Abstract
The ocean–atmosphere turbulent heat exchange plays a critical role in the energy and moisture budgets of the Tropical Atlantic Ocean (TAO) and in weather and climate forecasts. However, its estimation strongly depends on the choice of bulk parameterization, as direct in situ measurements are sparse. This study evaluates sensible (Hs) and latent (Hl) heat fluxes derived from three bulk parameterization schemes used operationally in models at the Brazilian Center for Weather Forecast and Climate Studies (CPTEC) of the National Institute for Space Research (INPE), Brazil: the Brazilian Atmospheric Model (BAM), the Modular Ocean Model version 6 (MOM6), and the Weather Research and Forecasting (WRF) model. Using daily in situ observations from seven Prediction and Research Moored Array in the Tropical Atlantic (PIRATA) buoys across the TAO during 1997–2023, we computed monthly mean fluxes and compared them against the Coupled Ocean–atmosphere Response Experiment (COARE) algorithm version 3.0b (COARE 3.0b) reference. COARE version 3.6 (COARE 3.6) and European Centre for Medium-Range Weather Forecast (ECMWF) Reanalysis 5th generation (ERA5) data were included as additional benchmarks. All offline schemes were forced with identical buoy data, isolating differences in internal physical assumptions. Hl is approximately one order of magnitude larger than Hs across all sites, and inter-scheme differences are substantially larger for Hl (±50 W∙m−2) than for Hs (±5 W∙m−2). All schemes reproduce the seasonal cycle linked to the Intertropical Convergence Zone (ITCZ) migration and trade-wind variability, with correlations generally exceeding 0.8 (p < 0.001) for most buoys. However, systematic magnitude biases remain. The Coordinated Ocean Research Experiments (CORE) bulk formulation implemented in MOM6 (MOM6-CORE) shows high temporal correlation (often r ≈ 1.0) but a persistent negative bias for both Hs and Hl (e.g., B1 Hl bias = −24.0 W∙m−2), indicating weaker turbulent exchange relative to COARE 3.0b. BAM overestimates Hs (by 1–3 W∙m−2) and underestimates Hl at most northern and southern sites, while the parametrization of the Yonsei University (YSU) implemented in the WRF model (WRF-YSU) amplifies Hs variability intermittently, particularly at the equator (B4). As expected, COARE 3.6 remains the closest to the reference (differences < 1 W∙m−2 for Hs and <7 W∙m−2 for Hl; r ≈ 0.99). ERA5 captures temporal variability well (r ≈ 0.7–0.9) but systematically overestimates Hl (positive bias up to +47.6 W∙m−2 at B7), implying stronger evaporative cooling. Buoy-specific regimes modulate skill. The choice of bulk formulation thus remains a first-order source of uncertainty in turbulent heat flux estimates over the TAO, with direct implications for mixed-layer heat budgets, SST evolution, and coupled ocean–atmosphere variability. MOM6-CORE provides the most consistent performance relative to the COARE reference and emerges as the most robust option for operational applications at CPTEC/INPE. The findings also provide guidance for improving the representation of ocean–atmosphere turbulent exchanges in MONAN (Model for Ocean-Land-Atmosphere Prediction), the new Brazilian Earth System Model under development for weather and climate prediction.
1. Introduction
Ocean–atmosphere turbulent exchanges of latent and sensible heat play a central role in the energy balance and dynamics of the global climate system, particularly in the tropics, where they drive convection and large-scale atmospheric circulation. In the TAO, these fluxes, alongside wind stress, critically regulate sea surface temperature (SST), ocean heat content, and hydrological cycle. Their influence spans scales from high-frequency turbulence to centennial variability, including the typical interannual climate phenomena of the Intertropical Convergence Zone (ITCZ) and the Atlantic Niño/Niña [1,2,3,4,5,6]. Small variations in these heat fluxes can significantly impact global energy and water budgets, altering ocean heat uptake and precipitation patterns [7,8]. Differences in bulk parameterization schemes may also influence simulated precipitation and top-of-atmosphere radiation biases [8], underscoring the importance of accurately estimating air–sea heat fluxes, particularly in the tropics.
Direct ocean–atmosphere heat flux measurements obtained from Eddy Covariance (EC) techniques are limited to a few moored-buoy sites and research cruises [9,10,11,12]. Consequently, long-term analyses rely on bulk parameterizations based on near-surface atmospheric and oceanic variables [13,14]. Over recent decades, these algorithms have been refined through improved atmospheric stability functions, treatments of strong winds and waves, and representation of cold-skin and warm-layer effects [15]. These developments led to the Coupled Ocean–atmosphere Response Experiment algorithm version 3.5 (COARE 3.5) algorithm, validated with ~16,000 h of EC observations [15,16]. Despite refinements (e.g., COARE 3.5 [16]), regional uncertainties persist because parameterizations perform differently across climatic regimes [17].
Although bulk aerodynamic schemes are based on Monin–Obukhov similarity theory (MOST), they differ in key empirical formulations and variable choices, leading to substantial differences in turbulent heat flux estimates under identical forcing conditions. While MOST provides a robust theoretical foundation, its implementation through heat and moisture transfer coefficients introduces important first-order uncertainties, especially in the unique convective regime of the TAO. Intercomparison and sensitivity studies highlight how near-surface temperature representation (skin versus bulk SST), wind speed adjustments, drag and transfer coefficients, convective gustiness, and surface currents influence turbulent fluxes and wind stress, producing climatologically meaningful divergence in offline experiments [18,19,20]. In coupled configurations, part of these differences is typically damped by the negative SST–flux feedback [19,20]. Consistently, evaluations of widely used global flux datasets against buoy observations still report substantial Hl uncertainty, with Root Mean Squared Error (RMSE) differences commonly spanning ~21–51 W∙m−2 across products [17], and systematic differences on an order of ~10–20% have been documented in algorithm evaluations, particularly for Hl [18]. Such discrepancies can result in divergent mixed-layer heat budgets and surface stress forcing in ocean-only and coupled simulations [14,17,21].
These uncertainties can lead to contrasting conclusions about climate variability and long-term trends. Different flux products may exhibit inconsistent or even opposite-sign trends over the same region and period [14,16,21,22], and in some cases may suggest no significant trend [13]. Large inter-product discrepancies have been documented despite advances in satellite observing systems and reanalyses, and these inconsistencies hinder closure of the ocean–atmosphere energy budget, particularly in the tropics [14,23]. Similar spreads have been reported in the Southwestern Atlantic and the Southern Ocean [22,24], emphasizing the need for regional evaluations using in situ observations and high-quality reanalyses. For example, Air-Sea Fluxes for the Global Tropical Oceans—TropFlux (a hybrid product) showed superior agreement with moored-buoy observations in the southeast Indian Ocean [25,26].
In the TAO, the Prediction and Research Moored Array in the Tropical Atlantic (PIRATA) has provided long-term, high-quality observational data since 1997 [27]. The PIRATA array currently comprises 18 moored buoys, measuring key atmospheric and oceanic variables (wind speed, air temperature, humidity, SST, atmospheric pressure, and incoming shortwave radiation) required for turbulent heat flux estimations [14]. These observations have supported the development and evaluation of blended products such as the Objectively Analyzed Air-Sea Fluxes (OAFlux) [28] and TropFlux, combining satellite, reanalysis, and buoy data to improve flux estimates in the tropics [29]. To date, however, no dedicated assessment has examined the implications of different parameterization schemes in these models for turbulent heat flux estimates in the TAO [28,30].
In Brazil, accurate estimates of ocean–atmosphere heat fluxes directly impact on the quality of the weather and climate forecasts produced by the Center for Weather Forecast and Climate Studies (CPTEC) of the National Institute for Space Research (INPE), Cachoeira Paulista, Brazil. Three models are operationally or pre-operationally run at CPTEC: (i) the Brazilian Atmospheric Model (BAM), (ii) the oceanic component of the Brazilian Earth System Model (BESM)—the Modular Ocean Model version 6 (MOM6) and (iii) the Weather Research and Forecasting (WRF) model. At the present, and unfortunately, no attempt has yet been made to assess the implications related to the use of different parameterization schemes in our models for the global ocean and, specifically, for the TAO—a region key to the weather and climate [31].
In this context, we analyzed ocean–atmosphere sensible and latent heat flux estimates produced by operational CPTEC/INPE models and compared them with the PIRATA OceanSITES Coupled Ocean–atmosphere Response Experiment algorithm version 3.0b (COARE 3.0b) heat flux products, the updated COARE algorithm in version 3.6 (COARE 3.6), and the European Medium-Range Weather Forecast (ECMWF) Reanalysis 5th Generation (ERA5) data between 1997–2023. The paper is organized as follows: Section 2 describes the study area, datasets, algorithms, and methods; Section 3 presents the results and discussion; and Section 4 summarizes the main conclusions.
2. Materials and Methods
This study evaluates the performance of turbulent ocean–atmosphere heat flux parameterizations used in operational CPTEC/INPE models in the TAO at seven PIRATA buoy sites during the 1997–2023 period. Sensible (Hs) and latent (Hl) heat flux estimates produced by these parameterizations are compared against the PIRATA OceanSITES COARE 3.0b flux products, the updated COARE 3.6 algorithm, and ERA5 reanalysis data. To ensure a consistent intercomparison, all offline bulk schemes were forced with the same in situ buoy observations, allowing the analysis to isolate differences associated with the internal physical assumptions and empirical formulations of each parameterization. The following sections describe the study area, datasets, bulk flux schemes, and statistical methods used in this work.
2.1. Study Area
The study domain comprises the TAO, between 20° N and 20° S, monitored by the PIRATA program [27]. The PIRATA buoys cover a wide range of weather, climate and oceanographic regimes across the TAO (Figure 1), from the warm tropical waters associated with the Atlantic Warm Pool (AWP) [3] to cooler zones influenced by coastal upwelling and the Atlantic Cold Tongue (ACT) [32]. Figure 1 summarizes the main oceanographic and atmospheric features that modulate, or are modulated by ocean–atmosphere exchanges in the TAO, including the mean position of the ITCZ, the subtropical high-pressure cells, and large-scale surface current systems. The main surface currents and subsurface undercurrents are also indicated: the North Equatorial Current (NEC), North Equatorial Countercurrent (NECC), Equatorial Undercurrent (EUC), the northern, central, and southern branches of the South Equatorial Current (nSEC, cSEC, sSEC), the South Equatorial Undercurrent (SEUC), the Brazil Current (BC), and the North Brazil Current (NBC). Figure 1 also indicates regions associated with the AWP and the ACT, which are linked to the Atlantic Meridional Mode and Atlantic Zonal Mode variability, as well as the coastal upwelling zones and the West African Monsoon (WAM) [32,33,34,35].
