Research Article  Open Access
Xiaoshan Sun, Renguang Wu, "Contribution of Wind Speed and SeaAir Humidity Difference to the Latent Heat FluxSST Relationship", OceanLandAtmosphere Research, vol. 2022, Article ID 9815103, 17 pages, 2022. https://doi.org/10.34133/2022/9815103
Contribution of Wind Speed and SeaAir Humidity Difference to the Latent Heat FluxSST Relationship
Abstract
This study investigates contributions of wind speed and seaair humidity difference (dq) terms to the seasonal change and time scale dependence in the relationship between surface latent heat flux (LHF) and sea surface temperature (SST) using daily data. Generally, the dq term is dominant in the SST effect on LHF in the midlatitude SST frontal zones and tropical Indowestern Pacific, and the wind speed term is dominant in the LHF effect on SST in the subtropical gyres and tropical Indowestern Pacific. The seasonal change in the dq term accounts for a larger SST effect in winter than in summer in the midlatitude frontal zones, and that of the wind speed term explains a larger LHF effect in summer than in winter in the subtropical gyres. In the tropical Indowestern Pacific, the dq term is dominant in the SST effect in summer, and the wind speed term is dominant in the LHF effect in winter. The contribution of the dq term to the SST effect increases with the time scale. The contribution of the wind speed term to the SST effect varies regionally: It is supplementary in the midlatitude frontal zones in winter and summer and in the Arabian Sea in summer, but it is opposite in the Philippine Sea in winter and summer and in the South China Sea and Bay of Bengal in summer. The contribution of the wind speed term to the LHF effect is confined to short time scales in most of the tropical Indowestern Pacific regions.
1. Introduction
Surface turbulent heat flux is an important quantity in atmosphereocean interactions. It is composed of latent heat flux (LHF) and sensible heat flux. LHF is much larger than sensible heat flux in most oceans [1, 2]. The direction of the influence of the atmosphere or ocean through surface turbulent heat flux can be determined by the relative magnitude of the simultaneous correlation coefficients of LHF with the sea surface temperature (SST) and the SST tendency (the change in SST with time). The SST influence on LHF is manifested in the simultaneous correlation between LHF and the SST as the atmospheric response to the SST forcing is quick. The LHF influence on the SST is through the SST tendency and thus is manifested in the simultaneous correlation between LHF and the SST tendency. The SST influence on LHF dominates if the magnitude of the correlation coefficient between LHF and SST is larger than the absolute value of the correlation coefficient between LHF and the SST tendency, whereas the LHF influence on the SST is dominant in the opposite case. This representation of the airsea relationship has been illustrated by simulations of a simple stochastic model driven by atmospheric or oceanic forcing [3–6]. In the case of oceanic forcing, a large positive correlation is obtained between LHF (positive denoting upward) and the SST, which denotes warmer water heating the overlying atmosphere by an enhanced LHF. In the case of atmospheric forcing, a large negative correlation is obtained between LHF and the SST tendency, which denotes an enhanced LHF leading to the SST decrease. Physically, in the case of oceanic forcing, oceanic processes play a major role in the SST change, and the induced SST anomalies, in turn, drive the atmosphere by modulating LHF. In the case of atmospheric forcing, atmospheric processinduced surface heat flux plays a larger role than oceanic processes in SST changes.
Previous studies have examined the features of the atmosphereocean relationship through diagnosis of the correlation or covariance of LHF and the SST or the SST tendency [4, 5, 7–18]. Spatial differences and seasonal changes have been detected in the relationship between LHF and SST/SST tendencies [4, 5, 8, 13–16]. In the North Pacific, the SST effect on LHF is larger than the LHF effect on the SST in the Kuroshio Extension region, whereas the LHF effect on the SST is dominant in the subtropical gyre region [4, 5]. In the Arabian Sea, the SST effect dominates in summer, but the LHF effect is larger in winter [5]. Along the Gulf Stream and Kuroshio Extension, the SST effect is larger in winter than in summer [4, 5, 13], which is attributed to seasonal changes in the SST gradient [4, 11, 13, 19].
One question is why the relationship between LHF and SST variations changes with location and season. This is associated with the relative roles of surface heat flux and oceanic processes in the SST change and the relative contributions of surface wind speed and the seaair humidity difference to the LHF change. The role of ocean advection in the SST change is determined by the SST gradient and ocean current. In regions with large mean SST gradients, such as the SST frontal zones along the Gulf Stream and Kuroshio Extension, highspeed ocean currents related to smallscale ocean eddies [20, 21] act on large mean SST gradients; thus, ocean advection may play a major role in SST changes, which leads to the large SST effect on LHF [4, 5, 11, 13]. In contrast, in regions with weak mean SST gradients, such as subtropical gyre regions, ocean advection is small, and surface heat flux may play a major role in SST changes [4, 22]. The role of LHF in SST changes depends upon both the surface wind speed and seaair humidity difference [11, 23]. The wind speed effect is mostly related to atmospheric processes, whereas the seaair humidity difference effect is associated with both the SST and atmospheric thermodynamics. This study attempts to investigate the relative contributions of wind speed and seaair humidity difference effects to the spatial and seasonal changes in the relationship between LHF and SST variations.
