An important stage in understanding processes which transfer signals within the climate system, is to identify where signals arise within the system. Also to identify where and if signals propagate in time and dimension (from the depths of the ocean to the upper reaches of the atmosphere, or vice versa).

The light blue arrows in the schematic below show atmospheric inputs to the oceans, the green arrows show potential pathways by which the oceans might provide feedback signals to the atmosphere. Since different mechanisms tend to operate over different spatial and temporal scales, mechanisms for the transport of both heat and moisture are to be examined.

Three types of data can be used to aid in “chasing” long term ocean-atmosphere feedback signals and underlying mechanisms, these are:

• In-situ Measurements

• Re-Analysis data (such as from NCEP, National Centres for Experimental Prediction)

• Coupled Ocean-Atmosphere Model Data

Real, in-situ, data should reveal feedback signals, but such data will have sparse spatial coverage and may not have long records at all levels for both the ocean and the atmosphere. Model data can be generated at any location and for all levels of the ocean and atmosphere. The limitation is that models can only incorporate known mechanisms, so model output may either inherently include these non-linear feedback signals or the models may be missing the multi-decadal oceanic signals and their mechanisms. Reanalysis data is a mixture of real and model data, merged over 40 to 50 years.

The following are preliminary analyses of in-situ data included to illustrate some of the transfer processes shown in the above schematic.

WITHIN the OCEANS

There are very few ocean locations with long, multi-decadal, records and even fewer locations with measurements of subsurface data. Ocean Weather Ship Mike (OWSM), situated off Norway, has the longest, continuous records of subsurface salinity and temperature, and is used here to illustrate where long-term signals may be found within the ocean.

Using the OWSM data, a persistence filter MONACLE(1,x,3,7), where x is a moving-month, was applied to the surface ocean data and to every depth. Each filtered depth series was then correlated to its corresponding surface seasonal series (e.g. season start month 8 surface data is correlated to each season 8 depth in turn).

Resulting season-depth correlation matrix was contoured; X-axis is season number and Y-axis is ocean depth. Colour scales, for the correlation, range from +1 (dark red) through to -1 (dark blue);

with the low correlation values (-0.3 to 0.3) shaded in grey. Left-hand plots show, top, temperature and bottom, density. For the right hand plots, all levels and all seasons are correlated against one surface season that marked by the green circle. For the oceans, data collection tends to vary seasonally, as well as with depth, so the line contours show the number of years used at each level and season. The colour contour plots are terminated at 1500m due to the low number of years having deep, winter data.

The right-hand plots show that long-term spring surface signals (season 4) propagate in time and depth in both the temperature and density fields. A portion of the long-term spring signal is retained later in the year below the surface and may still be found in subsurface waters in the autumn (seasons 8, 9 and 10). Some (negative) contour features, dark blue intrusions (bottom right of top-right plot) are likely to be data artefacts as they follow the line contours indicating low numbers.

AT the AIR-SEA INTERFACE

The following plots show the relationship between atmospheric and oceanic temperatures at the surface level.

A persistence filter MONACLE(1,x,2,7) was applied to the selected gridded 2?x2? SST (Sea Surface Temperature) and matching gridded 2?x2? surface air data. Each SST and air filtered seasonal time series was then correlated against each other, producing a square correlation matrix with the diagonal being the direct SST(x)-AIR(x) relationship. The full seasonal correlation matrix was contoured. To obtain a good match between SST (X-axis) data to air data (Y-axis), SST locations were selected close to land stations that also had long air temperature time series. Numerous data problems exist in generating full coverage and consistent results for many locations. For some SST locations a PCA (principal component analysis) was applied to surrounding locations to extract a single SST series to correlate with land air-temperature series. Colour scales are as above.

The following plots are for the two traditional end-point locations for the NAO (North Atlantic Oscillation); Iceland and the Azores. A diagonal line marks the 1-to-1 correlation (e.g. summer ocean temperature to summer air temperatures). Off Iceland the strongest correlation (red) between air temperatures and ocean temperatures occurs in the summer, off the Azores the relationship is positive for a longer period, peaking in the autumn (9) and in late spring (5).

ICELAND	AZORES

WITHIN the ATMOSPHERE

Daily radiosondes are launched by the weather offices/services of many countries in order to improve weather forecasting by collecting measurements throughout the air column. A persistence filter MONACLE(1,x,3,7) was applied to surface air data and to every upper air level for one location, Sable Island off the east coast of Atlantic Canada, situated approximately 300km from Halifax. The left hand plots shows each filtered upper air seasonal series correlated to its corresponding surface seasonal series and the right hand plot each level and season correlated to one surface seasonal series, that marked by the green dot. The season (X-axis) to level (Y-axis) correlation matrices are contoured with a logarithmic (base 10) Y-axis scale to obtain more evenly spaced grid-points (for contouring). Colour scales are as above.

SABLE ISLAND

The contour plots are for geopotential height, and both plots show positive correlations in autumn throughout the entire atmospheric column.