A new interhemispheric teleconnection increases predictability of winter precipitation in southwestern US

Reliable prediction of seasonal precipitation in the southwestern US (SWUS) remains a challenge with significant implications for the economy, water security and ecosystem management of the region. Winter precipitation in the SWUS has been linked to several climate modes, including the El Niño-Southern Oscillation (ENSO), with limited predictive ability. Here we report evidence that late-summer sea surface temperature and geopotential height anomalies close to New Zealand exhibit higher correlation with SWUS winter precipitation than ENSO, enhancing the potential for earlier and more accurate prediction. The teleconnection depends on a western Pacific ocean-atmosphere pathway, whereby sea surface temperature anomalies propagate from the southern to the northern hemisphere during boreal summer. Analysis also shows an amplification of this new teleconnection over the past four decades. Our work highlights the need to understand the dynamic nature of the coupled ocean-atmosphere system in a changing climate for improving future predictions of regional precipitation.

The paper claims that late-summer persistent sea surface temperature (SST) and geopotential height anomalies (GPA) in the southwestern Pacific exhibit higher correlation with southwestern US (SWUS) winter precipitation than the El Niño-Southern Oscillation (ENSO). Additionally, the authors suggest an amplification of this new teleconnection represented by the so-called New Zealand index (NZI) has occurred over the past four decades, and most likely is due to the expansion of the tropics. The most interesting aspect is perhaps the apparent strengthening of the correlation between SWUS precipitation and NZI, an amplification that is not observed in the ENSO teleconnection.
For reasons expressed in the overall comments, however, I cannot recommend acceptance of this paper at its present form.
Overall comments: If we step back and look at the manuscript as a whole, then basically what it's doing is some correlation of time series to discern various phase relationships between the time series. While this approach is fine as supplementary, supporting information to a physical argument, in this manuscript there is too little physical argument for why these phase relationships should be there. For example, there is well-developed physical theory for why ENSO directly modifies precipitation over western U.S. By comparison, why, physically, should precipitation in this region be modulated by SST changes in the other hemisphere, as claimed here?
The following figure [redacted] from Hoskins and Woollings (2015) displays the major "routes" of atmospheric teleconnection, comprising both long (zonal) and short (poleward) Rossby wave propagations. While newer modes of atmospheric teleconnection (other than ENSO-induced ones) have been found, the primary teleconnection routes remain unchanged. As this figure shows, these teleconnection routes do not include anything that crosses the hemispheres from New Zealand to SW US, as claimed by the authors.
From my understanding of the manuscript, there are two conflicting proposals concerning the cause-and-effect relationship: (1) that NZI modulates SWUS directly (inferred from Fig.1&2) or (2) that NZI leads ENSO and then ENSO modulates SWUS (inferred from Fig. 4). However, for (1), as noted in the left figure, there is no such route as an NZI-North America teleconnection existed and this paper has not provided any reason why there should be one. For (2), the ENSO correlation with the SWUS precipitation has declined as the authors noted (though it remains significant) and so, physically, it should not explain the increasing correlation of NZI-precipitation through ENSO.
The "Changing Pacific Dynamics" section of the manuscript is well written, but I found it not so relevant to the NZI-SWUS/P argument and somewhat distracting. The IPO/PDO modulation on ENSO and its teleconnection to SWUS has been proposed with both empirical and modeling evidence. Using similar arguments to explain the unexplained NZI-SWUS (tele)connection is premature.
Specific Comments: 1: The authors write -"here we report evidence that late-summer persistent sea surface temperature and geopotential height anomalies in the southwestern Pacific exhibit higher correlation with SWUS winter precipitation than ENSO." I feel this is not necessarily true since the correlation analysis is done using average precipitation corresponding to climate divisions where Niño 3.4 exhibits statistically significant correlation. More so, the NZI teleconnection strength ( Figure 3) varies considerably for different climate divisions. I suggest making the distinction clearer in the text.
2: "Notably, our analysis shows that the GPH results do not depend on the considered pressure level, further indicating the robustness of NZI and providing confidence on its physical underpinning. The similar correlations at various pressure levels also reveal that the GPH pattern has a barotropic structure"-I don't think this section in its entirety is evident from any of the figures. Where is the analyses for multiple pressure levels shown?
3: It would be interesting to see the lagged correlations between the NZI/ENSO indices and SWUS precipitation, with increasing lead times (perhaps starting a year prior) rather than averaging climate indices over the preceding three to six months (i.e. just Jul-Sep and Sep-Nov) and then presenting the best results. 4: Although the NZI yields relatively higher correlations with SWUS precipitation (Figure 1), the results from Figure S2 are not statistically robust enough to suggest the NZI can be utilized for useful predictions in a way that ENSO indices may not. Although a more compelling argument can be made for the period following 1970 (Figure 2b), specifically for fall; but I believe this will merit further research, including analysis o 5: The authors discuss the decadal variation of precipitation teleconnections and specifically point to a "periodicity of the order of 15 years," but then elect to show a moving correlation analyses using only a 30-yr moving window. Could the authors comment on why they chose the 30-yr window? It would help to show the moving correlation across different windows (maybe ranging from 10-30 with 5 years' increments). 6: Given the scarcity of data observation networks in the southern Hemisphere in comparison to the North, it will be of good value to add multiple reanalysis datasets for the most prominent analyses to serve as verification.
Reviewer #2 (Remarks to the Author): The subject manuscript is interesting and well written, and identifying climate teleconnections beyond the established teleconnections is important and useful. Although the paper has potential to be a useful contribution to the literature there are several issues that need to be addressed before the paper is suitable for publication. There are still several analyses that are needed to convincingly show the importance and reliability of the NZI. 2. Redmond and Koch (1991) performed one of the first studies of the predictive ability of ENSO indices. The authors should reference this highly cited paper -"90 the GPH results do not depend on the considered pressure level, further indicating the robustness of NZI and providing confidence on its physical underpinning." it is not unusual (actually expected) that multiple pressure levels vary coherently; multiple levels being involved does not indicate a robust association --one being correlated essentially brings the others along.
"198 We quantified the effect of non-stationarity on the prediction skill as the relative increase of RMSE when the calibration and prediction periods did not overlap (e.g. prediction in 1983-2015, calibration in 1951-1983) relatively to the case that the calibration-prediction periods completely overlapped (e.g. prediction in 1983-2015, calibration in 1983-2015). " it is difficult to understand this; producing two separate correlations is a pretty meager cross validation. Reviewer #1 (Remarks to the Author):

