As output from photovoltaic (PV) panels is closely related to sunlight levels, monitoring solar irradiance levels is crucial for system design and predictive purposes. With advances in PV technology, urban sites at northerly locations, where both horizontal and vertical solar irradiance make significant contributions, are becoming increasingly important. The aim of this paper is to compare solar irradiance for horizontal and vertical orientations and to assess the relative effectiveness of differing averages, from 10 min to 1 h, for use in short-term prediction of solar irradiance levels for a UK site. Vertical and horizontal solar irradiance observations were collected from a monitoring station on the roof of a five-storey building at an urban site in Newcastle-upon-Tyne, UK (latitude 55°N). 10-min data was collected for 13- or 15-day periods in two summers (1994, 1995) and two winters (1993, 1994). Although mean levels for horizontal and vertical irradiance were different, as expected, general patterns were very similar indicating the possibility of predicting vertical irradiance from horizontal at the same location. 10-min, 20-min, 30-min and 1-h averaging times were compared utilising autocorrelation coefficients and ARIMA models to assess the information lost when using longer averaging intervals. For consideration of short-term changes, 10-min averages were most informative whilst hourly averages were substantially poorer.