearly shows westerly winds on the equator before the TCs develop, at least as indicated by ACE"
Response: The occurrence of westerly winds disturbance on the south flank of TC is one of essential TCs genesis condition (Gray 1968, Emanuel 2003, hence, as the reviewers' saying, westerly winds near the equator occur before the TCs develop. In our study, we focus on the feedback of TCs to westerly wind. Actually, Figs. B1 and B2 are shown to explain this feedback. From Fig. B1 and B2, we can obtain three important messages as follows: 1. The mean value of TC-related anomalous westerlies is larger than that without TC occurrence.
2. The center of anomalous westerlies varies with the movement of TCs.
3. The ratio between the mean anomalous westerlies with TCs and that without TCs holds regardless of whether El Niño happens, which means the role of single TC to westerlies doesn't change with different year.
All these results can prove that TCs have a positive feedback to westerly wind. We check situation in all years from 1970 to 2016. All situations accord with the aforementioned results.

1) Comparing with single-factor models
A holdout method is first employed. And the N3.4 predicted by models is compared with observations There is a high correlation coefficient (~0.91) and sign consistency (92%) between the observed N3.4 and that predicted by the ACE+N3.4 model, far higher than is given by the single-factor N3.4 model (correlation coefficient of 0.75, sign consistency of 79%) and ACE model (correlation coefficient of 0.66, sign consistency of 65%) (Fig. B3a)

Response:
Thanks for your good question. As can be seen from Fig. B9

Response:
Thanks for your warm reminder. As the reviewer's saying, Nature Communication has wider audience than other special journal, it includes not only ENSO-related experts, but also TC-related experts. All researches need a wide discussion before estimating its value, we'd like to accept advices and doubts of experts from any field. Also, in our next work, we have verified that TCs can improve the current ENSO predictions indeed. These results also remind us that the climatic effect of TC activities has been ignored for a long time, it should be given more attention, even on other climatic events. Nature Communication has not only a broader audience also more experts in all fields, which is what we expect.

More materials
Thank for reviewer's all questions, they gave us many great reminders. Except for some aforementioned evidences, we also added the detailed discussion for the limitations in the manuscript (lines: 226-243): Although the primary focus here is on the feedback of WNP TC occurrence on westerlies and the thermocline, other related topics warrant further investigation. For example, the effect of the MJO and other factors on TC genesis and the possible physical mechanisms associated with the influence of WNP TCs on Hadley-like circulations and tropical westerlies still need to be verified in a fully coupled model. Previous studies18, 26, 30 have provided fundamental insights into simulating climatic TCs in a fully coupled model, but were restricted to TC climatic distributions (or to accumulated TCs). Currently, fully coupled models that perform well in simulating interannual variabilities of both TC activity and El Niño are rare, and remain an important challenge for researchers. Furthermore, the observed explained percentage of N3.4 intensity during the El Niño developing years in this work (Fig. 1) shows that the modulated strength of WNP TCs to El Niño intensity may reach ~20%. The absence of TCs can weaken El Niño intensity, and further research is needed to assess the extent of this modulation. The under-investigated relationship between TCs and ENSO could lead to an improved understanding of ENSO and help improve ENSO amplitude forecasts. How this relationship is incorporated into existing forecasting models still needs to be determined. Finally, these results suggest that the climatic effect of TC activity should be given more attention, including its effects on other climatic events.

Reviewer #1
The manuscript is improved. The ideas are interesting and fit with recent literature (though they go into much more depth here). Conclusions are still a bit speculative, and I suggest adding in some text on caveats/limitations based on the reviewers' points. I also suggest a careful edit for English. I recommend acceptance after these minor recommendations. Response:

Caveats/limitations
Thanks for the reviewer's warm reminder, as the reviewer's saying, more text on caveats/limitations should be added in the closing discussion paragraph indeed.
We have divided the original closing paragraph to two paragraphs: the key results in the first part and a detailed discussion including limitations in the second part. El Niño are rare, and remain an important challenge for researchers. In general, this under-investigated relationship between TCs and ENSO could lead to an improved understanding of ENSO and help improve ENSO amplitude forecasts. How this relationship is incorporated into existing forecasting models still needs to be determined.
Finally, these results suggest that the climatic effect of TC activity should be given more attention, including its effects on other climatic events.

A new physics-based empirical model for ENSO prediction
Good prediction skill is most useful proof to our current study. In this work, a new physically based empirical model for ENSO prediction model has been built on the basis of the preceding (three months earlier) WNP ACE and N3.4.

1) Comparing with single factor models
A holdout method is first employed. And the N3.4 predicted by models is compared with observations There is a high correlation coefficient (~0.91) and sign consistency (92%) between the observed N3.4 and that predicted by the ACE+N3.4 model, far higher than is given by the single-factor N3.4 model (correlation coefficient of 0.75, sign consistency of 79%) and ACE model (correlation coefficient of 0.66, sign consistency of 65%) (Fig. A1a). The root mean square error (RMSE) between the observed N3.4 and that predicted by the ACE+N3.4 model is 0.39C, which is smaller than that for the N3.

Reviewer #2
The It would be far better to try to publish this paper in a journal that specializes in ENSO-related research so that experts in the field can begin a discussion of the ideas presented here, before attempting to present it to a broader audience.

Response:
Thanks for the reviewer's helpful comments and suggestions. The manuscript has been revised carefully and more supporting materials have been added. More details and point-to-point responses to the reviewer's comments are listed as follows.

