DSM2 batches input data into files or "layers" in order to achieve the following goals:
For example, consider the two layers of channels in the figure below. The first layer defines seven channels and would have seven entries in the CHANNEL table. This might represent a "base" grid. The second layer changes the properties of Channel 2, adds a Channel 8 and removes Channel 7. The second layer will have only three entries, shown in red. These entries represent the changes relative to Layer 1, and presumably are thematically related.
Consider the above example using text input. We are going to create CHANNEL tables representing the channel connectivity, and assume the geometry is provided with CSDP style cross-sections listed in an XSECT table (child items are always associated with parent items in the same file).
The base data will be in a file channel_base.inp:
<file channel_base.inp> CHANNEL CHAN_NO LENGTH MANNING DISPERSION UP_NODE DOWN_NODE 1 18000 0.030 0.80 1 2 2 8000 0.040 0.80 2 3 3 18000 0.040 0.80 3 4 4 18000 0.040 0.80 4 5 5 18000 0.040 0.80 3 5 6 22000 0.040 0.80 5 6 7 14000 0.040 0.80 6 7 END XSECT CHAN_NO DISTANCE FILE 1 0.200 1_0_200.txt 1 0.800 1_0_800.txt 2 0.500 2_0_500.txt ... 7 0.900 7_0_900.txt END <eof channel_base.inp>
The revisions are in channel_revise.inp:
<file channel_base.inp> CHANNEL CHAN_NO LENGTH MANNING DISPERSION UP_NODE DOWN_NODE 2 8000 0.030 0.80 2 3 # Masks + Alters #3 9000 0.000 0.00 19 20 # Has no effect ^7 14000 0.040 0.80 6 7 # Masks + Deletes ... 8 16000 0.040 0.80 8 3 # Adds END XSECT CHAN_NO DISTANCE FILE 2 0.100 2_0_500.txt # Masks lower level x-sects 2 0.700 2_0_500.txt # 7 0.900 7_0_900.txt # Will be ignored 8 0.500 8_0_500.txt # END <eof channel_base.inp>
The two layers are managed by the model input file that is given directly to the model, in this case hydro_layering.inp. The two channel files are listed in a GRID include block that lists the layers in increasing priority.
<file hydro.inp> GRID channel_base.inp channel_revise.inp END <eof hydro_layering.inp>
Now lets consider the details...
Include blocks are input blocks in the master file that list other files. The data from these other files is "included" in the order listed. Priority is given to files read later, and these are assigned a higher "layer number"
Include blocks can only contain specific types of input data. For instance, a GRID input block only contains channel, gate, reservoir and transfer physical specifications (not boundary conditions attached to them). So the trick to using include blocks is knowing, say, that a CHANNEL table belongs in a file in a GRID include block and BOUNDARY_FLOW table belongs in a file in a HYDRO_TIME_SERIES block. In the reference documentation, the include blocks should be listed for each table in the Table Information section..
The only exception is the master file that is the one sent to the model on the command line (often named something like hydro.inp, qual.inp, ptm.inp). Data in this file always take precedence over other input.
Layer overriding occurs when the same data item is defined in multiple layers (files) in the same model. Files that are read later are given a higher "layer number" and take precendence over files read earlier. Within the same file it is an error to redefine an entry.
To use layering, you have to know what constitutes redefining an entry. Whether two items represent the same data item depends on the identifier for the table, which is some combination of columns that uniquely identify the item using a name or number. Identifiers for each table are listed in the reference documents. In the above example it is channel number CHAN_NO. The trickiest identifiers are in the output, because they involve two (NAME, VARIABLE) or three (NAME, VARIABLE,SOURCE_NAME) columns. In the reference documentation, the identifier is listed for each table in the Table Information section.
When parent-child tables are present in a file (e.g., Channels, Cross Section, Cross Section Layer), overriding is assessed at the level of the parent or top-level table. When you override on a top-level table, its child table information is completely replaced as well. So, for instance, the cross-section at Channel 2 Distance 0.500 in channel_base.inp in the example is completely ignored. The model makes no attempt to "mix it in" with the replacement version of Channel 2.
Child tables must be in the same file as their parent tables. This is a departure from earlier versions of DSM2, but is necessary to make layering well-defined.
Occasionally, the motivation for overriding an item is to eliminate it. You can do this on any top-level table by prepending a carat ^ at the beginning of the line. This will remove items on lower levels with the same identifier. Note that it doesn't matter what other data you put in the parent fields (you do need placeholders). Also you needn't add child information if the only reason for the parent entry is to delete it -- but sometimes it is nice to have the child data there if you are toggling back and forth.
Deleting data is quite different from commenting it out (using a # sign). Commenting data out on a high level would merely mean that the input reader would skip over the line. It would not affect any data with the same identifier on a lower level.