Small et al. [11] decomposed LHF into terms related to the surface wind speed and seaair humidity difference and found that the dominant driver of LHF variability changes with location. Atmospheric humidity is a predominant factor at high latitudes, and the wind speed effect is the most important driver of LHF changes in most of the tropics and subtropics [11]. Small et al.’s [11] analysis is based on monthly mean data that cannot resolve submonthly highfrequency variations in the ocean and atmosphere. In the tropical Indowestern Pacific region, large intraseasonal oscillations are prominent in both the atmosphere and ocean [14, 18]. In the Gulf of Mexico, anticyclonic loop current frontal eddies show variations on submonthly time scales [24, 25], and submonthly mesoscale heat flux convergence contributes to lateral advection [26]. Time smoothing may remove highfrequency atmospheric wind and humidity fluctuations that influence the upper ocean [27, 28]. The atmosphere changes with SST anomalies within a week, and the ocean mixed layer can respond to the changing atmosphere on submonthly time scales [29]. Using daily data, Sun and Wu [5] detected transitions from atmospheric forcing to oceanic forcing at submonthly time scales, which is not possible using monthly data. In this study, daily data are used to investigate the contributions of wind speed and seaair humidity difference effects to the spatial and seasonal changes in the relationship between LHF and SST variations.
The relationship between LHF and SST variations depends on the time scale [4, 5]. With the time scale increasing from 1 to 24 months, the oceanic contribution to the SST change increases, and the LHF influence is diminished [4]. In the tropical western North Pacific, the contribution of LHF to intraseasonal SST variations is larger on 3060day scales than on 1020day time scales [14, 18]. Sun and Wu [5] revealed that the transition from the LHF effect to the SST effect occurs on an approximately 20day time scale in the Gulf Stream and Kuroshio Extension and on an approximately 40day time scale in the Philippine Sea in summer. The present study intends to document the contributions of surface wind speed and seaair humidity difference effects to the time scale dependence of the relationship between LHF and SST variations based on daily data.
The present study compares the contributions of surface wind speed and seaair humidity difference effects to the correlation of LHF with the SST and the SST tendency in the midlatitude SST frontal zones, subtropical gyre regions, and tropical Indowestern Pacific regions. Our purpose is to understand the changes in the relationship between LHF and SST variations with season and time scale in the above regions. The rest of the paper is organized as follows. The observational datasets and methods are introduced in Section 2. The regional features of the contribution of the surface wind speed and seaair humidity difference to LHF and its relationship with the SST are shown in Section 3.1. In Section 3.2, we present the seasonality of the leadlag correlation between the driving factors of LHF and the SST/SST tendency at different locations. Section 3.3 focuses on the time scale dependence of the leadlag correlation at different locations. A summary and discussion are given in Section 4.
2. Materials and Methods
The daily observational SST data used in this study are from the National Oceanic and Atmospheric Administration (NOAA) Optimum Interpolation Sea Surface Temperature (OISST) v2.0 [30]. The OISST data have a 0.25° spatial resolution and are available from January 1, 1985, to December 31, 2018. The daily OISST data were obtained from the following website: https://psl.noaa.gov/data/gridded/data.noaa.oisst.v2.highres.html.
For the daily LHF, surface air specific humidity, and wind speed data, we used the objectively analyzed airsea flux (OAFlux) product [31] for the time period from 1985 to 2018. The spatial resolution of the OAFlux data is 1° grids. The OAFlux variables are interpolated to 0.25° grids to match the OISST data. The interpolation to high resolution does not alter the LHF information and its driving factors, as the signals of LHF variability tend to have larger spatial scales than the signals of SST variability.
We also used the daily LHF, surface air specific humidity and wind speed data from the Japanese Ocean Flux Data Sets with Use of Remote Sensing Observations version 3 (JOFURO3). The JOFURO3 is a thirdgeneration surface flux dataset estimated using satellite remote sensing techniques [32]. The dataset has been recently updated from V1.0 to V1.1. LHF, surface seaair humidity difference, surface wind, and SST variables are contained in the dataset and cover the time period from 1988 to 2017 on 0.25° grids. The JOFURO3 variables were downloaded from http://search.diasjp.net/en/dataset/JOFURO3_V1_1. The results obtained from the SST and JOFURO3 V1.1 were compared with those based on OAFlux. The results based on JOFURO3 and OAFlux showed similar features except for the correlation of the seaair humidity difference (dq) term and SST, which is generally larger based on OAFlux than on JOFURO3, and the negative correlation of the wind speed term and the SST tendency, which is larger based on JOFURO3 than on OAFlux. We present only figures based on OAFlux.
Surface turbulent heat flux is composed of LHF and sensitive heat flux. LHF makes much greater contributions than sensitive heat flux [1, 2]. Therefore, this study focuses on the decomposition of LHF and the effect of LHF on SST variations. LHF () is calculated using the following aerodynamic bulk formula: where , , and are the surface air density, wind speed at 10 m above the sea surface, and specific humidity at 2 m above the sea surface, respectively. represents the saturated specific humidity at the SST (). is the latent heat of vaporization for water, and is the bulk coefficient for moisture. According to Tanimoto et al. [23] and Small et al. [11], the LHF anomalies (Q_{L}^{’}) can be written using the climatological mean and anomalies: where overbars represent the climatological mean and primes denote the deviation from the climatological mean. The climatological calendar day mean is calculated based on the arithmetic mean of all same calendar day values in the analysis period. The meanings of the remaining symbols are the same as those in Formula (1). The first term on the righthand side of Equation (2) is the contribution of the wind speed (denoted as the wind speed term for convenience), and the second term is the contribution of the specific humidity difference between air and sea (denoted as the dq term for convenience). We focus on the first two terms on the righthand side of Equation (2), as the third and fourth terms are considered to be small and are negligible [11]. In this study, we analyze the correlation of the wind speed term and dq term with the SST and SST tendency to explain the relationship between LHF and SST variations. The quantity dq contains both and . The dq variation is highly correlated with in the subtropical gyre regions, but is of vital importance to the dq variation in the midlatitude frontal zones and the tropics within 015°N (not shown). In the present analysis, we show only the correlation of the dq term with the SST and SST tendency, and the correlation of is not presented for brevity.
