Template:GUM4

From Gsshawiki
Jump to: navigation, search


4.1 Units

With few exceptions the units used for input and output in GSSHA are in SI units. Units for particular processes are consistent with those found in the literature. For instance, hydraulic conductivity is specified in units of cm/h, while rainfall rates are specified in mm/h. The particular units expected of each input are given in the project file card descriptions. The discharge hydrograph may be output in cfs by specifying the QOUT_CFS flag.

Furthermore, the model was developed based on the Universal Tranverse Mecator (UTM) coordinate system, as described in Section 4.4.


4.2 Grid Size

Distributed models are used over a very wide range of grid sizes, from 10 to 1000 m. The selection of an appropriate grid size for a GSSHA model requires consideration of both the available data and computational effort required. Typical grid sizes range from 10 to 250 m. The selection of the grid size for a given watershed determines the total number of grid cells used to describe the watershed, setting the computational effort and memory required. Note that if the grid size is halved, the memory required and computational time increase by a factor of 4. In general, smaller grids are less sensitive to sub-grid variability for Hortonian runoff (Ogden and Julien, 1993). Therefore, smaller grids are generally “better”, if the data exists to assign relevant watershed characteristics to each grid cell. However, smaller grid-sizes do not guarantee superior model performance. As with the time step, the most appropriate grid size can be determined with a convergence study, where the effects of increasing or decreasing the grid size can be observed on model output.

4.3 Total Event Simulation Time

This project file card is used to specify the total GSSHA event simulation time in minutes. If the volume of water remaining on the surface at the end of the simulation is greater than 5% of the rainfall volume, the total simulation time is too short to capture the entire runoff hydrograph, and a warning is printed in the run summary (SUMMARY card) file. The TOT_TIME card is ignored for continuous simulations. For continuous simulations the end of an event is the time at which the outflow discharge falls below the discharge specified by the EVENT_MIN_Q project file card. For continuous events the EVENT_MIN_Q is used only for accounting purposes, except for the last event when it is also used to stop the simulation. For some simulations the EVENT_MIN_Q value may never be reached. In that case the user should specify the END_TIME card. The END_TIME card can be used to stop any long term simulation at any desired point. The simulation may end before the specified END_TIME if conditions dictating the end of the simulation are encountered before the specified END_TIME.

4.4 Coordinate System

GSSHA performs calculations on raster grids and data can be input with GRASS ASCII raster data. While any logical, and consistent, coordinate system can be used, the preferred coordinate system for raster based data is the UTM map projection. The UTM System breaks the entire earth into zones 6 degrees of longitude wide. The zones that cover the continental United States are:

ZONE WEST LIMIT EAST LIMIT
10 126° W 120° W
11 120° W 114° W
12 114° W 108° W
13 108° W 102° W
14 102° W 96°W
15 96° W 90° W
16 90° W 84° W
17 84° W 78° W
18 78° W 72° W
19 72° W 66° W


The standard specification of the UTM system includes (Davis et al., 1981):

  • The reference ellipsoid is Clarke 1866 in North America.
  • The origin of longitude is the central meridian.
  • The origin of latitude is the equator.
  • The unit of measure is the meter.
  • A false easting of 500,000 m is used for the central meridian of each zone.
  • The scale factor at the central meridian is 0.9996.
  • The zones are numbered beginning with 1 for the zone between 180° W and 174° W meridians and increasing to 60 for the zone between meridians 174° E and 180° E.
  • The latitude for the system varies from 80° N to 80° S.
  • In the southern hemisphere, a false northing of 10,000,000 m is used.
  • The scale error is 1/2500 on the central meridian.

The UTM coordinate system was chosen because of its global applicability and widespread acceptance. Note that watershed data that lie in two zones must be merged into one zone. The data should be merged into the zone that contains the majority of the watershed area. This is accomplished using a GIS.

4.5 Map Headers

Many spatially varied inputs for GSSHA can be input as ASCII GRASS GIS raster maps. Raster maps are defined as maps that assign attributes to areas, as opposed to vector maps, which assign attributes to lines or polygons. GSSHA requires that all raster grid cells are square.


Each raster map, including index maps, must have a header that conforms to the following example:

north:   4156000.0
south:   4135000.0
east:   601800.0
west:   575000.0
rows:   105
cols:   134

The entries for north, south, east, and west are the coordinates of the bounding rectangle that contains the entire watershed. The rows and cols entry in the header respectively contain the number of rows and columns in the watershed. Note that for this particular example header, that (north-south)/rows = 200 m, and (east-west)/cols=200 m. Therefore, the grid size of this particular model is 200 m, and the raster cells are square, as required by GSSHA. As the coordinate values in the map header are not used to determine geographic position on the globe, any values for north, south, east, and west may be used in the map headers provided that the grid size is accurate, the grids are square, and your maps are consistent. If the grid size calculated from the bounding rectangle, rows, and columns is not equal to the value specified in the project file using the GRID_SIZE project file card, GSSHA will not run. The header in each GRASS ASCII map must be IDENTICAL. This requirement forces the user to ensure that all maps are of the same geographic region and grid size.

The header is followed by rows and columns of space delimited data values. These values can be either integer or real, depending on the data type. For example, the mask, channel link, channel node, and index maps are by definition integer maps; maps containing the land surface and bedrock elevations contain real values.

