hurricane

The ocean/hurricane test group defines meshes, initial conditions, forward simulations, and validation for global, realistic ocean domains with regional refinement. These simulations are forced with time-varying atmospheric reanalysis data for tropical cyclone events and tides. The meshes contain refined regions in order to resolve coastal estuaries, making it possible to simulate the storm surge that results from a given hurricane event. Additionally, the simulations use tidal potential forcing and MPAS-Ocean’s wetting and drying scheme to simulate coastal inundation. The forward simulations can optionally be run with a subgrid-scale correction scheme that accounts for fine-scale bathymetric variation in partially-wet cells, improving the flood prediction accuracy. hurricane currently supports 3 meshes (DEQU120at30cr10rr2, DEVR45to5rr1, RRS6to18) and one storm, Hurricane Sandy.

These tests are configured to use the barotropic, single layer configuration of MPAS-Ocean. Each mesh is created to contain the floodplain, which is used to simulate coastal inundation using the wetting and drying scheme.

The time stepping options to run the simulations include the fourth order Runge-Kutta scheme (RK4), and two local time-stepping schemes. The first LTS scheme is based on a strong stability preserving Runge-Kutta scheme of order three and is called LTS3, see Lilly et al. (2023) for details. The second LTS scheme is based on a forward-backward Runge-Kutta scheme of order two and is called FB-LTS. Each test case in the ocean/hurricane test group has a counterpart for each LTS scheme which is identified by appending the test case name with _lts for LTS3 and _fblts for FB-LTS.

Shared config options

All hurricane test cases start the following shared config options. Note that meshes and test cases may modify these options, as noted below.

# options for spherical meshes
[spherical_mesh]

## config options related to the step for culling land from the mesh
# number of cores to use
cull_mesh_cpus_per_task = 18
# minimum of cores, below which the step fails
cull_mesh_min_cpus_per_task = 1
# maximum memory usage allowed (in MB)
cull_mesh_max_memory = 1000

# Elevation threshold to use for including land cells
floodplain_elevation = 10.0
# Resolution threshold to use for including land cells
floodplain_resolution = 10.0
# Floodplain region extent GEOJSON file
floodplain_geojson = floodplain.geojson
# Minimum depth to enforce outside floodplain region
min_depth_outside_floodplain = 1.0


# options for global ocean testcases
[global_ocean]

# The following options are detected from .gitconfig if not explicitly entered
author = autodetect
email = autodetect


# options for hurricane testcases
[hurricane]

## config options related to the initial_state step
# number of MPI tasks to use
init_ntasks = 512
# minimum of MPI tasks, below which the step fails
init_min_tasks = 512
# maximum memory usage allowed (in MB)
init_max_memory = 1000
# number of threads
init_threads = 1

## config options related to the forward steps
# number of MPI tasks to use
forward_ntasks = 1024
# minimum of MPI tasks, below which the step fails
forward_min_tasks = 1024
# maximum memory usage allowed (in MB)
forward_max_memory = 1000
# number of threads
forward_threads = 1


[hurricane_analysis]

# start and end dates for setting time period in timeseries plots
plot_min_date = 2012 10 24 00 00
plot_max_date = 2012 11 04 00 00

# runs to include in analysis
# others can be added as comma separated list in the format
#   run_name:path/to/file,other_run:path/to/other/file
#   - the name before the : is the label to be use in legend labels
#   - the path after the : is the path to the pointwiseStats.nc file
#     (not including file name)
analysis_runs = MPAS-O:./

plot_station_dems = False

Meshes

The process for creating hurricane meshes is described below in the mesh test case. hurricane currently supports 3 meshes, each with differing levels of refinement. The coarsest mesh (DEQU120at30cr10rr2) uses a global quasi-uniform horizontal resolution of 120 km with 30 km Atlantic refinement and floodplain refinement down to 2 km, and is primarily for testing purposes. The intermediate mesh (DEVR45to5rr1) uses a variable horizontal resolution of 45 km to 5 km scaled by bathymetry with floodplain refinement down to 1 km. The finest mesh (RRS6to18) is reproduced from the global_ocean test group. These meshes are designed to work with the barotropic, single layer hurricane configurations, and they do not include ice-shelf cavities.