The local correlation of LHF, the wind speed term and dq term with the SST and SST tendency is examined for all months and separately for months in winter and summer. The daily value of the SST tendency is the difference in the subsequent day SST minus the preceding day SST divided by 2. Northern winter (southern summer) extends from November to March in the following year (NDJFM), and northern summer (southern winter) covers from May to September (MJJAS). In Section 3.3, the correlation is calculated with respect to northern winter (southern summer) from December 15 to February 14 in the following year and northern summer (southern winter) from June 15 to August 15 to avoid the inclusion of signals in the other seasons in time smoothing.
In Section 3, the twotailed Student’s t test is used to examine the significance of the correlation coefficients of LHF, the wind speed term and dq term with the SST and SST tendency. The effective degree of freedom is employed to obtain the critical correlation coefficient at the 95% confidence level. The effective sample size () is estimated using the method of Preisendorfer [33]: where is the correlation coefficient and is the total number of days used to calculate the correlation coefficient. For the correlation coefficient with the SST tendency, is divided by 2 in view of the use of each daily value twice in the calculation of the SST tendency. Most grid correlation coefficients are significant due to the large number of days in the 34 years, and the critical correlation coefficient is small (ranging from 0.020 to 0.060 in the North Pacific and from 0.020 to 0.052 in the tropical Indowestern Pacific). In Section 4, the error margins of the correlation coefficients are estimated based on the maximum and minimum values of the correlation coefficients for a specific leadlag time. The maximum and minimum correlation coefficients are obtained based on a comparison of the four correlation coefficients calculated using variables including the errors, which are included in the heat flux and SST datasets. For details, refer to Sun and Wu [5].
3. Results
3.1. Regional Features of the Wind Speed and dq Term Contributions
In this subsection, we examine the spatial changes in the contribution of the wind speed and dq terms of LHF to the relationship between LHF and SST variations. First, we provide a comparison of the contributions of the wind speed and dq terms to LHF variations. The wind speed and dq terms of LHF are calculated based on Equation (2). The features in the North Pacific, North Atlantic, and Southern Ocean are generally similar. Therefore, we mainly describe features in one of the three regions.
Linear regressions of the wind speed and dq terms of LHF onto total LHF are shown in Figure 1. Positive values indicate a positive contribution of the term to LHF, while negative values mean that the term has a negative effect on LHF variation. The sum of the regression coefficients of the wind speed and dq terms may deviate from 1, which indicates contributions of the third and fourth terms of Equation (2). In the North Pacific, the distribution of the regression coefficient (Figures 1(a) and 1(b)) indicates that the wind speed term has a larger contribution at lower latitudes and that the dq term is dominant at higher latitudes. The sum is somewhat smaller than 1 at higher latitudes (Figure 1(c)), indicative of the effects of the third and fourth terms there. Similar features are observed in the North Atlantic and southern oceans. In the tropical North Indian Oceanwestern North Pacific, the regression coefficient of the wind speed term onto LHF decreases northward from the equator (Figure 1(a)) and that of the dq term is small near the equator and turns positive north of 10°N (Figure 1(b)). In comparison, the wind speed term is dominant in driving LHF near the equator, and both the wind speed and dq terms contribute to the LHF variation north of 10°N. The sum of the two terms is close to 1 (Figure 1(c)), confirming the negligible effects of the third and fourth terms in Equation (2).
Next, we compare the contributions of the wind speed and dq terms of LHF to the relationship between LHF and SST variations. The relative importance of the contributions of the wind speed and dq terms is assessed based on the absolute value of their simultaneous correlations with the SST and SST tendency. Figures 2 and 3 present the correlation coefficients of LHF, the wind speed term, and dq term with the SST and SST tendency in the North Atlantic and tropical North Indian Oceanwestern North Pacific. According to previous studies [4–6], a positive correlation between LHF and the SST illustrates the influence of the SST on LHF, and a negative correlation between LHF and the SST tendency indicates the influence of LHF on SST changes. In the SST frontal zone of the North Atlantic, the dq term has a large positive correlation, and the wind speed term has a small correlation with the SST (Figures 2(a), 2(c), and 2(e)). This result indicates a dominant contribution of the dq term to the positive LHFSST correlation, and the contribution of the wind speed term is small. Similar results are obtained in the SST frontal zones of the North Pacific and southern Indian Ocean (Figures S1a, S1c, and S1e). This finding is related to the strong currents and large SST gradient along the narrow SST frontal zones, where oceanic advection plays a major role in the SST change that in turn influences LHF by modulating the seaair humidity difference [4, 11]. In subtropical gyre regions, the dq and wind speed terms tend to have opposite correlations with the SST with comparable magnitudes (Figures 2(c), and 2(e); Figures S1c and S1e), leading to small LHFSST correlations (Figure 2(a); Figure S1a). The negative wind speedSST correlation in the subtropical gyre regions is likely related to atmospheric circulation changes. For example, when the subtropical high moves poleward, the wind speed is enhanced on the one hand, and on the other hand, the SST falls due to the cloudrelated downward shortwave radiation decrease and windrelated LHF increase, leading to a negative wind speedSST correlation. The wind speed term has a larger negative correlation with the SST tendency than the dq term in the subtropical regions, and thus, it has a larger contribution to the negative LHFSST tendency correlation there (Figures 2(b), 2(d), and 2(f); Figures S1b, S1d, and S1f). This result indicates the role of atmospheric processes in driving LHF and in turn the SST change.
