Relationships between the climate system and the environment are complex, in part due to non-linear processes in the system and to both natural and measurement noise. Climate indices such as the NAO (North Atlantic Oscillation), the AO (Arctic Oscillation), the PDO (Pacific Decadal Oscillation), the SOI (Southern Oscillation Index) and others all address the need to both simplify a complex system and to provide a simple representation of the response to changing physical processes. The next stage of translating such physical processes into either linear or non-linear bio-environmental responses is largely lacking.

So an alternative approach is to reverse the direction of the scientific search and find a non-linear physical signal, a "bulk parameterisation" of ocean processes, to which biology has already responded. Climate studies regard the oceans as holding the long-term memory for the land/atmosphere system. Regarding such memory as a reminder, or feedback, enables us to associate repeated or persistent signals in seasonal air temperatures with long-term bulk ocean signals.

This memory concept is applied via a filtering technique, named MONACLE, as outlined below.

THE APPROACH
Using this conceptual approach to ocean feedback, a simple filter has been developed which provides the potential to explore non-stationary systems via moving serial autocorrelation analysis. Applied to surface air temperature data, MONACLE is a 4-component filter of

In mathematical terms, this autocorrelation filter is a (Pearson) correlation coefficient, R, applied to the same variable (in this case seasonal air temperatures, Air), with itself lagged one year t-1, giving a new time series,

R[Air(ss,sl(x,t)),Air(ss,sl(x,t-n))]t = -y,+y

which can be expressed as a 4-component filter named: -

MONACLE* ( R[n; ss; sl; 2y+1] )

In non-mathematical terms, a filter/index M[1,9,2,11] represents a one-year lag, moving serial autocorrelation of mean air temperatures, calculated for a season starting in calendar month 9 of length 2 months, over an 11 year moving interval, centred at year 5.

In Nature's climate system a varied number of cycles would be interacting in varying and complex combinations with unknown noise (see simulation section). In a global analysis using available environmental series, MONACLE produced the best fits to marine data with geographically-varying filter parameters of season length and start month (see season start months in the world map). This geographical distribution is indicative of an underlying climatic-physical mechanism.

World map shows locations where climatic and long-term marine and aquatic environment records are being examined. Numbers (9, 2 etc) refer to the calendar month (September, February etc) for the start of the season used in each regional application of the MONACLE filter.

SUMMARY
The project seeks to identify information on ocean conditions within atmospheric records. The hypothesis is that such signals represent feedback within the climate system, between the oceans and the atmosphere.

With this approach, applied seasonally, it has been possible to obtain:

• visual comparisons between regional climate indices and long environmental time series (such as fish catch) around the North Atlantic.

• relationships which are not deemed accidental (via bootstrap re-sampling the majority are significant at greater than the 95% confidence limit).

• a regional approach for climate indictors which tracks abrupt changes in the system. The results indicate that "regime-shifts" may commonly occur in the marine environment over 20-70 year time scales.

In order to further this work we need to understand what physical mechanisms could be transferring these feedback signals.