Now an Australian-led team has come up with new estimates of ocean heat content that allow for sampling and instrumental biases. The revised estimates show 50% higher ocean warming and thermal expansion trends for 1961 to 2003 and are in better agreement with climate models.

"For the first time, we can provide a reasonable account of the processes causing the rate of global sea level rise over the past four decades – a puzzle that has led to a lot of scientific discussion since the 2001 IPCC report but with no signficant advances until now," said Catia Domingues of the Centre for Australian Weather and Climate Research. "Following the review of millions of ocean measurements, predominantly from expendable instruments probing the upper 700 m of the ocean, we were able to more accurately estimate upper-ocean warming, and the related thermal expansion and sea-level rise."

Domingues and colleagues from the Centre for Australian Weather and Climate Research, Antarctic Climate and Ecosystems Cooperative Research Centre, Australia, and Lawrence Livermore National Laboratory, U.S.A., added their observational estimates of upper ocean thermal expansion to other contributions to sea level rise. This gave a total sea-level rise of 1.5 ± 0.4 mm per year from 1961 to 2003.

"Our results are important for the climate modelling community because they boost confidence in the climate models used for projections of global sea level rise resulting from the accumulation of heat in the oceans," said Domingues. "These projections will, in turn, assist in planning to minimise the impacts and in developing adaptation strategies."

The ocean temperature measurements analysed by the team were from reversing thermometers, expendable bathy-thermographs (XBTs), conductivity-temperature-depth (CTD) measurements and Argo floats. More than 50% of the data came from XBTs, which measure temperature as they free fall to the bottom of the ocean. Scientists can calculate the depth of temperature readings by measuring the time since release of the instrument. This can introduce systematic errors, so Domingues and colleagues used a recent time-variable XBT fall-rate correction to minimize measurement bias from the devices.

Because of this removal of instrumental bias, between 1993 and 2003 the revised trends calculated for the study are about 40% smaller than previous estimates

The team, which reported its work in Nature, also used a statistical technique called reduced-space optimal interpolation to allow for sparse data coverage.