4.6 Watershed Mask

The required WATERSHED_MASK project file card is used to input the name of the file containing the mask map. The mask map is used to define the watershed boundaries within the rectangular grid and reduces memory requirements by an amount directly proportional to the ratio of mask area over the bounding rectangle area. The watershed mask is a map containing only 0s and 1s, where 0 and 1 represent grid cells that are outside and within the watershed, respectively. For instance, a watershed that is approximately the shape of Texas would have the following a mask map file:

north: yyyyyyy.y1
south: yyyyyyy.y2
east:  xxxxxx.x1
west:  xxxxxx.x2
rows:  15
cols:  12
0 0 0 0 1 1 1 0 0 0 0 0
0 0 0 0 1 1 1 0 0 0 0 0
0 0 0 0 1 1 1 0 0 0 0 0
0 0 0 0 1 1 1 0 0 0 0 0
0 0 0 0 1 1 1 1 1 1 0 0
0 0 0 0 1 1 1 1 1 1 1 1
0 0 0 0 1 1 1 1 1 1 1 1
0 0 0 0 1 1 1 1 1 1 1 1
1 1 1 1 1 1 1 1 1 1 1 1
0 1 1 1 1 1 1 1 1 1 1 1
0 0 0 1 1 1 1 1 1 1 1 0
0 0 0 0 1 0 1 1 1 1 0 0
0 0 0 0 0 0 1 1 1 0 0 0
0 0 0 0 0 0 1 1 1 0 0 0
0 0 0 0 0 0 0 1 0 0 0 0


If the watershed mask contains a 0 in a particular grid cell, GSSHA ignores data in all input maps for that particular grid cell. In the above map, there are a total of 15*12=180 grid cells. There are 85 grid cells with a 1 in the watershed mask. In this example, GSSHA would only allocate memory for 85 of the 180 grid cells for each map, representing a decrease of 53% of memory required to store the data for each map. Errors in watershed delineation (hence mask creation) will propagate through all data sets input to and output from GSSHA. The watershed should be delineated with care. WMS creates the watershed mask as part of the watershed delineation processes. GRASS users can use the r.watershed command to create the mask. During each simulation, the GSSHA model writes out the watershed mask with the internal number assigned to each grid cell in GRASS ASCII format in a file called “maskmap”. This file is useful for deciphering error messages from GSSHA that refer to the grid cell number where the problem occurred.

4.7 Elevation Map

Elevation data are input for every active grid cell in a GRASS ASCII map file specified with the ELEVATION project file card. The elevation data are perhaps the most important inputs for GSSHA modeling. The quality of the elevation data plays a major role in success of GSSHA simulations. Elevations in the grid are derived from digital elevation model data (DEM).

DEMs always contain errors. Large flat areas in the DEM may be due to the limited vertical resolution of elevation data from which the DEM was derived. Extensive flat areas usually cause problems for the 2-D explicit diffusive-wave overland flow routing used in GSSHA. Digital dams, pits, and depressions in the DEM may be artifacts of the interpolation scheme used to rasterize digitized contours, or due to coarse resolution in regions of concave topography.

In addition, the grid size used in a GSSHA model is normally coarser than the available DEM data. Elevation data in the grid must be somehow interpolated from the DEM data and lumped into larger areas. This process can introduce additional error in the elevation of grid cells. As a rule, the user must cross check the elevation values with in-field observations or topographic maps of the area. Digital topographic maps are often available and can be displayed as a background image in WMS.

One way to discover potential errors in the elevation data is to perform a simulation with the most basic GSSHA model: a single event with uniform rainfall, overland flow, a relatively short time step, and no other options. Surface depth output maps should be written frequently (see DEPTH and MAP_FREQ project file cards). If the simulation finishes without an error, the surface depth maps should be examined to determine where most water accumulates and whether such accumulations are justified by the topographic map of the watershed. Alternatively, the model may crash. The location of grid cells where problems occurred will be printed on the screen and also at the bottom of the run summary file.

Editing of the elevations in the grid is often necessary to impose the actual drainage trend observed in the topographic map. Digital dams, pits and depressions must be removed since they trap surface runoff that would otherwise contribute to the outlet discharge. Using grids with raw elevations requires shorter computational time steps, while properly prepared grids, particularly those with coarser resolution, allow use of longer time steps. If channel routing is performed, care should also be taken to ensure that overland flow runoff reaches grid cells that contain channel links. If the stream network is delineated independently from the DEM, e.g. from a digital line graph (DLG), then the elevations of the grid cells containing channel nodes should be checked to insure they are not higher than those of the surrounding cells. Otherwise the overland flow will not be correctly passed to the channels.

The cleandam.exe routine can be used to do much of the above. Cleandam.exe is described in a another section of the manual.

WMS offers the option of importing and displaying vector stream location files (DLG) to aid in stream channel delineation. WMS users can also automatically delineate streams from the DEM using the tools in the WMS software. GRASS users may use the r.watershed command to automatically delineate the streams. As either automatic delineation relies on the DEM to determine stream locations, there may be substantial differences from the DLGs. If an automated method is used to locate the stream locations from the DEM data and the grid resolution used in the GSSHA model is coarser than the available DEM data the stream may not fall in the lowest elevation grid cells. However the stream is located, the elevations of the cells containing the stream may need to be manually adjusted to ensure proper overland-channel interaction.

4.8 Optimizations

Senarath et al. (2000) demonstrated that the CASC2D model could be effectively parameratized by use of an automated calibration procedure, such as the shuffled complex evolution (SCE) method (Duan et al., 1992). Output that allows optimizations based on event peak discharge and event discharge volume can be produced by using the OPTIMIZE card. This card specifies a file that contains the peak discharge and discharge volume for each event in the rainfall file. Peak discharge (cms) and volume (m3) are written for the outlet and also at any locations in the stream network specified in the IN_HYD_LOCATION file. The last line contains the total outlet discharge volume for the entire simulation. The output format is:

Event # Peak at outlet Discharge volume at outlet Peak at first stream gage Discharge volume at first stream gage Peak at next stream gage Discharge volume at next stream gage