Lumped UZ Calculations (Column Classification)

Note: The column classification should probably be avoided today because the models have become more complex, MIKE SHE has become more efficient and computers have become faster.

Calculating unsaturated flow in all grid squares for large-scale applications can be time consuming. To reduce the computational burden MIKE SHE allows you to optionally compute the UZ flow in a reduced subset of grid squares. The subset classification is done automatically by the pre-process­ing program according to soil and, vegetation distribution, climatic zones, and depth to the groundwater table. The unsaturated flow conditions in two cells are identical when they have

·         identical soil- and vegetation characteristics, AND

·         identical boundary conditions.

If these two conditions are met, then the calculations need only be made in one of the cells and the results transferred to the other cell.

In practical terms, the first condition is usually not a serious restriction since most models are divided into several homogeneous soil zones. The second condition, however, is much more restrictive. Fluctuations in the groundwater table usually vary from cell to cell, and spatial variations in rainfall and the topography cause overland flow and infiltration to vary continuously across the domain.

However, if homogeneous zones can be defined based on

·            Topography,

·            Meteorology,

·            Vegetation,

·            Soil, and

·            Bypass characteristics,

then a representative cell for the zone can be defined and used for the UZ-calculations. If this is done, then the boundary conditions from the represent­ative cell (i.e. infiltration rate, evapotranspiration loss and groundwater recharge) can be transferred to the other cells within the zone.

Such an approximation does not introduce any water balance errors, but it can influence the dynamics of the simulation. However, an intelligent group­ing of the cells can reduce computational burden considerably.

The initial definition of homogeneous zones can be made using the depth to the groundwater table and the soil, vegetation, and rainfall distributions. It is often necessary to re-group the columns several times during the calibration phase, until the groundwater regime is reasonable calibrated. Also, when the groundwater table is shallow, smaller intervals are usually required.

Column classification can decrease the computational burden considerably. However, the conditions when it can be used are limited. Column classifica­tion is either not recommended or not allowed when

·         the water table is very dynamic and spatially variable because the classi­fication is not dynamic,

·         if the 2 layer UZ method is used because the method is fast and the ben­efit would be limited,

·         if irrigation is used in the model because irrigation zones are not a classi­fication parameter, and

·         if flooding and flood codes are used, since the depth of ponded water is not a classification parameter

If the classification method is used, then there are three options for the classi­fication:

·        Automatic classification

The automatic classification requires a distribution of groundwater eleva­tions (see Groundwater Depths used for UZ Classification). This can be either the initial depth to the groundwater based on the initial heads, or you can supply a .dfs2 map of the groundwater elevations. In both cases, you must supply a table of intervals upon which the classification will be based. The number of computational columns depends on how narrow the intervals are specified. If, for example, two depths are specified, say 1 m and 2 m, then the classification with respect to the depth to ground­water will be based on three intervals: Groundwater between 0 m and 1 m, between 1 m and 2 m, and deeper than 2 m.

One tip is to extract a map of the calculated potential head in the very upper saturated zone layer from a previous simulation. The map should represent the time of the year when the largest variations of the ground­water table are expected (deep groundwater in the hills and shallow groundwater close to the rivers). Repeat the procedure as calibration improves.

If the Linear Reservoir method is used for the groundwater, then the Interflow reservoirs are also used in the classification. However, since feedback to the UZ only occurs in the lowest Interflow reservoir of each subcatchment, the Interflow reservoirs are added to the Automatic Clas­sification in two zones - those that receive feedback and those that don’t.

·         Specified classification 

Alternatively a data file specifying Integer Grid Codes, where UZ compu­tations are carried out can be specified, with grid codes range from 2 up to the number of UZ columns (see Specified classification). The location of the computational column is specified by a negative code and the sim­ulation results are then transferred to all grids with the an equivalent pos­itive code. For example, if a grid code holds the value -2 a UZ computation will be carried out for the profile located in that grid. Simula­tion results will subsequently be transferred to all grid codes with code value 2. An easy way to generate a .dfs2 file to be used for specification of UZ computational columns is to let the MIKE SHE setup program gen­erate an automatic classification first, and subsequently extract the UZ classification grid codes. The extracted .dfs2 file can be edited in the 2D editor as desired and used to specify UZ computational grids.

·        Calculated in all Grid points (default)

For most applications you should specify that computations are to be carried out in all soil columns.

·         Partial Automatic 

Finally a combination of the Automatic classification and the Specified classification is available. If this option is chosen an Integer Grid Code file must be provide (see Partial automatic classification) with the follow­ing grid codes: In grid points where automatic classification should be used the grid code 1 must be given. In grid points where computation should be performed for all cells the grid code 2 must be given.