Nesting Algorithm Development ROMS 3.6 -r605 ROMS 3.6 -r605 ROMS 3.7 -rXXX ROMS 3.7 -rXXX NOAA...
-
Upload
austin-whitt -
Category
Documents
-
view
234 -
download
2
Transcript of Nesting Algorithm Development ROMS 3.6 -r605 ROMS 3.6 -r605 ROMS 3.7 -rXXX ROMS 3.7 -rXXX NOAA...
ean M od
arch C o m
r a i n - F o l l o w
M o d e l i n g
Nesting Algorithm Development ROMS 3.6 -r605 ROMS 3.7 -rXXX
NOAA HeadquartersWashington, D.C
February 28, 2012
Hernan G. ArangoIMCS, Rutgers University
Acknowledgements
Andrew M. MooreU. California Santa CruzAdjoint-Based Algorithms
Alexander F. ShchepetkinU. California Los AngelesNonlinear Kernel
David J. RobertsonIMCS, Rutgers UniversityCyber Infrastructure
Kate S. HedstromU. Alaska, FairbanksUser Community Forum
John C. WarnerUSGS, WHOISediment Transport, Nesting
Development Strategy
The nesting option is implemented in three phases:
• Phase I, released on April 2011 as version 3.5: Change ROMS numerical kernels (NLM, TLM, RPM, ADM) to allow different horizontal i- and j-ranges in DO-loops to permit operations on various nested grid classes (refinement, mosaics, and composite) and nesting layers (refinement and composite combinations).
This facilitates the computation of any horizontal operator (advection, diffusion, gradient, etc.) in the nesting overlap regions and avoids the need for cumbersome lateral boundary conditions on the model variables and their associated flux/gradient values. The advantage of this approach is that it is generic to any discrete horizontal operator. The overlap region is an extended section of the grid that overlays an adjacent grid.
The strategy is to compute the full horizontal operator at the contact points between nested grids instead of specifying boundary conditions.
Tile I- and J-Ranges
Istr = BOUNDS(ng) % Istr(tile)
IstrB = BOUNDS(ng) % IstrB(tile)
IstrM = BOUNDS(ng) % IstrM(tile)
IstrP = BOUNDS(ng) % IstrP(tile)
IstrR = BOUNDS(ng) % IstrR(tile)
IstrT = BOUNDS(ng) % IstrT(tile)
IstrU = BOUNDS(ng) % IstrU(tile)
Iend = BOUNDS(ng) % Iend(tile)
IendB = BOUNDS(ng) % IendB(tile)
IendP = BOUNDS(ng) % IendP(tile)
IendR = BOUNDS(ng) % IendR(tile)
IendT = BOUNDS(ng) % IendT(tile)
Jstr = BOUNDS(ng) % Jstr(tile)
JstrB = BOUNDS(ng) % JstrB(tile)
JstrM = BOUNDS(ng) % JstrM(tile)
JstrP = BOUNDS(ng) % JstrP(tile)
JstrR = BOUNDS(ng) % JstrR(tile)
JstrT = BOUNDS(ng) % JstrT(tile)
JstrV = BOUNDS(ng) % JstrV(tile)
Jend = BOUNDS(ng) % Jend(tile)
JendB = BOUNDS(ng) % JendB(tile)
JendP = BOUNDS(ng) % JendP(tile)
JendR = BOUNDS(ng) % JendR(tile)
JendT = BOUNDS(ng) % JendT(tile)
Istrm3 = BOUNDS(ng) % Istrm3 (tile) Istr-3Istrm2 = BOUNDS(ng) % Istrm2 (tile) Istr-2Istrm1 = BOUNDS(ng) % Istrm1 (tile) Istr-1IstrUm2 = BOUNDS(ng) % IstrUm2 (tile) IstrU-2IstrUm1 = BOUNDS(ng) % IstrUm1 (tile) IstrU-1
Iendp1 = BOUNDS(ng) % Iendp1 (tile) Iend+1Iendp2 = BOUNDS(ng) % Iendp2 (tile) Iend+2Iendp2i = BOUNDS(ng) % Iendp2i (tile) Iend+2 interior Iendp3 = BOUNDS(ng) % Iendp3 (tile) Iend+3
Jstrm3 = BOUNDS(ng) % Jstrm3 (tile) Jstr-3Jstrm2 = BOUNDS(ng) % Jstrm2 (tile) Jstr-2Jstrm1 = BOUNDS(ng) % Jstrm1 (tile) Jstr-1JstrVm2 = BOUNDS(ng) % JstrVm2 (tile) JstrV-2JstrVm1 = BOUNDS(ng) % JstrVm1 (tile) JstrV-1 Jendp1 = BOUNDS(ng) % Jendp1 (tile) Jend+1Jendp2 = BOUNDS(ng) % Jendp2 (tile) Jend+2Jendp2i = BOUNDS(ng) % Jendp2i (tile) Jend+2 interiorJendp3 = BOUNDS(ng) % Jendp3 (tile) Jend+3
Suffix:
R : tile RHO-points B : Boundary tile RHO- and V-pointsU : tile U-points M: Boundary tile PSI- and U-pointsV : tile V-points P : Nesting PSI-, U-, and V-points