In the tropical North Indian Oceanwestern North Pacific, the dq term has a large positive correlation (Figure 3(e)), and the wind speed term has a negative correlation with the SST (Figure 3(c)), leading to a small positive LHFSST correlation in most regions (Figure 3(a)). The negative wind speedSST correlation in the tropical North Indowestern Pacific region is likely associated with monsoon wind changes. For example, when the summer monsoon intensifies, the wind speed becomes enhanced on the one hand, and on the other hand, the SST falls due to the cloudrelated downward shortwave radiation decrease and windinduced LHF increase, leading to a negative wind speedSST correlation. The western Arabian Sea is an exception where the positive LHFSST correlation is relatively large (Figure 3(a)) due to the relatively small negative correlation of the wind speed term (Figure 3(c)). The wind speed term has a dominant contribution to the negative LHFSST tendency correlation, and the contribution of the dq term is small in most regions (Figures 3(b), 3(d), and 3(f)). This finding agrees with the dominant contribution of the wind speed term in the LHF decomposition (Figures 1(b), 1(d), and 1(f)).
In the SST frontal zones of the North Pacific, North Atlantic and southern Indian Ocean, the positive LHFSST correlation decreases from winter to summer (Figures S2a and S3a). This seasonal change is partly contributed by a decrease in the positive contribution of the dq term (Figures S2e and S3e) and partly by an increase in the negative contribution of the wind speed term (Figures S2c and S3c). In the subtropical gyre regions, the negative LHFSST tendency correlation increases from winter to summer (Figures S2b and S3b), which is mainly attributed to the increase in the contribution of the wind speed term (Figures S2d and S3d). Along the west coast of the Arabian Sea, the positive LHFSST correlation displays an obvious increase from winter to summer (Figures S2a and S3a), which appears to be associated with an increase in the positive contribution of the dq term (Figures S2e and S3e) and a decrease in the negative contribution of the wind speed term (Figures S2c and S3c). More details about the seasonal changes in the relationship between LHF and SST variations are presented in the next subsection.
3.2. Seasonality of the LeadLag Correlation at Different Locations
In this section, we compare the leadlag correlations of LHF, the wind speed term, and the dq term with the SST and SST tendency at different locations in winter and summer. The purpose is to understand the contributions of the wind speed and dq terms to seasonal changes in the relationship between LHF and SST variations. The relative importance of the contributions of the wind speed and dq terms is determined based on the magnitude and change with the time lag in their correlations in individual seasons. The leadlag correlations are analyzed for selected representative points, including 3 points in the midlatitude SST frontal zones of the Gulf Stream, the Kuroshio Extension, and the Agulhas Return Current; 3 points in the subtropical gyres; and 4 points in the tropical North Indian Oceanwestern North Pacific with each in the Arabian Sea, the Bay of Bengal, the South China Sea, and the Philippine Sea. The 3 points in the midlatitude SST frontal zones represent dominant oceanic forcing, and the 3 points in the subtropical gyres represent dominant atmospheric forcing [4, 5]. The 4 points in the tropical North Indian Oceanwestern North Pacific feature dominant atmospheric forcing with supplementary oceanic forcing, except for the Arabian Sea in summer with dominant oceanic forcing [5]. The locations of those points are the same as those in Sun and Wu [5], and those points in the North Atlantic and tropical North Indian Oceanwestern North Pacific are denoted by black diamonds in Figures 2 and 3. Following previous studies [4–6], a positive and symmetric leadlag correlation between LHF and the SST signifies an SST effect on LHF, and a negative and symmetric leadlag correlation between LHF and the SST tendency indicates the LHF effect on SST changes. In the following, we present only figures for the 3 points in the midlatitude SST frontal zones and the 2 points in the Arabian Sea and South China Sea. Short descriptions are provided for the 3 points in the subtropical gyres and the 2 points in the Bay of Bengal and Philippine Sea.
In the midlatitude SST frontal zones, the leadlag correlation between the dq term and SST has a symmetric structure with a large positive correlation at zero lag in both winter and summer (Figure 4). In comparison, the peak positive correlation is larger in winter than in summer. The leadlag correlation between the wind speed term and SST is small in both winter and summer and displays asymmetric features in summer. Apparently, the dq term makes a main contribution to the positive LHFSST correlation, and it has a larger value in winter than in summer. The above results illustrate that oceaninduced SST anomalies affect LHF by modulating the dq term in midlatitude SST frontal zones. The main reason for the larger correlation between the dq term and SST in winter is the stronger ocean currents and larger meridional SST gradient in winter than in summer, leading to a larger role of ocean advection in SST changes and, in turn, a larger influence of the SST on LHF by modulating the seaair humidity difference.
In the midlatitude SST frontal zones, the leadlag correlation between the dq term and SST tendency has an antisymmetric structure with positive and negative correlations when the SST tendency leads and lags, respectively (Figure 5). In comparison, the antisymmetric feature is more prominent in winter than in summer. The leadlag correlation between the wind speed term and SST tendency displays a symmetric feature with a large negative correlation at zero lag. In comparison, the simultaneous negative correlation is much larger in summer than in winter. The results mean that the negative LHFSST tendency correlation is mainly attributed to the wind speed contribution, which is larger in summer than in winter. The seasonal change in the above wind speed termSST tendency correlation occurs for two reasons. One reason is the smaller covariance between the wind speed and SST tendency in winter than in summer (not shown). The other reason is the smaller variance in the wind speed in summer than in winter (not shown), which is related to seasonal changes in the magnitude of atmospheric highfrequency fluctuations.
In the subtropical gyre regions, the wind speed termSST and the dq termSST correlations tend to be opposite and cancel each other at zero lag, leading to a small simultaneous LHFSST correlation in both winter and summer (figures not shown). This finding is consistent with Figures 2(a), 2(c), and 2(e). The wind speed termSST tendency correlation displays symmetric features with large negative values at a zero lag in both winter and summer, whereas the dq termSST tendency correlation displays an antisymmetric structure in both winter and summer (figures not shown). In comparison, the magnitude of the peak correlation coefficient is larger in summer than in winter. These results are consistent with Figures 2(b), 2(d), and 2(f). Overall, in the subtropical gyre regions, the wind speed term dominates the LHF effect on the SST change, and this effect is stronger in summer than in winter. The reasons for the seasonal changes are similar to those in the midlatitude SST frontal zones.
