The characterisation of the solar resource of a site is essential for different phases of solar energy projects. While only rough estimates of yearly levels of solar irradiation (global or direct, depending on the technology) are needed in their very early stages, the required depth of the assessment increases as the project advances, including long-term estimates that can only be obtained through a statistical analysis of a continuous and long-term database of solar radiation measurements. This paper provides the results of a statistical analysis of thirteen years of Global Horizontal Insolation (GHI) measurements and Direct Normal Insolation (DNI) measurements from Seville, Spain (37.4 degrees N, 6.05 degrees W) at different time resolutions, i.e. from annual to nearly instantaneous (5-s). In addition, a new methodology for gap-filling is proposed which keeps the frequency distribution of the original dataset and reduces the uncertainty of the aggregated values (hourly, daily, monthly, yearly) due to the gaps. Some relevant results of this analysis are: (a) the instantaneous values of GHI and DNI have bimodal distributions, although of different characteristics, in agreement with the results of some works developed in similar climate locations; (b) the frequency distributions of the instantaneous and 10-min clearness index (k(t)) and beam fraction index (k(b)) are almost identical, suggesting 10 min as a good time resolution for the simulation of Concentrated Solar Power (CSP) systems oriented to feasibility analyses; (c) the distributions of hourly k(i) and k(b) values, show significant differences with respect to the instantaneous ones; (d) the difference between the percentile 99 (P99) of the instantaneous GHI and its maximum value is very high, because of the enhancement effect due to the cloud reflection, while for the DNI the corresponding values are much closer. The comparison with the results of other locations of similar climates suggest that these results can be extrapolated, at least, to other locations of similar climates. Other, more site-specific, results are: (a) the number of typical overcast days in summer is extremely low, while it takes its maximum value in December, suggesting this month as the best for maintenance operations that require halting the operation of CSP plants; (b) the annual mean daily values are 4.98 kW h m(-2) for GHI and 5.68 kW h m(-2) for DNI, with a low inter-annual variability and a greater monthly variability which depends on the season. The monthly and yearly average values from Seville have been compared with three long-term databases derived from satellite images. The best concordance in GHI values is found with NASA's Surface Meteorology and Solar Energy (NASA SSE), but NASA SSE provides significantly higher DNI values compared to the Seville database. A comparison of one year of DNI and GHI measurements recorded at two locations, Durban (South Africa) and Abu Dhabi (United Arab Emirates), with high solar potential is also addressed. (C) 2016 Elsevier Ltd. All rights reserved.