Warmer and warmer

Are the heat waves really getting more extreme? This question popped up after the summer of 2003 in Europe, and yet again after this hot Russian summer. The European Centre for Medium-range Weather Forecasts (ECMWF), which normally doesn’t make much noise about climate issues, has since made a statement about July global mean temperature being record warm:

Consistent with widespread media reports of extreme heat and adverse impacts in various places, the latest results from ERA-Interim indicate that the average temperature over land areas of the extratropical northern hemisphere reached a new high in July 2010. May and June 2010 were also unusually warm.

Here, the ERA-Interim, also referred to as ‘ERAINT’, is the ECMWF’s state-of-the-art reanalysis. But the ERAINT describes the atmospheric state only since 1989, and in isolation, it is not the ideal data set for making inferences about long-term climate change because it doesn’t go all that far back in time. However, the statement also draws on the longer reanalysis known as the ERA40 re-analysis, spanning the time interval 1957-2002. Thus, taken into context of ERA40, the ECMWF has some legitimacy behind their statement.

The ERAINT reanalysis is a product of all suitable measurements fed into a model of the atmosphere, describing all the known relevant physical laws and processes. Basically, reanalyses represent the most complete and accurate picture that we can give for the day-to-day atmosphere, incorporating all useful information we have (satellites, ground observations, ships, buoys, aircrafts, radiosondes, rawinsondes). They can also be used to reconstruct things at finer spatial and temporal scales than is possible using met station data, based on physical rules provided by weather models.

The reanalyses are closely tied to the measurements at most locations where observations – such as 2-meter temperature, T(2m), or surface pressure – are provided and used in the data assimilation. Data assimilation is a way of making the model follow the observations as closely as possible at the locations where they are provided, hence constraining the atmospheric model. The constraining of the atmospheric model affect the predictions where there are no observations because most of the weather elements – except for precipitation – do not change abruptly over short distance (mathematically, we say that they are described by ‘spatially smooth and slowly changing functions’).

There are also locations – notably the in the Polar regions and over Africa – where ground-based measurements are sparse, and where much is left for the weather models to predict without observational constraints. In such regions, the description may be biased by model shortcomings, and different reanalysis may provide a different regional picture of the surface conditions. Surface variables such as T(2m) are strongly affected by their environment, which may be represented differently in different weather models (e.g. different spatial resolution implies different altitudes) and therefore is a reason for differences between reanalyses.

Page 1 of 3 | Next page