General Comment 1
The paper claims that late-summer persistent sea surface temperature (SST) and geopotential height anomalies (GPA) in the southwestern Pacific exhibit higher correlation with southwestern US (SWUS) winter precipitation than the El Niño-Southern Oscillation (ENSO). Additionally, the authors suggest an amplification of this new teleconnection represented by the so-called New Zealand index (NZI) has occurred over the past four decades, and most likely is due to the expansion of the tropics. The most interesting aspect is perhaps the apparent strengthening of the correlation between SWUS precipitation and NZI, an amplification that is not observed in the ENSO teleconnection.
For reasons expressed in the overall comments, however, I cannot recommend acceptance of this paper at its present form.
We thank the Reviewer for his/her critical and careful consideration of our work. We have performed extensive new analysis documenting the physical underpinning of the NZI teleconnection and its recent intensification. Point by point responses are provided below.

General Comment 2
If we step back and look at the manuscript as a whole, then basically what it's doing is some correlation of time series to discern various phase relationships between the time series. While this approach is fine as supplementary, supporting information to a physical argument, in this manuscript there is too little physical argument for why these phase relationships should be there. For example, there is well developed physical theory for why ENSO directly modifies precipitation over western U.S. By comparison, why, physically, should precipitation in this region be modulated by SST changes in the other hemisphere, as claimed here?
The following figure from Hoskins and Woollings (2015) displays the major "routes" of atmospheric teleconnection, comprising both long (zonal) and short (poleward) Rossby wave propagations. While newer modes of atmospheric teleconnection (other than ENSO-induced ones) have been found, the primary teleconnection routes remain unchanged. As this figure shows, these teleconnection routes do not include anything that crosses the hemispheres from New Zealand to SW US, as claimed by the authors.

