Dynamics of Upwelling Annual Cycle in the Equatorial Pacific Ocean

sustainability

Article Dynamics of Upwelling Annual Cycle in the Equatorial Paciﬁc Ocean

Li-Chiao Wang * and Jia-Yuh Yu

Department of Atmospheric Sciences, National Central University, Taoyuan 32001, Taiwan; [email protected] * Correspondence: [email protected]

Received: 30 July 2019; Accepted: 11 September 2019; Published: 15 September 2019

Abstract: A recent work proposed a simple theory based on the framework of Zebiak–Cane (ZC) ocean model, and successfully characterized the equatorial Atlantic upwelling annual cycle as a combination of the local wind-driven Ekman upwelling and nonlocal wind-driven wave upwelling. In the present work, utilizing the same simple framework, we examined the fidelity of the upwelling Pacific annual cycle using observations and simulations from the Coupled Model Intercomparison Project Phase 5 (CMIP5). We demonstrated that the theoretical upwelling annual cycles generally match the original upwelling annual cycles in the equatorial Pacific in both observations and CMIP5 simulations. Therefore, this simple formulation can be used to represent the upwelling annual cycle in the equatorial Pacific. Observationally, the equatorial Pacific upwelling annual cycle is dominated by the local wind-driven Ekman upwelling, while the remote wave upwelling is confined near the eastern boundary with little contribution. In CMIP5 simulations, though the theoretical-reconstructed upwelling well-reproduces the original upwelling, the contribution is totally different compared to the observation. The wave upwelling serves as the main contributor instead of the Ekman upwelling. We further demonstrated that such discrepancy is attributable to the bias of the central to eastern equatorial thermocline depth patterns. This amplified, westward-shift wave upwelling weakened the impacts of the Ekman upwelling, and contributes to the entire Pacific equatorial upwelling annual cycle substantially. This implies that a realistic simulation of the equatorial Pacific upwelling annual cycle in models is very sensitive to the careful simulation of the equatorial thermocline depth annual evolutions.

Keywords: upwelling; annual cycle; Ekman eﬀects; wave eﬀects

Key points: The linear framework of the Zebiak–Cane ocean model for upwelling annual cycle works well in • the Pacific through observed and Coupled Model Intercomparison Project Phase 5 (CMIP5) results. In observation, the equatorial Pacific upwelling annual cycle is dominated by local wind-driven • Ekman effects. In CMIP5 simulations, the amplified westward-shift wave upwelling overtakes Ekman effects and • plays a crucial role in the equatorial Pacific upwelling annual cycle.

1. Introduction Equatorial upwelling and its annual cycle play a vitally important role in cold tongue development. It pumps the subsurface water up and forms a tongue-like cold water at the surface, stretching from the coast of South America to the central Paciﬁc [1]. This band of cold water (so called cold tongue) restrains the atmospheric convection and severely impacts the entire tropical climatic system [2]. The

Sustainability 2019, 11, 5038; doi:10.3390/su11185038 www.mdpi.com/journal/sustainability Sustainability 2019, 11, 5038 2 of 10 equatorial upwelling generally peaks in early summer, and thereafter, the cold tongue sea surface temperature (SST) reaches a minimum immediately, revealing the important dynamic cooling effect in controlling the phasing of the equatorial cold tongue SST annual cycle [3,4]. The equatorial upwelling is also key to the marine ecosystem. The enhancing or weakening of the equatorial upwelling is tightly associated with nutrient concentration and primary productivities [5]. Great efforts have been made into investigating the variability of the equatorial upwelling. It is believed that the Ekman pumping driven by local winds is key to the variations of the equatorial upwelling [6–8]. A recent study pointed out that the turbulent momentum flux and the imbalance between circulation and pressure fields may be the main contributor of the equatorial upwelling [9]. However, the detailed fundamental dynamics of the equatorial upwelling annual cycle has not yet been completely refined. Recently, a linear framework for the equatorial upwelling annual cycle was developed by Wang et al. [10], based on the Zebiak–Cane (ZC) ocean model [[11], hereafter ZC87]. It is a linear combination of two processes: The local wind-driven Ekman upwelling and the non-local wind-driven wave upwelling. They also demonstrated in the observation that the Ekman upwelling plays a key role in the western equatorial Atlantic, whereas wave upwelling dominates the eastern equatorial Atlantic. Wang et al. [12] further explored the validity of the simple upwelling framework of the Coupled Model Intercomparison Phase 5 (CMIP5) simulations, and confirmed that this framework can be used to successfully characterize equatorial Atlantic upwelling annual cycles in CMIP5 simulations as well. However, this framework has not been verified in the Pacific. In the present work, we investigated the validity of the simple upwelling framework in the equatorial Pacific through observation and CMIP5 simulations. What is required to allow CMIP5 models to afford realistic simulations are also discussed in this work. The remainder of this paper is organized as follows. Section2 describes the dataset we used. In Section3, the theoretical-reconstructed equatorial Pacific upwelling annual cycles in observation and CMIP5 simulations are analyzed, and the dynamic reasons for the introduction of systematic bias into model simulations are explored. A conclusion (Section4) follows to summarize the major findings of this study.

