MIKE-SHE hydrological model: How to export SZ flow from a larger model to import into a smaller, nested model to inform SZ boundary conditions

Question

We are working on a MIKE-SHE coupled MIKE-Hydro model for a small wetland in South Africa. One of the major issues is that it has zero-flux boundaries around the model domain in the saturated zone (SZ), and therefore ignores any inflow from its boundaries in the SZ. This is unrealistic as this is a very fine-scale model, and there is very likely to be movement of water subsurface into the wetland and this may well be significant. We therefore need to correct the SZ boundary conditions.

One idea to do this was to use a nested approach, building a larger model for the catchment which would then aim to simulate the boundary conditions of the smaller model domain. The reason we didn't just work with a larger model to start with is that we thought it may mask the small scale effects we were wanting to investigate. We thought this was the best approach, however we have no idea how to output the SZ flow from the larger model, to input it into the boundary of the smaller model. Given the functionality of this model and all the grids we are able to export, it must be possible to do this. But how? Can we input a shapefile and ask for a timeseries (dfs0) file to be extracted for that domain from the large model? And then input this into the SZ boundary of the smaller model? Or is there another way to do this?

I enclose a map of the region of interest showing the nested approach, the smaller model domain is indicated by the yellow polygon, and the larger model domain by the blue polygon. The scale of the map is also indicated.

Answer 1

The most suitable approach would depend on what data is available. If there is groundwater level data close to the upslope boundaries of the target wetland or information about groundwater gradients in the landscape, or the gradient across the wetland itself, this data can be used to determine a head or gradient boundary at the edges of the wetland. This would not entail modeling the whole contributing area. If there is reason to believe that the groundwater gradient across the wetland and surrounding area more or less follows the gradient of the topography, this could be used to assign gradient boundaries around the wetland.

I am assuming there is not groundwater data like this for your site and that is why you want to model the surrounding area. It is hard to tell from the image what is being used to delineate the larger area being shown around the target wetland. For the larger model, I would suggest modelling the contributing catchment of the wetland down to the outlet of the wetland or a point further down the stream for which there is streamflow data. One could use a smaller surrounding area if there was groundwater level data to set the boundary conditions for it.

Groundwater flux timeseries can be obtained for a "cross section" line drawn in the gridded groundwater model output. This can be used to determine a timeseries of flux across the boundary into the target wetland area on the up-gradient side and out of the wetland on the down-gradient side, if a gradient exists. Alternatively a timeseries of heads can be extracted for given cross section lines to be input as time varying head boundaries. Values can be averaged to create dfs0 timeseries for different boundary segments. If there is head data from boreholes or piezometers to compare to the modelled values, using the head values in the boundary condition would be a good option.

The MIKE-SHE Manual v2, section "11.18.22 Outer boundary conditions", states that for the "Head" boundary option, one can input a "time varying dfs2 file of head, which is typically extracted from a regional results file. This can be done using the MIKE Zero Toolbox Extraction tool: 2D Grid from 3D files. MIKE SHE then interpolates in both time and space from the .dfs2 file to the local head boundary at each local time step." (DHI) This is not something I have attempted myself, so cannot give guidance on how to do this using the Toolbox.

When modelling the larger catchment you may also need to allow there to be a head, gradient, or flux boundary at the downstream end of the catchment to get reasonable results.