Some results and comments on the Tidal Flow Forum exercise

Two approaches are followed to estimate the fit between ‘observations’ and the results obtained with different numerical models within the frame of the Tidal Flow Forum exercise. The first approach (i.e. the computation of the RMS errors on the timeseries for a specific day) is shown to be insufficient because the RMS error computed over a short period of time is strongly dependent on the selected test period. Such RMS values are significant if and only if they are computed over a sufficiently long period of time, T. Hence, the model validation must be done on the basis of the results of long term simulations. The limit of the RMS values as T becomes arbitrarily large is related to the maximum error made on each constituent. Those maximum errors are easily calculated from the harmonic constants (computed versus observed) of the constituents. The results of the calculation of the maximum errors indicate that the M₂, S₂, N₂ and M₄ components are the largest contributors to the overall error for the area under consideration. On the basis of maximum errors, the TFF results obtained so far are almost of the same order of accuracy. The errors affecting the TFF results are not much larger than those affecting most reported results of similar models used in similar applications