The North Atlantic Oscillation (NAO) refers to the anomaly in the sea-level-pressure (SLP) difference between the sub-tropical high (near the Azores) and the subpolar low (near Iceland). While several operational definitions of the NAO exist, the monthly NAO based on empirical orthogonal functions (EOF) from the National Center for Environmental Information of the National Oceanic and Atmospheric Administration is used here. This definition better represents the larger-scale influence of the SLP patterns above the Northwest Atlantic and tends to be less noisy than the station-based definition. This time series is available online (https://www.ncdc.noaa.gov/teleconnections/nao/, last access: 4 March 2021). Because the monthly NAO index is normalized over the 1950–2000 period, this subindex is not re-normalized here.

The winter NAO, defined as the average of monthly values from December to March, is considered here (Fig. ). A positive phase of the NAO index is usually associated with an intensification of the Icelandic Low and the Azores High SLPs. Except for some years for which the SLP patterns are spatially shifted (e.g. 1999, 2000 and 2018), positive winter NAO years usually favour strong northwesterly winds, cold air and sea surface temperatures, and heavy ice conditions in the Northwest Atlantic . A predominance of a strongly positive winter NAO phase has persisted since 2012, including the record high of +1.61 in 2015 (the record low of -1.47 was in 2010). This recent positive phase of the NAO also corresponded to an intensification of the convection in the Labrador Sea .

The air temperature subindex consists of the average of the annual normalized anomalies at five coastal communities around the Northwest Atlantic: Nuuk (Greenland), Iqaluit (Baffin Island), Cartwright (Labrador), and Bonavista and St. John's in Newfoundland (see Fig. ). These sites were chosen because they are representative of both local (e.g. sea ice formation and SST) and remote (e.g. icebergs originating from Greenland and freshwater fluxes from the Canadian Arctic) effects on the NL shelf. While the data for Nuuk are obtained from the Danish Meteorological Institute , the air temperature data from the Canadian sites are from the second generation of Adjusted and Homogenized Canadian Climate Data (AHCCD) that accounts for shifts in station location and changes in observation methods . When necessary, they are updated using Environment Canada's National Climate Data and Information Archive.

Annual normalized anomalies since 1951 for these five sites are presented in Fig.  under the form of a stacked bar plot. Despite the fact that these sites are spread around the North Atlantic, they often exhibit consistency in the sign of their anomalies, especially in the periods of very cold or warm years. While 1972 appears as the coldest year of this time series, the period of the early 1990s exhibits a sustained period of cold air temperature, with 1991, 1992 and 1993 being respectively the sixth, second and third coldest years of the entire time series. This cold period was followed by the predominance of warmer-than-normal air temperatures at all sites from the mid-1990s to about 2013, with 2010 being the warmest year on record by a large extent. Except for 2015, which was the coldest year since 1993, and 2018 (slightly cold), recent years were close to normal.

Sea ice season duration and maximum cover area are estimated from ice cover products obtained from the Canadian Ice Service (CIS). The methodology is described in and briefly summarized here. The source CIS products consist of weekly geographic information system (GIS) charts covering the east coast for the period 1969–2020 and Hudson Bay for the period 1980–2020. The Hudson Bay charts include coverage of the northern Labrador shelf, and the east coast charts cover the southern Labrador shelf, as well as the Newfoundland shelf (Fig. 1). For each of the three regions, the seasonal maximum area of sea ice was determined, as well as the ice season duration (Fig. ). The former accounts for partial coverage (as opposed to sea ice extent), and the latter is obtained from a spatial average of the number of weeks with sea ice at every pixel, with zeros counted for areas where no ice was present but the 30 year climatology shows some. The normalized anomalies of these six time series (duration and area over three regions) are averaged into a single index presented in a stacked bar plot (Fig. d). Each stacked colour in this plot represents its respective contribution to the average. The numerical values of this sea ice subindex are reported in a colour-coded scorecard at the bottom of this panel. Negative anomalies, indicative of warmer conditions, have been coloured red and positive anomalies blue. This time series corresponds to the sea ice contribution to the NL climate index. The periods of maximum sea ice cover and season duration are found in the early 1970s, mid-1980s and early 1990s. Since the early 1990s the severity of the sea ice season has gradually decreased, reaching the lowest values in 2011 and 2010. With the exception of a rebound to near-normal values in 2014–2016, sea ice conditions have been weak over recent years.

