The term reasonableness check is used here to ensure that the gridded data set meets the expected range, type, or value based on its individual measurements. We want to examine how reasonably the NAPv1.0 can represent the CTD profiles closer than ∼ 6 km from each grid point. Although the data sets are not independent, this comparison shows how well the interpolation captured the real data set's magnitude, structure, and spatial variability.

The NAPv1.0 shows a good agreement with the CTD profiles for the hydrographic parameters. Approximately 62 %, 70 %, 74 %, and 65 % of the data points fall within the difference range of ±0.2 ∘C for summer, autumn, winter, and spring, respectively (Fig. 5a). The higher percentage of agreement in autumn and winter is due to the reduced number of CTD profiles in those seasons, which become more influential in the interpolation in those cases. The higher Θ difference, as expected, occurs in the upper ocean due to interannual variability (Mendes et al., 2013; Gonçalves Araújo et al., 2015) that is smoothed out in the interpolation process (Fig. S3a–b in the Supplement).

A good agreement between the NAPv1.0 and the CTD profiles is also observed for SA. Approximately 67 %, 74 %, 84 %, and 75 % of the data points fall within the difference range of ±0.04 g kg-1 for summer, autumn, winter, and spring, respectively (Fig. 5b). The SA differences over the water column are also higher in the upper levels due to the interannual variability and smaller towards the deep ocean (Fig. S3c–d in the Supplement), where the interannual variability tends to be relatively smaller (Dotto et al., 2016; Ruiz Barlett et al., 2018).

With regards to DO, approximately 65 %, 87 %, 87 %, and 80 % of the data points fall within the difference range of ±0.2 mL L-1 for summer, autumn, winter, and spring, respectively (Fig. 5c), which clearly suggests the influence of less data affecting the resemblance between the gridded product and the CTD stations. The NAPv1.0 shows relatively higher differences related to DO data, which is a non-conservative property. The largest differences are observed in the upper ocean, but also along the water column where it is not uncommon to observe differences of up to ±0.5 mL L-1 (Fig. S3e–f in the Supplement). The histograms previously presented (Fig. 3 and Figs.  S1 and S2 in the Supplement) show evidence of the weaker representation of DO compared to the other properties in autumn, winter, and spring.

The NAPv1.0 represents qualitatively well the surface thermal dichotomy regime of the NAP waters: a warm variety with Bellingshausen Sea and ACC influence to the west and a cold Weddell Sea-sourced water to the east (Fig. 6a). A surface thermal front that splits these two regimes near the Antarctic Peninsula, likely depicting the Peninsula Front (López et al., 1999; Sangrà et al., 2011), is more visible in summer (Fig. 6a–d). On the surface, the coastal Weddell Sea-sourced waters are saltier and denser compared to the warm waters to the west of the NAP (Figs. 6e–h and 7a–d). Thus, the Peninsula Front seems to be baroclinic (Sangrà et al., 2011). The DO is relatively supersaturated everywhere, with higher concentrations found close to the James Ross Island, the Gerlache Strait, and the northern end of the WAP (Fig. 7e–h). These regions are generally associated with high primary productivity, which could explain the relatively higher values of DO (Mendes et al., 2012; Detoni et al., 2015; Costa et al., 2020). Seasonally, the NAPv1.0 represents a general decrease in Θ, an increase in SA due to sea ice formation in the region, and consequently a γn gain from autumn to winter (Figs. 6 and 7a–d). In spring, Θ increases and SA and γn decrease slightly due to sea ice melting in the upper layers (Figs. 6 and 7a–d). The DO field is less represented in seasons other than summer because of less availability of in situ data (Fig. 7e–h).