[Redacted]
Thank you for the critical comment. Yes, it is indeed the interhemispheric nature of the NZI teleconneciton that makes it new and non-trivial. The new section The physical basis for the NZI teleconnection -The western Pacific pathway, accompanied with additional 10 figures (4

in the main text and 6 in the supplementary material) present details on the new analysis and results.
Below we respond to each comment separately.

General Comment 3
From my understanding of the manuscript, there are two conflicting proposals concerning the cause-and-effect relationship: (1) that NZI modulates SWUS directly (inferred from Fig.1&2) or (2) that NZI leads ENSO and then ENSO modulates SWUS (inferred from Fig. 4). However, for (1), as noted in the left figure, there is no such route as an NZI-North America teleconnection existed and this paper has not provided any reason why there should be one. For (2), the ENSO correlation with the SWUS precipitation has declined as the authors noted (though it remains significant) and so, physically, it should not explain the increasing correlation of NZI-precipitation through ENSO.
We thank the Reviewer for his/her comment. Fig. 5 which was designed to tell the story in a single graph, with details documented in further figures.

Concering (2):
We do not suggest that the connection of NZI to precipitation in SWUS occurs through the ENSO teleconnection. Actually, our argument is that the discovered teleconnection is separate from ENSO, and has been increasing in strength and decoupling from ENSO dynamics, especially during the last 3-4 decades.
The "Changing Pacific Dynamics" section (also Figure 4) of the previous version of the manuscript aimed to highlight this decoupling and suggest a propagation of NZI SST anomalies to the north Pacific; not specifically to the ENSO or the IPO region. Supplementary figure S5 and the GIF file supported this cascade. Fig.4-7) and the supplementary material (Figs. S5-S10) for details.

General Comment 4
The "Changing Pacific Dynamics" section of the manuscript is well written, but I found it not so relevant to the NZI-SWUS/P argument and somewhat distracting. The IPO/PDO modulation on ENSO and its teleconnection to SWUS has been proposed with both empirical and modeling evidence. Using similar arguments to explain the unexplained NZI-SWUS (tele)connection is premature.
The "Changing Pacific Dynamics" section (also Figure 4) of the previous version of the manuscript aimed to highlight this decoupling and suggest a propagation of NZI SST anomalies to the north Pacific; not specifically to the ENSO or the IPO region. Supplementary figure S5 and the GIF file supported this cascade.
However, we understand that the presentation of our arguments might not have been as clear as intended and we have added new analysis and discussion on this pointplease see the detailed response to your general comment 3.

Specific Comment 1
The authors write -"here we report evidence that late-summer persistent sea surface temperature and geopotential height anomalies in the southwestern Pacific exhibit higher correlation with SWUS winter precipitation than ENSO." I feel this is not necessarily true since the correlation analysis is done using average precipitation corresponding to climate divisions where Niño 3.4 exhibits statistically significant correlation. More so, the NZI teleconnection strength (Figure 3) varies considerably for different climate divisions. I suggest making the distinction clearer in the text.
We thank the Reviewer for his/her comment.
NZI is shown to be more correlated with precipitation than ENSO in almost all cl. divisions (see results in Figs 1 and 3). Also, based on Fig. 1, Niño 3.4 is shown to exhibit quite variable connection strengths in different cl. divisions, similarly to NZI. This is consistent with the literature, since ENSO signal to precipitation in the northern side of California, Nevada and Utah is significantly weaker than that in the south [1][2][3][4][5] . Note that Fig. 3 does not facilitate obtaining the latter variability of ENSO indices, since most ENSO correlation values in Fig. 3 are not statistically significant (especially for nonzero lead time) and are represented in white color, thus the variability is hidden (whereas in Fig. 1 figure below); moreover, working with the ENSO-correlated sites as we did, places inferences related to the NZI teleconnection strength on the conservative side. We have added a discussion in the manuscript; see lines 86-96.

