Warmer and warmer

In the book “A Vast Machine“, Paul N. Edwards discusses various types of data and how all data involve some type of modelling, even barometers and thermometers. It also provides an account on the observational network, models, and the knowledge we have derived from these. Myles Allen has written a review of this book in Nature, and I have reviewed it for Physics World (subscription required for the latter).

All data need to be screened though a quality control, to eliminate misreadings, instrument failure, or other types of errors. A typical screening criterion is to check whether e.g. the temperature estimated by satellite remote sensing is unrealistically high, but sometimes such screening may also throw out valid data, such as was the case of the Antarctic ozone hole. Such post-processing is done differently in analyses, satellite measurements, and reanalyses.

The global mean temperature estimated from the ERAINT, however, is not very different from other analyses or reanalyses (see figure below) for the time they overlap. We also see a good agreement between the ERA40 reanalysis, the NCEP/NCAR reanalysis, and the traditional datasets – analyses – of gridded temperature (GISTEMP, HadCRUT3v, NCDC).

Do the ERAINT and ERA40 provide a sufficient basis for making meaningful

inferences about extreme temperatures and unprecedented heat waves? An important point with reanalyses, is that the model used doesn’t change over the time spanned by the analysis, but reanalyses are generally used with caution for climate change studies because the number and type of observations being fed into the computer model changes over time. Changes in the number of observations and instruments is also an issue affecting the more traditional analyses.

Since the ERAINT only goes as far back as 1989, it involves many modern satellite-borne remote sensing measurements, and it is believed that there are less problems with observational network discontinuity after this date than in the earlier days. It may be more problematic studying trends in the ERA40 data, due to huge improvements in the observational platforms between 1958 and now. Hence, it is important also to look at individual long-term series of high quality. These series have to be ‘homogeneous’, meaning that they need to reflect the local climate variable consistently through its span, not being affected by changes in the local environment, instrumentation, and measurement practices.

Page 2 of 3 | Previous page | Next page