A portion of LWI Region 4 was identified as a coastal transition zone (CTZ), as characterized by both riverine and coastal hydrodynamics. To appropriately capture this breadth of complex hydraulic processes, a hybrid numerical model was proposed which coupled HEC-RAS (riverine), ADCIRC (coastal), and SWAN (wave) models. Specifically, the two-way coupled 2023 LCMP ADCIRC+SWAN coastal model was one-way coupled with a new HEC-RAS model. Special care was required in the exact placement of the handoff location between coastal and inland domains, to best leverage the capabilities of each model engine.

Performed model setup for this 2D HEC-RAS model (1400 sq mi domain); topobathymetry, gridded precipitation, spatial infiltration and Manning's N inputs (model setup done within a team of 3, with topobathy help from additional GIS personnel).

Calibrated and validated the model to observed gauge data from 94 available gauges for 5 historical storms.

Performed statistical analysis of 3 spatiotemporal wave characteristics from SWAN model outputs to develop domain of potential boundary condition handoff lines within the ocean.

Automated the following process end-to-end with Python; this was the process for evaluating 1 boundary condition line:

• cut bathymetric contour within this continuous domain of potential boundary lines

• adjust HEC-RAS model mesh to terminate at this handoff line

• sample stage timeseries at 10 equidistant points along this line from ADCIRC model, for each of 7 representative historical storms, identified through multivariate recurrence analysis (carried out by a partnering subconsultant)

• input these timeseries' as downstream boundary conditions for the 7 respective scenarios

• run these 7 HEC-RAS plans in parallel across compute cluster

• compare the results against observed gauge data and calculate 3 loss metrics

• aggregate these into a single trial score across all gauges, metrics, scenarios

Plugged the above evaluation process into an optimization algorithm, and allowed it to compute for 50 trials (testing 50 boundary condition lines and running 350 HEC-RAS simulations), during which it converged on a single optimal handoff location. Model assumptions notwithstanding, this line represents the best-fit handoff location for the average extreme event condition, given compound risk factors of precipitation and storm surge, within a precision of 6 horizontal Ft.

Built the hydrologic model for the LWI program's Whisky Chitto watershed using HEC-HMS. Delineated the watershed's roughly 350 subbasins from LiDAR data from the USGS. Performed sensitivity analysis to determine both the resolution of input LiDAR, and watershed delineation tool which performed best.

Set up hydrologic losses in the model using gridded Deficit and Constant method. This method requires four parameters, Initial Moisture Deficit, Maximum Moisture Deficit, Saturated Hydraulic Conductivity, and Imperviousness to be input in gridded format. Prepared these 4 model input grids, deriving the first three from soil parameters defined in the NRCS gSSURGO geodatabase, and taking Imperviousness from the NLCD. Performed sensitivity analysis evaluating different grid cell sizes for their effect on simulation run time, as well as the accuracy of simulation results.

Ten historical storms were selected for analysis for Region 4 of the LWI program. Prepared gridded precipitation data for the region's models from MRMS radar data, in formats accessible to the HEC-RAS and HEC-HMS numerical models. Filled grids of missing hours by linear interpolation. Clipped spatiotemporal dataset to each watershed's polygon boundary.

The update included building county-wide hydraulic models and prioritizing recommended projects. Worked on the building and stabilization of the HEC-RAS and HEC-HMS models. The new models included NOAA ATLAS 14 rainfall data, 2018 HGAC LiDAR and other updated flood impact data. The latest frequency storm tidal surge timeseries data from USACE Galveston District was used as the model's downstream boundary condition. One of the major challenges of a 1D/2D model is achieving stability. Myself and a colleague isolated each reach of the watershed and stabilized the model one reach at a time. This involved running many trial scenarios adjusting simulation timestep, initial conditions ramp up time, theta factor, baseflow, maximum compute iterations, Manning's N and cell size refinement, 1D river cross section spacing, and other parameters that affect model stability. I allowed the simulation to vary the timestep based on courant number to optimize the timestep for different simulation hours, and then hardcoded these optimized timesteps by hour to ensure consistent results between model runs.

Used Python to automate the MAAPNext level hydrology calculations (BDF Method), streamline the existing conditions model build, perform terrain modifications for proposed project modeling, compare proposed projects, and automate the creation of the report's 100+ GIS exhibits. Used Python, QGIS, ArcGIS, HEC-RAS 6, Civil3D with Subassembly Composer, and PCSWMM, to perform nuanced LiDAR processing to accurately model proposed detention ponds with side slopes, widened, dredged or benched channel sections, and hydraulic structures.

Was the modeler tasked with building and maintaining a SWMM hydraulic model for the city's MDP update in PCSWMM, an abstraction of the EPA's SWMM5 engine. Existing data utilized included geometry from an existing 1D XPSWMM model, TNRIS LiDAR, and Civil3D as-builts. Was able to import the as-built systems verbatim by way of LandXML. The model included storm sewer, open channel, bridges, and detention, with a total of 23,000 entities.

PCSWMM's native support of SQL, GIS, and IronPython was leveraged to streamline both the hydrology calculations and the hydraulic model build. Was involved as a beta tester for a new PCSWMM-integrated cloud computing service, which was utilized for model runs.

Worked with the community to help them understand model results, and refine the model based on their priorities. Later utilized the existing model to size in-line detention system as part of a separate design-build contract.