CALIBRATION OF A NUMERICAL-MODEL WITH A LIMITED NUMBER OF TIDAL CONSTITUENTS

The response of numerical models to small variations in depth, coeffic ient of bottom friction and the number of tidal harmonics is considere d with particular reference to a model of the Bristol Channel. The mod el is tuned by using an input of 23 largest gravitational and non-line ar constituents from a continental shelf model at its open boundary al ong the longitude 5 degrees west. In the process of tuning the drag co efficient: of the bottom friction, water depth and grid scales are adj usted. The modified version of the model has considerably improved the distribution of the M(2) and S-2 tidal constituents. Subsequently, th e model is tuned for the M(2) and S-2 tides alone to investigate the e ssential adjustments. The results suggest that some additional changes in the parameters are essential when the model is driven with an indi vidual constituent. Small constituents such as the S-2 and N-2 can als o be reproduced quite accurately in addition to the dominant constitue nt. However, the increase required in the drag coefficient of the bott om stress is very large when a small constituent is considered alone. In the case of the Bristol Channel model, it is found that the drag co efficient of friction should be increased by a factor of 1.20 and 4.33 for the M(2) and S-2 constituents, respectively. This factor increase s almost linearly with the ratio of the total variance in the tidal sp ectrum to the variance of the constituent under investigation when the ratio exceeds 7. The influence of all the constituents in the tidal s pectrum on the bottom stress of each constituent can be estimated by c omputer simulations. The adjustments required to improve results for t he principal constituents and non-linear constituents are contradictor y. The model results conform to analytical approximations. It is shown that non-linear constituents generated by a model driven by a single constituent are unreliable. Even those non-linear constituents that ap pear to be generated by the principal constituent under investigation in the model are not reliable. The results from this investigation pro vide some insight into the problem of choosing the value of drag coeff icient of bottom friction for large-scale (ocean-wide) models. These m odels cover regions where the tidal regime changes from semi-diurnal t o diurnal and bottom topography varies significantly in the continenta l shelf and coastal areas. Under those circumstances, a model may requ ire a spatially variable drag coefficient when driven by a limited num ber of constituents (e.g. M(2) or K-1). (C) 1996 Academic Press Limite d