Landscape and biophysical controls of lake productivity to inform evaluation of sockeye salmon (Oncorhynchus nerka) populations in data-limited regions

Landscape models are increasingly used to classify and predict the structure and productivity of data-limited aquatic ecosystems. One such suite of ecosystems is on the remote North and Central Coast (NCC) of British Columbia, where sockeye salmon (Oncorhynchus nerka) rear in more than 150 lakes. Given their remoteness and limited resources for assessment, limnological and population monitoring in many of these lakes has been periodic or absent, limiting understanding of the status of populations and their habitats. Lake photosynthetic rate (PR) estimates are foundational to models of sockeye salmon nursery lake productive capacity. Using data from 61 lakes across the NCC, we compared a suite of landscape and lake variables in an information theoretic framework producing a set of models relating these characteristics to lake PR. A categorical variable related to lake biogeochemistry—whether a lake is humic stained, clear, or glacially turbid—was the most important variable predicting lake PR and was included in all models. Lake surface area relative to upstream catchment size and lake perimeter-to-surface-area ratio were also important, with smaller upstream catchments yielding higher production, and high shoreline complexity correlated with lower productivity as measured by limnetic PR. Model-averaged predictions of PR from the four models with the lowest residual error were created for 96 lakes currently lacking limnological assessments. These landscape models represent a valuable starting point for evaluating lake-specific carrying capacities for data-poor sockeye salmon populations under Canada’s Wild Salmon Policy.