T : Nesting RHO-points
If not nesting grids, the additional boundary tile indices associated with nesting are set to:
IstrT = IstrR full range, starting I- direction (RHO-point)IendT = IendR full range, ending I- direction (RHO-point) JstrT = JstrR full range, starting J-direction (RHO-point)JendT = JendR full range, ending J-direction (RHO-point)
IstrP = Istr full range, starting I- direction (PSI-, U-point)IendP = Iend full range, ending I- direction (PSI-point)JstrP = Jstr full range, starting J-direction (PSI-, V-point)JendP = Jend full range, ending J-direction (PSI-point)
IstrB = Istr interior range, starting I- direction (RHO-, V-point)IendB = Iend interior range, ending I- direction (RHO-, V-point)JstrB = Jstr interior range, starting J-direction (RHO-, U-
point)JendB = Jend interior range, ending J-direction (RHO-, U-
point)
IstrM = IstrU interior range, starting I- direction (PSI-, U-point)JstrM = JstrV interior range, starting J-direction (PSI-, V-point)
Boundary Tile Indices
Boundary Tile Indices Locations
mod_param.F
get_bounds.F
Development Strategy
• Phase II, released on September 2011 as version 3.6: Major overhaul of ROMS lateral boundary conditions . All C-preprocessing options were eliminated and replaced with logical switches (LBC structure) that depend on the nested grid, if any.
This facilitates, in a generic way, the processing or not of lateral boundary conditions in applications with nested grids. In nesting applications, the values at the lateral boundary points are computed directly in the overlap region by the numerical kernel.
The logical switches allow different lateral boundary conditions types between active (temperature and salinity) and passive (biology, sediment, inert, etc.) tracers.
The lateral boundary condition switches for each state variable and boundary edge are now specified in ROMS input script file, ocean.in.
Lateral Boundary Conditions Structure
TYPE T_LBC logical :: acquire process lateral boundary data
logical :: Chapman logical :: clamped logical :: closed logical :: Flather logical :: gradient logical :: nested logical :: nudging logical :: periodic logical :: radiation logical :: reduced END TYPE T_LBC
TYPE (T_LBC), allocatable :: LBC(:,:,:)
For example, for free-surface gradient boundary conditions we have:
LBC(iwest, isFsur, ng) % gradientLBC(ieast, isFsur, ng) % gradientLBC(isouth, isFsur, ng) % gradientLBC(inorth, isFsur, ng) % gradient
u2dbc_im.F
For Example, in zetabc.F the western boundary conditions are:
IF ( DOMAIN (ng) % Western_Edge(tile) ) THEN
IF ( LBC (iwest, isFsur, ng) % radiation ) THEN…
ELSE IF ( LBC (iwest, isFsur, ng) % Chapman ) THEN…
ELSE IF ( LBC (iwest, isFsur, ng) % clamped ) THEN…
ELSE IF ( LBC (iwest, isFsur, ng) % gradient ) THEN…
ELSE IF ( LBC (iwest, isFsur, ng) % closed ) THEN
DO j = Jstr, JendIF ( LBC_apply (ng) % west ( j ) ) THEN ! Allows both specified
andzeta ( Istr-1, j, kout ) = zeta ( Istr, j, kout ) ! nested conditions
END IFEND DO
END IF
END IF
Lateral Boundary Conditions Code
Standard Input File: ocean.in
Development Strategy
• Phase III, to be released soon as version 3.7: includes the nesting calls to ROMS main time-stepping routines, main2d and main3d. The concept of nesting layers is introduced to allow applications with both composite grids and refinement grids. Several new routines are added to process the information that it is required in the overlap region, what information needs to be exchanged from/to another grid, and when to exchange it.