Q1. "Figure 3b clearly shows westerly winds on the equator before the TCs develop, at least as indicated by ACE"
Response: The occurrence of westerly winds disturbance on the south flank of TC is one of essential TCs genesis condition (Gray 1968, Emanuel 2003, hence, as the reviewers' saying, westerly winds near the equator occur before the TCs develop. In our study, we focus on the feedback of TCs to westerly wind. Actually, Figs. B1 and B2 are shown to explain this feedback. From Fig. B1 and B2, we can obtain three important messages as follows: 1. The mean value of TC-related anomalous westerlies is larger than that without TC occurrence.
2. The center of anomalous westerlies varies with the movement of TCs.
3. The ratio between the mean anomalous westerlies with TCs and that without TCs holds regardless of whether El Niño happens, which means the role of single TC to westerlies doesn't change with different year.
All these results can prove that TCs have a positive feedback to westerly wind. We check situation in all years from 1970 to 2016. All situations accord with the aforementioned results.

1) Comparing with single factor models
A holdout method is first employed. And the N3.4 predicted by models is compared with observations There is a high correlation coefficient (~0.91) and sign consistency (92%) between the observed N3.4 and that predicted by the ACE+N3.4 model, far higher than is given by the single-factor N3.4 model (correlation coefficient of 0.75, sign consistency of 79%) and ACE model (correlation coefficient of 0.66, sign consistency of 65%) (Fig. B3a)  To evaluate further the predicted ability, a running holdout method is employed.
The prediction skill of the ACE+N3.4 model for N3.4 is better than that with the holdout method (Fig. B3b) B6c) because of the anomalous absence of preceding TCs (Fig.B7a)

Q2. "and I see no evidence that external influences such as the MJO have been ruled out"
Response: Thanks your good questions. Fig. B7  removing MJO signal) is also hardly changed (Fig. B9c). This shows that the MJO signal might only slightly affect the modulation of TCs to ENSO intensity on interannual timescale. In our previous manuscripts, because of the inaccurate expression, we may make a misunderstand that MJO has no influence to this modulation. Hence, we change all expression in the related part (lines:201-203, 207-208, 212-213, 303-304). Also, this two figures have been added in the manuscript (Extended Data Figs. 10 and 11,.

Response:
Thanks for your good question. As can be seen from Fig. B10

Response:
Thanks for your warm reminder. As the reviewer's saying, Nature Communication has wider audience than other special journal, it includes not only ENSO-related experts, but also TC-related experts. All researches need a wide discussion before estimating its value, we'd like to accept advices and doubts of experts from any field. Also, in this work, we have verified that TCs can improve the current ENSO predictions indeed.
These results also remind us that the climatic effect of TC activities has been ignored for a long time, it should be given more attention, even on other climatic events. Nature Communication has not only a broader audience also more experts in all fields, which is what we expect.

More materials
Thank for reviewer's all questions, they gave us many great reminders. Except for some aforementioned evidences, we also added the detailed discussion for the limitations in the manuscript (lines: 326-339):

Reviewer #1
The authors addressed my minor concerns. The science is interesting and noteworthy, and I recommend publication after improvements to the presentation.
I suggest revising the text for flow and grammar, and add more specific text on implications for climate and prediction... avoid over general statements such as the last sentence of the text relating the importance of the "climatic effect of TCs" on "other climatic events". What does this even mean? The results have wide-ranging applications in large-scale dynamics and weather/climate interactions and predictability. I suggest the authors add more text highlighting these broader impacts of the work for a multi-disciplinary audience.

Response:
We sincerely thank reviewer for the time and effort in reviewing our manuscript. Firstly, we have revised the English language of the whole manuscript by asking a professional company, Stallard Scientific Editing. Secondly, according to the reviewer's reminder about the wide-ranging applications and implications of this work in the corresponding aspects, we have added the more detailed description in the discussion (lines:358-375 in the manuscript): "In general, although TC activity is usually considered a synoptic event, the cumulative effect of TCs can provide a significant cross-scale feedback to the largescale atmospheric and oceanic circulations that then affects the intensity of ENSO. TC activity thus help to improve ENSO forecasts. How this under-investigated relationship between TCs and ENSO is incorporated into existing forecasting models still needs to be determined. In addition, these results highlight the need for further study of the crossscale effect of short-timescale events on long-timescale events. Perhaps the cumulative effect of TCs in other basins (e.g., hurricanes over the North Atlantic) affects intraseasonal or longer-timescale climatic events in other regions by modulating the large-scale dynamic processes. The cumulative effect may be a critical link between synoptic and climatic events. If this cumulative effect of short-term events can be reproduced well in the existing forecasting models, the prediction skill of models would be improved. Furthermore, we can also apply this cumulative effect to cross-scale studies in other disciplines. For example, in oceanography, it may deepen our understanding of the inverse energy cascade from mesoscale eddies to large-scale circulation. All these aspects belong to the category of cross-scale dynamics, which deserves further in-depth study."

Reviewer #2
The authors have responded well to my comments on their revised paper, though I still have some reservations about statements on causality made on the basis of lagged correlations. It is time to let readers decide if it is plausible that tropical cyclones affect the progression of ENSO.

Response:
We sincerely thank reviewer for the time and effort in reviewing our manuscript. In order to show our results more objectively to the readers, we have given a clearer claim about limitation of causality in the discussion (lines: 350-354 in the manuscript): "Although the primary focus here is on the feedback of WNP TC occurrence on ENSO, other related topics warrant further investigation. For example, the effect of the MJO and other factors on TC genesis and the causality associated with the influence of WNP TCs on Hadley-like circulations and tropical westerlies still need to be further verified in a fully coupled model." In addition, we asked a professional company, Stallard Scientific Editing, to improve English.