An assessment of net primary productivity estimates using coupled physical-biogeochemical/earth system models in the Arctic Ocean

Net primary production (NPP) is the major source of energy for the Arctic Ocean (AO) ecosystem, as in most ecosystems. Reproducing current patterns of NPP is essential to understand the physical and biogeochemical controls in the present and the future AO. The Primary Productivity Algorithm Round Robin (PPARR) activity provides a framework to evaluate the skill and sensitivity of NPP as estimated by coupled global/regional climate models and earth system models in the AO. Here we compare results generated from 18 global/regional climate models and three earth system models with observations from a unique pan-Arctic data set (1959-2011) that includes in situ NPP (N=928 stations) and nitrate (N=678 stations). Models results showed a distribution similar to the in situ data distribution, except for the high values of integrated NPP data. Model skill of integrated NPP exhibited little difference as a function of sea ice condition (ice-free vs. ice-covered) and depth (shallow vs. deep), but performance of models varied significantly as a function of seasons. For example, simulated integrated NPP was underestimated in the beginning of the production season (April-June) compared to mid-summer (July and August) and had the highest variability in late summer and early fall (September-October). While models typically underestimated mean NPP, nitrate concentrations were overestimated. Overall, models performed better in reproducing nitrate than NPP in terms of differences in variability. The model performance was similar at all depths within the top 100 m, both in NPP and nitrate. Continual feedback, modification and improvement of the participating models and the resulting increase in model skill are the primary goals of the PPARR-5 AO exercise.