Specific Comment 2
"Notably, our analysis shows that the GPH results do not depend on the considered pressure level, further indicating the robustness of NZI and providing confidence on its physical underpinning. The similar correlations at various pressure levels also reveal that the GPH pattern has a barotropic structure"-I don't think this section in its entirety is evident from any of the figures. Where is the analyses for multiple pressure levels shown?
We have followed your suggestion and we now present the analysis for different pressure levels in Fig S1 in the supplementary material, which shows the barotropic structure of the teleconnection.

Specific Comment 3
It would be interesting to see the lagged correlations between the NZI/ENSO indices and SWUS precipitation, with increasing lead times (perhaps starting a year prior) rather than averaging climate indices over the preceding three to six months (i.e. just Jul-Sep and Sep-Nov) and then presenting the best results.
We have performed the suggested analysis and present the results in Fig. 4. Please see lines 143-161 for specific discussion of the results.

Specific Comment 4
Although the NZI yields relatively higher correlations with SWUS precipitation (Figure 1), the results from Figure S2 are not statistically robust enough to suggest the NZI can be utilized for useful predictions in a way that ENSO indices may not. Although a more compelling argument can be made for the period following 1970 (Figure 2b), specifically for fall; but I believe this will merit further research, including analysis.
In our study, we provide evidence for a new teleconnection and use ENSO as a "benchmark" teleconnection for comparison. We do not suggest excluding the use of ENSO indices as predictors for precipitation in SWUS and moreover, the NZI superiority relative to ENSO has appeared after the mid-1970s (e.g. see lines 29-31, 113-114, 127-141, and 254).

Specific Comment 5
The authors discuss the decadal variation of precipitation teleconnections and specifically point to a "periodicity of the order of 15 years," but then elect to show a moving correlation analyses using only a 30-yr moving window. Could the authors comment on why they chose the 30-yr window? It would help to show the moving correlation across different windows (maybe ranging from 10-30 with 5 years' increments).
Please note that correlation values based on 10 -15 years cannot be reliably evaluated as to their statistical significance, due to the very limited sample size so the 30-year moving average window is the only viable option.

Specific Comment 6
Given the scarcity of data observation networks in the southern Hemisphere in comparison to the North, it will be of good value to add multiple reanalysis datasets for the most prominent analyses to serve as verification.
We already used two different SST reanalysis datasets (see section Methods and Data). All results are almost identical based on both datasets, thus conclusions are robust. In the manuscript we now present results using both datasets to show robustness.

General Comment 1
The subject manuscript is interesting and well written, and identifying climate teleconnections beyond the established teleconnections is important and useful.
We thank the Reviewer for the encouraging comment.

General Comment 2
Although the paper has potential to be a useful contribution to the literature there are several issues that need to be addressed before the paper is suitable for publication. There are still several analyses that are needed to convincingly show the importance and reliability of the NZI.

Specific Comment 1
The authors should consider referencing: Both studies are now cited. We have also added a discussion on the different ENSO flavors in the manuscript; see lines 122-126.

Specific Comment 2
Redmond and Koch (1991) performed one of the first studies of the predictive ability of ENSO indices. The authors should reference this highly cited paper. We have added the suggested reference.

Specific Comment 3
The authors should provide an explanation for the selection of the 1950 through 2015 for analysis. Climate division data are available since 1895 and SST data go back to 1856. Why was the selected period chosen for this study?
Although reanalysis data are available since 1850s, SST data are not to be trusted in the period before 1950s due to the scarcity of the data observation networks, especially in the southern Hemisphere; see also [6][7]. Thus, we have decided to keep our analysis in 1950-2015 period. Please note also, that Referee 1 questioned even the accuracy of the pre-1970 southern hemisphere reanalysis SSTs which prompted us to report results from 2 reanalysis products.

