Polling the Ocean: Statistically Sound Inferences on the State of the Global Ocean

Few people seriously contest that human activities on land have become a major factor in shaping biodiversity on Earth. This includes habitat modifications and the extensive destruction of natural terrestrial ecosystems resulting from forest clearing to gain agricultural lands. However, in the marine realm, despite ample evidence of substantial impacts to habitats from bottom trawlers, serious problems with overfishing and bycatch, the fact that we likely have an analogous effect on the oceans is contested. Unfortunately, at present, an accurate representation of what happens in the world’s oceans or what is being done to global fisheries is not available, because the sample of assessed fish stocks often used to infer global status consists of highly valued, resilient target species that have been fished extensively for decades. To recall what we learnt in Introductory Statistics: for reliable, reproducible inferences on a population, we must either collect data on all its members (i.e., perform a census), or obtain representative samples (and not case studies, which may or may not represent a stratum, or well-defined subset, of the population in question). Here, we present the results of a stratified random sampling scheme for the global ocean, and validate its use for inferring two features of the global ocean: mean depth and global primary productivity, which are both well-known and uncontested quantities. Then we use this scheme to infer, from a rigorously defined, representative sample of the ocean, the state of fished resources in the global ocean.