In deeper layers, the dichotomy regime of the NAP is still evident, reflecting the presence of warm and salty waters associated with the CDW to the west of the NAP and the Warm Deep Water (WDW; a warm, salty, and poorly oxygenated intermediate water mass presented in the Weddell Sea and derived from the mixing between CDW and Winter Water) within the offshore zone of the Weddell Sea and Powell Basin (Figs. S4–S7 in the Supplement). Intrusions of mCDW are observed in the eastern Bellingshausen Sea through deep channels, reaching the Gerlache Strait and the western basin of the Bransfield Strait (Figs. S4a–S7a in the Supplement). After entering the Bransfield Strait, the mCDW warm signal follows the slope of the South Shetland Islands as the narrow Bransfield Current (Niiler et al., 1991; López et al., 1999; Zhou et al., 2006; Sangrà et al., 2011). However, given the lengthscales chosen, this current is not well represented in the NAPv1.0. Although the routes of mCDW inflow are relatively well known, few studies focused on the periodicity of these inflows into the NAP (e.g. Ruiz Barlett et al., 2018) or on the fast response of these inflows associated with the wind forcing (e.g. McKee and Martinson, 2020a). WDW is another water mass that composes the deep waters of the Bransfield Strait (Gordon et al., 2000). Although the NAPv1.0 suggests the existence of an intrusion of this water mass into the eastern basin at ∼ 61.5∘ S and 55∘ W (Fig. S4e in the Supplement), we still need to increase our comprehension of the contribution to the water mass mixture and the access routes of WDW into the Bransfield Strait via Powell Basin to resolve the local circulation in more detail (e.g. Thompson et al., 2009; Azaneu et al., 2017). The comprehension of these mechanisms is important to better resolve the heat, salt, and carbon fluxes into the NAP.

In all seasons, the Bransfield Strait is colder, fresher, denser, and oxygen-richer compared to the adjacent regions in deeper basins (Figs. S4–S7 in the Supplement) and is associated with the higher contribution of HSSW (Gordon et al., 2000; Dotto et al., 2016). Moreover, the water masses in the Bransfield Strait central basin are also slightly denser than in the eastern basin (e.g. Fig. S4g in the Supplement), suggesting a different origin of these waters (Gordon et al., 2000; van Caspel et al., 2018). The NAPv1.0 represents quite well a narrow cold and oxygen saturated path along the continental slope of the Weddell Sea at 500 m depth (Fig. S4a–d in the Supplement). At the Powell Basin, these water masses follow the bathymetry and enter the Bransfield Strait, thus connecting the Weddell Sea and the NAP (Heywood et al., 2004). Although slightly less visible than in summer, the narrow thermal signal along the continental slope of the Weddell Sea is observed entering the Bransfield Strait in autumn and spring (Figs. S5a–S7a in the Supplement). In winter, likely due to limited in situ data, this signal is not totally clear (Fig. S6 in the Supplement). These dense waters that flood the central NAP deep levels seem to follow the shortcuts of the canyons that cut across the Antarctic Peninsula continental shelf in the Bransfield Strait (Figs. S4–S7 in the Supplement). The higher DO in that deep region indicates a fast route of Weddell Sea shelf waters and hence a relatively small residence time (van Caspel et al., 2015, 2018). At deeper levels, the bathymetric constraints restrict the connection of the shelf waters from the Weddell Sea, and the dense waters that sink into the Bransfield Strait retain their quasi-pure properties (Figs. S4e–S7e in the Supplement; Gordon et al., 2000; Dotto et al., 2016). The horizontal maps clearly show the importance of the Bransfield Strait in trapping the shelf waters of the Weddell Sea, making it an in situ laboratory to study temporal changes of these water masses. Given the spatial scales and the smoothing following the gridding procedure, Bransfield Strait mesoscale and submesoscale eddies (Sangrà et al., 2011) are not observed in the NAPv1.0.