In winter, the dq term contributes to the positive SSTLHF correlation, and the wind speed term has a negative contribution in both the Arabian Sea and the South China Sea, leading to a small SSTLHF correlation (Figures 6(a) and 6(c)). In summer, the contribution of the dq term dominates the positive LHFSST correlation, and the contribution of the wind speed term is small in the Arabian Sea (Figure 6(b)). This finding validates the SST effect on LHF in the Arabian Sea in summer by modulating the seaair humidity difference. The SST effect in this region is related to the large SST gradient during summer [5]. The contribution of the dq term is also large in the South China Sea, but it is mostly cancelled by the negative contribution of the wind speed term, leading to a small SSTLHF correlation there (Figures 6(d)). This opposing wind speed term is distinct from that of the Arabian Sea. The features in the Bay of Bengal and Philippine Sea (figures not shown) are similar to those in the South China Sea.
In winter, the wind speed termSST tendency correlation is largely negative at zero lag, accounting for the simultaneous negative LHFSST tendency correlation in both the Arabian Sea and South China Sea (Figures 7(a) and 7(c)). The dq termSST tendency correlation displays an antisymmetric feature in the above regions. The above feature is also obtained in the South China Sea in summer (Figure 7(d)). Similar results are obtained in the Bay of Bengal and Philippine Sea in both winter and summer (figures not shown). In the Arabian Sea in summer, the antisymmetric dq termSST tendency correlation dominates the LHFSST tendency correlation (Figure 7(b)). The above results confirm the wind speedrelated LHF effect on the SST changes in the tropical Indowestern Pacific region except for the Arabian Sea in summer.
In summary, in the midlatitude frontal zones, the dq term in the SST effect is larger in winter than in summer. In subtropical gyre regions, the wind speed term in the LHF effect is larger in summer than in winter. In the Arabian Sea, the wind speed term dominates in winter in the LHF effect, and the dq term dominates in summer in the SST effect. In the other tropical Indowestern Pacific regions, the wind speed term in the LHF effect is larger in summer than in winter, and the dq term in the SST effect is larger in summer than in winter.
3.3. Time Scale Dependence of the LeadLag Correlation at Different Locations
The time scale dependence of the relationship between LHF and SST variations has been illustrated by correlation analysis based on timesmoothed variables in previous studies [4, 5]. In this section, we examine the correlations of the wind speed and dq terms with the SST and SST tendency for time scales from 2 to 90 days. These values are compared to the correlations of LHF with the SST and SST tendency to understand the contributions of the wind speed and dq terms to the time scale dependence of the relationship between LHF and SST variations. The relative importance of the contributions of the wind speed and dq terms is indicated in the magnitude and variation with the time scale of their correlations. The correlation is analyzed based on daily data for the selected points as in Section 3.2. We present figures only for the points located in the Gulf Stream, the Arabian Sea, and the Philippine Sea. We also present figures of one season selected for the correlation with the SST or the SST tendency. The results for other points are briefly discussed in the text without presenting figures.
In the Gulf Stream in winter, the positive correlation between the dq term and SST increases monotonously with time (Figure 8(e)), which accounts for the increase in the SST effect on LHF with time (Figure 8(a)). The wind speed term has a supplementary contribution, especially when the SST leads (Figure 8(c)). The increase in the correlation of the dq term with time is due to the time smoothing that reduces the variance in the seaair humidity difference, and the effect of time smoothing increases with time. Similar changes with time are obtained in the correlation with the SST in summer but with smaller correlation coefficients (figures not shown). In summer, the simultaneous negative correlation between LHF and the SST tendency and its change with time is mainly attributed to the wind speed term for time scales less than 30 days (Figures 8(b) and 8(d)). The correlation between the dq term and SST tendency maintains an antisymmetric feature for time scales up to approximately 50 days (Figure 8(f)). The results indicate that the wind speed contribution to the LHF effect is confined to short time scales. This feature may be related to the change in the variance in wind speed over time. With the increase in the time scale, the amplitude of the wind speed fluctuation decreases, as does its contribution to LHF and its effect on the SST change. In winter, the antisymmetric LHFSST tendency correlation is maintained up to a time scale of 90 days, which follows the correlation of the dq term, whereas the wind speed termSST tendency correlation is negative and increases with time (figures not shown). Similar contributions of the wind speed and dq terms to changes with time for the relationship between LHF and SST variations are observed in the Kuroshio Extension and the Southern Ocean (figures not shown). One difference to note is that the wind speed termSST correlation is larger in the Kuroshio Extension and the Southern Ocean than in the Gulf Stream.
In the Arabian Sea, in summer, the increase with time for the positive LHFSST correlation is contributed by both the wind speed and dq terms, with the latter being larger than the former (Figures 9(a), 9(c), and 9(e)). The positive wind speedSST correlation may be due to the SST effect on vertical mixing. An increase in the SST may destabilize the atmospheric boundary layer and enhance vertical mixing, resulting in an increase in the surface wind speed, which results in a positive wind speedSST correlation. The increase in the wind speedSST correlation with time is due to the effect of the time smoothing that reduces the variance in wind speed, and the effect of time smoothing increases with time. In winter, the three correlations are much smaller (figures not shown). In winter, the wind speed term appears to have a main contribution to the change with time for the negative LHFSST tendency correlation (Figures 9(b), 9(d), and 9(f)). The negative correlation near the zero lag displays an increase up to 2040 days and then a decrease with time (Figures 9(b) and 9(d)). This finding indicates that the atmospheric effect on the SST changes through windrelated LHF is confined to short time scales. In summer, the change with time for the LHFSST tendency correlation is related to both the wind speed and dq terms but with a larger contribution from the dq term than the wind speed term (figures not shown). The three correlations maintain antisymmetric features as the time scale increases.