2. Model and Data

The observational data including vertical velocity, SST, wind stresses, and 20 ◦C isotherm depth which represents the thermocline depth was obtained from the European Centre for Medium-Range Weather Forecasts (ECMWF) operational ocean analysis/reanalysis system [ORA-S3; [13], for the period 1959–2009]. The fine equatorial resolution of ORA-S3 is important in analyzing the ocean-atmosphere dynamics in the tropical Pacific [14,15]. It has also been frequently used in the equatorial area based on previous studies [10,12,16–18], indicating the reliability of this dataset. The annual cycles were defined in this work as the anomalies with respect to the climate annual mean of the entire data period. For the modeled data, we analyzed a series of historical experiment simulations from 25 CMIP5 models during the period around 1950–2005. We used the “historical” experiment and the first realization “r1i1p1” in each case. See Taylor et al. [19] for the relevant model descriptions and experiment designs. We mainly focused on the performance of CMIP5 multi-model ensemble (MME). The results of each individual and the average value of 25 models are also discussed. The data from all model outputs were interpolated onto a 1 1 grid. Detailed information about the analyzed models ◦ × ◦ is given in Table1. Sustainability 2019, 11, 5038 3 of 10

Table 1. List of the Coupled Model Intercomparison Phase 5 (CMIP5) models analyzed in the study.

Model Name Modeling Center

Sustainability 2019ACCESS1-0, 11, x FOR PEER REVIEW The Centre for Australian Weather and Climate Research, Australia3 of 10 ACCESS1-3 The Centre for Australian Weather and Climate Research, Australia bcc-csm1-1 Beijing Climate Center, China Meteorological Administration, China CCSM4 National Center for Atmospheric Research, US bcc-csm1-1-m Beijing Climate Center, China Meteorological Administration, China CESM1-BGC National Center for Atmospheric Research, US CanESM2 Canadian Centre for Climate Modelling and Analysis, Canada CESM1-CAM5CCSM4 National National Center Center for Atmospheric for Atmospheric Research, Research, US US CESM1-FASTCHEMCESM1-BGC National National Center Center for Atmospheric for Atmospheric Research, Research, US US CESM1-CESM1-CAM5WACCM National National Center Center for Atmospheric for Atmospheric Research, Research, US US CMCCCESM1-FASTCHEM-CESM Centro Euro-Mediterraneo National Center per for I Cambiamenti Atmospheric Climatici, Research, Italy US CMCCCESM1-WACCM-CM Centro Euro-Mediterraneo National Center per for I Cambiamenti Atmospheric Climatici, Research, Italy US CMCCCMCC-CESM-CMS Centro Centro Euro-Mediterraneo Euro-Mediterraneo per I Cambiamenti per I Cambiamenti Climatici, Climatici, Italy Italy FGOALSCMCC-CM-g2 LASG, Institute Centro of Atmospheric Euro-Mediterraneo Physics, Chinese per I Cambiamenti Academy of Climatici, Sciences, ItalyChina GFDL-ESM2MCMCC-CMS NOAA Centro Geophysical Euro-Mediterraneo Fluid Dynamics per I Cambiamenti Laboratory, Climatici, US Italy GISS-FGOALS-g2E2-R LASG, InstituteNASA Goddard of Atmospheric Institute Physics, for Space Chinese Studies, Academy US of Sciences, China GFDL-ESM2M NOAA Geophysical Fluid Dynamics Laboratory, US GISS-E2-R-CC NASA Goddard Institute for Space Studies, US GISS-E2-R NASA Goddard Institute for Space Studies, US HadGEM2-CC Met Office Hadley Centre, UK GISS-E2-R-CC NASA Goddard Institute for Space Studies, US HadGEM2-ES Met Office Hadley Centre, UK HadGEM2-CC Met Oﬃce Hadley Centre, UK IPSL-CM5AHadGEM2-ES-LR Institut Pierre Met- OSimonﬃce Hadley Laplace, Centre, France UK IPSL-CM5AIPSL-CM5A-LR-MR Institut Institut Pierre- Pierre-SimonSimon Laplace, Laplace, France France IPSL-CM5BIPSL-CM5A-MR-LR Institut Institut Pierre- Pierre-SimonSimon Laplace, Laplace, France France MRI-IPSL-CM5B-LRCGCM3 Meteorological Institut Research Pierre-Simon Institute, Laplace, Japan France NorESM1MRI-CGCM3-M Norwegia Meteorologicaln Climate ResearchCentre, Norway Institute, Japan NorESM1NorESM1-M-ME Norwegian Norwegian Climate Climate Centre, Centre, Norway Norway NorESM1-ME Norwegian Climate Centre, Norway 3. Results and Discussions 3. Results and Discussion 3.1. The Theoretical Formation of Upwelling Annual Cycles 3.1. The Theoretical Formation of Upwelling Annual Cycles The theory to characterize the upwelling annual cycle in the equatorial area proposed by Wang et al. [10] isThe based theory on the to ZC87 characterize model [11]. the T upwellinghe equatorial annual upwelling cycle annual in the cycle equatorial can be characterized area proposed by by a Wanglinear etcombination al. [10] is basedof a locally on the wind ZC87-driven model Ekman [11]. Theupw equatorialelling and upwellinga nonlocally annual wind cycle-driven can be upwelling:characterized by a linear combination of a locally wind-driven Ekman upwelling and a nonlocally wind-driven upwelling: 푊 = 푊푒 ∙ [1 − 푅(푥)] + 푊ℎ ∙ 푅(푥) (1) W = We [1 R(x)] + W R(x) (1) · − h·

푅(푥) R=(퐻x)1/=퐻(H푥1)/=H(50x)/=퐻(50푥)/=H1(x ) if= H1( ifx) H<(x50) < 50(2) (2)