To exemplify the representation of the water column of NAPv1.0, we selected six sections crossing different parts of the NAP and adjacent regions (Fig. 1a). To the southwest of the Bransfield Strait, along the section between Livingston Island and the Antarctic Peninsula, relatively warmer conditions are observed from the upper ocean to the bottom layers (Fig. 8a). An intrusion of mCDW is observed between 62.7–62.8∘ S, where the isotherms are sharply tilted towards the South Shetland Islands (Fig. 8a), SA is above 34.72 g kg-1 (Fig. 8b), and the DO is relatively low <6 mL L-1 (Fig. 8d). The tilt of the isotherms is a signature of the Bransfield Front, which in turn is associated with the mid-depth Bransfield Current (Niiler et al., 1991; López et al., 1999; Sangrà et al., 2011). In the section between King George Island and the Antarctic Peninsula the warm, salty, and deoxygenated mCDW is still present; however, it is located in slightly shallower levels (Fig. 8e–h). Towards the bottom, Θ decreases to the lowest values in the region, characterizing the Bransfield Strait central basin. In this section, the surface Peninsula Front is visible and associated with the cold regime from the western Weddell Sea continental shelf. In both sections, the dome-like shape of the isopycnals supports the cyclonic gyre presence in the region (Fig. 8c, g; Zhou et al., 2002; Sangrà et al., 2011; Collares et al., 2018). The circulation, mixing rates, and turbulent processes are poorly explored in the NAP, thus there is a need for further studies on these processes to better comprehend the local water mass transformation in the different depth layers (e.g. Brearley et al., 2017) as it ultimately impacts the physicochemical properties and biology of the region. The NAPv1.0 also distinguishes well the difference between the central and eastern basins of the Bransfield Strait (Fig. 8i–l). Colder, denser, and more oxygenated deep waters are observed in the former basin, whereas the latter is relatively warmer, lighter, and less oxygenated. This dichotomy occurs because the central basin has restricted connections with the adjacent regions due to bathymetric constrains and it also receives more contribution from HSSW than the eastern basin (Gordon et al., 2000; Dotto et al., 2016). Within the Bransfield Strait, more mCDW accesses the region in autumn coming from Livingston Island and reaching the eastern basin, increasing temperature and salinity in deep waters in all basins (Fig. S8 in the Supplement). Moreover, the Peninsula Front migrates toward the South Shetland Islands due to higher inflow of Weddell-sourced waters (Fig. S8e in the Supplement) until it vanishes in winter due to surface cooling (Fig. S9d in the Supplement). In spring, the hydrographic properties of the Bransfield Strait resemble more those of summer in subsurface levels (Fig. S10 in the Supplement). Although the winter DO suggests the presence of deep convection in the eastern basin, the distribution of the isopycnals refutes this idea, which reinforces the conclusion of Whitworth et al. (1994) that it is unlikely that deep convection occurs in Bransfield Strait. Due to lack of data from autumn to spring, we will focus on the summer conditions.

We now evaluate the NAPv1.0 in sections crosscutting the WAP at the WOCE-SR4 West (hereafter, SR4) and along the Scotia Arc – a route of Antarctic Bottom Water export (Fig. 1a; Franco et al., 2007). At the WAP (Fig. 9a–d), the NAPv1.0 shows high values of Θ and SA and lower values of DO, associated with CDW, intruding onto the continental shelf. Θ as high as 1 ∘C, sourced from upper CDW, floods the continental shelf in those regions (Moffat and Meredith, 2018) and are a heat source for the glacier retreating (Cook et al., 2016). Although the salinity values are high, they are not as high as the lower branch of CDW offshore (Fig. 9b). The isopycnals offshore tilt upwards near the continental slope, suggesting upwelling of CDW (additionally, several hypotheses have been developed to explain the origin of these warm waters, such as eddies shed by the ACC, current deflection by the cross-shelf-cutting canyons, or advections of the ACC onto the continental shelf; e.g. Prézelin et al., 2000; Martinson et al., 2008; Martinson and McKee, 2012; Couto et al., 2017; McKee and Martinson, 2020b). On the Weddell Sea side, the NAPv1.0 shows the main water masses of the region (Fig. 9e–h), including the Winter Water at the subsurface, the WDW with the largest (lowest) values of Θ and SA (DO), the Weddell Sea Deep Water, and the Weddell Sea Bottom Water, the latter of which is restricted to below 4000 m (Farhbach et al., 2004; Kerr et al., 2018c). Another important feature observed in the climatology is the downslope flow of the cold, fresh, dense, and recently ventilated waters at the continental slope (van Caspel et al., 2015). At the shelf break, the isotherms tilt suggesting the presence of the Antarctic Slope Front (Farhbach et al., 2004; Heywood et al., 2004). Once the Weddell Sea Deep Waters enters the Powell Basin, a fraction is exported across the Scotia Arc (Fig. 1b; Franco et al., 2007). This region is transitional between the Weddell Sea and the northward parts of the Southern Ocean and generally has a warm and weak density structure. The export of dense waters may occur at Philip Passage, where the upper parcel of Weddell Sea Deep Water can overflow the bathymetric constraint, as shown in Fig. 9i–l. On seasonal time scales, the NAPv1.0 represents the natural intraseasonal variability in the upper ocean as well as changes in the isopycnals tilting, suggesting small changes in the circulation across the sections. However, mainly in winter, the data is spatially sparse. For all seasons except summer, DO distribution is flawed (Figs. S11–S13 in the Supplement).

