Global Estimates of Marine Gross Primary Production Based on Machine Learning Upscaling of Field Observations

Approximately half of global primary production occurs in the ocean. While the large-scale variability in net primary production (NPP) has been extensively studied, ocean gross primary production (GPP) has thus far received less attention. In this study, we derived two satellite-based GPP models by training machine learning algorithms (Random Forest) with light-dark bottle incubations (GPP(LD)) and the triple isotopes of dissolved oxygen (GPP(17 Delta)). The two algorithms predict global GPPs of 9.2 +/- 1.3 x 10(15) and 15.1 +/- 1.05 x 10(15) mol O-2 yr(-1) for GPP(LD) and GPP(17 Delta), respectively. The projected GPP distributions agree with our understanding of the mechanisms regulating primary production. Global GPP(17 Delta) was higher than GPP(LD) by an average factor of 1.6 which varied meridionally. The discrepancy between GPP(17 Delta) and GPP(LD) simulations can be partly explained by the known biases of each methodology. After accounting for some of these biases, the GPP(17 Delta) and GPP(LD) converge to 9.5 similar to 12.6 x 10(15) mol O-2 yr(-1), equivalent to 103 similar to 150 Pg C yr(-1). Our results suggest that global oceanic GPP is 1.5-2.2 fold larger than oceanic NPP and comparable to GPP on land.

Automated summary

Global oceanic gross primary production (GPP) has been studied using two satellite-based models, showing that GPP(17 Delta) is 1.5-2.2 times larger than oceanic net primary production (NPP) and comparable to land GPP. The discrepancy between GPP(17 Delta) and GPP(LD) simulations can be partially explained by methodological biases.