Figure 1.
Geographic distribution of the PIRATA buoy array in the TAO (20° N–20° S). The seven buoys used in this study (B1–B7) are highlighted in yellow. The main surface currents and subsurface undercurrents are indicated by the black and gray arrows, respectively. The remaining atmospheric and oceanographic features are identified in the figure legend. Adapted from Foltz et al. [32], Speich et al. [34] and Schott et al. [33].
2.2. Datasets
The PIRATA array currently consists of 18 operational meteoceanographic buoys (Figure 1), seven of which (B1–B7; Table 1) belong to the PIRATA OceanSITES array [27] and were selected for this study (Table 1). These seven buoys were selected because they are equipped with the complete set of instruments required to estimate turbulent Hs and Hl heat fluxes, which are calculated by the Pacific Marine Environmental Laboratory (PMEL), of the National Oceanic and Atmospheric Administration (NOAA), Seattle—United States of America, using the COARE 3.0b algorithm.
Table 1.
PIRATA buoys used in this study, including their nominal geographic coordinates (latitude, longitude) and deployment dates (month/year).
The Hs and Hl products distributed by PMEL/NOAA are derived from in situ buoy observations, including SST (at ~1 m depth), near-surface air temperature and specific humidity (at about 3–4 m height), wind speed (at about 4 m height), precipitation, and downward shortwave and longwave radiation. Figure 2 summarizes the temporal availability of all observations used in this study.
Figure 2.
Temporal availability of PIRATA buoy observations used in this study for (a–g) B1 to B7 between 1997–2023. Colored horizontal segments indicate available monthly data for each variable: air temperature (AirT); sea surface temperature (SST); relative humidity (RH); zonal component of the wind (UWND); meridional component of the wind (VWND); wind speed (WSPD); sea surface salinity (SSS); sea level pressure (SLP); precipitation; shortwave radiation (SW Rad); longwave radiation (LW Rad); latent heat flux (Hl); sensible heat flux (Hs).
All PIRATA observations and heat flux products are quality controlled and publicly available by NOAA/PMEL “https://www.pmel.noaa.gov/tao/drupal/disdel/” (accessed on 3 June 2026). Due to logistical, technical, and maintenance limitations throughout the PIRATA program, the observational periods differ among buoys.
In addition to the in situ buoy data, we used ERA5 data [36] as an independent gridded flux product “https://cds.climate.copernicus.eu/datasets/reanalysis-era5-single-levels?tab=overview” (accessed on 3 June 2026). Specifically, we used the ‘monthly averaged reanalysis’ product of sensible and latent heat fluxes from the corresponding 0.25° × 0.25° grid box centered at the location of each of the seven PIRATA buoys. Positive (negative) flux values indicate heat transfer from the ocean to the atmosphere (from the atmosphere to the ocean).
In contrast to the offline bulk schemes forced by buoy observations, ERA5 fluxes are generated within the coupled Integrated Forecasting System (IFS), which includes two-way interactions among the atmosphere, wave, and ocean models. In this framework, the ECMWF Ocean Wave Model (ECWAM) provides sea-surface roughness and momentum loss, while the Nucleus for European Modelling of the Ocean (NEMO) provides the thermal structure. Within the ERA5 system, the aerodynamic roughness length is not prescribed from a fixed wind-profile assumption; instead, it is computed consistently with sea-state information from the coupled wave model. Stability corrections are then applied through established similarity functions for unstable and stable stratification regimes (e.g., Beljaars [37]). Consequently, ERA5 uncertainties reflect both errors in the assimilated atmospheric and oceanic state variables (e.g., SST and near-surface winds) and uncertainties associated with the internal coupled surface-exchange physics [37,38,39]. For instance, in the tropical belt (20° N–20° S), the SST analysis is relaxed towards products like the Operational Sea Surface Temperature and Ice Analysis (OSTIA) via ECMWF Ocean Data Assimilation system (NEMOVAR), which may introduce specific delays in capturing rapid regional changes.
ERA5 is therefore used here as an independent benchmark for evaluating the bulk parameterizations implemented at CPTEC/INPE. The dataset provides globally consistent monthly mean flux fields and is widely used in air–sea interaction studies [24,40,41,42,43].
2.3. Bulk Parameterizations
We evaluated five approaches for estimating ocean–atmosphere turbulent heat fluxes, including bulk flux algorithms applied offline to PIRATA buoy observations and surface-flux parameterizations used in operational weather and climate CPTEC/INPE models. The reference dataset corresponds to the (1) COARE 3.0b flux product provided operationally by NOAA/PMEL through the PIRATA OceanSITES framework [13,44]. According to PMEL documentation, these monthly values are calculated from daily flux estimates [45].
Using the same PIRATA meteorological and oceanic observations (Section 2.2), we computed four independent flux estimates offline: (2) COARE 3.6, the updated version of the COARE bulk algorithm; (3) BAM, the surface-flux parameterization used in the Brazilian Global Atmospheric Model [46]; (4) MOM6-CORE, based on the Coordinated Ocean Research Experiments (CORE) bulk formulation implemented in MOM6, described by Large and Yeager [47] and implemented in BESM [48]; and (5) WRF-YSU, the surface-layer parameterization coupled to the Yonsei University (YSU) planetary boundary layer scheme used in WRF [49,50]. The latter three parameterizations are currently used in operational or pre-operational models at CPTEC/INPE. Table 2 summarizes the main characteristics of the different parameterization schemes evaluated here.
Table 2.
Main characteristics and bibliographic references of the bulk parameterizations used to estimate the ocean–atmosphere sensible (Hs) and latent (Hl) heat fluxes evaluated in this study.
In general, Hs and Hl can be expressed by the standard bulk aerodynamic formulas given in Equations (1) and (2), respectively:
where is the air density, Cp is the specific heat of air at constant pressure, Le is the latent heat of vaporization, and U is the wind speed at 10 m height. Ch and Ce represent the turbulent transfer coefficients for heat and moisture, respectively. The temperature gradient is represented by the difference between SST and air temperature (Tair), while the humidity gradient is represented by the difference between saturation specific humidity at the sea surface (qs) and near-surface air specific humidity (qa).
These equations are based on MOST, in which turbulent fluxes are proportional to near-surface gradients modulated by atmospheric stability. Although all evaluated schemes share this theoretical framework, they differ in the empirical closures adopted for transfer coefficients, surface roughness lengths, and wind-speed dependence, leading to differences in the estimated fluxes. As noted by Cronin et al. [45], the effective forcing of turbulent exchanges depends on the wind speed relative to the ocean surface current. Because surface currents were not explicitly included in the offline calculations, part of the differences among schemes may also reflect the neglect of ocean surface advection and sub-monthly current variability, although the use of daily forcing partially captures the synoptic modulation of these effects.
The COARE algorithms (versions 3.0b and 3.6) were specifically developed for tropical ocean conditions characterized by light winds and strong convection. Based on MOST, they incorporate empirical formulations for transfer coefficients, surface roughness lengths, atmospheric stability, and a “gustiness” parameter that accounts for enhanced turbulent exchange under weak-wind conditions [20,51]. COARE 3.0b is a well-established and widely validated bulk flux algorithm used operationally by PMEL/NOAA to process turbulent heat fluxes from PIRATA moored buoy observations [20,44]. It is commonly adopted as a community reference for flux validation and intercomparison studies in the tropics [22,26], and previous evaluations against direct eddy-covariance measurements have shown that COARE performs among the most robust bulk aerodynamic formulations available [23].
In this study, the COARE 3.0b heat flux product provided by PMEL/NOAA was adopted as the reference dataset. The updated COARE 3.6 algorithm was applied offline to estimate Hs and Hl using the daily PIRATA observations. The source codes for both implementations are publicly available at the NOAA Physical Sciences Laboratory (NOAA-PSL), Boulder—United States of America, COARE Algorithm Repository “https://github.com/NOAA-PSL/COARE-algorithm” (accessed on 3 June 2026). Although COARE 3.6 includes refinements related to wave–flux interactions and sea-state-dependent roughness formulations, it was implemented here in a wind-driven configuration to ensure a controlled comparison with the other parameterizations. Heat fluxes were first estimated at the daily scale and subsequently averaged to monthly means in order to facilitate long-term intercomparison analyses while preserving the dominant variability at periods longer than one month. Figure 3 illustrates an example of the original daily and monthly averaged time series for the B4 buoy.
Figure 3.
Time series of daily (light gray) Hs (a) and Hl (b) from the PIRATA B4 buoy (0°, 23° W), provided by PMEL/NOAA and estimated using the COARE 3.0b algorithm. Black lines represent the monthly averaged values. Gaps in the gray and black lines represent the absence of buoy data at the displayed periods.
Here, COARE 3.0b serves as a community-standard reference rather than an absolute benchmark, since algorithm-based flux estimates still contain intrinsic physical uncertainties. Previous studies report uncertainties of approximately 5–10% for turbulent heat fluxes in tropical regimes [20,23], with estimated daily mean errors near 5 W∙m−2 for Hs and 12 W∙m−2 for Hl at reference mooring sites [40]. Consequently, the discrepancies reported here represent differences in algorithmic consistency among the CPTEC/INPE parameterizations relative to an established community reference, acknowledging the limited availability of direct eddy-covariance measurements in the TAO.
Conversely, the MOM6-CORE formulation employs a simplified framework in which neutral transfer coefficients are prescribed as empirical functions of the 10 m neutral wind speed. The scheme was designed for large-scale ocean climatologies, where the relative wind effect is often parameterized through adjustments in the drag coefficient, according to Large and Yeager [47]. In the offline implementation used here, ocean surface currents are not explicitly included, forcing the model to operate at the upper limit which may influence momentum transfer estimates. This approach also assumes a bulk SST interface and may include constraints on the neutral drag coefficient growth under strong wind conditions to regulate surface stress in global ocean simulations [19,48]. Further formulation differences exist in the parameterization of scalar roughness.
The BAM scheme follows the methodology of Zeng et al. [52], which differentiates the roughness lengths for heat and moisture from that of momentum, thereby modulating the partitioning of Hs and Hl under different stability conditions [47]. Zeng et al. [52] emphasize that under weak-wind conditions, neglecting ocean currents and using absolute wind speed may lead to an overestimation of the stability parameter, directly affecting the scalar exchange coefficients.