Most climatological products are built to represent the global ocean and its large-scale basins, such as the Southern Ocean. The NAP, on the other hand, is a relatively small region but highly dynamical and ecologically important to connect regional environments (Kerr et al., 2018a). In this context, we now compare the NAPv1.0 with some other available gridded products, widely used by the ocean community, to show how effective it is to have regional climatologies, especially in areas of intense spatial and temporal variability. For these comparisons, we chose to use the WOA and CARS climatologies and the SOSE state estimate previously described in Sect. 2. Although SOSE is based on a model, and one could argue that comparing it with a gridded product is not a fair comparison, we decided to used it because the data coverage of the Southern Ocean is relatively poor in space and time, and many studies have used SOSE as a benchmark to evaluate other ocean model outputs (e.g. Spence et al., 2017; Russell et al., 2018). In addition, SOSE has been used as initial conditions for simulations within the Bransfield Strait (e.g. Zhou et al., 2020). Despite the limitations of state estimate and ocean reanalysis (e.g. Mazloff et al., 2010; Azaneu et al., 2014; Dotto et al., 2014; Aguiar et al., 2017; Verdy and Mazloff, 2017), combining model and observations, such as in SOSE, leads to great improvements on the comprehension of the Southern Ocean dynamics (e.g. van Sebille et al., 2013; Abernathey et al., 2016; Rodriguez et al., 2016; Tamsitt et al., 2017).

The vertical water mass structure of the NAP and adjacent regions is well represented by the NAPv1.0 (Fig. 10a–b). The density distribution and thermohaline ranges agree well with the CTD casts (closer than 6 km from the grid points), as previously shown (Sect. 4.1). The summer-averaged WOA is colder and saltier at the surface, but the intermediate to deep ocean seems to have thermohaline range and density structure similar to observations (Fig. 10c–d). The summer-estimated CARS seems to represent well the NAP region, despite being a product with coarser resolution. It has the range of Θ and SA in agreement with the observations throughout the water column (Fig. 10e). A good agreement is also observed for the dense waters, where CARS has similar levels of γn compared to the measured data (Fig. 10f). Contrary to WOA, the NAPv1.0 and the CARS climatology do not show supercooled waters centred at SA 34.6–34.7 g kg-1 (and Θ ∼-2 ∘C), which is in agreement with the observations (Fig. 10b, d, f). The 2008–2018 summer averaged SOSE represents the upper ocean considerably fresher, whereas the water masses with γn<27.6 kg m-3 are colder than the observations (Fig. 10g). The dense water masses (γn>28.27 kg m-3) are not well represented in SOSE for the region analysed, being considerably warmer and saltier than the observations (Fig. 10h). However, due to higher temperatures, lighter waters flood the bottom layers of SOSE. The similarity of the NAPv1.0 and the WOA and CARS climatologies to the observations is because these data sets are totally fed by in situ data, whereas SOSE is based on a model run fed by in situ data.