In mosaic and composed grids, the information is exchanged between each sub-time step call in main2d or main3d. For example, the data donor grid and the mosaic/composite grids need to sub-time step the 2D momentum equations before any of them start solving and coupling the 3D momentum equations.
In refinement grids, the information at the contact points is processed at the beginning of the full time-step layer. The exchange between data donor and refinement grids is two-way.
main_3d.F
nesting.F
Nesting Configuration Types
Composite Grids
Case Grid Layer Donor West South East North
a1 1 2 2 LBC LBC LBC
2 1 1 LBC LBC 1 LBC
b1 1 2 LBC, 2 LBC LBC LBC
2 1 1 LBC LBC 1 LBC
c1 1 2 LBC, 2 LBC LBC LBC
2 1 1 LBC LBC, 1 1 LBC
Mosaic Class:
Ngrids = 2NestLayers = 1GridsInLayer = 2Ncontact = 2
Composite Class:
Ngrids = 2NestLayers = 1GridsInLayer = 2Ncontact = 2
Composite Overlap Class:
Ngrids = 2NestLayers = 1GridsInLayer = 2Ncontact = 2
a b c
Composite Grids: Complex Estuary
Grid Layer Donor West South East North
1 1 2, 3, 4, 5, 6 LBC, 2, 3 LBC LBC, 4, 5, 6 LBC
2 1 1 LBC LBC 1 LBC
3 1 1 LBC LBC 1 LBC
4 1 1 1 LBC LBC LBC
5 1 1 1 LBC LBC LBC
6 1 1 1 LBC LBC LBC
Composite Estuary Class:
Ngrids = 6NestLayers = 1GridsInLayer = 6Ncontact = 10
Refinement Grids
Case Grid Layer Donor West South East North
a1 1 n/a LBC LBC LBC LBC
2 2 1 1 1 1 1
b
1 1 n/a LBC LBC LBC LBC
2 2 1 1 1 1 1
3 2 1 1 1 1 1
4 3 2 2 2 2 2
Refinement Class:
Ngrids = 2NestLayers = 2GridsInLayer = 1 1Ncontact = 2
Multi-Refinement Class:
Ngrids = 4NestLayers = 3GridsInLayer = 1 2 1Ncontact = 6
a b
Composite-Refinement Grid Combination
Grid Layer Donor West South East North
1 1 2,3 LBC, 2 LBC LBC LBC
2 1 1 LBC LBC 1 LBC
3 2 1 1 1 1 1
Composite-Refinement Class:
Ngrids = 3NestLayers = 2GridsInLayer = 2 1Ncontact = 4
(Rancic and Purser 1996)
Cubed-Sphere grid
Regular grid mosaic
Refined grid mosaic
Mosaics Grids
Tripolar Grid
(Murray 1996, Griffies et al 2004)
Monterrey Bay Nested Grids
3:1 Ratio
5:1 Ratio
Tripolar Grid: ORCA R025(1/4o at the Equator, 1442x1021 points)
Orthographic Projection
Stereographic Projection
Contact Regions and Contact Points
integer :: Ncontact total number of contact regions
TYPE T_NGC
logical :: coincident coincident donor and receiver points, p=q=0logical :: interpolate perform vertical interpolation
integer :: donor_grid data donor grid numberinteger :: receiver_grid data receiver grid numberinteger :: Npoints
integer, pointer :: Idg (:) donor grid, cell I-left indexinteger, pointer :: Jdg (:) donor grid, cell J-bottom indexinteger, pointer :: Kdg (: , :) donor grid, cell K-index
integer, pointer :: Irg (:) receiver grid, I-contact pointinteger, pointer :: Jrg (:) receiver grid, J-contact point
real(r8), pointer :: Hweight (: , :) horizontal weightsreal(r8), pointer :: Vweight(: , : , :) vertical weights
END TYPE T_NGC
TYPE (T_NGC), allocatable :: Rcontact (:) RHO-points TYPE (T_NGC), allocatable :: Ucontact (:) U-points TYPE (T_NGC), allocatable :: Vcontact (:) V-points
Contact Points Structure
Contact Points Interpolation
Hweight (1, :) = (1 - p) * (1 - q)Hweight (2, :) = p * (1 - q)Hweight (3, :) = p * qHweight (4, :) = (1 - p) * q