Similarly, the WRF-YSU surface-layer scheme combines MOST with the wind-dependent Charnock relation for momentum roughness. Its exchange coefficients are linked to the non-local mixing formulation of the YSU planetary boundary layer (PBL) scheme and depend on the prescribed relationship between scalar and momentum roughness lengths [50,51,54]. Since WRF-YSU is commonly applied in atmospheric-only simulations, the assumption of a stationary lower boundary (i.e., ocean surface currents are neglected) is a standard operational proxy, which may explain systematic biases in regions with strong advection like the PIRATA sites.
Detailed discussions of the physical foundations and performance of these bulk flux algorithms are available in previous studies (e.g., [14,15,18,20,55]), including comparative assessments by Brunke et al. [18,55], analyses of wave effects by Edson et al. [15], evaluations of flux products by Yu [14], and regional analyses by Bonino et al. [20] and Santini et al. [10].
Here, using identical daily PIRATA buoy observations to force each scheme offline allows the analysis to isolate the effects of the internal parameterization formulations, including gustiness, transfer coefficients, and roughness length representations. Consequently, the differences in flux magnitude and variability primarily reflect the intrinsic physical assumptions of each scheme rather than inconsistencies in the forcing data.
2.4. Time Series and Statistical Analysis
For each of the seven buoy sites, the five bulk parameterization schemes were forced using daily PIRATA observations provided by PMEL/NOAA. Heat fluxes were first computed at the daily scale to preserve the non-linear relationship between meteorological forcing and surface exchanges, and subsequently averaged to produce monthly means. This procedure minimizes biases associated with calculating fluxes directly from monthly averaged atmospheric variables and is consistent with previous studies that recommend daily data for evaluating systematic differences among bulk flux products and reanalyses [17].
Monthly COARE 3.0b heat fluxes provided by PMEL were adopted as the reference dataset. The performance of each parameterization was evaluated using Taylor diagrams, which summarize three complementary statistics: (i) the standard deviation (std), representing the amplitude of variability; (ii) the Pearson correlation coefficient (r) relative to the COARE 3.0b reference; and (iii) the centered root-mean-square error (RMSE), which quantifies deviations after removal of the mean bias [56]. This approach provides an integrated assessment of how well each bulk parameterization estimated here reproduces the temporal variability of turbulent heat fluxes observed by the PIRATA OceanSITES buoys.
3. Results and Discussion
Here, we evaluate Hs and Hl time series produced by the bulk parameterizations BAM, MOM6-CORE, and WRF-YSU, as well as ERA5 and COARE 3.6, using COARE 3.0b as the reference dataset over 1997–2023. This intercomparison provides an integrated assessment of how consistently the parameterizations used in CPTEC/INPE operational systems represent turbulent ocean–atmosphere exchanges in the Tropical Atlantic when forced by the same PIRATA observations. In the TAO, the seasonal evolution of turbulent heat fluxes is tightly linked to the ITCZ migration and trade-wind variability, which also drive the seasonal equatorial cold tongue and associated SST gradients [3]. Recently, Steffen et al. [57] also described the importance of the ITCZ to the short-period changes in the upper ocean at the western part of TAO. From a mixed-layer heat-budget perspective, these fluxes exert first-order control on SST and upper-ocean heat content, primarily through evaporative cooling associated with Hl and secondarily through sensible exchange [23,58,59].
3.1. Time Series Analysis
The COARE 3.0b reference dataset exhibits a clear seasonal cycle, with maxima generally occurring during austral winter (June, July, August—JJA) and minima during austral summer (December, January, February—DJF) (Figure 4 and Figure 5), consistent with the migration of the ITCZ and the seasonal strengthening of the trade winds [3,13,32,54,60]. At all buoy sites, Hl was approximately one order of magnitude larger than Hs, confirming that ocean heat loss is primarily governed by evaporation, while Hs remains comparatively weak because of the small ocean–atmosphere temperature contrast typical of tropical regions [61,62,63]. This result highlights the dominant role of Hl in the mixed-layer heat budget. The diurnal cycle of SST and near-surface Tair in the TAO is generally small (<0.5 °C) [64], so intra-diurnal variability contributes less to monthly turbulent fluxes than the synoptic and seasonal variations captured by our daily estimates. However, because bulk fluxes depend nonlinearly on SST, neglecting the diurnal cycle may still introduce biases in monthly mean fluxes, particularly under weak-wind conditions [65].
Figure 4.
Monthly mean time series of Hs at PIRATA buoy sites B1 to B7 (a–g). Solid lines represent COARE 3.0b (black), COARE 3.6 (green), ERA5 (gray), BAM (orange), MOM6-CORE (red), and WRF-YSU (blue). Gaps in the colored lines represent the absence of buoy data at the displayed periods.
Figure 5.
Same as Figure 4, but for Hl, from B1 to B7 (a–g). Gaps in the colored lines represent the absence of buoy data at the displayed periods.
In the Tropical North Atlantic, buoys B1, B2, and B3 exhibit a pronounced annual cycle in turbulent heat fluxes (Figure 4a–c and Figure 5a–c), largely controlled by the seasonal trade-wind regime and the meridional migration of the ITCZ [66]. During boreal winter (DJF), strengthened trades and the advection of relatively dry air over warm SST enhance the ocean–atmosphere humidity gradient and turbulent mixing, increasing Hl and, to a lesser extent, Hs. In this period, COARE 3.0b Hl typically reaches ~110–160 W∙m−2, while Hs varies around ~5–15 W∙m−2. At interannual to longer time scales, anomalies in surface turbulent fluxes and ocean heat transport contribute comparably to upper-ocean heat-content variability in the Tropical North Atlantic [67], indicating that differences among products may lead to meaningful heat-budget discrepancies.
A robust feature is the systematically lower Hs and Hl produced by MOM6-CORE, particularly during high-flux seasons at the northern and southern buoy sites, indicating weaker turbulent coupling under similar atmospheric conditions. In contrast, BAM generally remains closer to COARE 3.0b for Hl, but exhibits stronger Hs variability at several buoy sites, suggesting greater sensitivity to local thermal gradients and atmospheric stability. WRF-YSU is characterized less by a persistent offset than by intermittent amplifications of variability, especially for Hs at B4 (Figure 4d) and at some southern buoys. More generally, differences among bulk formulations tend to increase under very weak or very strong winds, when latent heat exchange becomes highly sensitive to transfer coefficients and effective wind-speed definitions [68]. At the same time, interannual turbulent-flux anomalies are also influenced not only by wind variability, but by changes in ocean–atmosphere humidity and temperature gradients, which can play an equally important role in modulating surface exchange [69].
ERA5 exhibits differences in both magnitude and variability, with a tendency toward stronger Hl at several sites. The systematically higher Hl suggests product-dependent tendencies in near-surface humidity, atmospheric stability, or turbulent exchange under tropical marine conditions [70,71]. In this context, recent work also indicates that improvements in near-surface specific humidity fields may occur alongside an increase in latent heat flux difference, and that deficiencies in latent heating aloft can be associated with a cooler free troposphere, mechanisms that can feed back onto surface flux partitioning in coupled systems [72]. In contrast, COARE 3.6 remains the closest to COARE 3.0b across the PIRATA network. While Hs is nearly indistinguishable at several buoys, Hl presents enhanced peak magnitudes, particularly during high-flux periods in the northern and southern sectors. These differences are consistent with the sensitivity of bulk formulations to the near-surface thermal structure, since the use of bulk SST instead of skin temperature can modify the ocean–atmosphere gradients that control both Hs and Hl [13].
At the equatorial buoy B4, the seasonal cycle is weaker than at the off-equatorial sites, and intra-seasonal variability becomes relatively more important (Figure 4d and Figure 5d). This region is characterized by persistent trade winds, weak air–sea temperature gradients, and fewer cold-air incursions, although intermittent gusts and convective activity still occur [45,66,73]. Hl varies between ~50–100 W∙m−2, while Hs remains small (<5 W∙m−2). A modest increase in Hl from boreal spring to mid-year is associated with the strengthening of the southerly trades as the ITCZ migrates northward and the South Atlantic Subtropical High (SASTH) intensifies [2,74]. However, this wind-driven tendency is partially offset by the seasonal development of the Atlantic Cold Tongue, whose upwelling and enhanced vertical mixing cool the equatorial SST by up to ~6 °C and reduce the ocean–atmosphere humidity gradient, thereby limiting evaporation [2,73,75]. The difference time series further indicate that B4 is a site where Hs discrepancies can become episodically large for some schemes, particularly WRF-YSU (Figure 4d), whereas Hl differences remain comparatively moderate relative to the off-equatorial buoys.
The Tropical South Atlantic buoys B5, B6, and B7 exhibit maximum turbulent fluxes during austral winter (JJA) and weaker fluxes during austral summer, following the seasonal displacement of the ITCZ and the modulation of the SASTH (Figure 4e–g and Figure 5e–g). Strengthening trade winds during winter enhance turbulent mixing and surface cooling, leading to higher fluxes in this sector [76]. During these periods, COARE 3.0b indicates Hl values generally between ~70–200 W∙m−2, while Hs varies around ~5–20 W∙m−2, with noticeable interannual differences in peak intensity and duration. Consistent with previous buoy-based heat-budget studies, latent and shortwave fluxes dominate the surface heat balance in the Tropical Atlantic, whereas sensible and longwave components show comparatively smaller variability [77]. Moreover, net surface heat flux tends to be directed into the ocean at buoy locations, with weaker net heating away from the equator where wind-driven evaporation (latent heat loss) is enhanced, particularly near the southeastern PIRATA buoy site (10° S, 10° W) [77]. In the southern sector, the largest and most systematic differences among parameterizations are associated with Hl (Figure 5e–g): BAM, WRF-YSU, and MOM6-CORE generally produce weaker evaporative cooling than COARE 3.0b, whereas ERA5 tends to produce stronger cooling, with discrepancies intensifying during high-flux periods.