Now, to present the water mass structure, we show the representation of these products along the several sections at the NAP and adjacent regions. We start showing the representation of the climatologies in the central area of the NAP, i.e. the Bransfield Strait, because it has a transitional regime between cold and warm conditions and it is highly dynamical. WOA unveils the upper ocean considerably colder, and despite having similar ranges of Θ as the NAPv1.0 in the mid- to deep ocean, the distinction between the Bransfield Strait's central and eastern basins is not well represented (Fig. 11a and Fig. S14a in the Supplement). The SA distribution in WOA is slightly saltier than the NAPv1.0 and it does not show a clear spatial variability (Fig. 11b). Consequently, the WOA is generally denser than NAPv1.0, mainly in the eastern basin (Fig. 11c and Fig. S15a in the Supplement). The Θ climatology in CARS is in closer agreement with the NAPv1.0, both in the upper ocean and deeper layers (Fig. 11d). Although the deep Θ is not as low as the NAPv1.0 (Fig. S14g in the Supplement), a clear thermal distinction is observed between the central and eastern basin of the Bransfield Strait. On the other hand, the SA field is slightly saltier in CARS than the NAPv1.0 and the spatial distribution is less variable, as observed by the flatness of the isohalines (Fig. 11e). In general, CARS is slightly denser than NAPv1.0 for the Bransfield Strait (Fig. S15g in the Supplement). Interestingly, however, is the representation of DO in CARS, which resembles the NAPv1.0 in terms of magnitude and spatial distribution (Fig. 11g). In both products, the DO is higher in the central basin (although slightly overestimated in CARS climatology). SOSE is considerably warmer in most of the water column in both basins (Fig. 11h and Fig. S14m in the Supplement) when compared to the NAPv1.0 (Fig. 8i). As discussed previously, SOSE shows higher salinity than the NAPv1.0 in both basins below ∼ 250 m depth (Fig. 11i). However, the deep density structure in the Bransfield Strait is lighter than NAPv1.0 because of the higher Θ values in SOSE (Fig. 11j and Fig. S15m in the Supplement). Regarding DO, SOSE is oversaturated in the upper layer and highly deoxygenated below ∼ 100 m (Fig. 11k) compared to NAPv1.0 (Fig. 8l).

Regarding the other sections in the Bransfield Strait (Fig. 1a), no product captured the Peninsula and Bransfield fronts as well as NAPv1.0, likely due to their coarser resolution (Fig. 12a, b, f, g, k, l). The mid-depth inflow of mCDW is better represented in CARS, where it is restricted to the South Shetland Islands side. In WOA and SOSE, the mCDW seems to be present in the whole meridional extent of the Bransfield Strait (Fig. 12a, b, f, g, k, l), which creates a band of higher temperatures in levels deeper than 200 m in both products compared to NAPv1.0 (Fig. S14 in the Supplement). As previously noticed, the temperature range of CARS is in better agreement with our regional climatology (Fig. S14 in the Supplement). In terms of γn, WOA and SOSE are considerably lighter than NAPv1.0 below ∼ 300 m (Figs. 13a, b, f, g, k, l and Fig. S15 in the Supplement) because of their higher temperature (Fig. S14 in the Supplement). For instance, the isopycnals of 28.27 and 28.4 kg m-3 are found deeper in the Livingston–Antarctic Peninsula and King George–Antarctic Peninsula sections in the WOA (Fig. 13a–b). In SOSE, the 28.27 and 28.4 kg m-3 isopycnals are respectively absent in the former and latter sections (Fig. 13k–l). CARS is more in agreement with NAPv1.0 in terms of magnitude and distribution of the isopycnals within the Bransfield Strait despite its larger resolution (Fig. 13f–g). However, in general, the dome-like structure of the isopycnals in the Bransfield Strait, associated with the cyclonic circulation, is not apparent in the other products analysed, which suggests a misrepresentation of the circulation pattern of the strait (Fig. 13).