The time scale dependence of the leadlag correlations in the Philippine Sea has a smaller seasonal change than that in the midlatitude regions and the Arabian Sea. In winter, the positive correlation between the dq term and SST increases monotonously with time (Figure 10(c)). The correlation of the wind speed term has a counteracting effect, which also increases with time (Figure 10(c)). The net effect leads to a positive LHFSST correlation whose magnitude increases with time (Figure 10(a)). In summer, the three correlations show similar changes with time, but the positive correlation of the dq term and the negative correlation of the wind speed term are larger, and their cancelling effect leads to a small positive LHFSST correlation (figures not shown). In winter, the correlation of the wind speed and dq terms with the SST tendency tends to be opposite except for the correlation near the zero lag (Figures 10(d) and 10(f)). The negative LHFSST tendency correlation displays a nearly symmetric feature and first increases up to 20 days and then decreases with time (Figure 10(b)), which is mainly due to the contribution of the wind speed term (Figure 10(d)). Similar results are obtained in summer (figures not shown). The changes with the time scale of the leadlag correlations in the Bay of Bengal and South China Sea are similar to those in the Philippine Sea except that the opposing effect of the wind speed term to the LHFSST correlation is absent in winter.
In summary, in the midlatitude frontal zones during summer and the tropical North Indowestern Pacific region during winter and summer, the wind speed term in the LHF effect tends to dominate over a short time scale, and the dq term in the SST effect tends to be dominant over a long time scale, with the wind speed term supporting the Arabian Sea and cancelling the Philippine Sea, the Bay of Bengal, and the South China Sea. In the midlatitude frontal zones during winter and the Arabian Sea during summer, the dq term dominates the SST effect on a time scale up to 90 days. In the subtropical gyre regions during winter and summer, the wind speed term dominates the LHF effect on the time scale up to 90 days.
4. Conclusion
Previous studies revealed the spatial and seasonal changes and time scale dependence of the relationship between LHF and SST variations. Using daily data, this study decomposes LHF into the wind speed term and dq term. The correlation of LHF, the wind speed term, and the dq term with the SST and SST tendency is compared to understand the changes in the relationship between LHF and SST variations in the midlatitude SST frontal zones, subtropical gyre regions, and tropical North Indian Oceanwestern North Pacific. The use of the daily data enables a clear illustration of the effect of time smoothing on the relationship, in particular, at submonthly time scales and the detection of the switch in the dominant contribution from the wind speed term to the dq term at short time scales. The present study reveals that the relative contributions of the wind speed term and dq term vary with region and season and depend upon the time scale. The main results are summarized in Table 1.

The SST effect on LHF is well manifested in the dq term in the midlatitude SST frontal zones and tropical Indowestern Pacific. The LHF effect on the SST change comes mainly from the wind speed term in the subtropical gyre regions and tropical Indowestern Pacific. In subtropical gyre regions, the opposing contributions of the wind speed and dq terms lead to a weak LHFSST correlation. The smaller opposing wind speed effect in the Arabian Sea accounts for the larger LHFSST correlation compared to those in the other tropical Indowestern Pacific regions.
Obvious seasonal changes are detected in the contributions of the wind speed and dq terms. In the midlatitude frontal zones, the larger SST effect in winter than in summer is validated by the increase in the contribution of the dq term from summer to winter. The wind speed term accounts for the LHF effect on the SST change, and the wind speed effect is larger in summer than in winter. In subtropical gyre regions, the larger LHF effect on the SST changes in summer than in winter is attributed to seasonal changes in the wind speed term. The correlations of the wind speed and dq terms with the SST tend to cancel each other in both winter and summer, leading to a small SST effect on LHF in most of the subtropical gyre regions. In the tropical Indowestern Pacific, the SST effect on LHF mainly occurs through modulation of the dq term, whose contribution is larger in summer than in winter. The LHF effect on the SST change in winter and summer comes mainly from the wind speed term. In winter, the correlation of the wind speed term with the SST tends to cancel that of the dq term, leading to a weak LHFSST relationship. In summer, the correlation of the wind speed term with the SST is small in the Arabian Sea but is opposite to that of the dq term in the other tropical Indowestern Pacific regions. This accounts for the weaker SST effect on LHF in the Bay of Bengal, South China Sea, and Philippine Sea than in the Arabian Sea.
The contributions of the wind speed and dq terms display distinct time scale dependences. The SST effect on LHF through the dq term increases monotonically with time in most regions. The contribution of the wind speed term to the SST effect also increases with time, but its effect varies regionally. The wind speed term is supplementary in the midlatitude frontal zones in both winter and summer and in the Arabian Sea in summer. The wind speed term opposes the dq term in the Philippine Sea in winter and summer, in the South China Sea and Bay of Bengal in summer, and in the subtropical gyre regions in both winter and summer. The contribution of the wind speed term to the LHF effect first increases and then decreases with time in the midlatitude SST frontal zones in summer; in the subtropical gyre regions in both winter and summer; in the Arabian Sea in winter; and in the Bay of Bengal, South China Sea, and Philippine Sea in both winter and summer. This feature signifies that the LHF effect on the SST changes through the wind speed effect tends to be confined to short time scales.
Seasonal changes in the magnitude of the wind speed effect on SST changes are likely related to the variance in highfrequency atmospheric wind changes. Atmospheric wind fluctuations tend to be larger in winter than in summer in the midlatitude regions, which may partly explain the larger correlation of the wind speed term with the SST tendency there. Seasonal changes in the contribution of the seaair humidity difference to LHF are attributed to the SST gradient. The SST gradient in the midlatitude SST frontal zones is larger in winter than in summer. Accordingly, oceanic advection plays a larger role in the SST changes, and in turn, the SST has a larger influence on LHF through the dq term in winter than in summer. The SST gradient is also a factor for the large SST effect on LHF in the Arabian Sea in summer. The increase with time of the correlation of the dq term and SST is attributed to the time smoothing that reduces the highfrequency atmospheric noise. With the increase in the time scale, the smoothing reduces the highfrequency noise more efficiently, and the correlation between the dq term and SST becomes larger.
