Global length-weight relationships of deep-pelagic fishes: Coverage, biases and variability.

Length-weight relationships (LWRs) are morphometric tools that allow body mass to be estimated from fish length, providing input for fisheries, ecological and biodiversity studies. For deep-pelagic fishes, however, LWR data remain fragmented and unavailable for most species. Here, we present a global synthesis of LWRs for deep-pelagic fishes, consolidating available data into a single harmonised dataset and mapping taxonomic and geographic gaps. We provide the first LWR estimates for seven species and provide additional morphometric information, including extensions of previously reported size ranges and revised estimates based on freshly collected specimens for 23 species. With these additions, 427 LWRs are now available for 227 species, about 15% of recognised deep-pelagic diversity. Coverage remains strongly biased. Small-bodied, vertically migrating species dominate, especially within the Myctophidae, whereas resident species with stable vertical distributions, slower growth and longer lifespans are poorly represented. Geographic coverage is likewise uneven, with the North Atlantic extensively sampled, but the South Atlantic, Indian Ocean and large areas of the Pacific still severely undersampled. On average, the allometric exponent b was close to isometry (median 3.09), though most individual relationships departed significantly from it, and it was only weakly structured across taxa. Variation among families and body shapes was modest and broadly overlapping because the apparent body-shape gradient in b largely tracked taxonomy. The dominant signal was intraspecific. Within a species, b varied almost as much as it did among species, often by more than a full unit across independent studies. Consequently, a single LWR should not be regarded as a fixed species trait. Length-to-mass conversion should preferentially rely on locally derived relationships wherever they exist, and when relationships must be borrowed from elsewhere, this additional uncertainty should be explicitly acknowledged.