Principles of training Rivers through Tidal Estuaries. 519 



the model, whilst at the close of the experiment it had shifted, as 

 shown, to the Havre channel. Accretion occurred behind the 

 training walls in the model ; and some silting up took place in the 

 Villerville Channel and along the foreshore in front of Trouville, 

 owing apparently to the preference of the main channel for the 

 other outlets, and the diminished capacity of the estuary resulting 

 from accretion. 



This arrangement of training walls was further investigated by 

 working the model for about G300 tides more, with the results shown 

 on Plate 4, fig. 2. The chief features of the estuary in the. model 

 showed only slight changes from the state previously recorded (Plate 

 4, fig, 1), with the exception of the main channel which had shifted 

 again to the central outlet ; whilst the northern foreshore above low 

 water extended over part of the former site of the channel. The two 

 conditions of the estuary, represented by Plate 4, figs. 1 and 2, have 

 therefore the interest of exhibiting in the model a shifting channel, 

 such as actually exists at the present time in the Seine estuary below 

 Berville. 



Scheme F. The last experiment was made on an arrangement of 

 training walls inserted in the model, making the trained channel 

 expand as gently as practicable between Aizier and the sea, whilst 

 retaining the natural width at the outlet (Plate 4, fig. 3) . This is the 

 form of channel which theory indicates as the most suitable;* for 

 whilst it facilitates the influx of the flood tide, it prevents, as far as 

 possible, the abrupt changes in the velocity of a river in passing from 

 its estuary to the sea, which are so prejudicial to uniformity of depth 

 in a channel. It was therefore of interest to ascertain what results 

 would be produced by this theoretical arrangement of training walls 

 in the model, which, in order to leave the outlet free, and thus avoid 

 favouring a progression of the foreshore outside, had to provide a 

 wide channel near Honfleur compared w^th the restricted width 

 available at Quillebeuf. The direction of the channel between Aizier 

 and Quillebeuf, together with the cliffs bordering the river at Quille- 

 beuf and Tancarville Points, determined the maximum width obtain- 

 able at Quillebeuf, and the direction of the channel from Aizier to 

 Tancarville ; and the extension of the training walls in the model 

 from this point was regulated by the necessity of passing close to 

 Honfleur at the south, and not impeding the approach to Havre on 

 the north. The effects produced in the model by working with this 

 arrangement of training walls for about 7300 tides are indicated on 

 the chart (Plate 4, fig. 3). The southern training wall was kept 

 above high-water level all the way to its termination at Honfleur in 

 the model, but the northern training wall was gradually reduced in 

 height from nearly opposite Honfleur towards Havre. The trained 



* ' Rivers and Canals,' L. F. Vernon-Harcourt, p. 236. 

 VOL. XLV. 2 M 



