Résumé
Pond polyculture, an ancient fish farming technique, is being studied anew as demand for sustainable aquaculture products is increasing. It takes advantage of species’ trophic interactions and behavioural characteristics. This study aimed to identify species compositions that use resources more efficiently and recommend management practices for extensive polyculture by modelling food webs and calculating performance indicators. We monitored experiments with 10 polyculture scenarios in three regions of France and analysed them using the Ecopath model. Including all scenarios, net yields of the scenarios did not differ with or without additional feed, or with or without fertilisation and liming (201-635 and 87-435 kg ha-1, respectively). The biomass of grass carp stocked was positively correlated with net fish yield (r = 0.58). Mean net fish yield did not differ between scenarios with < 10% or ≥ 10% of common carp stocked (300 and 377 kg ha-1, respectively). Ecotrophic efficiency (EE) was high for macroinvertebrates (0.90-0.99) and zooplankton (0.55-0.99) but low for phytoplankton and detritus (≤ 0.29 and ≤ 0.19, respectively). The Finn cycling index ranged from 5-25%, was moderately correlated with net fish yield (r = 0.52) and strongly correlated with the EE of phytoplankton (r = 0.90) and detritus (r = 0.89). This study provides insights into energy flows among species during a given period, which could be improved by modelling temporal dynamics of the system. This study indicates that nutrient recycling and productivity in pond systems can be increased by defining the percentages of species more precisely.