Nitrogen is a key element in aquatic environments and an important pond management variable. In aquaculture systems, nitrogen accumulation eventually leads to a deterioration of the system. The interactions between various N-species are complex and difficult to integrate. Modelling can improve our ability to evaluate this complex system. This paper integrates existing knowledge about nitrogen transformations in fish ponds into a model that calculates the amount of various N-compounds in the water column and in the sediment. The model is also used to gain insight into the relative importance of transformation processes between the various N-compounds. The model was divided into three modules: fish, phytoplankton and sediment-water. The fish module is based on physiological and bio-energetic principles. The phytoplankton dynamics module is based on physico-chemical principles of alga growth. The water–sediment module is based on the bacterial transformations and chemical fluxes of N-species across the water–sediment interface. Relationships and parameters were taken from the literature, except for a few parameters that were estimated by fitting model predictions to observed data. The model was implemented in Turbo Pascal (7.0) using a fixed time step of 1 h and it was calibrated using a set of data from an earthen fish pond stocked with Colossoma macropomum. The validation was performed using data from earthen ponds stocked with Oreochromis niloticus. The difference between the calibrated and validated model was related to the fish species. All concentrations of the various N-species present were simulated well, except the N retained in organic matter in the sediment (average relative error -0.34). Sensitivity analysis revealed that the concentrations of inorganic-N compounds, both in the water column and in the sediment, are more affected by changes in specific parameters included in the fish and phytoplankton modules than other forms of nitrogen in the pond. The model works well, except for organic matter accumulation in the sediment. Further research should concentrate on a better understanding of the bottom organic matter dynamics, to make the model a more comprehensive predictive tool.