4. APPLICATION OF THE GEOSPATIAL STREAMFLOW AND MIKE BASIN
4.3. GeoSFM application in flood forecasting in the Zambezi Basin 94
4.3.2. Rainfall-runoff parameterisation data in GeoSFM 99
4.3.2.2. Land cover 103
The USGS Global Land Cover Characteristics (GLCC) database, derived from 1 km Advanced Very High Resolution Radiometer (AVHRR) data, was used to generate the flow velocity and hydrographs for the entire Zambezi River Basin. Considering that land surface influences the flow velocity and hence the runoff generation and overland flow processes, the non-uniform velocity grids approach (USGS, 2000a) is applied by the GeoSFM model to generate the unit hydrographs representing the response of each basin to rainfall input events. The land cover and soil water holding capacity were used in combination to estimate the Soil Conservation Service (SCS) runoff curve numbers in GeoSFM as follows:
Unit hydrograph: As a semi-distributed hydrologic model, GeoSFM requires a single input
value of precipitation and other forcing data for each sub-basin during each modelling time step. A unit hydrograph is developed to simulate the typical response of the basin to a uniformly distributed rainfall input event for each sub-basin. According to Ramírez (2000), three types of synthetic unit hydrographs are possible: (1) those relating hydrograph characteristics (time to peak, peak flow, etc.) for watershed characteristics (Snyder, 1938; others); (2) those based on a dimensionless unit hydrograph (Soil Conservation Service, 1972), and (3) those based on models of watershed storage (Clark, 1943). GeoSFM uses the Soil Conservation Service approach and generates watershed storage hydrographs to describe a typical response for each sub-basin to rainfall events. The default approach for estimating overland velocity from land cover uses the Manning’s Equation, with values of hydraulic radius assigned to each cell based on drainage area, as shown in Equation 4.11 (Asante, et al., 2007b).
∗ ∗ √ Equation 4.11
where RH = hydraulic radius; HILLSLOPE = average elevation change divided by the average flow length from each cell to the basin outlet; MANNINGN = Manning roughness for the dominant land cover in the sub-basin; Velocity = Average overland velocity in the sub-basin. Cells with drainage areas greater than 5 000 km2 were assumed to be river cells and velocities
ranging from 0.3 m s-1 to 1.5 m s-1 are directly assigned, based on drainage and slope (USGS,
2000a). For non-river cells, Manning’s roughness values were estimated based on the land cover type related to the soils group as identified by the Anderson Code (Asante et al., 2007a). Therefore the distribution of discharge at the catchment outlet is given by the probability density function (PDF) of travel times in the basin (Equation 4.12), as described in USGS (2000a) and which is the time taken to cover the distance of a certain (given) flow length.
t Equation 4.12
where
t
iis the travel time from a given grid cell to the basin outlet in days; li is the flow length
in m from a given grid cell to the basin outlet and v
1 is the average overland velocity in m 3 s-1
for the basin.
SCS Runoff Curve Number: The RCN is used by GeoSFM to generate a unit hydrograph
and simulate the typical response of the sub-basin to a uniformly distributed rainfall event for each sub-basin. The SCS curve numbers assigned to the different soil hydraulic classes were described in Section 4.2.1. Figure 4.9 shows a sample of simulated hydrograph results based on the capability of each selected sub-basin to respond to the rainfall events. The approach is represented in GeoSFM model by the hypothetical unit response of each sub-basin in terms of runoff volume and timing to a unit input of rainfall, in daily time scale. The odd shapes of the hydrographs may be explained by taking into the consideration that way the model is integrating different responses over large areas.
It can be seen in Figure 4.9 that the Mazoe (185), Tete (180), and Marromeu (235) sub-basins are releasing, on average, 40% of the effective rainfall received during the first day. These sub-basins have smaller drainage areas, with higher peak flow than the median and large sub-basins (such as the Upper Zambezi (227), Cuando (230), Lupata (162), Luangwa (151), Barotse (148), Manyame (175) and Revubue (144) sub-basins) which release between 15% and 35% of the rainfall during the first day. In this application the responses were used to identify sub-basins with similar responses at the calibration stage. Therefore, in this study, sub-basins with similar responses were assumed to have similar parameters.
227) 175) 151) 180) 185) 144) 191) 235)
Figure 4.9: A dimensionless unit hydrograph and cumulative mass curve. Graphs generated using various soil classes and land cover types for each selected sub-basin namely: Cuando (230), Barotse (148), Upstream Zambezi (227), Kafue (169), Lupata (162), Luangwa (157), Manyame (175), Luia (151), Tete (180), Mazoe (185), Revubue (144), Shire (191) and Marromeu (235)
Maximum cover: GeoSFM used the GLCC data to represent the impervious area
GeoSFM, the default MAXCOVER is 1 for areas which represent water bodies and MAXCOVER=0 for impervious areas. The model uses the MAXCOVER results to determine the excess amount of precipitation (in (mm) millimetres) on each sub-basin being modelled where the precipitation cannot be infiltrated into the soil layer or used by evapotranspiration processes. As the upper soil layer becomes wetter, a larger percentage of the basin acts as if it was impervious, contributing to surface runoff generation (in mm) in the model. For this study all the existing water bodies (such as lakes, rivers, and large wastelands) were parameterised using the land cover land use classification data. On areas classified as impervious in the land cover maps, runoff (in mm) is generated directly by GeoSFM. The approach generates many uncertainties – because the GLCC data and the modelling structure are not considering the processes of expansion and contraction of the areas covered by the water (USGS, 2000b; Asante et al., 2007b). Another shortcoming of this approach relates to the spatial resolution of the land cover data, which makes it impossible to represent all existing water bodies in the Zambezi Basin – this may result in an indeterminate amount of direct runoff not being included in the calculations.