Modeling of Water Temperature, Dissolved Oxygen, and Fish Growth Rate in Stratified Fish Ponds using Stochastic Input Variables

A computer model has been developed to simulate water temperature, dissolved oxygen, and fish growth in a stratified fishpond using stochastic weather variables as input. The model consists of generated weather variables and calculated water quality and fish growth rate. The weather variables are generated using Monte Carlo methods, and include solar radiation, air temperature, and wind speed and direction. The water quality parameters are state variables that include water temperature, dissolved oxygen (DO), phytoplankton (in terms of chlorophyll a, Chla), and total ammonia nitrogen (TAN). Water temperature and DO are predicted at three depths in the water column and the other state variables are assumed to be uniformly distributed. Fish growth rates are predicted under the effects of weather variables and water quality for various pond fertilization treatments. The model has been calibrated and validated using data from the Pond Dynamics/Aquaculture Collaborative Research Program (PD/A CRSP) database. To evaluate the model's performance for different pond management strategies and different locations, model simulations were compared to data collected from 36 fishponds with 11 fertilization treatments in Thailand, Rwanda, and Honduras sites.The comparisons of simulations and observed data indicate that the model is capable of predicting water temperature and DO stratification and fish growth for simulations up to six months long. The simulated results indicate that water temperature and DO are affected by weather variables, especially solar radiation. Changes in Chla and DO are affected by environmental conditions and fish grazing. The stochastically generated weather variables have little influence on fish growth. Fish growth rate is affected by changes in Chla and fertilization rate because the model assumes that phytoplankton is the preferred food for tilapia.The current model is limited by the uncertainty of available weather data and the corresponding limitati