Specific Comment 4
It would be helpful and interesting to compare the spectral frequencies of precipitation in the southwestern U.S. with the spectral frequencies of ENSO and NZI. What are the common spectral frequencies and do they change during the period analyzed?
We thank the Reviewer for his/her suggestion. As shown, in the following figure, due to the limited sample size, results are not very reliable.
To determine uncertainty we iteratively repeated the analysis after having disregarded each time a single value from the series. As expected, the highest uncertainty appears in higher frequencies.
As a general remark, for ENSO, important frequencies are those in the range of 0.15 to 0.35 yr -1 (periodicity 6.6 to 2.8 years), while for precipitation and NZI, lower frequencies are more important (higher time scales of periodicity; see also lines 108, 114, and Fig S2). Please note that due to the uncertainty of the results, we have decided not to include this figure in the manuscript.

Specific Comment 5
Why was the region of the US used for the analysis restricted to the climate divisions with significant ENSO correlations? Also, why not look at precipitation correlations for the entire western U.S.? Does NZI result in a western U.S. precipitation dipole similar to ENSO?
Precipitation amounts over climate divisions which are located in the northern CA, NV and UT shows insignificant correlation with both NZI and ENSO indices. Indeed, precipitation in the northern side is known not to be related to Pacific SSTs, and to exhibit very different variability than in the rest part of SWUS see Figs 1, 3, S10 and [1][2][3][4][5]. To account for this natural disassociation of the northern region, we define the regionally averaged SWUS precipitation as the area-weighted average precipitation amount over climate divisions for which ENSO (Niño 3.4 Fig 1 herein)  We have also produced as suggested correlations of NZI and winter precipitation in the entire western US, and have added 2 new figures in the manuscript. Please see Fig. S9-10 and discussion in lines 194-199.

Specific Comment 6
A possible useful twist to this analysis would be to perform a couple of different analyses -(1) perform a principal components analysis of western U.S. precipitation and correlate the significant PC score time series with SSTs and see if the NZI index shows up in one of the correlation fields?
(2) perform a spectral analysis of western U.S. precipitation and then correlate time series of significant spectral frequencies with SSTs and see if the NZI shows up.
We thank the Reviewer for his/her suggestion.
(1) We have performed the analysis. Results are provided in the supplementary material (see Figure S10). Concerning the PC2, SSTs in the ENSO region exhibit statistically insignificant correlation, while NZI exhibits a correlation value on the order of -0.6.
(2) As in the case of specific comment 4, the sample size is unfortunately too short to perform a reliable spectral analysis.

Specific Comment 7
Something that would really strengthen this paper is some analysis and explanation describing how the NZI affects atmospheric circulation and results in an effect on precipitation in the western U.S. How do changes in the southwestern Pacific move across the tropics and influence the mid-latitude atmospheric circulation of the Northern Hemisphere? (Figs. 4-7) and in the supplementary material (Figs. S5-S10).

Specific Comment 8
Do long-term trends in NZI influence the results? There is a long-term warming trend in the NZI that is not seen in NINO3.4 SSTs and this warming trend may be influencing the results. The authors should consider removing long-term trends in all data sets and re-computing the results.
To ensure that possible long-term warming does not affect our conclusions, we have reproduced the results of Fig. 1 c-f, after detrending SST and GPH time series at each grid point. As shown in the following figure, results are identical to those presented in Fig. 1 in the manuscript. Reviewer #3 (Remarks to the Author):

General Comment 1
The authors present a not-well-known teleconnection and even less well known predictive measure for seasonal precipitation in the Southwest U.S. I believe the results will be useful to researchers and forecasters. Methods and analyses are appropriate, a few suggestions/questions appear below.
We thank the Reviewer for the encouraging comment.

