In general, the internal validation indicated that SIENA is comparable to the structure of the inland sample cities. Fig. 5 plots the percentage of urban metrics that do not significantly differ between SIENA and the sample cities for each urban component (a ¼ 0.05, 90% confidence interval).The modelled terrain imitates the mean terrain values of theinland sample cities well, indicating that the Axminster area provides a good representation of the terrain in the sample cities. Only urban metrics representing slope in the urban surroundings deviate from the sample city mean. The t-test also showed strong agreement between SIENA and the mean of the sample cities for all of the transport metrics analysed. Associations between main roads/railways and the terrain are, however, not statistically significant for higher altitude. SIENA also slightly over predicted minor road density close to the city centre. For land cover analysed at the city level, city size was comparable to the mean of the inland sample cities but the number of land cover classes, the number of patches, Shannon’s Diversity Index and Shannon’s Evenness Index
at city level all differed. These differences can be explained by the land cover model. The number of land cover classes was constrained in SIENA to the seven main land cover classes, while the sample cities also include a number of minor land cover classes. This also accounted for the differences in Shannon’s Diversity and Evenness Indexes. The difference in the number of patches was a result of the use of a 25 25 m grid resolution for SIENA, and probably also an effect of the smoothing filter used to generalise the data. Similar results were seen at land cover class level. No