comparison with field observations (e.g., Amein and Chu 1975) has demonstrated 

 that the time step can be selected in accordance with the resolution of the 

 available field data. For example, if boundary condition data are provided at 

 hourly intervals, time steps of 1 or 0.5 hr are adequate. The use of small 

 time steps does not improve accuracy of the solution because the resolution in 

 the field data controls the final answer. Because the numerical procedure 

 used in DYNLETl is implicit, the computation is unconditionally stable, and 

 large time steps are possible. DYNLETl overrides the time step if it is 

 larger than the interval between two consecutive time -dependent data inputs. 



91. A computation time step of the same magnitude as the interval 

 between data inputs is satisfactory. Numerical accuracy is not seriously 

 affected by the time-step size, because accuracy also depends on the tolerance 

 value. For tidal inlets, a computation time step of 1,800 sec is recommended 

 unless the input values are specified at shorter intervals or more frequent 

 outputs are desired. 



92. Number of iteration steps . The number of iteration steps executed 

 by the model to reach convergence depends on the tolerance values selected for 

 the variables. For natural channels, convergence is usually attained within 

 three to five iterations. Additional iterations may be required at start-up, 

 where cross -sectional properties vary sharply between cross sections or if 

 there are rapid changes in water surface elevation and flow rate with time. A 

 limit of 20 iteration steps is recommended. If the solution does not converge 

 within a reasonable number of iterations, one of the following situations may 

 be the cause : 



a. The time step may be too large for the problem. 



b. The tolerance values may be too small for the given physical 

 situation. 



c. There is an error in cross-sectional geometrical data, in the 

 friction coefficients, or in the transition loss coefficients. 



d. The model fails because it is applied to a physical situation 

 violating the assumptions underlying its development. 



93. External boundary conditions . External boundary forcing and the 



boundary type must be identified. Presently, DYNLETl accepts four types of 



external boundary conditions: 



a. Type 1 . Values of water surface elevation as a function of 

 time are tabulated in the input file. 



37 



