The number of points created was based on previous experience of sampling in urban areas where, on average, for every site successfully visited, access to a further 1·5 would 3-Methyladenine not be possible. One hundred and thirty points would therefore ensure that at least 50 sites would be surveyed in each landcover class, irrespective of land ownership (which was determined after sites were accessed but prior to analysis). During the survey, all 520 points were visited, and data were recorded from 375 sites, exceeding the desired 50 per landcover category (Table 1). At each survey site within a landcover class, a 5 × 5 m quadrat was centred on the GPS coordinates of the sample point. The area of individual Shrub, Tall Shrub and Tree patches is highly skewed across the city, with the majority being small in size (median = 66·25 m2; mean = 211·30 m2). A quadrat of 5 × 5 m provides sufficient area for a representative sample of vegetation in larger patches, but not so large as to excessively cover smaller patches (18·2% of patches across the city are 1 cm at 1·30 m above-ground level; Condit 1998), woody vegetation (e.g. bushes, small tree saplings with a DBH ≤1 cm), water, litter, hard surface (e.g. tarmac) and buildings was estimated, along with canopy cover. In addition, trees present within the quadrat were identified to species or genus, and the DBH and crown height were measured (using a clinometer). Where quadrats fell across two landcover categories, or in small vegetation patches that did not cover the entire 5 × 5 m area, they were always classified according to the central GPS point. During the survey, this only occurred between Shrub/Tall Shrub/Tree patches and Herbaceous Vegetation/hard surface/buildings (e.g. single trees in a residential street, surrounded by hard surface). In such instances, carbon storage was only estimated for the areal extent of the defining landcover class (e.g. if a Herbaceous Vegetation quadrat was partially covered by tree canopy, only the carbon density of the quadrat outside the canopy area would be calculated to prevent overestimation when scaling up, using the GIS landcover categories, to generate city-wide estimates). In October 2009, at the end of the growing season, above-ground herbaceous vegetation was harvested from within 25 × 25 cm quadrats at 56 sites across the city. The quadrat locations were randomly generated in the GIS from within the Herbaceous Vegetation landcover category. All standing crop was removed at ground level using a cut-throat razor, bagged and removed for carbon analysis.