DEQU120at30cr10rr2

The quasi-uniform 120 km mesh with regional refinement (DEQU120at30cr10rr2) uses the following resolutions: (1) global horizontal resolution of 120 km, (2) 30 km refinement in the Atlantic Ocean, (3) 10 km refinement along the Mid-Atlantic Bight, and (4) 2 km refinement along the coastal floodplain. The floodplain is defined by a 10 m elevation threshold. This mesh is primarily for testing purposes.

DEVR45to5rr1

The variable-resolution 45 km to 5 km mesh with regional refinement (DEVR45to5rr1) uses the following resolutions: (1) global horizontal resolution ranging from 45 km over the deep ocean to 5 km over the shallow ocean, (2) 1 km refinement along the Mid-Atlantic Bight and the coastal floodplain. The floodplain is defined by a 40 m elevation threshold, and only where refinement exceeds 4 km. This mesh is designed for predicting hurricane flooding along the US East Coast, such as the flooding caused by hurricanes Sandy and Irene.

RRS6to18

The RRS6to18 mesh is a high-resolution global mesh with Rossby-radius-scaling of horizontal resolution from 18 km down to 6 km, designed for E3SMv3. This mesh is reproduced here from the global_ocean test group. The hurricane version of this mesh includes a floodplain with an extent prescribed by a GEOJSON region file. Additionally, the floodplain is constrained by a 20 m elevation threshold, and only where refinement exceeds 16 km. The use of this mesh for flooding simulations is experimental.

Test cases

mesh test case

The mesh test case uses the mesh step from the global_ocean test group to generate the global mesh based on a specified mesh resolution function. Next, bathymetry/topography data is interpolated onto the mesh from the NASA Shuttle Radar Topography Mission 15 arcsecond (SRTM15+) data product. This interpolation step is necessary, because the topography in the floodplain is used to set a mask for the cell culling process. The land cells above the floodplain_elevation are then culled from the mesh. The floodplain can be further constrained by a refinement threshold floodplain_resolution, or a region GEOJSON file floodplain_geojson. Finally, the bathymetry is re-interpolated onto the mesh since this data is not carried over from the cell culling process.

If either LTS option is selected for the mesh test case, an additional step is carried out after the mesh culling. This step appropriately flags the cells of the mesh according to a user defined criterion in order to use time-steps of different sizes on different regions of the mesh. The parallel partitioning is modified accordingly to achieve proper load balancing.

init test case

The init test performs steps to set up the vertical mesh, initial conditions, atmospheric forcing, and parameterized wave and bottom drag, and prepares the station locations for timeseries output.

interpolate atmosphere forcing step

The CFSv2 reanalysis wind vector components and atmospheric pressure fields for the storm event are interpolated onto the horizontal mesh at hourly intervals. These are read in and used to update the atmospheric forcing in the forward run.

create pointstats file step

In order to perform validation of the forward simulation, timeseries data is recorded at mesh cell centers which are closest to observation stations. This step reads in the observation station locations and finds the cells closest to them. A file is created that is the input to the pointWiseStats analysis member for the forward run.

compute topographic wave drag step

The reciprocal of the e-folding time, r_inv, from the HyCOM model, is computed in this step. See Buijsman et al. (2016) for details on the computation. This coefficient is needed to account for the topographic wave drag tendency in the model.

initial state step

The initial state step runs MPAS-Ocean in init mode to create the initial condition file for the forward run. The vertical mesh is set up for a single layer case and the ssh with a thin layer on land for wetting and drying cases.

If the subgrid option is selected, the Digital Elevation Model (DEM) and Land Use/Land Cover (LULC) tiles are processed to create look-up tables for the forward step corrections, and the DEM tiles are averaged to create the coastal and floodplain topography.

If either LTS option is selected for the init test case, the modified partitioning done in the mesh step is used to run MPAS-Ocean init mode.

sandy test case

The sandy test case is responsible for the forward model simulation and analysis.

forward step

The forward step runs the model simulation of the storm. The simulation begins with a spinup period, where the tides and atmospheric forcing are ramped to their full value to avoid shocking the system.

If the subgrid option is selected, look-up tables are used to make the DEM and LULC corrections in the forward mode.

If either LTS option is selected for the sandy test case, the LTS scheme is used to advance the solution in time rather than the default RK4 scheme.

analysis step

The analysis step plots the timeseries data at each observation station to compare the modeled and observed data. Both NOAA and USGS station data are used for the validation.