The EkmanThe Ekman and wave and wave upwelling upwellingss are represented are represented as W ase andWe Wandh, respeWh, respectively.ctively. “x” here “x” heredenotes denotes longitude.longitude. The wave The wave upwelling upwelling here hereis calculated is calculated as -흏 ash/ -흏ht /(h tdenotes (h denotes the the depth depth of the of the thermocline) thermocline).. In In otherother words, words, the the waving waving upwelling upwelling represents represents the the negative negative thermocline thermocline tendency tendency.. H Handand H1H refe1 referr to to thethe mean mean climat climaticic thermocline thermocline depth depth and and the theconstant constant mixed mixed layer layer depth, depth, respectively. respectively. This Thislinear linear frameworkframework of equatorial of equatorial upwelling upwelling essential essentiallyly follows follows ZC87, ZC87, except except that this that formulation this formulation takes takes the the zonalzonal variations variations in the in climatic the climatic mean mean of the of thermocline the thermocline depth, depth, which which is R(x is),R into(x), account. into account. Thus, Thus, the the functionfunction R(x) alsoR(x) depends also depends on longitude on longitude.. See Wang See Wang et al. et[10] al. and [10] the and supplementary the Supplementary information Information for for moremore details. details. RegardingRegarding the casethe case in the in the equatorial equatorial Pacific, Pacific, it it was was generally generally believed believed that that upwellings upwelling weres were mostly mostlydriven driven by by local local winds winds through through the the Ekman Ekman pumping pumping processes processes [6– 8[6].– In8]. thisIn this work, work, we we first first verified verifiedif the if formulation the formulation still held still in he theld in equatorial the equatorial Pacific Pacific upwelling upwelling annual annual cycle. Next, cycle. we Next, investigated we investigatedthe dynamics the dynamics of theessential of the essential controls controls on the on equatorial the equatorial Pacific Pacific upwelling upwelling annual annual cycle. cycle. Both the Both observationalthe observational and and modeled modeled results results will will be discussed. be discussed.

3.2. Validation in Observation To obtain the theoretical equatorial vertical velocity in the Pacific using this formulation, we used: (1) The observed surface wind stress to calculate the Ekman upwelling We, (2) the observed thermocline depth to calculate Wh; and (3) the observed climate mean depth of thermocline H(x) to calculate R(x). We then made a comparison between the original W from ORA-S3 data and the theoretical-estimated W from the formulation (Figure 1). Sustainability 2019, 11, 5038 4 of 10

3.2. Validation in Observation To obtain the theoretical equatorial vertical velocity in the Paciﬁc using this formulation, we used: (1) The observed surface wind stress to calculate the Ekman upwelling We, (2) the observed thermocline depth to calculate Wh; and (3) the observed climate mean depth of thermocline H(x) to calculate R(x). We then made a comparison between the original W from ORA-S3 data and the theoretical-estimated

WSustainabilityfrom the 2019 formulation, 11, x FOR PEER (Figure REVIEW1). 4 of 10

Figure 1. The annual cycles of the ( a) observed, ( b) theoretical, (c) Ekman and (d) wave upwelling in the equatorial PaciﬁcPacific (averaged(averaged inin 33°◦ N–3N–3°◦ S) (units: m/day). m/day).

The original equatorial upwelling annual cycle was found to be dominated by a strong component The original equatorial upwelling annual cycle was found to be dominated by a strong expanding all over the year in the central to eastern Pacific. The negative value (downwelling) peaked component expanding all over the year in the central to eastern Pacific. The negative value around May–June, while the positive value (upwelling) took place at the beginning and the end of (downwelling) peaked around May–June, while the positive value (upwelling) took place at the the year. This seasonal pattern was also found in previous work [20]. These features were both beginning and the end of the year. This seasonal pattern was also found in previous work [20]. These captured by the theoretically estimated W (Figure1b), the pattern correlation between the original and features were both captured by the theoretically estimated W (Figure 1b), the pattern correlation theoreticalbetween the estimated original verticaland theoretical velocity reachedestimated 0.81, vertical indicating velocity that reached the simple 0.81, formulation indicating of that (1) wasthe alsosimple successful formulation in the of Pacific. (1) was also successful in the Pacific. With our simple formulation, we further examined the contributions to the 50 m vertical velocity from thethe EkmanEkman upwellingupwelling andand thethe wavewave upwelling.upwelling. The Ekman upwelling (Figure 11c)c) showedshowed aa consistent pattern as the main dominated signal in the original vertical velocity. It was dominated in the central to eastern PacificPacific with a clear prevailing component whichwhich had a phasing very similar to the observedobserved original original upwelling upwelling (Figure (Figure1a), except1a), except that the that highest the variabilityhighest variability in the Ekman in the upwelling Ekman startedupwelling 2–3 started months 2–3 earlier months than earlier that in than the original that in upwelling.the original Theupwelling. pattern correlationThe pattern between correlation the observedbetween the vertical observed velocity vertical in Figure velocity1a and in theFigure derived 1a and Ekman the derived upwelling Ekman reached upwelling 0.72, demonstrating reached 0.72, thatdemonstrating the Ekman that upwelling the Ekman determined upwelling to adetermined large extent to the a large upwelling extent annualthe upwelling cycle in annual the equatorial cycle in Pacificthe equatorial Ocean. InPacific contrast, Ocean. the In wave contrast, upwelling the wave played upwelling a minor roleplayed in the a minor equatorial role Pacificin the equatorial upwelling annualPacific cycleupwelling (Figure annual1d). The cycle pattern (Figure correlation 1d). The between patternthe correlation observed between vertical velocitythe observed and the vertical wave upwellingvelocity and was the about wave 0.11, upwelling implying was that about the wave 0.11, eimplyingffects contributed that the verywave little. effects contributed very little.Indicated by Wang et al. [10,12], this linear combination derived from the ZC87 model is naturally the bestIndicated way to by capture Wang contributions et al. [10,12], from this the linear Ekman combination and wave upwellings derived from in the the equatorial ZC87 model Atlantic. is Thenaturally results the in thebest present way studyto capture further contribution verify that thiss from formulation the Ekman worked and just wave as well upwellings in the equatorial in the Pacific.equatorial Moreover, Atlantic. it confirmedThe results the in dominantthe present role study of the further wind-driven verify th Ekmanat this processes formulation in the worked upwelling just annualas well in cycle the inequatorial the equatorial Pacific. Pacific. Moreover, it confirmed the dominant role of the wind-driven Ekman processes in the upwelling annual cycle in the equatorial Pacific. Since the Ekman upwelling is associated with the local wind stress forcing, we further used a shallow-water model [21] (hereafter denoted as SWM) to discern the dynamics of the wave upwelling near the eastern boundary. We found that the SWM successfully captures the annual variations of the equatorial thermocline (Figure 2a,b, shadings), and this was in high agreement with several studies [22,23]. The pattern correlation between the observational and the SWM-modeled thermocline depth was 0.78. The essential point here, however, is that the notable consistency of the simulated wave upwelling with the observed wave upwelling as estimated by the observed thermocline tendency (Figure 2a,b, contours), showed a pattern correlation as high as 0.79. Sustainability 2019, 11, 5038 5 of 10

