Coastal geometry rather than upwelling intensity predicts larval fish distribution and composition in the southern Humboldt Current System

In eastern boundary upwelling systems, wind-driven upwelling supports coastal production but also carries larvae offshore. Hence, the processes that supply resources and those that keep larvae where they can use them need not coincide. We asked whether larval distribution in the southern Humboldt Current System follows wind forcing or coastal geography. We sampled larval fish assemblages at 160 stations between 29°S and 42°30′S during three austral winter surveys. We modelled larval abundance, grouped by adult habitat (epipelagic, mesopelagic, and demersal), against upwelling intensity, distance to shore, and bathymetry using spatially explicit models. Upwelling intensity was the weakest predictor of the three. Added to a spatial random field, it worsened the fit for epipelagic and mesopelagic larvae. It improved it only marginally for demersal larvae, even though the index varied strongly along the coast. The descriptors that predicted abundance were static coastal properties and differed among groups: distance to shore for epipelagic larvae, bathymetry for demersal larvae, and, for mesopelagic larvae, no measured covariate. Species richness showed no latitudinal trend, but assemblage composition was strongly structured along the coast. The two components of beta diversity ran in opposite directions: species turnover increased southward and peaked between 38 and 42°S, where the Humboldt Current gives way to the Cape Horn Current, whereas the nestedness component increased northward and peaked between 30 and 32°S, within the region of persistent upwelling. Comparable richness therefore arose in different ways at either end of the same coast. In an upwelling-driven system, coastal features that change little between years predicted larval distribution better than upwelling intensity.