Using materials in new applications (in particular the built environment) requires that durability predictions be made. Often, the only way this can be done within economically favourable timescales is to use accelerated ageing tests. Extrapolating these tests to in-service temperatures requires that some average of the climatic temperature be used. The normally published mean temperature for a location does not in general adequately represent the time-temperature envelope and its use in calculations based on Arrhenius type relationships may lead to serious underestimation of degradation rates. This paper shows how an equivalent temperature, T*, can be calculated from the numerical integration of a single annual temperature cycle, accounting for both daily and monthly temperature variations. The difference between the average temperature and T* is shown to be strongly correlated to the daily and monthly temperature ranges, to be significant for most locations and sometimes as high as 10degreesC. (C) 2004 Kluwer Academic Publishers.