A semi-empirical model of boreal-forest gross primary production, evapotranspiration, and soil water — calibration and sensitivity analysis

Simple approaches to predicting ecosystem fluxes are useful in large-scale applications because existing data rarely support justified use of complex models. We developed a model of daily ecosystem gross primary production (P), evapotranspiration (E), and soil water content (theta), which only requires standard weather data and information about the fraction of absorbed radiation. We estimated the parameters of the model for two boreal Scots pine eddy-covariance sites (Hyytiala and Sodankyla). The model predicted P and E adequately for Hyytiala for both calibration and additional test years. The model calibrated for Hyytiala slightly overestimated P and E in Sodankyla, but its performance levelled with the model calibrated for Sodankyla in a dry year. Sensitivity analysis of the model implied that drought prediction is sensitive, not only to key E submodel parameters, but also to P submodel parameters. Further improvement and calibrations of the model could benefit from forest sites with varying canopy and different species structures.