Fig. 6

Random forest regression model predicted conditional differences in bathymetric depth ratio across the study area. Maps show predicted differences when: sea surface zooplankton concentrations are 1 SD below or above their summer average (top left); the moon is 0% and 100% illuminated (top right); sea surface temperatures are 1 SD below or above their summer average (bottom left); and sea surface oxygen concentrations are 1 SD below or above their summer average (bottom right). Relative differences were calculated as the difference between predictions (e.g., 0% and 100% illumination) in a given spatial cell. Negative (red) values represent greater predicted depth with greater lunar illumination, temperature, zooplankton concentration, or oxygen concentration. Non-focal spatial covariates were fixed at July 30, 2020 values and the remaining covariates were fixed at median values