3.2. Difference Patterns Across PIRATA Sites
Figure 6 and Figure 7 show the monthly differences in the time series of Hs and Hl relative to COARE 3.0b across the seven PIRATA sites. Overall, Hs differences remain relatively small (±5 W∙m−2), whereas Hl differences are substantially larger (±50 W∙m−2), reflecting the stronger sensitivity of Hl to near-surface humidity gradients and exchange efficiency. In bulk formulations, these differences reflect distinct representations of near-surface state variables and stability-dependent transfer coefficients [13,54]. More fundamentally, MOST-based bulk flux estimates depend on scheme-specific formulations of roughness lengths and atmospheric stability corrections, which directly affect the transfer coefficients (CH and CE) and, consequently, Hs and Hl [54]. As a result, inter-product differences of ~10 W∙m−2 (reaching ~30 W∙m−2 in some regions) are common even among widely used flux products [13], and changing turbulent-flux parameterizations can alter the global ocean surface heat budget by about 10%, making persistent differences of ~10 W∙m−2 in Hl climatologically meaningful [19]. Previous buoy-based evaluations report typical RMSE differences of ~30 W∙m−2 for Hl and ~7 W∙m−2 for Hs [17], while average differences among major datasets are generally <30 W∙m−2 for Hl and <10 W∙m−2 for Hs [78].
Figure 6.
Differences in monthly mean Hs estimates relative to COARE 3.0b at PIRATA buoy sites B1 to B7 (a–g). Solid lines represent the difference between the various bulk formulations in respect to COARE 3.0b. COARE 3.6 to COARE 3.0b (green), ERA5 to COARE 3.0b (gray), BAM to COARE 3.0b (orange), MOM6-CORE to COARE 3.0b (red), and WRF-YSU to COARE 3.0b (blue). Gaps in the colored lines represent the absence of buoy data at the displayed periods.
Figure 7.
Same as Figure 6, but for Hl, from B1 to B7 (a–g). Gaps in the colored lines represent the absence of buoy data at the displayed periods.
BAM exhibits modest positive Hs differences but predominantly negative Hl differences at most northern and southern buoys, with departures intensifying during high-flux periods. This behavior reflects the stronger sensitivity of Hl to near-surface humidity gradients and the CE, making it more responsive than Hs to stability corrections, roughness-length formulations, and effective wind-speed treatments in bulk schemes [8,13]. This pattern is also consistent with sensitivities reported for classical bulk-aerodynamic formulations based on continuous stability functions and prescribed roughness relationships for momentum and scalar quantities constrained largely by Tropical Ocean Global Atmosphere Program (TOGA)/COARE conditions and supplemented by higher-wind datasets [52,79]. Since the parameterization of exchange coefficients at high winds is often constrained by limited observational datasets, sometimes including coastal or shallow-water conditions, their application to open-ocean regimes should be interpreted cautiously [52,80]. Near-surface humidity differences remain a dominant control on Hl variability, particularly during boreal winter and spring [81,82].
MOM6-CORE exhibits persistent negative differences in both Hs and Hl across nearly all buoys, particularly during high-flux periods at the off-equatorial sites. While Hs differences remain relatively small, Hl differences are larger and more sustained, indicating systematically weaker turbulent coupling relative to COARE 3.0b. This behavior is consistent with reduced effective exchange associated with differences in transfer coefficients, stability corrections, and roughness-length parameterizations within MOST-based formulations [13,18]. In CORE-type bulk formulations, neutral transfer coefficients are typically prescribed as empirical functions of the neutral 10 m wind speed, and some implementations constrain the growth of the drag coefficient (CD) under strong winds [19,47]. Compared to COARE-type algorithms, another important distinction is the treatment of SST and weak-wind conditions: CORE formulations generally rely on bulk SST and may impose minimum wind-speed thresholds for scalar transfer, whereas COARE/ECMWF-type includes skin-temperature physics and cool-skin/warm-layer corrections [19,44]. These structural differences are particularly relevant under weak-wind, high-insolation conditions, such as those intermittently observed at the equatorial PIRATA sites.
WRF-YSU shows the most regime-dependent behavior among the evaluated schemes. For Hs, the model shows intermittent months with markedly positive differences, particularly at the equatorial buoy (Figure 6d) and at one southern site (Figure 6g), indicating that short-lived enhancements in exchange are more strongly expressed in the sensible component for this scheme. This behavior is consistent with the known challenges of representing turbulent exchange under weak-wind and convective regimes, where convective gustiness strongly influences the computed fluxes [8,18]. In WRF-YSU, momentum roughness length is parameterized as a wind-dependent quantity, while scalar roughness lengths may differ from momentum roughness, directly affecting the transfer coefficients controlling Hs and Hl [50,54]. For Hl, the scheme frequently exhibits strong negative differences at several buoys, indicating weaker evaporative cooling relative to COARE 3.0b. In addition, some WRF configurations include modifications to the Charnock relation under strong winds, reducing CD and adjusting enthalpy transfer, which can further alter turbulent coupling depending on the atmospheric regime. Overall, these results reinforce that the dominant controls on turbulent ocean–atmosphere flux variability across the PIRATA sites depend not only on wind variability, but also on the scheme-dependent representation of humidity and temperature gradients, particularly for Hl [69,81].
COARE 3.6 remains the closest to zero differences for both Hs and Hl, as expected given the shared physical framework with COARE 3.0b. The COARE family was specifically developed from the Liu–Katsaros–Businger (LKB) framework for tropical-ocean conditions, incorporating MOST stability functions, low-wind gustiness effects, and optional cool-skin and warm-layer corrections that improve the representation of near-surface thermodynamics [44,45]. The main advances in COARE 3.6 involve a refined treatment of wave-dependent surface roughness and whitecap effects based on recent open-ocean observations [51]. Accordingly, Hs differences are generally negligible, while Hl exhibits small but structured deviations, with slightly enhanced peaks during high-flux seasons at some off-equatorial sites. These differences are physically consistent with refinements in near-surface thermodynamic treatments (e.g., interface temperature and stability adjustments). In particular, incorporating skin-temperature physics tends to reduce evaporation and turbulent heat loss, leading to slightly warmer ocean surface conditions compared to formulations based solely on bulk SST [20]. Previous studies show that neglecting skin-temperature and diurnal SST effects can introduce turbulent flux differences of several W∙m−2, locally exceeding ~30 W∙m−2 under strong tropical insolation [58,83], reinforcing the importance of near-surface thermal structure even in analyses based on monthly averages.
ERA5, in contrast, stands out by exhibiting predominantly positive Hl differences at many buoys, often larger and more persistent than those from the model-based schemes, indicating systematically stronger evaporative cooling relative to COARE 3.0b. Hs differences are generally smaller but also tend to be positive, suggesting overall stronger turbulent exchange. In the tropical Atlantic, excessive Hl can imply enhanced surface cooling and anomalous moisture transfer to the atmosphere [17]. Nevertheless, ERA5 performance is known to vary regionally, and previous buoy-based evaluations have shown comparatively good skill in other tropical basins [26]. In the Tropical Atlantic, overestimation of Hl, as reported in some buoy-based reanalysis evaluations, implies excessive surface cooling and enhanced moisture transfer from the ocean to the atmosphere [17], with potential impacts on coupled variability. Buoy-based assessments further indicate that a substantial fraction of net heat flux differences is associated with latent heat loss errors linked to differences in transfer coefficients and stability-dependent formulations [66]. These discrepancies likely result from both differences in the assimilated atmospheric/oceanic fields and the surface-layer parameterizations used in the reanalysis system [14,84]. Nevertheless, ERA5 performance is not uniformly poor. In at least one independent regional evaluation (Tropical Indian Ocean moorings), ERA5 was found to outperform several other products in correlation and RMSE relative to in situ values [26], highlighting that skill and differences can be regime and basin dependent.
3.3. Skill and Uncertainty in Heat Flux Estimates
Skill and uncertainty were evaluated using scatter density plots and Taylor diagrams for monthly Hs and Hl estimates from BAM, MOM6-CORE, WRF-YSU, ERA5, and COARE 3.6 relative to COARE 3.0b during 1997–2023. Statistical metrics for each buoy and parameterization are summarized in Table 3, while Figure 8, Figure 9 and Figure 10 present the scatter density distributions and Taylor-diagram performance.
Table 3.
Monthly statistical metrics (1997–2023) for each parameterization relative to the COARE 3.0b reference at the PIRATA buoys, including mean flux, difference (parameterization—COARE 3.0b), RMSE, correlation coefficient, p-value, and upper and lower confidence intervals between the COARE 3.0b and parameterizations.
Figure 8.
Scatter plots of monthly Hs estimates at the seven PIRATA buoys in the TAO. Panels (a–g) correspond to buoys B1–B7, respectively. Colored markers represent ERA5, WRF-YSU, BAM, COARE 3.6, and MOM6-CORE estimates relative to the COARE 3.0b reference.
Figure 9.
Same as Figure 8, but for Hl, from B1 to B7 (a–g). Gaps in the colored lines represent the absence of buoy data at the displayed periods.
Figure 10.
Taylor diagrams of monthly Hs (circles) and (triangles) estimates at the seven PIRATA buoys in the TAO. Panels (a–g) correspond to buoys B1–B7. Colored markers represent ERA5, WRF-YSU, BAM, COARE 3.6, and MOM6-CORE estimates relative to the COARE 3.0b reference. Radial distance indicates normalized standard deviation, while concentric curves denote correlation coefficients.
Part of the dispersion among products reflects not only differences in bulk formulations and near-surface variables, but also the effects of temporal and spatial averaging, which can suppress high-frequency wind variability and reduce flux variance in gridded products relative to buoy observations [13,29,45], particularly relevant for gridded reanalyses and NWP-based products [29]. Consequently, satellite- and reanalysis-based flux estimates may exhibit systematic differences associated with the incomplete representation of short-lived extreme events characterized by strong winds and sharp near-surface thermodynamic (temperature and humidity) gradients [13].
Overall, the model-based schemes reproduce a substantial fraction of the monthly variability, with correlations generally exceeding 0.8 (Figure 10). However, systematic differences remain evident (Table 3). BAM tends to overestimate Hs (1 to 3 W∙m−2) while underestimating Hl (−5 to −25 W∙m−2) at most buoys (although Hl becomes slightly positive at B4), suggesting different sensitivities of sensible and latent exchange to stability and roughness formulations. MOM6-CORE displays the most consistent negative bias for both Hs and Hl across all sites (e.g., B1 Hs = −1.3 W∙m−2, Hl = −24.0 W∙m−2), indicating systematically weaker turbulent exchange relative to COARE 3.0b. WRF-YSU generally overestimates Hs (positive differences, +1 to +3 W∙m−2) and underestimates Hl (negative differences, often −10 to −30 W∙m−2) at northern and southern sites, while opposite signs occur at B4 and B7, highlighting its sensitivity to local wind and stability regimes.