Since the Ekman upwelling is associated with the local wind stress forcing, we further used a shallow-water model [21] (hereafter denoted as SWM) to discern the dynamics of the wave upwelling near the eastern boundary. We found that the SWM successfully captures the annual variations of the equatorial thermocline (Figure2a,b, shadings), and this was in high agreement with several studies [22,23]. The pattern correlation between the observational and the SWM-modeled thermocline depth was 0.78. The essential point here, however, is that the notable consistency of the simulated wave upwelling with the observed wave upwelling as estimated by the observed thermocline tendency

Sustainability(Figure2a,b, 2019 contours),, 11, x FOR PEER showed REVIEW a pattern correlation as high as 0.79. 5 of 10

FigureFigure 2. 2.TheThe equatorial equatorial Pacific Pacific annual annual cycles cycles (average (averagedd in in3° 3N–3°◦ N–3 S)◦ S) of ofthe the thermocline thermocline changes changes (shadings,(shadings, units: units: m) m) and and the the observed observed wave wave upwellin upwellingg estimated estimated by bythermocline thermocline tendencies tendencies (contours, (contours, units:units: m/day) m/day) in in(a) ( aobservation) observation and and (b ()b shallow-water) shallow-water model model (SWM) (SWM) simulations. simulations. 3.3. Validation in CMIP5 Simulations 3.3. Validation in CMIP5 Simulations Before contrasting the upwelling annual cycles in the equatorial Pacific between the observation Before contrasting the upwelling annual cycles in the equatorial Pacific between the observation and CMIP5 simulations through this linear framework, we first took a glance at how the CMIP5 and CMIP5 simulations through this linear framework, we first took a glance at how the CMIP5 upwelling annual cycle performed in longitude–time evolutions (Figure3). Compared to the observation upwelling annual cycle performed in longitude–time evolutions (Figure 3). Compared to the (Figure1a), not all CMIP5 models yielded realistic upwelling annual cycle patterns. Their performances observation (Figure 1a), not all CMIP5 models yielded realistic upwelling annual cycle patterns. Their are actually diverse. The weak semi-annual signal in the observational wave upwelling seemed to be performances are actually diverse. The weak semi-annual signal in the observational wave upwelling enlarged and shifted westward in CMIP5 simulations. Yet, a large part of them still reproduced an seemed to be enlarged and shifted westward in CMIP5 simulations. Yet, a large part of them still annual signal in the central Pacific. reproduced an annual signal in the central Pacific. We then applied this simple framework to the CMIP5 simulations of the Pacific equatorial upwelling annual cycle (Figure4) and compared them with the observation. Suggested from Figure1, in the observed upwelling annual cycle, the Ekman upwelling served as a dominant contributor, whereas the wave upwelling contributed very little. In contrast, the CMIP5 MME pictured a fairly different story. The pattern correlation between the original and theoretical-reconstructed upwelling (Figure4a,b) from the MME was 0.7, a bit smaller than observational results but still agreeing that the simple framework did a good job in characterizing the equatorial upwelling annual cycle pattern in model simulations. However, the wave upwelling (Figure4d) served as a main contributor instead of the Ekman upwelling (Figure4c) here. The pattern correlation between the original and the Ekman/wave upwelling was 0.35/0.75. Surprisingly, compared to the theoretically-reconstructed upwelling, the wave upwelling had an even higher consistency with the original upwelling. This somehow implies that in CMIP5 MME, the contribution of the local wind-driven Ekman upwelling was minor and overtaken by the remote wave upwelling. There are two more points worthy of note about the observation and the CMIP5 MME: First, the wave upwelling in observation was weak and confined to the eastern boundary (Figure1d), while the wave upwelling in the MME was much

Figure 3. The upwelling annual cycles of the equatorial Pacific (averaged over 3° N–3° S) originally simulated by 25 CMIP5 models. Sustainability 2019, 11, x FOR PEER REVIEW 5 of 10

Figure 2. The equatorial Pacific annual cycles (averaged in 3° N–3° S) of the thermocline changes (shadings, units: m) and the observed wave upwelling estimated by thermocline tendencies (contours, units: m/day) in (a) observation and (b) shallow-water model (SWM) simulations.