In the adjacent regions of the NAP, where the oceanographic regime is less dynamical, WOA, CARS, and SOSE represent the vertical structure of the water masses relatively better than within the central NAP (Figs. 12 and 13). All of them successfully represent the core of CDW offshore, the upward tilting of the isotherms and isopycnals towards the continental slope, and the CDW accessing the WAP continental shelf (Figs. 12 and 13). In the Weddell Sea, all products represent the core of the WDW and the thermal and density distinction of Weddell Sea Deep Water and Weddell Sea Bottom Water. However, only CARS can represent the downslope flow of cold and dense waters at the continental slope, in agreement with NAPv1.0 and observations (Figs. 12 and 13). Along the Scotia Arc, despite all products roughly agreeing regarding the distribution of the warmer water parcel (Fig. 12), only CARS and SOSE represent the export of lighter Weddell Sea Deep Water through the Philip Passage (Fig. 13), in agreement with NAPv1.0 (Fig. 9k) and observations (Franco et al., 2007).

One must keep in mind that WOA and CARS are global climatologies, and their relatively low-resolution (i.e. 1/4 and 1/2∘ respectively) does not properly represent some local and regional environments of the Southern Ocean like in the NAP (for instance, the distance between the South Shetland Islands and the Antarctic Peninsula is slightly more than 100 km). Despite having lower resolution, CARS seemed to better represent the hydrographic properties in the Bransfield Strait than WOA. Although beyond the scope of this work, this discrepancy could be associated with the interpolation methods that are distinct for each climatology. On the other hand, SOSE is a state estimate where a model is constrained by observations. Although its output products are good to study and represent large-scale processes, SOSE did not seem fit to be used in regional seas, at least in the NAP, where complex dynamics and contrasting regimes set the local water mass structure and variability. In this sense, simulations of regional models could be significantly impacted by the use of those large-scale climatologies and reanalysis products.

In summary, the NAPv1.0 is robust in the representation of many described dynamic features of the region and, to our knowledge, its hydrographic fields closely follow the observations better than any other available product. Hence, the NAPv1.0 can be used to represent the mean state conditions as well as to feed ocean regional models of the highly transitional and dynamic NAP environments. However, a few caveats can be identified in the NAPv1.0.

In situ data are spatially and temporally sparse in the Southern Ocean. The Argo and MEOP data sets could have the potential to overcome this uneven sampling; however, a few regions are still undersampled, such as the western Weddell Sea (Fig. 2). To avoid creating larger errors (and unreal hydrographic values) in the interpolation procedure, we decided to avoid those regions by masking the final products.

CTD data are still the only data that can be used to measure deeper (>2000 m) levels in the region. Considering that most of the hydrographic cruises were conducted in summer (Fig. 3), deeper regions are not well represented in other seasons. In those cases, most of the climatology is developed to the upper ocean (<2000 m) given the measurement restrictions of the Argo and MEOP data sets. To avoid creating larger errors (and unreal hydrographic values) in the interpolation procedure, we decided not to interpolate where data are not available. Because of this choice, autumn, winter, and spring showed a lower representativeness than summer. This might be solved in the near future with the development of deep-Argo floats (Roemmich et al., 2019) and could potentially sample many deep parts of the adjacent regions of the NAP (e.g. Weddell Sea, Bellingshausen Sea, and Drake Passage).

Another limitation of our climatology regards the DO field. As most of the CTD measurements were conducted in summer, the greatest representation of this field is restricted to that season. In addition, a considerable number of DO measurements were conducted by the GOAL, restricted to the Bransfield and Gerlache straits during summer (Fig. 3c). The development of biogeochemical Argo floats could increase the amount of available DO data in the future for different seasons (Roemmich et al., 2019), which would improve the representation of the different climatologies.

The grid scale (∼ 10 km) is similar or even higher than the first baroclinic Rossby radius of the region (Chelton et al., 1998), which limits the representation of some mesoscale and submesoscale features. In addition, the smoothing radius chosen in the objective interpolation and the mean state associated with the interpolation filter out most of the small dynamic features, such as eddies associated with the Peninsula Front (e.g. Sangrà et al., 2011) or at the WAP continental slope (Martinson and McKee, 2012).

Finally, the NAPv1.0 product brings a novel tool and opens plenty of possible future applications to better understand the regional circulation and hydrography along the NAP, a recognized marine climate hotspot (Kerr et al., 2018a). This climatology will be useful not only to observers and modellers, but also to ecologists and anyone interested in the oceanographic conditions and behaviour of the NAP region.