Value (Irg, Jrg) = Hweight (1,:) * F2d(Idg ,Jdg )+ Hweight (2,:) * F2d(Idg+1,Jdg )+ Hweight (3,:) * F2d(Idg+1,Jdg+1)+ Hweight (4,:) * F2d(Idg ,Jdg+1)
If coincident contact points between data donor and data receiver grids, p = q = 0.0,
Hweight (1, :) = 1.0 Value (Irg, Jrg) = Hweight (1, :) * F2d( Idg, Jdg )Hweight (3, :) = 0.0Hweight (4, :) = 0.0Hweight (5, :) = 0.0
3 (Idg+1, Jdg+1, Kdg-1)
2
4
(Idg, Jdg, Kdg-1) 1 Irg
Jrg
q
1- q
1- p
p
3 Kdg-1
7 Kdg
56
4
1 2
8
5
Suffix: dg = donor grid rg = receiver grid
Contact Regions Metadata: Additional NetCDF File
Nested Grids: Mosaic Class
Ngrids = 2NestLayers = 1GridsInLayer = 2Ncontact = 2
Nested Grids: Composite Class
Ngrids = 2NestLayers = 1GridsInLayer = 2Ncontact = 2Donor Grid = blueReceiver Grid = red
Nested Grids: Composite Class
Ngrids = 2NestLayers = 1GridsInLayer = 2Ncontact = 2Donor Grid = redReceiver Grid = blue
Nested Grids: Composite Overlap Class
Ngrids = 2NestLayers = 1GridsInLayer = 2Ncontact = 2Donor Grid = blueReceiver Grid = red
Nested Grids: Composite Overlap Class
Ngrids = 2NestLayers = 1GridsInLayer = 2Ncontact = 2Donor Grid = redReceiver Grid = blue
Nested Grids: Composite Estuary Class
Ngrids = 6NestLayers = 1GridsInLayer = 6Ncontact = 10Donor Grids = redReceiver Grid = blue
Nested Grids: Composite Estuary Class
Ngrids = 6NestLayers = 1GridsInLayer = 6Ncontact = 10Donor Grid = blueReceiver Grids = red
Nested Grids: Refinement Class
Ngrids = 2NestLayers = 2GridsInLayer = 1 1Ncontact = 2Donor Grid = blueReceiver Grid = red
Nested Grids: Refinement Class
Ngrids = 2NestLayers = 2GridsInLayer = 1 1Ncontact = 2Donor Grid = redReceiver Grid = blue
Nested Grids: Multi-Refinement Class
Ngrids = 4NestLayers = 3GridsInLayer = 1 2 1Ncontact = 6Donor Grid = blueReceiver Grid = red
Ngrids = 4NestLayers = 3GridsInLayer = 1 2 1Ncontact = 6Donor Grid = redReceiver Grid = green
Nested Grids: Multi-Refinement Class
Ngrids = 4NestLayers = 3GridsInLayer = 1 2 1Ncontact = 6Donor Grid = redReceiver Grid = blue
Ngrids = 4NestLayers = 3GridsInLayer = 1 2 1Ncontact = 6Donor Grid = greenReceiver Grid = red
Nested Grids: Composite-Refinement Class
Ngrids = 3NestLayers = 2GridsInLayer = 2 1Ncontact = 4Donor Grid = blueReceiver Grids = red
Nested Grids: Composite-Refinement Class
Ngrids = 3NestLayers = 2GridsInLayer = 2 1Ncontact = 4Donor Grids = redReceiver Grid = blue
Python Based GUI
Python Based GUI
Remarks
• ROMS nesting capabilities are unique and allows complex estuary and coastal configurations with unlimited number of composite and refined grids
• Coincident composite and mosaic grids produce identical solutions when compared to one large continuous grid
• Complete overhaul of ROMS numerical kernels: NLM, TLM, RPM, and ADM
• Development divided in three phases. Phase I released as ROMS 3.5 on April 25, 2011. Phase II released as ROMS 3.6 on Sep 23, 2011. The final Phase III will be released as ROMS 3.7 soon…
• A Python-based GUI is currently under development to facilitate building input scripts in complex configurations
• We need to start looking to more sophisticated grid generation tools
Warner, J.C., W.R. Geyer, and H.G. Arango, 2010: Using composite grid approach in complex coastal domain to estimate estuarine residence time, Computer and Geosciences, 36, 921-935, doi:10.1016/j.cageo.2009.11.008.
Reference