Rainfall estimation in south-west Tasmania using satellite images and phytosociological calibration

Abstract

NOAA AVHRR digital data are used to map precipitation in southwest Tasmania, Australia. The technique uses multiple linear regression analyses between precipitation and satellite data to obtain yearly and seasonal averages over a three-year period. These maps are later transformed into long term average maps by drawing on linear relations between 3-year and long term averages for various precipitation stations in the study area. Rms model residuals (measured-model) range from 0.45 mmday⁻¹ for the summer regression to l.20mmday⁻¹ for the winter regression. A significant fraction of this error may be attributed to limitations in the rain gauge measurements. The predicted precipitation is further verified by reference to variation in the alpine floras of the study area which have a known relationship with soil conditions that vary with precipitation. Single axis ordination scores for the floras show strong relationships with predicted precipitation both on a seasonal and annual basis, especially in subsets related to substrate and region.