Sensitivity of GIS patterns to data resolution: a case study of forest fragmentation in New Zealand

Spatial pattern plays an influential role in the ecological processes of ecosystems, and landscape pattern metrics computed from remotely sensed data offer a way to quantify the correlation between pattern and process. However, the resolution of geographic data affects the landscape metrics obtained from a GIS, with consequent implications for the interpretation of biological effects studied at landscape scales. Here, we studied the effect of data resolution on estimates of three metrics of forest cover commonly used in the landscape ecology literature: percent forest cover, forest edge density, and mean fractal dimension of forest patches. Estimates of each metric were computed for six landscapes (30 × 30 km) in the North Island of New Zealand at 10 different data resolutions with pixels ranging from 30 to 1000 m. All three metrics exhibited significant changes in value as a result of changing resolution, and the sensitivity of the fragmentation metrics to data resolution was impacted in a non-linear manner by the amount of forest cover in a landscape. In landscapes with low forest cover, changing pixel size altered estimates of percent forest cover by as much as 75%. Extrapolation to correct for effects of changing resolution and different landscapes seems a likely solution in the case of some, but not all, metrics. The scaling problem hinders efforts to correlate spatial pattern with ecosystem process and the subsequent conclusions concerning biodiversity and conservation policy.