NPZfc: An ecological relation based fish catch prediction model using Artificial Neural Network

Quantifying interactions of organisms of the various trophic levels is very important in understanding the dynamics of aquatic ecosystems. With regards to fish, as both ecologically and commercially important parts of an ecosystem, predicting their catch in relation to primary producers provides insight into sustainable management. The present paper describes a novel model NPZ f c , for nutrients, phytoplankton, zooplankton and fishes, which can predict planktivorous fish catch. Unlike the existing models which deal with the interactions within the system through mathematical equilibrium, the proposed model uses artificial neural network (ANN) to automatically learn inter-dependencies between different related variables and predict the fish catch of a water body. The efficiency of the model was increased by refining the input variables. Here biomass of plankton species population (phyto-plankton and zooplankton) were specifically selected from feeding ecology studies of target fish species as input variable. The study of two of the commercially important fish species, Etroplus suratensis and Nematalosa nasus in Chilika lagoon showed that the model can predict with high accuracy from limited input data. The root mean square error (RMSE) is found to be very satisfactory, ranging from 3.53% to 11.5% for E. suratensis and from 1.63% to 2.22% for N. nasus . Higher accuracy and better predictive ability with a smaller dataset makes this ANN-based NPZ f c model more conducive.