Figure 1.

Figure 2.

Figure 3.

The North Atlantic subarctic gyre, a region of relatively cold surface water, experiences large variability at interannual to interdecadal time scales, both in the atmosphere and in the ocean. A major component in the atmospheric variability is the North Atlantic Oscillation (NAO), which is associated with atmospheric pressure deviations of one sign near Iceland and southern Greenland and the opposite sign near the Azores (40°N). The NAO modulates the dominant westerly winds. These large changes in atmospheric circulation have an important impact on European and eastern North American climates. The ocean surface temperature is variable at decadal time scales, with evidence of propagation of the signal from the southwest to the northeast part of the gyre. The processes of propagation and the nature of the coupling between the ocean and the atmosphere are not yet well defined.
     The surface waters of the Nordic Seas are renewed rather rapidly due to the presence of a deep water formation north of the ridges between Greenland and Scotland. Dense, cold water flows south across the ridge and plunges into the deep North Atlantic, entraining upper ocean water along the way. Winter convection also renews the intermediate waters of the North Atlantic in the central Labrador Sea. These surface waters are replaced mostly by warm and salty waters originating from the Gulf Stream via the North Atlantic Current circulation. To a lesser extent, surface waters are replaced by cold, fresh water from the Arctic, brought to the area by narrow currents flowing mostly along the shelf breaks of Greenland and Labrador and often covered by sea ice in winter. The air-sea heat exchange rapidly (a few years) cools the warm North Atlantic water as it flows to the north and around the subarctic gyre. This results in the entrainment of deeper water rich in nutrients into the surface layer. Thus, the North Atlantic subarctic gyre is a productive biological area where large phytoplankton blooms occur in late spring and summer. These blooms correspond to high chlorophyll concentrations visible by satellite imagery (for example in the recent SeaWifs ocean color images). High production depletes the surface layers of carbon and, therefore, of dissolved carbon dioxide. This, together with a cooling effect on dissolved carbon dioxide, contributes to a disequilibrium in carbon dioxide across the air-sea interface so that the ocean can absorb large amounts of carbon dioxide from the atmosphere.
     Scientific investigation from merchant vessels traveling between Iceland and Newfoundland was initiated in 1985 on ships of the Icelandic company EIMSKIP by Fred Dobson of the Bedford Institute of Oceanography in Halifax, Nova Scotia. The initial program, which lasted until 1993, was designed to investigate the seasonal and interannual variability in heat content of the upper ocean between Newfoundland and Iceland. About four times a year, roughly 24 expendable bathythermograph (XBT) probes are launched by the ship's officers during one crossing. A continuous plankton recorder (CPR) was towed from the same vessel to investigate the distribution of large plankton species and concentration of phytoplankton. The CPR was operated by the Sir Alister Hardy Foundation for Ocean Sciences of Plymouth, UK. This ship route is labeled AX2 by the international World Ocean Circulation Experiment (WOCE). In late 1993, the sampling rate was increased. Since then, additional XBTs have been launched during 22 transits by a scientific observer in the southern region where eddy variability is expected to be larger. The observer also collected water samples to measure salinity, analyze surface nutrients, and study plankton. For a while, surface samples were collected for oxygen isotope and tritium concentrations in the surface water. The ship was equipped in early 1994 with a thermosalinograph (TSG) to continuously measure temperature and salinity. Recently, it has served as a platform for rawinsondes with 263 launched in 1998. Rawinsondes record wind, temperature, and humidity profiles, information that is important for meteorological forecasts in the Northern Atlantic.
      The Icelandic company EIMSKIP owns both the MV Skogafoss and its replacement, the MV Godafoss. The XBT data have been transmitted in real time since early 1994, and hourly averages of the TSG data have been transmitted in real time since the spring of 1996. The data return is somewhat lower than was anticipated, but the overall project has developed as initially envisioned. Thermal sampling in the autumn has been less regular, in particular between late August 1997 and January 1998. We have had to reject more than ten percent of the XBT data because the XBT wire had touched the ship's hull and corrupted the electric circuit. This happens more often during winter storms. The sampling is less dense north of 56°N, with little data north of 60°N. The TSG salinity data are often accurate to within 0.015 practical salinity units (PSU) but were of low quality before March 1996 because of poor weather and sporadic sampling. Figure 1 indicates the route of the vessel between Iceland and Newfoundland by showing the locations of all XBTs taken between December 1993 and January 1999.