The number of icebergs drifting south of 48∘ N in the Northwest Atlantic (see dashed cyan box in Fig. ) has been monitored by the International Ice Patrol (IIP) of the US Coast Guard since 1900 . These icebergs mostly originate from western Greenland and may carry with them important amounts of freshwater . The entire time series is presented in Fig. . The 121-year average annual number is 495, and the 1991–2020 average is 771. An iceberg count above 1500 has been observed in some years, including in 2014, 2019, and between the early 1980s and mid-1990s. The all-time record of 2202 was reached in 1984. Only 2 years (1966 and 2006) in the 121-year time series reported no icebergs south of 48∘ N. Years with low iceberg numbers on the Grand Banks generally correspond to higher than normal air temperatures, lighter than normal sea ice conditions and warmer than normal ocean temperatures on the NL shelf. The normalized anomalies of this time series are provided in the form of a scorecard below the main panel of Fig. . This time series, starting from 1951, corresponds to the icebergs contribution to the NLCI.

Sea surface temperatures (SSTs) used here are a blend of data from Pathfinder version 5.3 (1982–2020), the Maurice Lamontagne Institute (1985–2013) and the Bedford Institute of Oceanography (1997–2020). Monthly anomalies are computed as the average of available daily anomalies at the pixel level within each geographical region, chosen to be the Northwest Atlantic Fisheries Organization (NAFO) Divisions of Fig. , except that here they are cropped at the shelf break. Details of the processing are in with the extended spatial coverage as in .

Figure  presents annual normalized anomalies averaged over the ice-free season, in which the contribution of each region is weighted according to its open-water area. The ice-free season varies from as short as June to September in NAFO Division 2G to as long as March to November in NAFO Division 3P (seasonal information in the legend of Fig. ). This time series corresponds to the SST contribution to the NLCI (normalized anomalies are colour coded at the bottom of the figure). This figure shows the colder than average conditions that prevailed in the early 1990s, with 1991 and 1992 being the coldest years of this time series. This period was followed by a predominance of warmer than average conditions that lasted until about 2014. In recent years, the period 2015–2019 (except for 2016) was colder than normal (defined as SST̃<-0.5SD or blue colours at the bottom of Fig. ). The year 2020 was, however, back to above-normal SST for the first time since 2014.

Station 27 (47∘32.8′ N, 52∘35.2′ W) is located in the Avalon Channel just outside St. John's harbour, NL (Fig. ). It is one of the longest hydrographic time series in Canada with frequent (near-monthly basis) conductivity–temperature–depth (CTD) observations since 1946. Station 27 was integrated into DFO's AZMP in 1999. In addition to sampling during these traditional hydrographic surveys, this station has been seasonally equipped with an automatic CTD profiling system installed on a surface buoy (type Viking) since 2017 seefor further information.

Station occupations were first combined into monthly averaged temperature (T) and salinity (S) profiles from which the climatological annual cycle was extracted (Fig. ). This figure shows the seasonal warming of the top layer (∼20 m), with temperature peaking in August before being mixed during the fall (Fig. a). Also visible in the temperature field is the cold intermediate layer (CIL), a prominent feature of the NL ecosystem . The CIL is defined here as the water below 0 ∘C and delineated with a thick black contour in Fig. a. This layer originates in the winter as a cold surface layer which becomes isolated from the surface after the appearance of a seasonally heated surface layer during the spring (April–May). The CIL thus remains below the surface throughout most of the year, while its top boundary slowly deepens from about 50 to 100 m as the heat from the surface layer penetrates deeper into the water column. While in deeper areas of the NL shelf a third warmer layer is present beneath the CIL, at Station 27 the CIL generally extends down to the bottom (∼176 m). The summer CIL core temperature is defined as the minimum temperature of the monthly averaged profile for June, July and August (see below).

The surface salinity at Station 27 is generally lowest (S<31) between early September and mid-October (Fig. b). These low near-surface salinities, generally from early summer to late fall, are prominent features of the salinity cycle on the Newfoundland shelf and are largely due to the melting of coastal sea ice upstream and carried over by the Labrador coastal current. Below the surface, the salinity at any depth increases in the spring and peaks in the summer before decreasing in the fall as a consequence of vertical mixing.

The data from Station 27, including the recent greater coverage obtained from Viking buoy automatic casts, contribute to three subindices of the NL climate index: vertically averaged T and S and the CIL core temperature (Fig. ). The focus here is on the period starting in 1951 for which the seasonal coverage includes at least 8 months per year. Averages from 1980 and 1981 are excluded because of insufficient seasonal coverage (4 and 7 months, respectively). Data from 2020 have also been excluded because there were no station occupations during the first 6 months of the year due to the Covid-19 pandemic (first occupation occurred on 14 July, the latest start since 1946). In order to account for possible changes in seasonal coverage, the annual anomalies have been calculated as the average of monthly normalized anomalies.

The vertically averaged (0–176 m) temperature exhibits decadal-like cycles (Fig. a). The period between the mid-1980s and mid-1990s struck as the coldest decade of the last 70 years. It was followed by a gradual warming trend that lasted 2 decades and peaked in 2011, the warmest year of this time series. The period from the mid-1960s and the early 1970s was also marked by sustained warmer than normal temperatures. The end of this period coincided with the freshest anomaly on record at Station 27 observed in 1970 (Fig. b), an event coinciding with the Great Salinity Anomaly in the North Atlantic . Similarly, the recent warmer than normal period (2010–2013) was followed by the second freshest anomaly on record observed in 2018. This most recent fresh anomaly on the NL shelf coincides with a large-scale salinity anomaly in the subpolar gyre , but no clear causal relationship has been established between the two. The saltiest anomaly at Station 27 (1990) occurred during the cold period of the late-1980s and early 1990s.

