Comparisons of One-way Nesting Schemes

Various one-way nesting techniques so far presented are tested in order to decide the best technique from a practical standpoint. Nesting schemes tested here are: (1) Schemes with special care to avoid over-specification as a boundary value problem. i) The scheme with simple extrapolation (referred to as SE-scheme) ii) The scheme of SundstrOm (1973) (SS-scheme) iii) The scheme modified by taking acc o unt of gradient value of coarse mesh model in extrapolation (SG-scheme) iv) The scheme of Okamura (1975) (SO-cheme) (2) The scheme of Orlanski (1976) with Sommerfeld's radiation condition (O-scheme) (3) The scheme of Perkey & Kreizberg (1976) in which the tendency is smoothed in the lateral boundary layer (T-scheme) (4) The scheme with Newtonian damping in the lateral boundary layer i) Davies' method (1976) (D1-scheme) ii) The scheme utilizing si m ple Newtonian damping (Do-scheme) Numerical tests for comparison are done under the fol l o wing conditions: (1) Tests in which the phase velocity error is given in the outer domain during the forecasting period. (Test 1) (2) Tests in which isolated disturbance is given in the inner fine mesh domain at intial time. (Test 2)Results of the test are as follows: (1) T-scheme is the best one because of its least deviation from the control run and the computationally stable results. (2) Dr-scheme and Do-scheme are the second best. Little difference is found between the results of the two schemes. Damping coefficient used in these tests se e ms to cause more error than in T-scheme because of its inadequately large value. (3) Among the SE-, SS-, SO-, o-, and SG-schemes, SG-scheme gives the least error. The error in that scheme is the same order as that in D1-scheme within 24 hours foreca s t. But, for longer time integration,, the results of SG-scheme tend to be unsta b l e. (4) When O-scheme is adopted, results are very sensitive to the way in which the phase velocity of the disturbances near the boundary is estimated.