General Comment 2
The fall NZI to winter SW precipitation relationship would be more meaningful if you were able to show how anomalous atmospheric circulation that actually produces above normal precipitation in the Southwest US evolves from a distant, un-connected SST and GPH anomaly over New Zealand.

Specific Comment 1
"72 SWUS, which is defined as the area-weighted average precipitation amount over climate divisions for which ENSO (Niño 3.4) exhibits statistically significant correlations (significance level α =0.05)" not standard definition of SW U.S. Explain why the ENSO connected region is adopted for purpose of testing NZI connection.
Precipitation amount over climate divisions which are located in the northern CA, NV and UT shows insignificant correlation with both NZI and ENSO indices. Indeed, precipitation in the northern side is known not to be related to Pacific SSTs, and to exhibit very different variability than in the rest part of SWUS see Figs 1, 3, S10 and [1][2][3][4][5]. To account for this natural disassociation of the northern region, we define the regionally averaged SWUS precipitation as the area-weighted average precipitation amount over climate divisions for which ENSO (Niño 3.4) exhibits statistically significant correlations (significance level α =0.05 ; Figs 1a-b and 1h). Notice that defining the average precipitation based on the significance of NZI-correlations does not affect the results (see Fig.1 herein), while it is not in the conservative side for comparing NZI and ENSO teleconnection strengths. We have added a discussion in the manuscript; see lines 86-96.

Specific Comment 2
"86 precipitation are shown in Figs 1c-d, for two lead times: Jul-Sept and Sept-Nov" lead times usually refer to the time lag between predictor and predictand. the Sept-Nov predictor is problematic in context of prediction because it overlaps with the Nov-Mar precipitation predictand.
We do not explicitly suggest that NZI averaged over Sep-Nov should be used as predictor for winter precipitation. Our investigation considers different periods to be complete. Indeed, for predicting purposes, one should consider NZI in a period that does not overlap with the period of interest of precipitation. We have changed our wording in the lines 98 and 449. Also, below we present results similar to those of Fig. 1 Fig 1), so conclusions are robust.

Specific Comment 3
"90 the GPH results do not depend on the considered pressure level, further indicating the robustness of NZI and providing confidence on its physical underpinning." it is not unusual (actually expected) that multiple pressure levels vary coherently; multiple levels being involved does not indicate a robust association --one being correlated essentially brings the others along.
We have deleted the sentence and we now present the analysis for different pressure levels in Fig  S1 in the supplementary material, which shows the barotrpic structure of NZI.

Specific Comment 4
"198 We quantified the effect of non-stationarity on the prediction skill as the relative increase of RMSE when the calibration and prediction periods did not overlap (e.g. prediction in 1983-2015, calibration in 1951-1983) relatively to the case that the calibration-prediction periods completely overlapped (e.g. prediction in 1983-2015, calibration in 1983-2015). " it is difficult to understand this; producing two separate correlations is a pretty meager cross validation.
We thank the Reviewer for his/her comment. We have changed our wording and added a statement to explain our results more effectively -see lines 237-240, and 246-248.