The CIL subindex of Station 27 is also presented in Fig. c. This time series is the average normalized anomalies of the summer (June–August) CIL core temperature (minimum temperature of the monthly mean profile). The striking feature in this figure is the anomalously warm CIL anomaly present from the early 1960s to the mid-1970s. After the prevalence of a warm CIL in the early 2010s (with 2010 and 2011 being the warmest years since the 1970s), there has been a recent period of a return to near-normal conditions (roughly 2014–2017) that receded in 2018 and 2019. The CIL subindex of Station 27 has not been calculated for 2020.

As mentioned above, the CIL is a prominent feature of the NL shelf. It is found almost everywhere in the subsurface during the summer. In order to highlight the influence of the CIL on the shelf as a whole, a proxy for its volume is established using the area (in km2) of water below 0 ∘C along Seal Island (SI), Bonavista Bay (BB) and Flemish Cap (FC) hydrographic sections (Fig. ). These sections were selected because they have been systematically surveyed since the early 1950s before being formerly standardized by the International Commission for the Northwest Atlantic Fisheries in 1976 . Since 1999, these hydrographic sections have been monitored by DFO as part of the AZMP.

Figure  shows the summer temperature along section SI during 2 extreme years, the warm 1965 and the cold 1990. The 1991–2020 climatology is also presented. For each sampled summer, the CIL area (e.g. the area of the section delimited by the thick black contour in Fig. ) was calculated. For example, in 1990, the area of the CIL was 26.9 km2, while it was only 1.5 km2 in 1965. This shows the amplitude of the interannual variability of the CIL and its potential influence on the ecosystem. In 1990, most of the sea floor along section SI was in direct contact with the CIL and water below -1 ∘C. In 1965, when the CIL was small and fragmented, none of the sea floor was in contact with the CIL, and the bottom conditions were in consequence several degrees warmer than in 1990.

The normalized anomalies of the CIL area for sections SI, BB and FC are presented in Fig. . This figure highlights again the warmer conditions (negative anomalies) of the 1960s and the colder conditions of the mid-1980s and early 1990s. The average of the normalized anomalies are shown in a scorecard at the bottom of the main panel and correspond to the CIL area contribution to the NLCI.

Canada has been conducting random stratified trawl surveys in NAFO sub-areas 2 and 3 of the NL shelf since 1971 . Since 1980, temperature (and salinity since 1990) measurements have been available for most of these fishing sets thanks to trawl-mounted CTD observations. The scientific trawl surveys target NAFO Subdivision 3Ps (south coast of Newfoundland) and Division 3LNO (Grand Banks) during the spring surveys and Divisions 2H (northern Labrador), 2J (southern Labrador), 3K (eastern Newfoundland) and 3LNO (Grand Banks) during the fall (see map Fig. ). These surveys, combined with other available data from multiple sources (see below), are used to provide large-spatial-scale oceanographic information of the NL shelf, including information on the bottom habitat parameters of numerous commercial species e.g..

The method used to derive the bottom temperature was introduced by and is briefly summarized here. First, all available annual profiles of temperature (scientific trawl surveys, AZMP hydrographic campaigns, surveys from other DFO regions, international oceanographic campaigns, expendable bathythermographs, Argo program, etc.) are vertically averaged in 5m bins and vertically interpolated to fill missing bins. Then, for each season (April–June for spring and September–December for fall), all data are averaged on a regular 0.1∘ × 0.1∘ (latitudinal × longitudinal) grid to obtain one seasonal profile per grid cell. Since this grid has missing data in many cells, each depth level is horizontally linearly interpolated. For each grid point deeper than 10 m, the bottom observation is considered as the data at the closest depth to the GEBCO 2014 Grid bathymetry (version 20141103) to a maximum 50 m difference. In order to only focus on the shelf, observations deeper than 1000 m are clipped. This method is applied for all years between 1980 and 2020 from which the 1991–2020 climatology is also derived. In order to match the scientific surveys schedule, the normalized anomalies of bottom temperature are calculated separately for NAFO Divisions 3Ps and 3LNO in the spring and 2H, 2J, 3K and 3LNO in the fall. The time series of the bottom temperature for both seasons is presented in a stacked bar plot in Fig. . This figure shows the cold phase from the mid-1980s to the mid-1990s, followed by a warmer phase that peaked in 2011. A scorecard at the bottom of this figure presents the mean normalized anomalies. The latter corresponds to the bottom temperature contribution to the NLCI.