A 22-group Ecopath model representing the southern Humboldt (SH) upwelling system in the year 1970 is constructed. The model is projected forward in time and fitted to available time series of relative biomass, catch and fishing mortality for the main fishery resources. The time series cover the period 1970 to 2004 and the fitting is conducted using the Ecopath with Ecosim (EwE) software version 5.1. The aim is to explore the relative importance of internal (trophic control) and external (fishing, physical variability) forcing on the dynamics of commercial stocks and the Southern Chilean food web. Wide decadal oscillations are observed in the biomass of commercial stocks during the analyzed period. Fishing mortality explains 21% of the variability in the time series, whereas vulnerability (v) parameters estimated using EwE explain an additional 20%. When a function affecting primary production (PP) is calculated by Ecosim to minimize the sum of squares of the time series, a further 28% of variability is explained. The best fit is obtained by using the fishing mortality time series and by searching for the best combination of v parameters and the PP function simultaneously, accounting for 69% of total variability in the time series. The PP function obtained from the best fit significantly correlates with independent time series of an upwelling index (UI; rho = 0.47, p<0.05) and sea surface temperature (SST; rho = -0.45, p<0.05), representing environmental conditions in the study area during the same period of time. These results suggest that the SH ecosystem experienced at least two different environmentally distinct periods in the last three decades: (i) from 1970 to 1985 a relatively warm period with low levels of upwelling and PP, and (ii) from 1985 to 2004 a relatively cold period with increased upwelling and PP. This environmental variability can explain some of the changes in the food webs. Fishing (catch rate) and the environment (bottom-up anomaly in PP) appear to have affected the SH both at the stock and at the food web level between 1970 and 2004. The vulnerability setting indicates that the effects of external forcing factors may have been mediated by trophic controls operating in the food web. (C) 2013 Elsevier B.V. All rights reserved.