Evaluation of a simulation model forpredicting early morning oxygen depletion in tropical brackishwater tilapia ponds

An early-warning simulation model developed by C. E. Boyd and colleagues to predict critical early-morning dissolved oxygen (DO) levels in fishponds was evaluated. DO lost or gained by diffusion was the most uncertain factor in solving the predictive equation, but appeared to vary consistently for all ponds according to wind conditions. Simple linear regression comparison between measured and predicted DO values at dawn, adjusted afterwards for diffusion, resulted in r² = 0.94 (P < 0.01). The absolute difference between actual and predicted early morning DO values was ≤ 0.34 mg·lˉ¹, averaging 0.26 mg·lˉ¹ or 15%, but values < 1 mg·lˉ¹ varied by as much as 50%. If diffusion could be estimated more accurately, the model would be acceptable for general management purposes. A simple graphical projection method, plotting DO decline during the first hours of darkness versus time, and requiring no other information, was also evaluated, and found to predect DO depletion much less precisely under prevailing conditions, and with less advance warning.