Specific Comment 5
Relevant references and findings that are not cited: line 182 and elsewhere "northeastern Pacific" might better be called "tropical and subtropical western North Pacific" line 114 and legend to Figure S2 the 15 year periodicity is stated rather prominently but not really demonstrated (spectra). the record is relatively short to make a strong claim, which could be construed as characterizing a SWUS precipitation cycle.  The ocean-atmosphere bridge mechanism-this is an interesting idea. it seems useful to note that GCM experiments exploring seasonal atmospheric predictability (e.g. Hoerling, Kumar and others) have found much greater influence from SST in the tropical Pacific than from the extra-tropics. Your results propose an influence from SST outside the tropics it seems.
line 256 "and thus, they CAN be used towards increasing precipitation predictability." i'd suggest saying "they MAY be used.....", since we don't know how long these stronger than ENSO linkages will hold up.
lines 59, 270 you refer to rainguage observations, which sounds like you use individual station precipitation records. but the divisional data are averages of precipitation from a set of stations whose membership has changed over time (depending on availability of date, etc). and "rainguage" might be interpreted as only including rain, but this is winter precipitation and important parts (higher elevations) of the SWUS divisions are dominated by snow during the winter.
line 177 and legend, Figure 5 " connecting 500 mb GPH anomalies in the area east TO Australia ...." do you mean "east OF Australia" ? We thank the Reviewer for his/her comment. Please note that results in Fig. 6 refer to a regional jet stream over a relatively narrow longitudinal zone centered at the west coast (220°E-260°E). It is not the zonally-averaged jet steam over the entire longitude. Zonal wind over this west coast area cannot be assumed to affect the stationary waves across the northern Pacific.
The teleconnection highlighted by the Reviewer (short-wave train) suggests how SST anomalies in the western Pacific alter the atmospheric pressure in the US west coast. Pressure anomalies affect the local wind and regional storm tracks and ultimately the winter precipitation. The regional jet stream we discussed in Figure 6 can be viewed as a result of this short-wave train teleconnection.
To address your concern and to make clear that we refer to the local upper wind field in Fig. 6, and also highlight the role of the atmospheric pressure anomalies in the physical process we have revised the wording appropriately (see lines 187-191, 209-211, 503-504).
line 182 and elsewhere "northeastern Pacific" might better be called "tropical and subtropical western North Pacific" Thanks for the suggestion but we would prefer to retain the original simpler terminology.
line 114 and legend to Figure S2 the 15 year periodicity is stated rather prominently but not really demonstrated (spectra). the record is relatively short to make a strong claim, which could be construed as characterizing a SWUS precipitation cycle.
We thank the Reviewer. We have deleted the statement in the text and caption of Fig. S2. Figure 4 to the eye, you are unfairly diminishing the positive correlations, especially in the ENSO core region of the late fall-winter maps. the color scale underplays positive correlations in the +0.4 to +0.5 category, compared to negative correlations -0.4 to -0.5. the faint salmon colored positive correlations hardly show up compared to the lighter blue negative correlations.
We thank the Reviewer. The colorscale used in Fig. 4 (and adopted for all correlation maps of the manuscript) is not intended to give unfair perception to the eye, especially since we present the real numbers of correlation coefficients throughout the manuscript. The correlations of ENSO are not higher than +0.5 in Figs 4e-f (we added a sentence in line 152-153), and this is why they hardly show up. Following your comment however, we tried using red color instead of salmonno real difference is illustrated (see Fig. R1 below). In contrast, in Fig S5 for    Please note that we present results and discussion on the relation of SOI and precipitation in two sections of the manuscript, and many figures/tables: Figs 2-3, S3 and Table S1. Specifically, in Fig. 1a,b, we chose to compare only SST-based indices and in particular only ENSO and NZI in order to demonstrate how and which SWUS climatic regions were chosen for our analysis.
The ocean-atmosphere bridge mechanism: this is an interesting idea. it seems useful to note that GCM experiments exploring seasonal atmospheric predictability (e.g. Hoerling, Kumar and others) have found much greater influence from SST in the tropical Pacific than from the extratropics. Your results propose an influence from SST outside the tropics it seems.
We thank the Reviewer. We have added a discussion in the conclusions (lines 258-261) and such analysis is planned to be pursued in the near future.
line 256 "and thus, they CAN be used towards increasing precipitation predictability." i'd suggest saying "they MAY be used.....", since we don't know how long these stronger than ENSO linkages will hold up.
We thank the Reviewer and we agree with the suggestion, which we have incorporated.
lines 59, 270 you refer to rainguage observations, which sounds like you use individual station precipitation records. but the divisional data are averages of precipitation from a set of stations whose membership has changed over time (depending on availability of date, etc). and "rainguage" might be interpreted as only including rain, but this is winter precipitation and important parts (higher elevations) of the SWUS divisions are dominated by snow during the winter.