Sampling and Variability Issues Related to Anthropogenic Warming of the World Oceans

In recent work we have examined the increase in ocean heat content over the past 40 years, and attributed it to anthropogenic forcing. In this presentation we show a number of issues relating to that previous work, including an evaluation of natural climate variability in the coupled climate models (PCM and HadCM3) we used and how it compares to observations, the possible influence of sampling on our results, an estimate of the uncertainty in model heat uptake (compared to the signal strength), and the possible effects of seasonally biased observations. We find that PCM and HadCM3 show levels of natural variability similar to that observed at time scales longer than 0.15 cy/yr, except for having overly weak variability in basin-averaged temperatures between 300 and 500 m. This latter detail is not due to weak variability at a given point, but rather because observations show more coherence in variability at this depth than the models. Model estimates suggest that the observed sampling is sufficient to estimate basin-averaged temperatures to about +/- 0.1 C; this is highly dependent on basin and depth, but sufficient to reliably detect ocean warming in all basins by the 1980s. CMIP2 results suggest ocean basin heat uptake in various climate models can vary by a wide margin (2 to 8); synthetically constructed ocean warming fingerprints corresponding to the "maximum" and "minimum" heat uptake ocean models suggest that even in the case of minimum heat uptake, the anthropogenically forced model signal falls outside the expected range of natural internal climate variability in the surface layers. Finally, an analysis of ocean temperature trends by season, the representation of the seasonal cycle, and detectability of the warming signal with depth indicate that the results are unaffected by possible seasonal sampling biases in the observations.