W. Huang et al. / Ocean Engineering 30 (2003) 22752295
2287
Table 3
Statistical summary of comparison between model predictions and observations of water levels
Output Station
Input NOAA
Training
Verification
Correlation
Root-mean-
station
period
period
value (r)
square error (m)
P2
Montauk
12/1999
04/1999
0.985
0.0667
P8
Battery
11/2000
03/2000
0.977
0.0529
5.2. Model testing in predicting water levels over yearlong periods
After the model had been satisfactorily trained and verified over a short period,
the model is capable of predicting long-term water levels at the coastal inlets of
Long Island South Shore. Water level data from NOAA stations at Montauk and
Battery are available from the 1940's to the present. Therefore, using the RNN--
WL neural network model, long-term water levels over several decades in the inlets
of Long Island's south shore can be predicted. These long-term model predictions
will be very helpful in studying the long-term circulation and shoreline change in
the region. Comparisons between model predictions and observations for yearlong
periods are given in Fig. 8 for Station P2, and Fig. 9 for Station P8. Results indicate
very good correlation (above 0.95) between model predictions and available yearlong
data (excluding missing data gaps). This good correlation using data over yearlong
Fig. 8. Comparison of water levels between model predictions and observations at station P2 of Shinnec-
ock Inlet for year 2000. The distance between input NOAA station at Montauk and the inlet is about
60 km.
Previous Page