3.3. Validation in CMIP5 Simulations

SustainabilityBefore contrasting2019, 11, 5038 the upwelling annual cycles in the equatorial Pacific between the observation6 of 10 and CMIP5 simulations through this linear framework, we first took a glance at how the CMIP5 upwelling annual cycle performed in longitude–time evolutions (Figure 3). Compared to the stronger, shifting westward and occupying the central Paciﬁc (Figure4d). Second, the MME Ekman observation (Figure 1a), not all CMIP5 models yielded realistic upwelling annual cycle patterns. Their upwelling (Figure4d) actually showed a quite similar pattern and amplitude as the observational performances are actually diverse. The weak semi-annual signal in the observational wave upwelling Ekman upwelling (Figure1c), but it failed to take control in the original upwelling annual cycle seemed to be enlarged and shifted westward in CMIP5 simulations. Yet, a large part of them still (Figure4a). Evidently, the original output upwelling in CMIP5 MME was systematically biased. The reproduced an annual signal in the central Pacific. reason will be discussed in the next section.

Sustainability 2019, 11, x FOR PEER REVIEW 6 of 10

We then applied this simple framework to the CMIP5 simulations of the Pacific equatorial upwelling annual cycle (Figure 4) and compared them with the observation. Suggested from Figure 1, in the observed upwelling annual cycle, the Ekman upwelling served as a dominant contributor, whereas the wave upwelling contributed very little. In contrast, the CMIP5 MME pictured a fairly different story. The pattern correlation between the original and theoretical-reconstructed upwelling (Figure 4a,b) from the MME was 0.7, a bit smaller than observational results but still agreeing that the simple framework did a good job in characterizing the equatorial upwelling annual cycle pattern in model simulations. However, the wave upwelling (Figure 4d) served as a main contributor instead of the Ekman upwelling (Figure 4c) here. The pattern correlation between the original and the Ekman/wave upwelling was 0.35/0.75. Surprisingly, compared to the theoretically-reconstructed upwelling, the wave upwelling had an even higher consistency with the original upwelling. This somehow implies that in CMIP5 MME, the contribution of the local wind-driven Ekman upwelling was minor and overtaken by the remote wave upwelling. There are two more points worthy of note about the observation and the CMIP5 MME: First, the wave upwelling in observation was weak and confined to the eastern boundary (Figure 1d), while the wave upwelling in the MME was much stronger, shifting westward and occupying the central Pacific (Figure 4d). Second, the MME Ekman upwelling (Figure 4d) actually showed a quite similar pattern and amplitude as the observational Ekman upwelling (Figure 1c), but it failed to take control in the original upwelling annual cycle (FigureFigure 4a). 3. Evidently,The upwelling the original annual cycles output of theupwelling equatorial in PaciﬁcCMIP5 (averaged MME was over systematically 3 N–3 S) originally biased. The Figure 3. The upwelling annual cycles of the equatorial Pacific (averaged over 3°◦ N–3°◦ S) originally reasonsimulated will be by discussed 25 CMIP5 in models. the next section. simulated by 25 CMIP5 models.

FigureFigure 4. 4.Same Same asas FigureFigure1 ,1, but but for for the the CMIP5 CMIP5 multi-model multi-model ensemble ensemble (MME). (MME).

Figure 5 displays the pattern correlations between the originally simulated and the Ekman/wave/theoretically estimated equatorial upwellings, which are presented as blue, green and yellow colors, respectively, in light of 25 CMIP5 individual simulations. Concerning the theoretical upwellings (yellow bars), a great part of the models adequately reproduced the upwelling annual cycle in the equatorial Pacific; the mean correlation between the originally simulated and theoretically estimated upwellings applied by all CMIP5 models was nearly 0.6. The observation and the MME reached 0.81 and 0.7, respectively. These results imply that this linear combination framework can be used to characterize the equatorial upwelling annual cycle pattern. Sustainability 2019, 11, 5038 7 of 10

Figure5 displays the pattern correlations between the originally simulated and the Ekman/wave/theoretically estimated equatorial upwellings, which are presented as blue, green and yellow colors, respectively, in light of 25 CMIP5 individual simulations. Concerning the theoretical upwellings (yellow bars), a great part of the models adequately reproduced the upwelling annual cycle in the equatorial Paciﬁc; the mean correlation between the originally simulated and theoretically estimated upwellings applied by all CMIP5 models was nearly 0.6. The observation and the MME reached 0.81 and 0.7, respectively. These results imply that this linear combination framework can be Sustainabilityused to characterize 2019, 11, x FOR the PEER equatorial REVIEW upwelling annual cycle pattern. 7 of 10

FigureFigure 5. AA bar bar chart chart showing showing pattern pattern correlations correlations between between the the original original and and theoretical theoretical values values of of the equatorialequatorial upwelling upwelling annual cycles from 25 CMIP5 models and the observation. The The blue, blue, green, green, and and yellow bars represent the correlations between the original and the Ekman, wave, and theoretical upwellings,upwellings, respectively.