These systematic differences are dynamically important because persistent changes in evaporative cooling directly affect the net surface heat budget. A formulation that is systematically less evaporative implies greater ocean heat gain and therefore tends to favor warmer SSTs [20]. In coupled or forced-ocean simulations, differences in turbulent exchange can accumulate through the net surface heat flux (Qnet) and influence SST evolution and air–sea coupling on seasonal timescales, even when Hs differences are relatively small [77]. Consistent with this, Valdivieso et al. [77] showed that differences among flux formulations can produce substantial changes in Qnet, highlighting that variations in turbulent coupling remain climatically relevant even when individual flux-component differences appear modest on average.
The Taylor diagrams indicate generally strong correlation strengths (r ≥ 0.8) (Figure 10, Table 3), implying that a large fraction of temporal variability is captured despite systematic differences. However, buoy B6 in the southwestern Tropical Atlantic exhibits comparatively lower skill, as evidenced by the broader scatter (Figure 8 and Figure 9) and weaker correlations. This behavior is likely associated with the influence of austral-winter cold-air incursions and frontal systems in this area, which produce abrupt changes in air–sea temperature and humidity gradients and generate intense turbulent exchange events that are difficult for bulk formulations to represent [10]. Monthly averaging inherently smooths synoptic and diurnal variability, potentially masking larger discrepancies that become more evident at shorter timescales.
At the equatorial buoy B4, WRF-YSU exhibits the largest std for Hs (std > 1.5), indicating an overestimation of variability relative to the reference. Agreement also tends to degrade near the equator, where low-wind conditions increase the sensitivity of bulk flux estimates to small differences in the input variables and reduce correlations among products [78]. Part of the spread observed in the scatter plots and Taylor diagrams therefore reflects both the weak-wind sensitivity of bulk formulations and the reduced wind variability represented in gridded products relative to mooring observations [29,78].
As expected, COARE 3.6 shows near-reference behavior, with scatter densities closely aligned with the 1:1 line for both Hs and Hl, and Taylor-diagram statistics approaching close agreement (r ≈ 0.99 and std ≈ 1) across all buoys, reflecting the shared physical framework with COARE 3.0b. The small remaining differences are mainly associated with refinements introduced in COARE 3.6, including updated treatments of wave effects, gustiness, and near-surface thermodynamic processes such as cool-skin and warm-layer corrections, which primarily affect Hl under weak-wind, high-insolation conditions [51,85].
ERA5 exhibits moderate-to-strong correlations and std close to the reference, indicating that temporal variability is well represented. However, the scatter density plots reveal a systematic tendency to overestimate turbulent heat exchange, particularly Hl, implying stronger evaporative cooling than COARE 3.0b at many sites. This behavior is consistent with previous evaluations reporting positive differences in both flux components, with larger departures for Hl than for Hs (e.g., [21]). Similar assessments have also documented substantial RMSE differences for ERA5 (~27–32 W∙m−2 for Hl and ~5–8 W∙m−2 for Hs) in the Tropical Indian Ocean [26] despite relatively high correlations, highlighting that a realistic representation of temporal variability does not necessarily imply accurate flux magnitudes [26]. Overall, these results highlight that a realistic representation of temporal variability does not necessarily imply accurate flux magnitudes, as systematic differences related to near-surface wind, humidity, and stability remain non-negligible.
The discrepancies observed in ERA5 reflect the combined uncertainties of both the assimilated state variables and the internal coupling physics of the forecasting system. In the tropical belt, the NEMO ocean analysis is relaxed toward SST products such as OSTIA, which may exhibit delays of up to 60 h in capturing rapid SST changes [36]. These limitations can affect the representation of ocean–atmosphere gradients and, consequently, turbulent heat flux estimates. Part of this uncertainty can be reduced through multi-product approaches. Bentamy et al. [17] showed that ensemble-based estimates improve agreement with buoy observations, achieving RMS differences typically below ~25 W∙m−2 for Hl and ~5 W∙m−2 for Hs, often outperforming individual products. This provides context for interpreting Figure 8, Figure 9 and Figure 10, as individual parameterizations and reanalyses exhibit distinct systematic behaviors, whereas combining complementary products can reduce uncertainties in turbulent heat flux estimates, particularly for Hl, which dominates both the flux magnitude and the error budget of the tropical surface heat balance [17,77]. Therefore, although the dynamic consistency of the IFS enables ERA5 to reproduce much of the observed temporal variability, the magnitude of turbulent heat fluxes in the TAO remains sensitive to limitations in representing rapid changes in regional SST and near-surface thermodynamic gradients. This reinforces the value of ERA5 as a benchmark for assessing the total discrepancies of gridded flux products compared to point-based PIRATA observations.
4. Conclusions
This study assessed the performance of three bulk flux parameterizations used operationally at CPTEC/INPE against the COARE 3.0b reference at seven PIRATA moorings in the Tropical Atlantic Ocean during the period 1997–2023. It is important to note that COARE 3.0b is itself a bulk-algorithm estimate rather than a direct observation. COARE 3.6 and ERA5 were included as additional benchmarks. All offline schemes were forced with identical daily buoy observations, isolating differences due to internal physical assumptions. Monthly means are used, but the original daily estimates made previously to the monthly means are expected to preserve the original, nonlinear ocean–atmosphere heat exchange processes happening in situ in the study area.
Hl is approximately one order of magnitude larger than Hs at all PIRATA buoy sites, and differences among parameterizations are larger for Hl (±50 W∙m−2) than for Hs (±5 W∙m−2), reflecting the stronger sensitivity of latent exchange to humidity gradients and stability-dependent transfer coefficients [18,77]. All schemes reproduce the seasonal cycle forced by the ITCZ migration and trade winds variability, with correlations generally exceeding 0.8 (p < 0.001) for most buoys; however, systematic magnitude biases remain.
Among the model-based parameterizations, MOM6-CORE showed the most consistent representation relative to COARE 3.0b, combining high correlations and realistic variability across most buoy sites. In contrast, BAM and WRF-YSU exhibit stronger regime-dependent behavior, including intermittent amplification of Hs and larger Hl deviations under weak-wind and convective conditions [29,45,78]. As expected, COARE 3.6 remains the closest to the reference due to its shared physical framework with COARE 3.0b. ERA5 reproduces a large fraction of the temporal variability but retains systematic magnitude differences, particularly for Hl [14,84]. These results demonstrate that, even under identical observational forcing, differences in roughness-length formulations, effective wind treatment, and non-neutral transfer coefficients remain a first-order source of uncertainty in ocean–atmosphere turbulent heat flux estimates.
Specific environmental regimes occurring at each buoy site affect the ocean–atmosphere heat flux estimates performance of each parametrization in different ways, partially explaining the spatial variability in the models’ skill. At the southwestern Tropical Atlantic buoy site (B6), episodic atmospheric synoptic forcing, including austral-winter air-mass intrusions and frontal passages, broadens the scatter and reduces skill, particularly for Hs, highlighting the difficulty of representing short-term, high-gradient events with bulk aerodynamic formulations. Near the Equator (B4 site), weaker wind variability and greater sensitivity to stability and convective gustiness lead to lower agreement and larger differences in variance among schemes, consistent with the known degradation of bulk-flux performance under weak-wind conditions and gridded products may under-represent wind variance [29,78].
By prescribing identical observational forcing, the present intercomparison isolates the effects of parameterization and enables a more mechanistic interpretation of differences among schemes. Future studies using higher-frequency (daily or hourly) observations would help assess how these parameterizations respond to abrupt changes in ocean–atmosphere temperature and humidity gradients and wind conditions, which are partially smoothed in the monthly mean framework adopted here. Consequently, the generally high skill metrics reported here for monthly means may underestimate performance differences under weather-scale extreme conditions, as suggested by the lower skill at B6 site. As noted by Santini et al. [10] for the Southwestern Atlantic Ocean, bulk parameterizations tend to exhibit larger biases relative to direct Eddy-Covariance measurements, especially in regions of strong SST gradients or during the passage or formation of transient atmospheric systems. Overall, these results demonstrate that the choice of bulk formulation remains a first-order source of uncertainty in turbulent heat flux estimates over the TAO, with implications for mixed-layer heat budgets, SST evolution, and coupled ocean–atmosphere variability.
Among the CPTEC/INPE operational parameterizations evaluated here, MOM6-CORE provides the most consistent performance relative to the COARE 3.0b reference, while COARE 3.6 remains a valuable benchmark for process-based evaluation of operational models and reanalysis products. More broadly, this work contributes to the ongoing development and implementation of the new Brazilian Earth System Model, MONAN (Model for Ocean-Land-Atmosphere Prediction), thereby supporting efforts to improve the representation of ocean–atmosphere interaction processes in next-generation weather and climate prediction systems for Brazil and Latin America.
Author Contributions
Conceptualization, L.S. and R.B.d.S.; methodology, L.S. and R.B.d.S.; software, L.S.; validation, L.S. and R.B.d.S.; formal analysis, L.S. and R.B.d.S.; investigation, L.S. and R.B.d.S.; resources, R.B.d.S.; data curation, L.S.; writing—original draft preparation, L.S., R.B.d.S., R.M., B.T.S., G.F. and L.P.P.; writing—review and editing, L.S., R.M., R.B.d.S., B.T.S., G.F. and L.P.P.; visualization, L.S.; supervision, R.B.d.S.; project administration, R.B.d.S.; funding acquisition, R.B.d.S. All authors have read and agreed to the published version of the manuscript.
Funding
This research was funded by the Brazilian agency CNPq for SOAC Multiescala—Multiscale Study of the Ocean–atmosphere-Cryosphere Coupled System (CNPq 406663/2022-0) and was part of the MSc. Dissertation of L.S. supported by the Graduate Program in Meteorology of INPE. L.P.P. is partly funded by CNPq through the Scientific Productivity Fellowship Program (CNPq 303981/2023-7) and the Antarctic Modeling and Observation System (ATMOS) Project.
Data Availability Statement
PIRATA data, PIRATA flux and ERA5 data can be found on the websites cited in the Materials and Methods and Acknowledgements sections. Furthermore, the original contributions presented in the study are included in the article, and further inquiries can be directed to the corresponding author.
