Climatic trends

A 10,000-year long simulation has been made with the CSIRO Mark 2 coupled global atmospheric-oceanic model for present climatic conditions. The annual mean output from the model has been used to calculate global distributions of climatic trends. These trends were derived by linear regression using a least squares fit to a given climatic time series for a selected trend duration. Typically, this information cannot be obtained from the limited observational record, hence the simulation provides a documentation of many climatic trend characteristics not previously available. A brief examination of observed climatic trends is given to demonstrate the viability of the trend analysis. This is followed by a range of global trend distributions for various climatic variables and trend durations. At any one time only relatively small regions of the globe have trends significant at the 95% level. Markedly different trend patterns occur for a given trend duration computed for different times within the simulation. Decadal and multi-decadal trend patterns revealed consistent relationships for El Niño/Southern Oscillation (ENSO)-related climatic variables. It was found that within a given duration trend, noticeable shorter term counter-trends can exist, with the latter being much stronger. In general, a strong trend is indicative of a short duration, thus highlighting the danger of extrapolating such trends. Examination of time series of climatic trends emphasised the dominance of decadal variability and the essential residual nature of, especially longer term, trends. Rainfall trends over Australia are used to indicate the almost continent-wide changes that can occur in trend patterns within a few decades, in agreement with observation. The outcome emphasises that any changes in current, observed climatic trends should not automatically be attributed to greenhouse forcing. Importantly, it is noted that for conditions associated with naturally occurring climatic variability, the global mean of any climatic trend distribution should be zero or near zero. Departures from this situation imply the existence of an external forcing agency. Thousand year trends could be readily identified within the simulation, but the variations from millennium to millennium indicate the occurrence of secular variability. A probability density function distribution of 30-year duration trends within a selected millennium revealed a near-Gaussian outcome. This, together with other analyses, supports the conclusion that stochastic processes dominate the climatic variability within the simulation.