However,However, the contributions from the Ekman and wa waveve upwellings are are not that well well captured captured in in modelmodel simulations. TheThe patternpattern correlations correlations between between the the original original and and the the wave wave upwelling upwelling (green (green bars) bars)are mostly are mostly higher thanhigher that than between that between the original the and original the Ekman and upwellingthe Ekman (blue upwelling bars), displaying (blue bars), an displayingopposite relationship an opposite to relationship the observation. to the Inobservat nearlyion. half In of nearly the models, half of thethe correlationsmodels, the betweencorrelations the betweenoriginal andthe theoriginal wave and upwellings the wave even upwellings overtake theeven correlations overtake the between correlations the original between and theoreticallythe original andreconstructed theoretically values, reconstructed suggesting va thatlues, the suggesting Ekman upwellings that the almostEkman lose upwellings their controls almost on lose their their own controlsequatorial on upwelling their own annualequatori cycles.al upwelling annual cycles.

3.4. The Dynamical Bias in CMIP5 Simulations 3.4. The Dynamical Bias in CMIP5 Simulations To investigate the origin of the systematic bias in the equatorial upwelling annual cycles in CMIP5 To investigate the origin of the systematic bias in the equatorial upwelling annual cycles in simulations, we needed to investigate the dynamics of the wave upwelling as estimated by thermocline CMIP5 simulations, we needed to investigate the dynamics of the wave upwelling as estimated by tendency, which was originated from the thermocline. Figure6a,b contrasts the annual evolution of thermocline tendency, which was originated from the thermocline. Figure 6a,b contrasts the annual the equatorial thermocline depth in terms of observation and CMIP5 MME. Apparently, they are not evolution of the equatorial thermocline depth in terms of observation and CMIP5 MME. Apparently, comparable regardless of amplitude or phase. The observational thermocline pattern was weak in they are not comparable regardless of amplitude or phase. The observational thermocline pattern the central equatorial Pacific, taking effects merely near the eastern boundary (Figure6a). On the was weak in the central equatorial Pacific, taking effects merely near the eastern boundary (Figure other hand, the MME thermocline pattern did not take place near the eastern boundary. Instead, the 6a). On the other hand, the MME thermocline pattern did not take place near the eastern boundary. strongest area of MME thermocline pattern shifts westwards from the eastern boundary near 90 W to Instead, the strongest area of MME thermocline pattern shifts westwards from the eastern boundary◦ the central equatorial Pacific near 140 W, and its amplitude is nearly three times the strength of the near 90° W to the central equatorial Pacific◦ near 140° W, and its amplitude is nearly three times the observational thermocline depth pattern (Figure6b). Demonstrated from the di erences in thermocline strength of the observational thermocline depth pattern (Figure 6b). Demonstratedff from the differencesannual evolution in thermocline between the annual MME evolution and observation, between the the CMIP5 MME MME and tendsobservation, to have the an overestimationCMIP5 MME tendsof thermocline to have an depth overestimation strength in of the thermocline central equatorial depth strength Pacific (Figure in the 6centralc). The equatorial MME bias Pacific in the (Figure annual 6c).evolution The MME of thermocline bias in the annual depth evolution was also directlyof thermo transferredcline depth into wasthe also wave directly upwelling, transferred which into wasthe waveestimated upwelling, as thermocline which was depth estimated tendency as thermocline (Figure6d). depth tendency (Figure 6d). As discussed before in Figure 4, the westward-shifting, strengthened wave upwelling occupies the central equatorial Pacific (160° W–100° W), overlapping the impact area of the Ekman upwelling. Hence, the Ekman upwelling with an annual signal and the wave upwelling with a semi-annual signal contend with each other and biased the total upwelling annual cycle. We may further discover that although the MME Ekman upwelling exhibits relatively weak variance, it perfectly mimics the annual evolution pattern of the observed Ekman upwelling (Figures 1c and 4c). The correlation between them reached 0.76. On the other hand, the pattern correlation between the observed and MME original vertical velocity was low (0.43), which was also largely associated with the diverse upwelling annual cycle patterns in CMIP5 individual simulations (Figure 3). It explains that biased wave upwelling contributes substantially to the entire Pacific equatorial upwelling annual cycle. The strengthened westward-shift wave upwelling, originating from the biased equatorial thermocline annual evolutions, overtakes the impact of the Ekman upwelling and consequently serves as a main contributor to the entire Pacific equatorial upwelling annual cycle in CMIP5 simulations. Sustainability 2019, 11, 5038 8 of 10 Sustainability 2019, 11, x FOR PEER REVIEW 8 of 10

