CNR - ISAC, Unit UOS of Padua

Corso Stati Uniti 4, I-35127 Padova - Italy
Tel: +39 049 8295906 e-mail:
European Commission dg12

"Improved understanding of past climatic variability from early daily European instrumental sources" (IMPROVE)

Coordinator: Dr. D. Camuffo.
Contract N°: ENV4-CT97-0511 funded by the European Commission
programme Environment and Climate.

The Research Project

The EU research project Improved understanding of past climatic variability from early daily European instrumental sources (IMPROVE) has been developed with the funding of the European Commission, DG XII. The study involved European regions from the Baltic to the Mediterranean and from the Atlantic to Eastern Europe.

IMPROVE's general objectives were to:

As far as these fundamental questions are concerned, IMPROVE has contributed in a number of ways by:

  1. producing seven new highly reliable series, covering a very long time interval, nearly three centuries. Another important aspect is the high (daily) resolution of the series, which allows more insight into past variability, as monthly averages smooth out and mask many important climatic features;
  2. analysing and correcting errors and inhomogeneities in the long series. Metadata have been considered as important as data, since metadata are fundamental not only to correct, make homogeneous and interpret data, but also to distinguish apparent climatic changes, due to variations in observational methodology, from real climatic changes. This not only helped to produce more accurate series, but also clarified the question of the quality of the existing ones. In actual fact, the long series are affected by a huge number of problems, and not only casual reading errors. Reading errors are mainly randomly distributed and they disappear when the averages of many series are calculated. The real problem is the presence of systematic errors that vary in the course of time, the changes in measuring style due to national or international regulations or the simple evolution of technology, that affected the observations in the same way and simultaneously in all sites. All homogeneity assessment techniques would perform poorly if all station time series were affected similarly by common factors, and changes in measuring style cannot be assessed with statistical analysis, but only with historical research of all the metadata. Some illustrative case studies reported in this issue will clarify the situation. The better the quality of the existing data, the better the response we obtain.
    At present, attention is being mainly devoted to forecasting future scenarios. To this aim, the quality of data is essential and a number of subjective and relatively objective criteria exist for testing monthly data and sources of inhomogeneity (e.g. Jones et al., Reviews of Geophysics, 37:2, 173-199, 1999). However, the evaluation of the data quality, as deduced from the statistical analysis of the data, appears more optimistic than the error deduced by looking at the instruments capability and the field practice. For instance, in general a first contribution ±0.2°C corresponds to the accuracy of standard thermometers (WMO Report No.8, Geneva, 1983) but the overall error in the field is greater. In the sunny countries, the largest error is reached on clear, calm days, when the screen may over-heat by +2.5°C and on clear nights when it may under-cool by -0.5°C. The correction based on metadata describing observational errors and problems requires a long and unrewarding work to improve the quality of long series. One of the aims of IMPROVE is to encourage critical revision and improvement of the quality of existing series, while providing, at the same time, examples of typical errors to be removed and identifying the procedures needed to amend them.
  3. The actual warming rate has been proven to be at such a slow rate that temperature changes, over years (i.e. 0.006°C/yr) and even decades (i.e. 0.06°C/decade), are in most cases smaller than the instrumental resolution and can hardly be directly detected. This occurs especially with temperature averages, for which the climate signal is often below the noise and/or the instrumental limit. However, looking at the frequency distribution of extreme events, things appear differently. Extreme events depart so much from the average, that they can be easily detected, and in computing their signal to noise ratio the observational error becomes less relevant. IMPROVE has clearly demonstrated that this approach is very promising. An analysis of the distribution of extreme events has also shown that recent warming is characterised by an increase in frequency of the hottest days, in association with a decrease in frequency of the coldest.

Research in cooperation with:


IMPROVE CD with the data
Camuffo, D. and Jones, P.D. 2002: Improved Understanding of Past Climatic Variability from Early Daily European Instrumental Sources. Kluwer Academic Publisher

Climatic Change, Vol 53 Nos. 1-3 (2002) IMPROVE (special issue):

Papers in other journals or in other issues:



Active project