Basic Model: SWAT

SWAT (Soil Water Assessment Tool, Version 2005; Arnold et al. 1993) accounts for most of hydrological processes relevant in the research area, including snow hydrology and channel processes, while maintaining simple approaches (Figure 5-2), so that hydrological complexity and data scarcity are considered at the same time. Selected equations are presented within this section, while others can be found in the manuals (Neitsch et al. 1999; Neitsch et al. 2004).

SWAT2005 has proven capable of simulating hydrological processes in high mountain areas (Cao et al. 2006; Fontaine et al. 2002), as well as in semi-arid zones (Menking et al. 2003;

Hernandez et al. 2000). SWAT has also been applied to a small headwater catchment in the High Atlas (Chaponnière et al. 2008).

Figure 5-2: Simplified schematic diagram of the water movement as simulated by SWAT (modified after Arnold et al. 1993)

The following Table 5-1 provides an overview of the SWAT routines used in this study.

Several other routines (sediment, nutrients) as well as parameters that regulate daily flow (surface runoff delay, etc.) have not been considered due to the scope of this study, which is to assess water availability on a monthly timescale. Therefore the respective parameters have been set to default values. The derivation of the parameters listed in Table 5-1 is lined out in section 5.3.

Surface Runoff (Curve Number)

Infiltration Soil storage

• Evapotranspiration (Penman-Monteith)

• Lateral Flow (Kinematic Wave)

• Percolation

Shallow aquifer

Linear storage

Deep aquifer

Streamflow

• Variable storage

• Transmission Losses Transmission losses

in tributary channels

Alluvial aquifer

Linear storage

Precipitation

Rain Snowfall

Snow Cover

Snow melt

Degree Day Approach

Table 5-1: SWAT-routines and required parameters. Parameters are described in Appendix 3, values are determined in section 5.3. For details on the model routines see (Neitsch et al. 1999)

SWAT routine Process description Associated parameters and variables

Interception

indirectly modeled as initial abstractions in Curve Number method

see Surface runoff

Surface runoff NRCS Curve Number climate data: PCP CN2, HYDGRP

Lateral flow Kinematic storage model SOL_K, HRU_SLOPE, SLSOIL Groundwater flow Linear storage model RCHRG_DP, ALPHA_BF Potential

evapotranspiration Penman-Monteith

climate data: PCP, TMP, WND, RH, RAD CHTMX, GSI, SOL_ALB, SOL_ZMAX,

Losses Effective conductivity Channel width and channel length CH_K1, CH_K2, ALPHA_BNK Crop growth Heat units concept climate data: TMP

ALAI, BLAI, T_BASE, T_OPTLAI

Within SWAT water fluxes are calculated on a daily basis for hydrological response units (HRU), elementary spatial units, considering land cover, soil type slope and indirectly elevation. Within each HRU precipitation occurs either as rain or snow, depending on temperature. Snowfall is added to the HRUs snowpack, which starts melting, when temperature rises. Snowmelt is quantified by a degree-day approach (Fontaine et al. 2002).

Snowmelt and precipitation either infiltrate or generate surface runoff according to the NRCS Curve Number method (US National Resource Conservation Service; NRCS 1986), an empiric rainfall-runoff relationship, taking into account land-cover and antecedent soil moisture condition. Hence infiltration is not directly modeled but regarded as rainfall minus surface runoff (Neitsch et al. 1999):

[-]

Interception is calculated as a by-product of the Curve Number approach. It is therefore independent from the development of the vegetation. Due to the little vegetation cover in the study area, the error can be considered small. It has been noted that the Curve Number method gives a faulty representation when saturation-excess runoff is the dominant runoff type (Easton et al. 2008; Garen & Moore 2005). Since infiltration-excess of the Hortonian type is the dominant runoff-generating process in the catchment (see section 2.3), the procedure appears to be adequate.

Infiltrating water is distributed between the different soil layers according to saturated conductivity and soil water content. The permanent wilting point is calculated as:

[%] hydraulic conductivity. If the saturated hydraulic conductivity of the lower layer is lower than that of the upper layer lateral subsurface flow is quantified by a kinematic wave approach accounting for slope, slope length and saturated conductivity (Sloan & Moore 1984).

Evapotranspiration is calculated using a modified Penman-Monteith approach (Allen et al.

1989a). The approach has been chosen as required data are available in a daily resolution and regional characteristics (aridity, frost, advective conditions) constrain the application of simpler approaches. Due to diurnal distributions of the required climatic variables, the daily calculation of evapotranspiration might result in faulty values, but given the data availability there exist no alternatives. Furthermore in the research area evapotranspiration is usually limited by water availability and not by evaporative demand; therefore the error should be small. While transpiration is calculated directly by the Penman-Monteith approach by using the respective plant parameters, potential evapotranspiration is determined by applying the Penman-Monteith approach to a fictional well-watered Alfalfa grass of 50 cm height. Based on the acquired values the plant specific actual transpiration, evaporation from interception storage and snow are satisfied. The remainder is the evaporative demand from the soil.

This is satisfied by calculating the evaporative demand for a given soil depth (Neitsch et al.

Then the evaporative demand for a soil layer is calculated (Neitsch et al. 1999):

[-]

ESCO represents a conceptual parameter that allows the user to calibrate soil evaporation. If the soil layers are above field capacity, water might as well leave the soil profile downwards, entering the shallow aquifer. Transfer through the vadose zone is simulated by a mean residence time. A fraction of the percolating water is diverted to a deep aquifer and permanently lost from the system. The shallow aquifer is represented by linear storage.

Lateral flow and baseflow directly enter the channel, whereas surface runoff is routed through tributary channels accounting for transmission losses via effective conductivity (Lane 1983).

These transmission losses enter the shallow aquifer directly. The remaining surface runoff enters the primary channel as well. Discharge is then routed using the variable storage concept (Williams 1969), again accounting for transmission losses that are entering bank storage.