Acknowledgments
The authors are thankful to the NOAA/PMEL (https://www.pmel.noaa.gov/tao/drupal/disdel, accessed on 3 June 2026); The ERA5 “atmospheric monthly averaged reanalysis products” (https://cds.climate.copernicus.eu/datasets/reanalysis-era5-single-levels?tab=overview, accessed on 3 June 2026); and CPTEC/INPE for the data used here. We thank the PIRATA Project, NOAA, the Brazilian Navy, the Institute de Researche pour Développement (IRD) and Météo-France for the deployment and maintenance of the PIRATA moorings. We also acknowledge the suggestions to the work or to the text made by Paulo Kubota, Vinicio Santos, Marcelo Santini and the anonymous reviewers of this manuscript.
Conflicts of Interest
The authors declare no conflicts of interest.
Abbreviations
The following abbreviations are used in this manuscript:
| ACT | Atlantic Cold Tongue |
| AWP | Atlantic Warm Pool |
| BAM | Brazilian Atmospheric Model |
| BC | Brazil Current |
| BESM | Brazilian Earth System Model |
| CE | Exchange coefficient for moisture |
| CH | Exchange coefficient for heat |
| CORE | Coordinated Ocean-ice Reference Experiments |
| COARE | Coupled Ocean–Atmosphere Response Experiment |
| CPTEC | Centro de Previsão de Tempo e Estudos Climáticos |
| CD | Drag coefficient |
| EC | Eddy Covariance |
| ECMWF | European Centre for Medium-Range Weather Forecasts |
| ERA5 | ECMWF Reanalysis (5th generation) |
| EUC | Equatorial Undercurrent |
| Hl | Latent heat flux |
| Hs | Sensible heat flux |
| INPE | National Institute for Space Research |
| ITCZ | Intertropical Convergence Zone |
| LSM | Land Surface Model |
| MOM | Modular Ocean Model |
| MONAN | Model for Ocean-Land-Atmosphere Prediction |
| MOST | Monin–Obukhov Similarity Theory |
| NBC | North Brazil Current |
| NEC | North Equatorial Current |
| NECC | North Equatorial Countercurrent |
| NOAA | National Oceanic and Atmospheric Administration |
| NWP | Numerical Weather Prediction |
| nSEC | Northern branches of the South Equatorial Current |
| cSEC | Central branches of the South Equatorial Current |
| sSEC | Southern branches of the South Equatorial Current |
| OAFlux | Objectively Analyzed Air–sea Fluxes |
| PIRATA | Prediction and Research Moored Array in the Tropical Atlantic |
| PMEL | Pacific Marine Environmental Laboratory |
| Qnet | Net surface heat flux |
| RMSE | Root Mean Square Error |
| RMSD | Root Mean Square Difference |
| SEC | South Equatorial Current |
| SEUC | South Equatorial Undercurrent |
| TAO | Tropical Atlantic Ocean |
| TOGA | Tropical Ocean Global Atmosphere |
| WAM | West African Monsoon |
| WHOI | Woods Hole Oceanographic Institution |
| WRF | Weather Research and Forecasting |
| YSU | Yonsei University |
References
- Fontaine, B.; Janicot, S.; Roucou, P. Coupled Ocean–atmosphere Surface Variability and Its Climate Impacts in the Tropical Atlantic Region. Clim. Dyn. 1999, 15, 451–473. [Google Scholar] [CrossRef] [Scilit]
- Xie, S.-P.; Carton, J.A. Tropical Atlantic Variability: Patterns, Mechanisms, and Impacts. In Earth’s Climate: The Ocean–Atmosphere Interaction; American Geophysical Union: Washington, DC, USA, 2004; Volume 147, pp. 121–142. [Google Scholar]
- Cabos, W.; de la Vara, A.; Koseki, S. Tropical Atlantic Variability: Observations and Modeling. Atmosphere 2019, 10, 502. [Google Scholar] [CrossRef] [Scilit]
- Zhang, H.; Ma, X.; Zhao, S.; Kong, L. Advances in Research on the ITCZ: Mean Position, Model Bias, and Anthropogenic Aerosol Influences. J. Meteorol. Res. 2021, 35, 729–742. [Google Scholar] [CrossRef] [Scilit]
- Aguedjou, H.M.A.; Chaigneau, A.; Dadou, I.; Morel, Y.; Baloïtcha, E.; Da-Allada, C.Y. Imprint of Mesoscale Eddies on Air-Sea Interaction in the Tropical Atlantic Ocean. Remote Sens. 2023, 15, 3087. [Google Scholar] [CrossRef] [Scilit]
- Liu, H.; Zhang, L.; Capotondi, A.; Wang, X.; Fan, H.; Chen, B. Formation Mechanisms of the Central and Eastern Atlantic Niño. npj Clim. Atmos. Sci. 2025, 8, 48. [Google Scholar] [CrossRef] [Scilit]
- Harris, E.A.; Marsh, R.; Grist, J.P. Tracing Oceanic Sources of Heat Content Available for Atlantic Hurricanes. J. Geophys. Res. Oceans 2023, 128, e2022JC019407. [Google Scholar] [CrossRef] [Scilit]
- Reeves Eyre, J.E.J.; Zeng, X.; Zhang, K. Ocean Surface Flux Algorithm Effects on Earth System Model Energy and Water Cycles. Front. Mar. Sci. 2021, 8, 642804. [Google Scholar] [CrossRef] [Scilit]
- Grachev, A.A.; Bariteau, L.; Fairall, C.W.; Hare, J.E.; Helmig, D.; Hueber, J.; Lang, E.K. Turbulent Fluxes and Transfer of Trace Gases from Ship-Based Measurements during TexAQS 2006. J. Geophys. Res. Atmos. 2011, 116, D13110. [Google Scholar] [CrossRef] [Scilit]
- Santini, M.F.; Souza, R.B.; Pezzi, L.P.; Swart, S. Observations of Air–Sea Heat Fluxes in the Southwestern Atlantic under High-Frequency Ocean and Atmospheric Perturbations. Q. J. R. Meteorol. Soc. 2020, 146, 4226–4251. [Google Scholar] [CrossRef] [Scilit]
- Sun, Y.; Sude, B.; Lin, X.; Geng, B.; Liu, B.; Ji, S.; Jing, J.; Zhu, Z.; Xu, Z.; Liu, S.; et al. Quality Evaluation for Measurements of Wind Field and Turbulent Fluxes from a UAV-Based Eddy Covariance System. Atmos. Meas. Tech. 2023, 16, 5659–5679. [Google Scholar] [CrossRef] [Scilit]
- Pezzi, L.P.; Souza, R.B.; Farias, P.C.; Acevedo, O.; Miller, A.J. Air-Sea Interaction at the Southern Brazilian Continental Shelf: In Situ Observations. J. Geophys. Res. Oceans 2016, 121, 6671–6695. [Google Scholar] [CrossRef] [Scilit]
- Cronin, M.F.; Gentemann, C.L.; Edson, J.; Ueki, I.; Bourassa, M.; Brown, S.; Clayson, C.A.; Fairall, C.W.; Farrar, J.T.; Gille, S.T.; et al. Air-Sea Fluxes with a Focus on Heat and Momentum. Front. Mar. Sci. 2019, 6, 430. [Google Scholar] [CrossRef] [Scilit]
- Yu, L. Global Air-Sea Fluxes of Heat, Fresh Water, and Momentum: Energy Budget Closure and Unanswered Questions. Annu. Rev. Mar. Sci. 2019, 11, 227–248. [Google Scholar] [CrossRef] [Scilit]
- Edson, J.B.; Jampana, V.; Weller, R.A.; Bigorre, S.P.; Plueddemann, A.J.; Fairall, C.W.; Miller, S.D.; Mahrt, L.; Vickers, D.; Hersbach, H. On the Exchange of Momentum over the Open Ocean. J. Phys. Oceanogr. 2013, 43, 1589–1610. [Google Scholar] [CrossRef] [Scilit]
- Zhang, Z.-L.; Hui, F.-M.; Vihma, T.; Granskog, M.A.; Cheng, B.; Chen, Z.-Q.; Cheng, X. On the Turbulent Heat Fluxes: A Comparison among Satellite-Based Estimates, Atmospheric Reanalyses, and in-Situ Observations during the Winter Climate over Arctic Sea Ice. Adv. Clim. Change Res. 2023, 14, 347–362. [Google Scholar] [CrossRef] [Scilit]
- Bentamy, A.; Piollé, J.F.; Grouazel, A.; Danielson, R.; Gulev, S.; Paul, F.; Azelmat, H.; Mathieu, P.P.; von Schuckmann, K.; Sathyendranath, S.; et al. Review and Assessment of Latent and Sensible Heat Flux Accuracy over the Global Oceans. Remote Sens. Environ. 2017, 201, 196–218. [Google Scholar] [CrossRef] [Scilit]
- Brunke, M.A.; Fairall, C.W.; Zeng, X.; Eymard, L.; Curry, J.A. Which Bulk Aerodynamic Algorithms are Least Problematic in Computing Ocean Surface Turbulent Fluxes? J. Clim. 2003, 16, 619–635. [Google Scholar] [CrossRef] [Scilit]
- Brodeau, L.; Barnier, B.; Gulev, S.K.; Woods, C. Climatologically Significant Effects of Some Approximations in the Bulk Parameterizations of Turbulent Air–Sea Fluxes. J. Phys. Oceanogr. 2017, 47, 5–28. [Google Scholar] [CrossRef] [Scilit]
- Bonino, G.; Iovino, D.; Brodeau, L.; Masina, S. The Bulk Parameterizations of Turbulent Air–Sea Fluxes in NEMO4: The Origin of Sea Surface Temperature Differences in a Global Model Study. Geosci. Model Dev. 2022, 15, 6873–6889. [Google Scholar] [CrossRef] [Scilit]
