Fine Scale ANUClimate Data for Ecosystem Modeling and Assessment of Plant Functional Types

Abstract:

High resolution spatially extended values of climate variables play a central role in the assessment of climate and projected future climate in ecosystem modeling. The ground based meteorological network remains a key resource for deriving these spatially extended climate variables. We report on the production, and applications, of new anomaly based fine scale spatial interpolations of key climate variables at daily and monthly time scale, across the Australian continent. The methods incorporate several innovations that have significantly improved spatial predictive accuracy, as well as providing a platform for the incorporation of additional remotely sensed data. The interpolated climate data are supporting many continent-wide ecosystem modeling applications and are playing a key role in testing optimality hypotheses associated with plant functional types (PFTs). The accuracy, and robustness to data error, of anomaly-based interpolation has been enhanced by incorporating physical process aspects of the different climate variables and employing robust statistical methods implemented in the ANUSPLIN package. New regression procedures have also been developed to estimate “background” monthly climate normals from all stations with minimal records to substantially increase the density of supporting spatial networks. Monthly mean temperature interpolation has been enhanced by incorporating process based coastal effects that have reduced predictive error by around 10%. Overall errors in interpolated monthly temperature fields are around 25% less than errors reported by an earlier study. For monthly and daily precipitation, a new anomaly structure has been devised to take account of the skewness in precipitation data and the large proportion of zero values that present significant challenges to standard interpolation methods. The many applications include continent-wide Gross Primary Production modeling and assessing constraints on light and water use efficiency derived from the TERN OzFlux network. Redundancy analysis (RDA) has also been used to explain plant trait variance in tropical forest dynamics. Aggregates of ANUClimate data, including annual mean solar radiation, temperature and moisture index, performed well predicting biophysical trait observations (35% of variance).