Flood models are a valuable part of a hydrologist’s toolkit: they are used to map flood risk, assess flood damage, invest in future flood control strategies, and forecast flooding to ensure timely evacuation and short-term mitigation measures. Different modeling approaches may be employed depending on the scale and purpose of the model [59].

Empirical Method Models

Utilize in situ measurements, survey data, and remote sensing data to produce geographic information system (GIS) data and statistical models about flood extents and depths. Results from these models may be used alone or as inputs or validation data for more detailed hydrodynamic models. Traditionally, data from hydrologic gaging stations (such as the USGS StreamStats database) have been utilized in empirical models. However, the high-quality resolution of modern satellite imagery has made remote sensing data a powerful flood modeling tool in recent years. Synthetic aperture radar (SAR) is valuable for flood inundation mapping, while Shuttle Topography Radar Mission (STRM) and Light Detecting and Ranging (LiDAR) digital elevation models provide high-quality topographic data. LiDAR data can have accuracy on the order of a few centimeters, but there is a tradeoff between accuracy and data processing time. In general, higher accuracy requires more LiDAR collection points and increased data preprocessing to remove noise, meaning greater expense [59]. 

1D Hydrodynamic Models

Simulate the movement of water using mathematical equations that capture physical laws. These models can vary considerably in complexity and scope. 1D models use data captured at specific cross-sections along the centerline of a river or stream to model channel and surface floodplain flow. The inputs to this type of model include upstream flows (either calculated using a hydrologic model or determined using empirical data), channel bed elevations and slopes, channel roughness, floodplain elevations and roughness, and the dimensions of any in-stream structures such as bridges or culverts. While 1D models are computationally efficient, the choice of cross-section location is a decision that has significant ramifications for modeling outcomes, since the cross-sections are in effect snapshots of the characteristics of the overall channel. These simplified models also do not take lateral wave movement into consideration. Many designs (including most of FEMA’s flood insurance studies) use a 1D modeling approach utilizing HEC-RAS, a free software developed by the U.S. Army Corps of Engineers. 

2D Hydrodynamic Models 

Represent floodplain flow as a two-dimensional surface flow, with the assumption that flow depth on the floodplain is minimal. Unlike the 1D cross-sectional approach, 2D models use either a rectangular (structured mesh) or a triangular (unstructured mesh) grid to spatially resolve the individual calculations made by the computer. More advanced models also allow flexible meshes, which enable different sizes of grid cells within the model depending on the complexity of the surface being modeled at a given location. 2D models are often coupled with 1D models, with the 1D model used to model channel flow and the 2D model used to model the floodplain. They are the most common type of model used in medium-scale flood mapping and flood risk estimation. HEC-RAS can be run as a 2D model, a 1D model, or a coupled 1D/2D model. 2D models require many of the same inputs as required for the 1D model, but over a defined spatial extent rather than at individual cross-sections. In addition, 2D models must include the location of breaklines: barriers to flow over the floodplain such as levees, roads, or natural embankments. 2D models are highly computationally expensive. General industry guidance is that 2D models not be used for areas larger than 1000 km2 where the required resolution is less than 10m [59].

3D Hydrodynamic  Models  

Are the most computationally expensive hydrodynamic models. In most cases, a 3D model is unnecessary for flood extent mapping, especially given the spatial availability of data for appropriate model calibration and validation. However, modeling of 3D features such as vertical vortices or spiraling flow at river bends may be appropriate during high-velocity turbulent flow such as during a flash flood, dam breach, or ice jam flood. 

Simplified Conceptual Models

The final modeling approach, typically are used for large, data-sparse floodplains. These models are highly simplified representations of physical processes with shorter runtimes than traditional hydrodynamic models. An example of this type of model is a “planar model” that derives flood depth by intersecting a high-resolution digital elevation model with planes at fine elevation intervals to create a relationship between the volume stored and the flood elevation, then calculating the flood extent based on flood volume and depth. Simplified conceptual models are appropriate for modeling large areas (over 2000 km2 in area) or doing probabilistic analysis that requires many model runs. They are robust in situations with well-defined flow paths but are not appropriate in areas with complex topography or in situations where flow velocity is important [59].

[59] J. Teng, et al., “Flood inundation modelling: A review of methods, recent advances,” Environmental Modeling and Software, vol. 90, pp. 201–216, 2017.