- Tang, R.; Wang, Y.; Jiang, Y.; Liu, M.; Peng, Z.; Hu, Y.; Huang, L.; Li, Z.-L. A Review of Global Products of Air-Sea Turbulent Heat Flux: Accuracy, Mean, Variability, and Trend. Earth-Sci. Rev. 2024, 249, 104662. [Google Scholar] [CrossRef] [Scilit]
- Moura, R.; Casagrande, F.; de Souza, R.B. An Overview of Air-Sea Heat Flux Products and CMIP6 HighResMIP Models in the Southern Ocean. Atmosphere 2025, 16, 402. [Google Scholar] [CrossRef] [Scilit]
- Yu, L.; Jin, X.; Weller, R.A. Role of Net Surface Heat Flux in Seasonal Variations of Sea Surface Temperature in the Tropical Atlantic Ocean. J. Clim. 2006, 19, 6153–6169. [Google Scholar] [CrossRef] [Scilit]
- Moura, R.; de Souza, R.B.; Casagrande, F.; da Silva Lindemann, D. Air–Sea Heat Fluxes Variations in the Southern Atlantic Ocean: Present-Day and Future Climate Scenarios. Int. J. Climatol. 2024, 44, 3136–3153. [Google Scholar] [CrossRef] [Scilit]
- Cyriac, A.; McPhaden, M.J.; Phillips, H.E.; Bindoff, N.L.; Feng, M. Seasonal Evolution of the Surface Layer Heat Balance in the Eastern Subtropical Indian Ocean. J. Geophys. Res. Oceans 2019, 124, 6459–6477. [Google Scholar] [CrossRef] [Scilit]
- Pokhrel, S.; Dutta, U.; Rahaman, H.; Chaudhari, H.; Hazra, A.; Saha, S.K.; Veeranjaneyulu, C. Evaluation of Different Heat Flux Products Over the Tropical Indian Ocean. Earth Space Sci. 2020, 7, e2019EA000988. [Google Scholar] [CrossRef] [Scilit]
- Bourlès, B.; Araujo, M.; McPhaden, M.J.; Brandt, P.; Foltz, G.R.; Lumpkin, R.; Giordani, H.; Hernandez, F.; Lefèvre, N.; Nobre, P.; et al. PIRATA: A Sustained Observing System for Tropical Atlantic Climate Research and Forecasting. Earth Space Sci. 2019, 6, 577–616. [Google Scholar] [CrossRef] [Scilit]
- Yu, L.; Weller, R.A. Objectively Analyzed Air–Sea Heat Fluxes for the Global Ice-Free Oceans (1981–2005). Bull. Am. Meteorol. Soc. 2007, 88, 527–540. [Google Scholar] [CrossRef] [Scilit]
- Praveen Kumar, B.; Vialard, J.; Lengaigne, M.; Murty, V.S.N.; McPhaden, M.J. TropFlux: Air-Sea Fluxes for the Global Tropical Oceans—Description and Evaluation. Clim. Dyn. 2012, 38, 1521–1543. [Google Scholar] [CrossRef] [Scilit]
- Dey, D.; Sil, S.; Jana, S.; Pramanik, S.; Pandey, P.C. An Assessment of TropFlux and NCEP Air-Sea Fluxes on ROMS Simulations over the Bay of Bengal Region. Dyn. Atmos. Oceans 2017, 80, 47–61. [Google Scholar] [CrossRef] [Scilit]
- Utida, G.; Cruz, F.W.; Etourneau, J.; Bouloubassi, I.; Schefuß, E.; Vuille, M.; Novello, V.F.; Prado, L.F.; Sifeddine, A.; Klein, V.; et al. Tropical South Atlantic Influence on Northeastern Brazil Precipitation and ITCZ Displacement during the Past 2300 Years. Sci. Rep. 2019, 9, 1698. [Google Scholar] [CrossRef] [Scilit] [PubMed]
- Foltz, G.R.; Brandt, P.; Richter, I.; Rodríguez-Fonseca, B.; Hernandez, F.; Dengler, M.; Rodrigues, R.R.; Schmidt, J.O.; Yu, L.; Lefevre, N.; et al. The Tropical Atlantic Observing System. Front. Mar. Sci. 2019, 6, 206. [Google Scholar] [CrossRef] [Scilit]
- Schott, F.A.; McCreary, J.P., Jr.; Johnson, G.C. Shallow Overturning Circulations of the Tropical-Subtropical Oceans. In Earth’s Climate: The Ocean–Atmosphere Interaction; American Geophysical Union: Washington, DC, USA, 2004; Volume 147, pp. 261–304. [Google Scholar] [CrossRef] [Scilit]
- Speich, S.; Araujo, M.; Balmaseda, M.; Bourlès, B.; Foltz, G.; McPhaden, M.J.; Rodrigues, R.R. An Introduction to the Tropical Atlantic Ocean Observing System: Past and Present. CLIVAR Exch. Trop. Atl. Ocean Obs. Syst. TAOS 2022, 82, 3–8. [Google Scholar] [CrossRef]
- Faria, A.L.S.; de Freitas, R.A.P.; de Souza, R.B.; Wallner-Kersanach, M.; Evangelista, H.; Moura, R.; Contreira-Pereira, L.; Mirlean, N.; Seus-Arrache, E.R.; da Costa Machado, E. The Role of the Intertropical Convergence Zone on the Aerosol Transport Pathways From Africa Towards the Western Tropical Atlantic Ocean. Int. J. Climatol. 2025, 45, e8910. [Google Scholar] [CrossRef] [Scilit]
- Hersbach, H.; Bell, B.; Berrisford, P.; Hirahara, S.; Horányi, A.; Muñoz-Sabater, J.; Nicolas, J.; Peubey, C.; Radu, R.; Schepers, D.; et al. The ERA5 Global Reanalysis. Q. J. R. Meteorol. Soc. 2020, 146, 1999–2049. [Google Scholar] [CrossRef] [Scilit]
- Beljaars, A.C.M. The parametrization of surface fluxes in large-scale models under free convection. Q. J. R. Meteorol. Soc. 1995, 121, 255–270. [Google Scholar] [CrossRef] [Scilit]
- Janssen, P.A.E.M. Wave-Induced Stress and the Drag of Air Flow over Sea Waves. J. Phys. Oceanogr. 1989, 19, 745–754. [Google Scholar] [CrossRef] [Scilit]
- Brutsaert, W. Evaporation into the Atmosphere: Theory, History and Applications; Springer Science & Business Media: Dordrecht, The Netherlands, 2013; ISBN 978-94-017-1497-6. [Google Scholar]
- Nagy, H.; Mohamed, B.; Ibrahim, O. Variability of Heat and Water Fluxes in the Red Sea Using ERA5 Data (1981–2020). J. Mar. Sci. Eng. 2021, 9, 1276. [Google Scholar] [CrossRef] [Scilit]
- Kong, B.; Liu, N.; Fan, L.; Lin, L.; Yang, L.; Chen, H.; Wang, Y.; Zhang, Y.; Xu, Y. Evaluation of Surface Meteorology Parameters and Heat Fluxes from CFSR and ERA5 over the Pacific Arctic Region. Q. J. R. Meteorol. Soc. 2022, 148, 2973–2990. [Google Scholar] [CrossRef] [Scilit]
- Mayer, J.; Haimberger, L.; Mayer, M. A Quantitative Assessment of Air–Sea Heat Flux Trends from ERA5 since 1950 in the North Atlantic Basin. Earth Syst. Dyn. 2023, 14, 1085–1105. [Google Scholar] [CrossRef] [Scilit]
- Shevchenko, G.V.; Lozhkin, D.M. Spatial and Temporal Variability of a Latent Heat Flux in the Northwest Pacific Ocean Based on the ERA5 Reanalysis Data. Phys. Oceanogr. 2024, 31, 387–397. [Google Scholar]
- Fairall, C.W.; Bradley, E.F.; Hare, J.E.; Grachev, A.A.; Edson, J.B. Bulk Parameterization of Air–Sea Fluxes: Updates and Verification for the COARE Algorithm. J. Clim. 2003, 16, 571–591. [Google Scholar] [CrossRef] [Scilit]
- Cronin, M.F.; Fairall, C.W.; McPhaden, M.J. An Assessment of Buoy-Derived and Numerical Weather Prediction Surface Heat Fluxes in the Tropical Pacific. J. Geophys. Res. Oceans 2006, 111, C06038. [Google Scholar] [CrossRef] [Scilit]
- Figueroa, S.N.; Bonatti, J.P.; Kubota, P.Y.; Grell, G.A.; Morrison, H.; Barros, S.R.M.; Fernandez, J.P.R.; Ramirez, E.; Siqueira, L.; Luzia, G.; et al. The Brazilian Global Atmospheric Model (BAM): Performance for Tropical Rainfall Forecasting and Sensitivity to Convective Scheme and Horizontal Resolution. Weather Forecast. 2016, 31, 1547–1572. [Google Scholar] [CrossRef] [Scilit]
- Large, W.G.; Yeager, S.G. The Global Climatology of an Interannually Varying Air–Sea Flux Data Set. Clim. Dyn. 2009, 33, 341–364. [Google Scholar] [CrossRef] [Scilit]
- Veiga, S.F.; Nobre, P.; Giarolla, E.; Capistrano, V.; Baptista, M., Jr.; Marquez, A.L.; Figueroa, S.N.; Bonatti, J.P.; Kubota, P.; Nobre, C.A. The Brazilian Earth System Model Ocean–Atmosphere (BESM-OA) Version 2.5: Evaluation of Its CMIP5 Historical Simulation. Geosci. Model Dev. 2019, 12, 1613–1642. [Google Scholar] [CrossRef] [Scilit]
- Hong, S.-Y.; Noh, Y.; Dudhia, J. A New Vertical Diffusion Package with an Explicit Treatment of Entrainment Processes. Mon. Weather. Rev. 2006, 134, 2318–2341. [Google Scholar] [CrossRef] [Scilit]
- Skamarock, W.C.; Klemp, J.B. A Time-Split Nonhydrostatic Atmospheric Model for Weather Research and Forecasting Applications. J. Comput. Phys. 2008, 227, 3465–3485. [Google Scholar] [CrossRef] [Scilit]
- Fairall, C.W.; Blomquist, B.; Bariteau, L.; Edson, J.B. COARE-Algorithm/References/Coare36_readme_1.Pdf at Master · NOAA-PSL/COARE-Algorithm·GitHub. Available online: https://github.com/NOAA-PSL/COARE-algorithm/blob/master/References/coare36_readme_1.pdf (accessed on 3 June 2026).