Some questions remain to be answered. One is why the contribution of the wind speed effect on the SST changes is confined to short time scales (less than 4060 days). Another is the spatial changes in the correlation between the SST and wind speed term of LHF: positive in the SST frontal zones and the Arabian Sea and negative in the subtropical gyre regions and the Bay of Bengal, South China Sea, and Philippine Sea. The first issue may be related to the predominant period of intraseasonal atmospheric wind variations. The second issue may be related to the regional changes in the coherence of atmospheric wind and moisture variations. Further investigation is needed to address the above issues.
Data Availability
The NOAA OISST v2.0 data were obtained from ftp://ftp.cdc.noaa.gov/Datasets/noaa.oisst.v2.highres/. The OAFlux flux data were obtained from ftp://ftp.whoi.edu/pub/science/oaflux/data_v3/daily/turbulence/. The JOFURO flux data were obtained from http://dtsv.scc.utokai.ac.jp/jofuro/.
Conflicts of Interest
There are no conflicts of interest to declare.
Authors’ Contributions
XS did the analysis. RW acquired the funding and supervised the research. XS prepared the draft. Both contributed to the revising of the paper.
Acknowledgments
This study is supported by the National Natural Science Foundation of China (grant 41721004).
Supplementary Materials
Figure S1. Correlation of LHF, the wind speed term, and the seaair humidity difference (qsQA) term with the SST and SST tendency in all months. Figure S2. Correlation of LHF, the wind speed term, and the seaair humidity difference (qsqa) term with the SST and SST tendency in NDJFM. Figure S3. Correlation of LHF, the wind speed term, and the seaair humidity difference (qsqa) term with the SST and SST tendency in MJJAS. (Supplementary Materials)
References
 E. L. Andreas, R. E. Jordan, L. Mahrt, and D. Vickers, “Estimating the Bowen ratio over the open and icecovered ocean,” Journal of Geophysical Research, Oceans, vol. 118, no. 9, pp. 4334–4345, 2013. View at: Publisher Site  Google Scholar
 S. Park, C. Deser, and M. A. Alexander, “Estimation of the surface heat flux response to sea surface temperature anomalies over the global oceans,” Journal of Climate, vol. 18, no. 21, pp. 4582–4599, 2005. View at: Publisher Site  Google Scholar
 J. J. Barsugli and D. S. Battisti, “The basic effects of atmosphere–ocean thermal coupling on midlatitude variability,” Journal of the Atmosphere Sciences, vol. 55, no. 4, pp. 477–493, 1998. View at: Publisher Site  Google Scholar
 S. P. Bishop, R. J. Small, F. O. Bryan, and R. A. Tomas, “Scale dependence of midlatitude air–sea interaction,” Journal of Climate, vol. 30, no. 20, pp. 8207–8221, 2017. View at: Publisher Site  Google Scholar
 X. Sun and R. Wu, “Seasonality and time scale dependence of the relationship between turbulent surface heat flux and SST,” Climate Dynamics, vol. 56, no. 910, pp. 3173–3186, 2021. View at: Publisher Site  Google Scholar
 R. Wu, B. P. Kirtman, and K. Pegion, “Local air–sea relationship in observations and model simulations,” Journal of Climate, vol. 19, no. 19, pp. 4914–4932, 2006. View at: Publisher Site  Google Scholar
 D. R. Cayan, “Latent and sensible heat flux anomalies over the northern oceans: driving the sea surface temperature,” Journal of Physical Oceanography, vol. 22, no. 8, pp. 859–881, 1992. View at: Publisher Site  Google Scholar
 J. P. Duvel and J. Vialard, “IndoPacific Sea surface temperature perturbations associated with intraseasonal oscillations of tropical convection,” Journal of Climate, vol. 20, no. 13, pp. 3056–3082, 2007. View at: Publisher Site  Google Scholar
 C. Frankignoul, A. Czaja, and B. L’Heveder, “Air–sea feedback in the North Atlantic and surface boundary conditions for ocean models,” Journal of Climate, vol. 11, no. 9, pp. 2310–2324, 1998. View at: Publisher Site  Google Scholar
 C. Frankignoul and K. Hasselmann, “Stochastic climate models, part II application to seasurface temperature anomalies and thermocline variability,” Tellus, vol. 29, no. 4, pp. 289–305, 1977. View at: Publisher Site  Google Scholar
 R. J. Small, F. O. Bryan, S. P. Bishop, and R. A. Tomas, “Air–sea turbulent heat fluxes in climate models and observational analyses: what drives their variability?” Journal of Climate, vol. 32, no. 8, pp. 2397–2421, 2019. View at: Publisher Site  Google Scholar
 J.S. von Storch, “Signatures of air–sea interactions in a coupled atmosphere–ocean GCM,” Journal of Climate, vol. 13, no. 19, pp. 3361–3379, 2000. View at: Publisher Site  Google Scholar
 X. Sun and R. Wu, “Spatial scale dependence of the relationship between turbulent surface heat flux and SST,” Climate Dynamics, vol. 58, no. 34, pp. 1127–1145, 2022. View at: Publisher Site  Google Scholar
 R. Wu, X. Cao, and S. Chen, “Covariations of SST and surface heat flux on 10–20 day and 30–60 day time scales over the South China Sea and western North Pacific,” Journal of Geophysical Research – Atmospheres, vol. 120, no. 24, pp. 12486–12499, 2015. View at: Publisher Site  Google Scholar
