The quality of fish produced in ponds needs to be ensured. Indeed, pond is often strongly connected to an agricultural watershed, and pesticides are a main health and environmental issue of concern. In this context, the purpose of this study is to highlight the management practices which could impact the pesticide contamination profiles in edible fish and to give recommendations for better practices. A principal component analysis, coupled to a hierarchical cluster analysis, was performed to evaluate temporal evolution of contamination profiles and to assess variability among fish species and among sites according to watershed characteristics. The explicative variables correspond to muscular concentrations of pesticides (azoxystrobin, clomazone, diflufenican, carbendazim, isoproturon, metazachlor, napropamid) in three species of fish (Perca fluviatilis, Cyprinus carpio and Rutilus rutilus), caught in five ponds during two sampling campaigns. Management data are added variables in order to discuss about parameters suspected to be implicated in the contamination profiles recorded. This work shows that high amounts of pesticides applied, short crop rotation durations and bare soil practices led to contamination of sediments and fish and were associated to a ``bad'' management of watershed. Breeding fish that had low masses and establishing the fishing period at the end of winter seemed to be ``bad'' management of pond. Aggravating topological parameters were big watershed coupled to small pond and high proportions of sand soils in the watershed. Reducing amounts of pesticide used (e.g. policy agency plans, farmer acceptance), favouring long-term rotations and inter-cultures, adapting pond creation and fish farming practices to watershed management and topography all could reduce pesticide levels in edible fish and contribute to a better sustainability of the extensive fish farming in pond.