- Zeng, X.; Zhao, M.; Dickinson, R.E. Intercomparison of Bulk Aerodynamic Algorithms for the Computation of Sea Surface Fluxes Using TOGA COARE and TAO Data. J. Clim. 1998, 11, 2628–2644. [Google Scholar] [CrossRef] [Scilit]
- McDougall, T.J.; Barker, P.M.; Holmes, R.M.; Pawlowicz, R.; Griffies, S.M.; Durack, P.J. The Interpretation of Temperature and Salinity Variables in Numerical Ocean Model Output and the Calculation of Heat Fluxes and Heat Content. Geosci. Model Dev. 2021, 14, 6445–6466. [Google Scholar] [CrossRef] [Scilit]
- Yang, Y.; Hu, X.-M.; Gao, S.; Wang, Y. Sensitivity of WRF Simulations with the YSU PBL Scheme to the Lowest Model Level Height for a Sea Fog Event over the Yellow Sea. Atmos. Res. 2019, 215, 253–267. [Google Scholar] [CrossRef] [Scilit]
- Brunke, M.A.; Zhou, M.; Zeng, X.; Andreas, E.L. An Intercomparison of Bulk Aerodynamic Algorithms Used over Sea Ice with Data from the Surface Heat Budget for the Arctic Ocean (SHEBA) Experiment. J. Geophys. Res. Oceans 2006, 111, C09001. [Google Scholar] [CrossRef] [Scilit]
- Taylor, K.E. Summarizing Multiple Aspects of Model Performance in a Single Diagram. J. Geophys. Res. Atmos. 2001, 106, 7183–7192. [Google Scholar] [CrossRef] [Scilit]
- Steffen, B.T.; de Souza, R.B.; de Freitas, R.A.P.; Noernberg, M.A.; Parise, C.K. Synoptic Ocean–Atmosphere Coupling at the Intertropical Convergence Zone and Its Vicinity in the Western Tropical Atlantic Ocean. Atmosphere 2026, 17, 101. [Google Scholar] [CrossRef] [Scilit]
- Foltz, G.R.; Grodsky, S.A.; Carton, J.A.; McPhaden, M.J. Seasonal Mixed Layer Heat Budget of the Tropical Atlantic Ocean. J. Geophys. Res. Oceans 2003, 108, 3146. [Google Scholar] [CrossRef] [Scilit]
- Thiam, M.; de Coetlogon, G.; Wade, M.; Sarr, M.; Diop, B. Air–Sea Feedback in the Northeastern Tropical Atlantic in Boreal Summer at Intraseasonal Time-Scales. Q. J. R. Meteorol. Soc. 2025, 151, e4982. [Google Scholar] [CrossRef] [Scilit]
- Martín-Rey, M.; Polo, I.; Rodríguez-Fonseca, B.; Losada, T.; Lazar, A. Is There Evidence of Changes in Tropical Atlantic Variability Modes under AMO Phases in the Observational Record? J. Clim. 2018, 31, 515–536. [Google Scholar] [CrossRef] [Scilit]
- Bunker, A.F. Computations of Surface Energy Flux and Annual Air–Sea Interaction Cycles of the North Atlantic Ocean. Mon. Weather Rev. 1976, 104, 1122–1140. [Google Scholar] [CrossRef] [Scilit]
- Chou, S.-H.; Nelkin, E.; Ardizzone, J.; Atlas, R.M. A Comparison of Latent Heat Fluxes over Global Oceans for Four Flux Products. J. Clim. 2004, 17, 3973–3989. [Google Scholar] [CrossRef] [Scilit]
- Bourras, D. Comparison of Five Satellite-Derived Latent Heat Flux Products to Moored Buoy Data. J. Clim. 2006, 19, 6291–6313. [Google Scholar] [CrossRef] [Scilit]
- Wen, C.; Xue, Y.; Kumar, A.; Behringer, D.; Yu, L. How Do Uncertainties in NCEP R2 and CFSR Surface Fluxes Impact Tropical Ocean Simulations? Clim. Dyn. 2017, 49, 3327–3344. [Google Scholar] [CrossRef] [Scilit]
- Clayson, C.A.; Bogdanoff, A.S. The Effect of Diurnal Sea Surface Temperature Warming on Climatological Air–Sea Fluxes. J. Clim. 2013, 26, 2546–2556. [Google Scholar] [CrossRef] [Scilit]
- Bigorre, S.P.; Plueddemann, A.J. The Annual Cycle of Air-Sea Fluxes in the Northwest Tropical Atlantic. Front. Mar. Sci. 2021, 7, 612842. [Google Scholar] [CrossRef] [Scilit]
- Grist, J.P.; Josey, S.A.; Marsh, R.; Good, S.A.; Coward, A.C.; de Cuevas, B.A.; Alderson, S.G.; New, A.L.; Madec, G. The Roles of Surface Heat Flux and Ocean Heat Transport Convergence in Determining Atlantic Ocean Temperature Variability. Ocean Dyn. 2010, 60, 771–790. [Google Scholar] [CrossRef] [Scilit]
- Hsu, C.-W.; DeMott, C.A.; Branson, M.D.; Reeves Eyre, J.; Zeng, X. Ocean Surface Flux Algorithm Effects on Tropical Indo-Pacific Intraseasonal Precipitation. Geophys. Res. Lett. 2022, 49, e2021GL096968. [Google Scholar] [CrossRef] [Scilit]
- Leyba, I.M.; Solman, S.A.; Saraceno, M. Trends in Sea Surface Temperature and Air–Sea Heat Fluxes over the South Atlantic Ocean. Clim. Dyn. 2019, 53, 4141–4153. [Google Scholar] [CrossRef] [Scilit]
- Martens, B.; Schumacher, D.L.; Wouters, H.; Muñoz-Sabater, J.; Verhoest, N.E.C.; Miralles, D.G. Evaluating the Land-Surface Energy Partitioning in ERA5. Geosci. Model Dev. 2020, 13, 4159–4181. [Google Scholar] [CrossRef] [Scilit]
- Song, X.; Ning, C.; Duan, Y.; Wang, H.; Li, C.; Yang, Y.; Liu, J.; Yu, W. Observed Extreme Air–Sea Heat Flux Variations during Three Tropical Cyclones in the Tropical Southeastern Indian Ocean. J. Clim. 2021, 34, 3683–3705. [Google Scholar] [CrossRef] [Scilit]
- Kroll, C.A.; Jnglin Wills, R.C.; Kornblueh, L.; Niemeier, U.; Schneidereit, A. Parameterization Adaption Needed to Unlock the Benefits of Increased Resolution for the ITCZ in ICON. Atmos. Chem. Phys. 2025, 25, 16915–16943. [Google Scholar] [CrossRef] [Scilit]
- Hummels, R.; Dengler, M.; Brandt, P.; Schlundt, M. Diapycnal Heat Flux and Mixed Layer Heat Budget within the Atlantic Cold Tongue. Clim. Dyn. 2014, 43, 3179–3199. [Google Scholar] [CrossRef] [Scilit]
- Foltz, G.R.; McPhaden, M.J.; Lumpkin, R. A Strong Atlantic Meridional Mode Event in 2009: The Role of Mixed Layer Dynamics. J. Clim. 2012, 25, 363–380. [Google Scholar] [CrossRef] [Scilit]
- Wade, M.; Caniaux, G.; du Penhoat, Y. Variability of the Mixed Layer Heat Budget in the Eastern Equatorial Atlantic during 2005–2007 as Inferred Using Argo Floats. J. Geophys. Res. Oceans 2011, 116, C08006. [Google Scholar] [CrossRef] [Scilit]
- Johns, W.; Speich, S. Tropical Atlantic Observing System (TAOS) Review Report; CLIVAR: Qingdao, China, 2021. [Google Scholar]
- Valdivieso, M.; Haines, K.; Balmaseda, M.; Chang, Y.-S.; Drevillon, M.; Ferry, N.; Fujii, Y.; Köhl, A.; Storto, A.; Toyoda, T.; et al. An Assessment of Air–Sea Heat Fluxes from Ocean and Coupled Reanalyses. Clim. Dyn. 2017, 49, 983–1008. [Google Scholar] [CrossRef] [Scilit]
- Pinker, R.T.; Bentamy, A.; Katsaros, K.B.; Ma, Y.; Li, C. Estimates of Net Heat Fluxes over the Atlantic Ocean. J. Geophys. Res. Oceans 2014, 119, 410–427. [Google Scholar] [CrossRef] [Scilit]
- DeCosmo, J.; Katsaros, K.B.; Smith, S.D.; Anderson, R.J.; Oost, W.A.; Bumke, K.; Chadwick, H. Air-Sea Exchange of Water Vapor and Sensible Heat: The Humidity Exchange Over the Sea (HEXOS) Results. J. Geophys. Res. Oceans 1996, 101, 12001–12016. [Google Scholar] [CrossRef] [Scilit]
- Gulev, S.; Taylor, P.K. Intercomparison and Validation of Ocean–Atmosphere Energy Flux Fields; Final Report of the Joint WCRP/SCOR Working Group on Air–Sea Fluxes (SCOR Working Group 110); World Meteorological Organization: Geneva, Switzerland, 2000. [Google Scholar]
- Zhou, X.; Ray, P.; Barrett, B.S.; Hsu, P.-C. Understanding the Bias in Surface Latent and Sensible Heat Fluxes in Contemporary AGCMs over Tropical Oceans. Clim. Dyn. 2020, 55, 2957–2978. [Google Scholar] [CrossRef] [Scilit]
- Zhou, X.; Ray, P.; Barrett, B.S.; Hsu, P.-C. Systematic Improvement in Simulated Latent and Sensible Heat Fluxes over Tropical Oceans in AMIP6 Models Compared to AMIP5 Models with the Same Resolutions. Atmos. Res. 2022, 274, 106214. [Google Scholar] [CrossRef] [Scilit]
- Clayson, C.A.; Edson, J.B. Diurnal Surface Flux Variability Over Western Boundary Currents. Geophys. Res. Lett. 2019, 46, 9174–9182. [Google Scholar] [CrossRef] [Scilit]
- Brunke, M.A.; Wang, Z.; Zeng, X.; Bosilovich, M.; Shie, C.-L. An Assessment of the Uncertainties in Ocean Surface Turbulent Fluxes in 11 Reanalysis, Satellite-Derived, and Combined Global Datasets. J. Clim. 2011, 24, 5469–5493. [Google Scholar] [CrossRef] [Scilit]
- Fairall, C.W.; Thompson, E.J.; Bariteau, L.; Wick, G.A.; Minnett, P.J.; Szczodrak, M.; Jessup, A.T.; Witte, C. An Updated Treatment of the Oceanic Cool Skin in the COARE Bulk Flux Algorithm. J. Geophys. Res. Oceans 2026, 131, e2025JC023539. [Google Scholar] [CrossRef] [Scilit]
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2026 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.









