Fig. 3: Comparing the spatial variability of the dominant multiple linear regression coefficients between remote sensing and Earth system models.
From: Global decline in net primary production underestimated by climate models
Zonal averages ± standard deviations of the multiple linear regression coefficients for a, d sea surface temperature (SST), b, e chlorophyll-a concentrations (CHL) and c, f mixed layer depth (MLD) for the a–c Eppley-VGPM, Behrenfeld-VGPM, Behrenfeld-CbPM, Westberry-CbPM, Lee-AbPM and Silsbe-CAFE NPP algorithms and d–f the ensemble of CMIP6 Earth system models. Only pixels/grid points where the multiple linear regression analysis was significant are included in the zonal averages. The shaded region in panels d–f represents the range of coefficients as estimated from the remote sensing algorithms zonal averages (panels a–c).
