28.02.2013 Views

A Semi-Implicit, Three-Dimensional Model for Estuarine ... - USGS

A Semi-Implicit, Three-Dimensional Model for Estuarine ... - USGS

A Semi-Implicit, Three-Dimensional Model for Estuarine ... - USGS

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

4. Finite-difference Formulation 83<br />

where the overbar ( ) on a layer height or density variable is used to represent a spatial average in the x-direction between adjacent<br />

values; <strong>for</strong> example, hi + 1⁄ 2,<br />

j, k<br />

= ( h<br />

ijk , ,<br />

+ h ) ⁄ 2<br />

i + 1,<br />

j, k . Also, hi + 1⁄ 2,<br />

jk , – 1⁄ is defined to be the average of layer heights<br />

2<br />

hi + 1⁄ 2,<br />

jk , – 1 and hi + 1⁄ 2,<br />

jk , . The average values of the layer heights only are needed in the computations involving the surface and<br />

bottom layers where the heights are permitted to vary horizontally. The substitution un + 1<br />

n + 1<br />

i + 1⁄ 2,<br />

j, k = ( U⁄ h)<br />

i + 1⁄ 2,<br />

j, k has been made in<br />

n + 1<br />

the vertical diffusion term because the dependent variable used in the model is the layer volumetric transport Uk rather than<br />

n + 1<br />

the average layer velocity uk . The layer velocity uk is available in the computer code at time level n − 1 and there<strong>for</strong>e is used<br />

directly in the diffusion term.<br />

The finite-difference equation similar to 4.26 <strong>for</strong> y-momentum is<br />

n + 1<br />

, + , k =<br />

Vij 1⁄ 2<br />

+<br />

, + , k g Δt<br />

-----<br />

Δy<br />

V ˆ ij 1⁄ 2<br />

– hn i, j+ 1⁄ 2,<br />

k<br />

ρi j 1<br />

2<br />

ρn i j 1⁄ 2<br />

n<br />

⎛ , + ⁄ , 1⎞<br />

⎜--------------------- ⎟ ⋅ ( ζn + 1<br />

i, j+ 1 – ζn + 1<br />

ij , + ζn – 1<br />

i, j+ 1 – ζn – 1<br />

ij , )<br />

⎝ ⎠<br />

, + , k<br />

n + 1<br />

n + 1 n – 1<br />

n – 1<br />

⎛ , + , k⎞<br />

⎜ ------------------------------------------------------ ⎟<br />

⎝ ⎠<br />

⎛ n ( V ⁄ h)<br />

i, j+ 1⁄ 2,<br />

k – 1 – ( V ⁄ h)<br />

i, j+ 1⁄ 2,<br />

k<br />

Δt⎜A Vi<br />

-----------------------------------------------------------------------------<br />

, j + 1⁄ 2,<br />

k – 1⁄ 2<br />

n + 1<br />

⎝<br />

hi, j+ 1⁄ 2,<br />

k – 1⁄ 2<br />

vi, j+ 1⁄ 2,<br />

k – 1 – vi j 1⁄ 2<br />

⋅<br />

+<br />

n – 1<br />

hi, j+ 1⁄ 2,<br />

k – 1⁄ 2<br />

n + 1<br />

n + 1<br />

n – 1 n – 1<br />

n ⎛( V⁄ h)<br />

i, j+ 1⁄ 2,<br />

k – ( V⁄ h)<br />

i, j+ 1⁄ 2,<br />

k + 1 vi, j+ 1⁄ 2,<br />

k – vij , + 1⁄ 2,<br />

k + 1⎞⎞<br />

– AVi ⋅ -----------------------------------------------------------------------------<br />

, j+ 1⁄ 2,<br />

k + 1 ⎜ + ------------------------------------------------------ ⎟⎟.<br />

(4.27)<br />

⁄ 2 ⎝ n + 1<br />

n – 1 ⎠⎠<br />

, ⁄ , + ⁄<br />

, ⁄ , + ⁄<br />

hi j+ 1<br />

2 k 1<br />

2<br />

hi j+ 1<br />

2 k 1<br />

2<br />

Here the overbar represents a spatial average in the y-direction between adjacent values; <strong>for</strong> example,<br />

hi j 1⁄ 2<br />

, + , k ( hi, j, k+<br />

hi, j+ 1,<br />

k)<br />

⁄ 2<br />

= ; also hi, j+ 1⁄ 2,<br />

k – 1⁄ h 2 i, j+ 1⁄ 2,<br />

k – 1 + hi j 1⁄ 2<br />

= ( , + , k)<br />

⁄ 2 . Hereinafter the use of overbars is dropped<br />

<strong>for</strong> convenience, and a layer height or density variable possessing a half-integer subscript in any one of the three spatial indices is<br />

considered to be an average of the nearest adjacent values. 35<br />

n + 1<br />

Because the surface layer height is time dependent, hi, j, 1 is unknown <strong>for</strong> the evaluation of equations 4.26 and 4.27. This<br />

value is estimated in the computations by extrapolating in time using the second-order <strong>for</strong>mula<br />

h ˆ n + 1 n n – 1 n – 2<br />

ij1 , , = 3( hij1 , , – hi, j, 1)<br />

+ hij1 , , . (4.28)<br />

n + 1<br />

Using this estimate of hi, j, 1 in equations 4.26 and 4.27 has worked well in numerical testing of the model and is more accurate<br />

n + 1 n<br />

than, say, choosing hi, j, 1 ≈ hij1 , , . The estimate of h ˆ n + 1<br />

i, j, 1 is made first during each time step calculation immediately be<strong>for</strong>e the<br />

n – 2<br />

array that stores values of hi, j, 1 is rewritten; there<strong>for</strong>e, the storage of an extra 3-D array is not required to evaluate equation 4.28.<br />

In equations 4.26 and 4.27, the boundary shear stress terms <strong>for</strong> the wind and bottom friction are boundary conditions <strong>for</strong> the<br />

surface and bottom layers. The wind stress is specified as a <strong>for</strong>cing function at the free surface by substituting<br />

and<br />

n 1 U h<br />

--<br />

2<br />

⁄ ( ) n + 1<br />

n + 1<br />

i + 1⁄ 2,<br />

j, 0–<br />

( U⁄ h)<br />

i + 1⁄ 2,<br />

j, 1<br />

-------------------------------------------------------------------n<br />

+ 1<br />

⁄ , , ⁄<br />

u n – 1 n – 1<br />

⎛ i + 1⁄ 2,<br />

j, 0 – ui + 1⁄ 2,<br />

j, 1⎞<br />

( τxs) + ------------------------------------------------ i + 1<br />

⎜ ⎟<br />

⁄ 2 j<br />

⋅<br />

⎝ n – 1 ⎠<br />

⁄ , , ⁄<br />

1 n<br />

, , ⁄ 2<br />

= ---------------------------ρn<br />

i + 1⁄ 2,<br />

j, 1<br />

AVi + 1⁄ 2 j 1 , , ⁄ 2<br />

hi + 1<br />

2 j 1 2<br />

hi + 1<br />

2 j 1 2<br />

35 In the 3-D model, higher order averages such as the one in equation 4.10 in the 1-D model are not used.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!