Data on the distribution and abundance of mid-trophic level organisms (MTLOs) in the pelagic open-ocean ecosystem are normally sparse or absent. Consequently, ecosystem models are limited in their ability to support decision-making for issues ranging from fisheries management to ecosystem resilience to climate change. We used acoustic data collected at 38 kHz frequency across the Southern Ocean (SO) between 2008 and 2014 to develop explanatory and predictive models for acoustic backscatter, a proxy for MTLO abundance in the epi- and mesopelagic zones. Boosted regression trees and generalised additive mixed models were used to develop simple predictive models for backscatter in the epi- and mesopelagic zones, using sea surface temperature, time of day (day/night) and depth. The resulting models predicted backscatter reasonably well in the Pacific sector of the SO, and in an independent dataset in the Indian sector of the SO. Our predictive models may provide a tool for inferring abundance and distribution of MTLOs in other parts of the SO.