Marine phytoplankton account for approximately half of the production of organic matter on earth, support virtually all marine ecosystems, constrain fisheries yields, and influence climate and weather. Despite this importance, long-term trajectories of phytoplankton abundance or biomass are difficult to estimate, and the extent of changes is unresolved. Here, we use a new, publicly-available database of historical shipboard oceanographic measurements to estimate long-term changes in chlorophyll concentration (Chl; a widely used proxy for phytoplankton biomass) from 1890 to 2010. This work builds upon an earlier analysis (Boyce et al., 2010) by taking published criticisms into account, and by using recalibrated data, and novel analysis methods. Rates of long-term chlorophyll change were estimated using generalized additive models within a multi-model inference framework, and post hoc sensitivity analyses were undertaken to test the robustness of results. Our analysis revealed statistically significant Chl declines over 62% of the global ocean surface area where data were present, and in 8 of 11 large ocean regions. While Chl increases have occurred in many locations, weighted syntheses of local- and regional-scale estimates confirmed that average chlorophyll concentrations have declined across the majority of the global ocean area over the past century. Sensitivity analyses indicate that these changes do not arise from any bias between data types, nor do they depend upon the method of spatial or temporal aggregation, nor the use of a particular statistical model. The wider consequences of this long-term decline of marine phytoplankton are presently unresolved, but will need to be considered in future studies of marine ecosystem structure, geochemical cycling, and fishery yields.