FigureFigure 6. 6.The The equatorial equatorial Pacific Pacific annual annual anomalies anomalies with with respect respect to to the the climate climate annual annual mean mean (averaged (averaged in 3 N–3 S) of the thermocline depth (units: m) in (a) observation and (b) CMIP5 MME and the in ◦3° N–3°◦ S) of the thermocline depth (units: m) in (a) observation and (b) CMIP5 MME and the departures of the CMIP5 MME from the observation in annual anomalies of (c) thermocline depth and departures of the CMIP5 MME from the observation in annual anomalies of (c) thermocline depth (d) wave upwelling (units: m/day), respectively. and (d) wave upwelling (units: m/day), respectively. As discussed before in Figure4, the westward-shifting, strengthened wave upwelling occupies 4. Conclusions the central equatorial Pacific (160◦ W–100◦ W), overlapping the impact area of the Ekman upwelling. Hence,The the EkmanPacific upwellingequatorial withupwelling an annual is a signal substantial and the dynamic wave upwelling characteristic with a semi-annualof tropical climatic signal contendsystems with and eachmarine other ecosystems. and biased Its theannual total cycle, upwelling especially annual in the cycle. central We to may eastern further equatorial discover region, that althoughplays a crucial the MME role Ekman in the upwellingdevelopment exhibits of Pacific relatively cold weaktongue. variance, Understanding it perfectly the mimics detailed the dynamics annual evolutionof the Pacific pattern equatorial of the observed upwelling Ekman annual upwelling cycle is therefore (Figures1 can and important4c). The correlationtarget in scientific between research. them reachedIn this work, 0.76. Onwe further the other confirmed hand, the the pattern validity correlation of the linear between framework the observed developed and previously MME original for the verticalequatorial velocity upwelling was low [10] (0.43), in the which case of was the also Pacific. largely associated with the diverse upwelling annual cycle patternsThe theoretical in CMIP5 upwelling individual annual simulations cycles estimate (Figured 3through). It explains this framework that biased generally wave upwelling agree with contributesthe original substantially upwelling annual to the entirecyclesPacific in both equatorial observational upwelling and modeled annual cycle.simulations. The strengthened This simple westward-shiftformulation was wave proved upwelling, to successfully originating work from thenot biasedonly in equatorial the Atlantic, thermocline but also annual in the evolutions, Pacific. In overtakesobservation, the impactthe Ekman of the upwelling Ekman upwelling serves as and a consequentlydominant contributor serves as ato main the contributorequatorial toPacific the entireupwelling Pacific annual equatorial cycle, upwelling while the annualwave process cycle in cont CMIP5ributes simulations. very little. This remarkably matches what previous studies have suggested. The CMIP5 simulations, however, pictured a fairly different story: 4.The Conclusions contribution of wave upwelling overtakes the role of the Ekman upwelling, dominating the equatorialThe Pacific Pacific equatorial upwelling upwelling annual is cycle. a substantial dynamic characteristic of tropical climatic systems and marineUnlike ecosystems. the observational Its annual wave cycle, upwelling especially confined in the centralonly near to easternthe eastern equatorial boundary, region, the plays CMIP5 a crucialsimulated role inwave the developmentupwelling was of Pacificstrengthened cold tongue. and sh Understandingifted westward the to detailedthe central dynamics Pacific. of From the Pacificintermodal equatorial statistics, upwelling Li et al. annual [24] argued cycle is that therefore the aforementioned an important target too strong in scientific upwelling research. could In be thisattributed work, we to a further too shallow confirmed thermocline the validity along of the the eq linearuatorial framework Pacific found developed in CMIP5 previously models [24]. for the We equatorialfurther demonstrated upwelling [10 that] in thesuch case a significant of the Pacific. bias may be attributable to the bias in the central to easternThe theoreticalequatorial upwellingthermocline annual depth cycles patterns. estimated This throughbiased remote this framework wave upwelling generally weakens agree with the theimpacts original of the upwelling local wind-driven annual cycles Ekman in both upwelling, observational and contributed and modeled substantially simulations. to the Thisentire simple Pacific equatorial upwelling annual cycle. Thus, a realistic simulation of the equatorial Pacific upwelling Sustainability 2019, 11, 5038 9 of 10 formulation was proved to successfully work not only in the Atlantic, but also in the Pacific. In observation, the Ekman upwelling serves as a dominant contributor to the equatorial Pacific upwelling annual cycle, while the wave process contributes very little. This remarkably matches what previous studies have suggested. The CMIP5 simulations, however, pictured a fairly different story: The contribution of wave upwelling overtakes the role of the Ekman upwelling, dominating the equatorial Pacific upwelling annual cycle. Unlike the observational wave upwelling confined only near the eastern boundary, the CMIP5 simulated wave upwelling was strengthened and shifted westward to the central Pacific. From intermodal statistics, Li et al. [24] argued that the aforementioned too strong upwelling could be attributed to a too shallow thermocline along the equatorial Pacific found in CMIP5 models [24]. We further demonstrated that such a significant bias may be attributable to the bias in the central to eastern equatorial thermocline depth patterns. This biased remote wave upwelling weakens the impacts of the local wind-driven Ekman upwelling, and contributed substantially to the entire Pacific equatorial upwelling annual cycle. Thus, a realistic simulation of the equatorial Pacific upwelling annual cycle is very sensitive to a careful simulation of the equatorial thermocline depth annual patterns.

Supplementary Materials: The following are available online at http://www.mdpi.com/2071-1050/11/18/5038/s1, Figure S1. The weighting function R(x) used to estimate the equatorial Pacific upwelling annual cycle. Author Contributions: L.-C.W. designed the study, analyzed the data and wrote the paper. J.-Y.Y. reviewed the paper and provided suggestions. Funding: This study was supported by the Ministry of Science and Technology (MOST), Taiwan, under grants 108-2111-M-008-001-MY3. J.-Y. Yu was supported by MOST105-2111-M-008-025-MY3. Acknowledgments: The authors would like to thank the editor and the anonymous reviewers for their careful review of the manuscript and detailed suggestions to improve the manuscript. The authors are grateful to F.-F. Jin and C.-R. Wu for useful communications on this topic. The data used in this study is listed in Section2 at http://apdrc.soest.hawaii.edu/data/data.php. Conflicts of Interest: The authors declare no conflict of interest.

