Consistent estimates of gross primary production of Finnish forests : comparison of estimates of two process models

We simulated Gross Primary Production (GPP) of Finnish forests using a landsurface model (LSM), JSBACH, and a semi-empirical stand-flux model PRELES, and compared their predictions with the MODIS GPP product. JSBACH used information about plant functional type fractions in 0.167 degrees pixels. PRELES applied inventory-scaled information about forest structure at high resolution. There was little difference between the models in the results aggregated to national level. Temporal trends in annual GPPs were also parallel. Spatial differences could be partially related to differences in model input data on soils and leaf area. Differences were detected in the seasonal pattern of GPP but they contributed moderately to the annual totals. Both models predicted lower GPPs than MODIS, but MODIS still showed similar south north distribution of GPP. Convergent results for the national total GPP between JSBACH and PRELES, and those derived for comparison from the forest ghg-inventory, implied that modelled GPP estimates can be realistically up-scaled to larger region in spite of the fact that model calibrations may not originate from the study region, or that a limited number of sites was used in the calibration of a model.