Coverage of structural options within modelling tools

All of the tools reviewed have some flexibility in how a catchment model can be set-up. This adds a layer of complexity when trying to compare approaches across tools. In this review an effort has been made to document and consider the main structural options available and approaches considered ‘typical’ for each; however, it should be noted that not all potential options have been covered.

This page clarifies the focus of coverage for each modelling tool reviewed in the wiki. This is perhaps most important for MIKE-SHE which has the largest number of structural options. 

WRSM-Pitman

WRSM-Pitman and SPATSIM-Pitman are both based on the original Pitman model structure (Pitman, 1973), but have diverged from one another in various ways. A major addition in both tools has been the separate representation of groundwater, allowing for groundwater withdrawals and impacts on streamflow to be explicitly considered. Two different methods have been developed, named for their developers: the Hughes method (Hughes, 2004) and the Sami method (Sami in Bailey and Pitman, 2016).

The WRSM-Pitman tool includes the option to implement:

• the original Pitman model algorithm (lumped subsurface storage)
• the Hughes groundwater algorithm
• the Sami groundwater algorithm

In descriptions on this wiki WRSM-Pitman is described with the Sami groundwater algorithm, for the following reasons:

• The SPATSIM-Pitman tool includes the Hughes groundwater representation and not the Sami method. The SPATSIM-Pitman formulation of the Hughes method has evolved since its incorporation into WRSM-Pitman, and it has received more testing as applied in SPATSIM-Pitman.  
• The WRSM-Pitman tool was used for the Water Resources of South Africa 2012 Study (WR2012) (Bailey and Pitman, 2015) in which it was applied to all quaternary catchments in South Africa, calibrated where possible, using the Sami groundwater option.

Another addition in both tools has been development of daily timestep versions. However these have been rarely applied to date and there has been less testing and experience in calibrating them. As such, WRSM-Pitman is described with a monthly timestep.

SPATSIM-Pitman

In this wiki, SPATSIM-Pitman is described with the Hughes groundwater algorithm (Hughes, 2004) and a monthly timestep.  A daily timestep version exists, but has been rarely applied to date and there has been less testing and experience in calibrating it.

ACRU4

There is currently not an equivalent theory manual for ACRU4 comparable to the ACRU3 theory manual (Schulze, 1995) describing the available algorithms, so the ACRU3 manual was used as a reference. It should be noted that some of the process representation options presented in the ACRU3 theory documentation, such as modelling shallow groundwater tables, are only relevant when using subcatchments as the unit of calculation rather than HRUs, and are not available in ACRU4.

SWAT2012

In SWAT2012 (ArcSWAT2012 & QGIS version) there is choice of two infiltration and rain-event runoff generation algorithms to use (Neitsch et al., 2011):

• the Soil Conservation Service Curve Number (SCS-CN) method
• the Green-Ampt Mein-Larson method, only applicable when using sub-daily rainfall inputs (hourly, half-hourly)  

In this wiki, SWAT2012 is described with the SCS-CN method for calculating runoff generation during rain events for the following reasons:

• It is less common to access sub-daily rainfall data required for the Green-Ampt Mein-Larson method
• The SCS-CN approach has been used in SWAT versions since the tool originated and is the option more commonly applied at present, most likely due to data availability

The choice of which option is applied in a model has relevance to several aspects of its process representation. The SCS-CN method is an empirical equation originally developed to predict runoff generation alone, rather than being part of a comprehensive catchment water balance model. If the SCS-CN approach is used, canopy interception is not explicitly modelled. This means the impact of canopy interception is implicitly considered in the calculations of other processes in the model, which changes how processes like infiltration and ET from soil can be viewed conceptually. The Green-Ampt Mein-Larson option does allow canopy interception to be calculated as a separate process in SWAT2012.  

MIKE-SHE

Schematic diagram of process representation algorithm options in MIKE-SHE and their compatibilities

MIKE-SHE includes a wider diversity of process algorithms, several with associated levels of spatial discretisation, than the other tools reviewed. It allows users to build models with very different levels of complexity. For most processes there are also representation algorithm options that are more conceptual, often needing fewer parameters, and those that are more physics-based. In any MIKE-SHE model a computational grid must be defined and several process algorithms will be solved by grid cell. However, parameters and climate inputs can be specified for zones, polygons of areas considered to have relatively uniform properties, rather than for each grid cell individually. Options are provided to model the flow of surface runoff across the landscape using a sequence of larger zones in a subcatchment (“simple overland flow routing”), rather than from grid cell to grid cell. Groundwater storage and flow can also be represented by linear reservoirs in subcatchments rather than in the full 3D finite-difference grid.

A certain representation option for one process or component may only be compatible with a particular set of other structure and algorithm choices. A tree of the major structural options and compatibilities in MIKE-SHE is shown to the right.

To facilitate comparisons across tools within this review, MIKE-SHE is described for two model set-up approaches at opposite ends of the complexity spectrum:

• using all the semi-distributed, more conceptual options (simpler calculation)
• using all the fully-distributed, more physical options (full complexity).  

These approaches are highlighted in the options tree diagram and the algorithm options for each are also listed in a table below. 

It is important to note that these particular combinations of options are not enforced by the MIKE-SHE tool: other combinations can be used, as highlighted in the diagram. (See tool documentation for more details.)

MIKE-SHE set-up option combinations used in inter-comparisons

Process	MIKE-SHE, semi-distributed, more conceptual	MIKE-SHE, fully-distributed, more physical
Surface runoff (overland flow) routing	Simple overland flow routing (across series of zones within subcatchments)	Finite difference solution of diffuse wave (flow between grid cells)
Evapotranspiration	“Two-layer” method (by grid cell)	Kirstensen & Jensen method (by grid cell)
Soil water (unsaturated zone) storage & movement		Richard’s equation method or gravity flow method (by grid cell layer)
Interflow & groundwater processes	Linear reservoirs (within subcatchments)	Finite difference solution of Darcy’s Law (between 3D grid elements)