References

1. Xie, S.-P. Ocean–atmosphere interaction in the making of the Walker circulation and equatorial Pacific cold tongue. J. Clim. 1998, 11, 189–201. [CrossRef] 2. Li, G.; Xie, S.-P. Tropical Biases in CMIP5 Multimodel Ensemble: The Excessive Equatorial Pacific Cold Tongue and Double ITCZ Problems. J. Clim. 2014, 27, 1765–1780. [CrossRef] 3. Clement, A.C.; Seager, R.; Cane, M.; Zebiak, S.E. An ocean dynamical thermostat. Oceanogr. Lit. Rev. 1997, 6, 556. [CrossRef] 4. An, S.-I.; Kim, J.-W.; Im, S.-H.; Kim, B.-M.; Park, J.H. Recent and future sea surface temperature trends in tropical Pacific warm pool and cold tongue regions. Clim. Dyn. 2012, 39, 1373–1383. [CrossRef] 5. Dandonneau, Y.; Deschamps, P.-Y.; Nicolas, J.-M.; Loisel, H.; Blanchot, J.; Montel, Y.; Thieuleux, F.; Bécu, G. Seasonal and interannual variability of ocean color and composition of phytoplankton communities in the North Atlantic, equatorial Pacific and South Pacific. Deep. Sea Res. Part II Top. Stud. Oceanogr. 2004, 51, 303–318. [CrossRef] 6. Weingartner, T.J.; Weisberg, R.H. On the Annual Cycle of Equatorial Upwelling in the Central Atlantic Ocean. J. Phys. Oceanogr. 1991, 21, 68–82. [CrossRef] 7. Richter, I.; Xie, S.-P. On the origin of equatorial Atlantic biases in coupled general circulation models. Clim. Dyn. 2008, 31, 587–598. [CrossRef] 8. Hagos, S.M.; Cook, K.H. Development of a Coupled Regional Model and Its Application to the Study of Interactions between the West African Monsoon and the Eastern Tropical Atlantic Ocean. J. Clim. 2009, 22, 2591–2604. [CrossRef] 9. Caniaux, G.; Giordani, H.; Redelsperger, J.-L.; Guichard, F.; Key, E.; Wade, M. Coupling between the Atlantic cold tongue and the West African monsoon in boreal spring and summer. J. Geophys. Res. Space Phys. 2011, 116.[CrossRef] Sustainability 2019, 11, 5038 10 of 10

10. Wang, L.-C.; Jin, F.-F.; Wu, C.-R.; Hsu, H.-H. Dynamics of Upwelling Annual Cycle in the Equatorial Atlantic Ocean. Geophys. Res. Lett. 2017, 44, 3737–3743. [CrossRef] 11. Zebiak, S.E.; Cane, M.A. A model El Niño–Southern Oscillation. Mon. Weather Rev. 1987, 115, 2262–2278. [CrossRef] 12. Wang, L.-C.; Jin, F.-F.; Wu, C.-R. Dynamics of simulated Atlantic upwelling annual cycle in CMIP5 models. J. Geophys. Res. Ocean. 2017, 122, 3737–3743. [CrossRef] 13. Balmaseda, M.A.; Vidard, A.; Anderson, D.L.T. The ECMWF Ocean Analysis System: ORA-S3. Mon. Weather Rev. 2008, 136, 3018–3034. [CrossRef] 14. Zhai, F.; Hu, D. Revisit the interannual variability of the North Equatorial Current transport with ECMWF ORA-S3. J. Geophys. Res. Ocean. 2013, 118, 1349–1366. [CrossRef] 15. Jin, F.-F.; Boucharel, J.; Lin, I.-I. Eastern Pacific tropical cyclones intensified by El Niño delivery of subsurface ocean heat. Nature 2014, 516, 82–85. [CrossRef][PubMed] 16. Munoz, E.; Kirtman, B.; Weijer, W. Varied representation of the Atlantic Meridional Overturning across multidecadal ocean reanalyses. Deep. Sea Res. Part II Top. Stud. Oceanogr. 2011, 58, 1848–1857. [CrossRef] 17. Nnamchi, H.C.; Li, J.; Kucharski, F.; Kang, I.-S.; Keenlyside, N.S.; Chang, P.; Farneti, R. Thermodynamic controls of the Atlantic Niño. Nat. Commun. 2015, 6, 8895. [CrossRef] 18. Nnamchi, H.C.; Li, J.; Kucharski, F.; Kang, I.-S.; Keenlyside, N.S.; Chang, P.; Farneti, R. An Equatorial–Extratropical Dipole Structure of the Atlantic Niño. J. Clim. 2016, 29, 7295–7311. [CrossRef] 19. Taylor, K.E.; Stouffer, R.J.; Meehl, G.A. An Overview of CMIP5 and the Experiment Design. Bull. Am. Meteorol. Soc. 2012, 93, 485–498. [CrossRef] 20. Wyrtki, K. An Estimate of Equatorial Upwelling in the Pacific. J. Phys. Oceanogr. 1981, 11, 1205–1214. [CrossRef] 21. McGregor, S.; Holbrook, N.J.; Power, S.B. Interdecadal Sea Surface Temperature Variability in the Equatorial Pacific Ocean. Part I: The Role of Off-Equatorial Wind Stresses and Oceanic Rossby Waves. J. Clim. 2007, 20, 2643–2658. [CrossRef] 22. Bunge, L.; Clarke, A.J. Seasonal Propagation of Sea Level along the Equator in the Atlantic. J. Phys. Oceanogr. 2009, 39, 1069–1074. [CrossRef] 23. Ding, H.; Keenlyside, N.S.; Latif, M. Seasonal cycle in the upper equatorial Atlantic Ocean. J. Geophys. Res. Ocean. 2009, 114.[CrossRef] 24. Li, G.; Du, Y.; Xu, H.; Ren, B. An Intermodel Approach to Identify the Source of Excessive Equatorial Pacific Cold Tongue in CMIP5 Models and Uncertainty in Observational Datasets. J. Clim. 2015, 28, 7630–7640. [CrossRef]

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).