Charlotte Begouen Demeaux, Emmanuel Boss, Toby K. Westberry

We recently found a significant bias while validating frequently used ocean color algorithms retrieving the spectral diffuse attenuation coefficient (Kd(λ))(Begouen and Boss, 2022). Here we modify existing algorithms for Kd(λ) to remove the observed bias at Kd(490), and evaluate the impact on global and regional estimates of net primary production (NPP) using two different primary production models. The new parametrization results in improved retrievals of Kd(490) by all algorithms. The new coefficients are validated using measurements not included in the training dataset and are found to perform significantly better at a different wavelength (412nm) than the one used for the new parametrization (490nm) and perform reasonably well in Case-2 waters. Since the new coefficients were developed with a dataset encompassing larger proportions of the ocean’s variability, they are better suited to compute Kd(λ) in regions that weren’t present in the original algorithm’s dataset and are therefore appropriate for global Kd(λ) estimation. Using the new Kd parameterization results in a global increase of NPP of ≈ 25 − 30%, mostly driven by the previous overestimation of Kd(λ) (underestimation of light penetration) in the clear, subtropical gyres. Subtropical gyres show the largest increase (70%) in the VGPM model. Their large surface area and the magnitude of the bias in Kd(λ) in the old parameterization causes the observed difference in global NPP estimates. Our results suggest that the oceanic carbon uptake is larger than previously thought, which will be most relevant to the oceanic carbon dioxide budget once humanity slows the increase of atmospheric CO2.