For calibration of hydraulic models you should consider some important points: 1) Is flow hydrograph estimated accurately for you desired storm event? the wrong estimation of this model input can lead to the huge amounts of underestimations or overestimations...
2) manning roughness coefficient along the river bed is one the most important ones...however by reducing or increasing this parameter you can change the water surface and flood extents and compare this results by observed datasets...
3) time step interval and mesh size are the most important ones and you shouldn't ignore them.. Now in HEC-RAS 2D-5.04 an option added that tries to force model to decrease the courant number under 1...
4) some numerical coefficients must be checked before running the model..
in user manual you can find more information about this model and its parameters...
also, at the following link you can find more about this model:
If you have trusted measured data (e.g. flow hydrograph, waterlevel, water depth, flood extent, etc), you can use roughness(friction) values to caliberate your model results according to the measured data.
It is important to note that the number of cross sections used and the grid size of your topography can affect your model output (among others).
you need to precise what is the objective of your hydrodynamical modelling using 2D approach. HEC RAS 2D may be less reliable in hydraulically complex areas due to poor ability to address spatial shifts in water velocity.
Its all of create the boundaries for your study area, get the measurement of Cross section of rivers boundaries/roughness coefficient/levels/discharges and run the program.