Developing fundamental data layers to support environmental modelling in New Zealand: progress and problems

Abstract

As part of a programme to develop fundamental data layers suited to spatial modelling of environmental processes, we used standard geostatistical techniques to model mean monthly soil temperatures over the South Island of New Zealand. These methods help to account for the majority of topographic variability at high spatial resolutions where topographic influences dominate spatial patterns of climate. Standard climate sites do not provide a good sample for understanding full topographic variation of many climatic parameters. Interpolation was often hampered by the closure, from 1986, of long-term climate stations recording soil temperature. By adding purposefully sampled data we gained useful knowledge about topographic variation in soil temperature, but the data were not well suited to some interpolation methods. Partial thin plate splines produced the most accurate long-term mean monthly and specific month/year soil temperature surfaces, but multiple linear regression also provided a simple and robust method for soil temperature interpolation because the data were not strongly spatially dependent.