 R. Wu and J. L. Kinter III, “Atmosphereocean relationship in the midlatitude North Pacific: seasonal dependence and eastwest contrast,” Journal of Geophysical Research, vol. 115, no. D6, p. D06101, 2010. View at: Publisher Site  Google Scholar
 R. Wu, B. P. Kirtman, and K. Pegion, “Surface latent heat flux and its relationship with sea surface temperature in the National Centers for Environmental Prediction Climate Forecast System simulations and retrospective forecasts,” Geophysical Research Letters, vol. 34, no. 17, p. L17712, 2007. View at: Publisher Site  Google Scholar
 R. Wu and T. You, “Summer intraseasonal surface heat fluxsea surface temperature relationship over northern tropical IndoWestern Pacific in climate models,” Journal of Geophysical Research – Atmospheres, vol. 123, no. 11, pp. 5859–5880, 2018. View at: Publisher Site  Google Scholar
 K. Ye and R. Wu, “Contrast of local air–sea relationships between 10–20day and 30–60day intraseasonal oscillations during May–September over the South China Sea and western North Pacific,” Climate Dynamics, vol. 45, no. 1112, pp. 3441–3459, 2015. View at: Publisher Site  Google Scholar
 A. Bellucci, P. J. Athanasiadis, E. Scoccimarro et al., “Airsea interaction over the Gulf Stream in an ensemble of HighResMIP present climate simulations,” Climate Dynamics, vol. 56, no. 78, pp. 2093–2111, 2021. View at: Publisher Site  Google Scholar
 D. B. Chelton, M. G. Schlax, M. H. Freilich, and R. F. Milliff, “Satellite measurements reveal persistent smallscale features in ocean winds,” Science, vol. 303, no. 5660, pp. 978–983, 2004. View at: Publisher Site  Google Scholar
 D. B. Chelton and S.P. Xie, “Coupled oceanatmosphere interaction at oceanic mesoscales,” Oceanography, vol. 23, no. 4, pp. 52–69, 2010. View at: Publisher Site  Google Scholar
 C. D. Roberts, M. D. Palmer, R. P. Allan et al., “Surface flux and ocean heat transport convergence contributions to seasonal and interannual variations of ocean heat content,” Journal of Geophysical Research, Oceans, vol. 122, no. 1, pp. 726–744, 2017. View at: Publisher Site  Google Scholar
 Y. Tanimoto, H. Nakamura, T. Kagimoto, and S. Yamane, “An active role of extratropical sea surface temperature anomalies in determining anomalous turbulent heat flux,” Journal of Geophysical Research, vol. 108, no. C10, p. 3304, 2003. View at: Publisher Site  Google Scholar
 K. A. Donohue, D. R. Watts, P. Hamilton, R. Leben, M. Kennelly, and A. LugoFernández, “Gulf of Mexico loop current path variability,” Dynamics of Atmospheres and Oceans, vol. 76, pp. 174–194, 2016. View at: Publisher Site  Google Scholar
 J. Jouanno, J. Ochoa, E. PallàsSanz et al., “Loop current frontal eddies: formation along the Campeche Bank and impact of coastally trapped waves,” Journal of Physical Oceanography, vol. 46, no. 11, pp. 3339–3363, 2016. View at: Publisher Site  Google Scholar
 D. A. Putrasahan, I. Kamenkovich, M. Le Hénaff, and B. P. Kirtman, “Importance of ocean mesoscale variability for airsea interactions in the Gulf of Mexico,” Geophysical Research Letters, vol. 44, no. 12, pp. 6352–6362, 2017. View at: Publisher Site  Google Scholar
 K. Hasselmann, “Stochastic climate models part I. Theory,” Tellus, vol. 28, no. 6, pp. 473–485, 1976. View at: Publisher Site  Google Scholar
 J. M. Wallace, C. Smith, and Q. Jiang, “Spatial patterns of atmosphereocean interaction in the northern winter,” Journal of Climate, vol. 3, no. 9, pp. 990–998, 1990. View at: Publisher Site  Google Scholar
 C. Deser and M. S. Timlin, “Atmosphere–ocean interaction on weekly timescales in the North Atlantic and Pacific,” Journal of Climate, vol. 10, no. 3, pp. 393–408, 1997. View at: Publisher Site  Google Scholar
 R. W. Reynolds, T. M. Smith, C. Liu, D. B. Chelton, K. S. Casey, and M. G. Schlax, “Daily highresolutionblended analyses for sea surface temperature,” Journal of Climate, vol. 20, no. 22, pp. 5473–5496, 2007. View at: Publisher Site  Google Scholar
 L. Yu and R. A. Weller, “Objectively analyzed air–sea heat fluxes for the global icefree oceans (1981–2005),” Bulletin of the American Meteorological Society, vol. 88, no. 4, pp. 527–540, 2007. View at: Publisher Site  Google Scholar
 H. Tomita, T. Hihara, S. Kako, M. Kubota, and K. Kutsuwada, “An introduction to JOFURO3, a thirdgeneration Japanese ocean flux data set using remotesensing observations,” Journal of Oceanography, vol. 75, no. 2, pp. 171–194, 2019. View at: Publisher Site  Google Scholar
 R. W. Preisendorfer, “Principal component analysis in meteorology and oceanography,” in Developments in Atmospheric Sciences, C. Mobley, Ed., pp. 402–418, Elsevier Science Publishers, Amsterdan, 1988, https://www.scienceopen.com/document?vid=50203fc2dfae439a8edbe017cffb846b. View at: Google Scholar
Copyright
Copyright © 2022 Xiaoshan Sun and Renguang Wu. Exclusive Licensee Southern Marine Science and Engineering Guangdong Laboratory (Zhuhai). Distributed under a Creative Commons Attribution License (CC BY 4.0).