     Individual temperature sections indicate some mesoscale variability and variability of the larger gyre scale structure. Figure 2a south of 55°N is quite typical, with cold water trapped in the upper 200 meters close to the shelf edge off Newfoundland and warmer water 600-800km offshore. The warmer water reflects the presence of the North Atlantic current which loops anticyclonically in the southern Labrador Sea (called the Northwest Corner). The North Atlantic Current is also often characterized by higher surface salinity. The temperature minimum near 55°N (900km on Figure 2a) is common and has been associated with an outflow of water from the Labrador Sea. These hydrographic features remain evident in the average section (see Fig. 2b), with isotherms deepening north of 55°N. This deepening is associated with the crossing into the cyclonic circulation of the Irminger Sea (see article by Mork and Pérez-Brunius). The average March winter mixed-layer depth is plotted (dashed line) and indicates a deepening of the mixed layer north of 55°N, where the section crosses the Irminger Current. It might be added that these differences in water masses have been observed to include different biological characteristics, in particular in the assemblage of different copepod species.
      Although sampling has not been sufficient to follow the variability associated with the Northwest Corner of the North Atlantic Current, and it does not resolve the fine scales associated with the Labrador Current, it has been instrumental in resolving the large-scale changes of the temperature structure and surface salinity. The temperature variability is illustrated on a time-latitude plot at 100 meters. Figure 3a indicates a warming of most of the section since early 1994, with most warming occurring in late 1995-early 1996 and late 1997-early 1998. The exception near 52°N corresponds to the Northwest corner where the variability does not seem to correlate with the other parts of the section. What is seen at 100m depth is typical of other depths in the upper ocean with variability that decreases with depth more slowly north of 55°N than south of it. The greatest temperature variability near 56°N-60°N seems to correspond at least in part with changes in the isotherm slope across the Irminger Current. Surface salinity (Figure 3b) also increases north of 55°N, with low salinity before 1997 and higher salinity after. This increase does not seem to occur farther south, even in the Labrador Current, which is the site of a large seasonal variability of the anomalies. This suggests that the waters in the Labrador Current are subject to other factors such as varying fresh water supply from the north. An earlier analysis of observations from 1993 to 1995, which were years of large ice cover, suggests a lower winter near-surface salinity than for later years which were ice-free in the southern part of the Labrador Current. The salinity and temperature variability near 52°N reflects the meandering of the North Atlantic Current where it turns east at the northwest corner.
      The absence of salinity profiles makes it difficult to investigate changes in the density of the upper water column in this region, where the watermass properties evolve on similar time scales to that of the surface. It is, nonetheless, plausible to assume that salinity changes observed at the surface take place in the upper water column as well. Together, these temperature and salinity changes should contribute noticeably to the surface sea level increase of more than 8cm which has been observed with satellite altimetry. Changes in the heat exchanged with the atmosphere have contributed to the upper ocean warming, although available data suggest that this can explain only a third of the warming. The remaining warming has to result from change in the circulation of the ocean. Longer time series and further analysis of additional data available for the subarctic gyre, either from other ships of opportunity (i.e. the Nuka Arctica), or from in situ or satellite observations (for example, temperature and salinity profiles from profiling floats or sea level from altimetry) will be instrumental in better understanding the variability and the role of the ocean on climate variability in the northern North Atlantic sector. The ships of opportunity programs are instrumental in providing the basic sampling from which such analyses can be made.

Acknowledgments
The company EIMSKIP and the ships' masters and crews have been extremely welcoming, the main reason for the success of this sampling program. Cooperation between scientists from different institutions is also key. I am indebted to NOAA for its support, and, in particular, AOML (Warren Krug), Norfolk (Jim Farrington), and my colleagues in France and at LDEO. Logistical help by the Institute for Marine Research in Reykjavik (Iceland) and by Tom Rossby at URI were also very important.