Continental Scale Water Balance Model

3rd WaterNet/Warfsa Symposium 'Water Demand Management for Sustainable Development', Dar es Salaam, 30-31 October 2002
A Continental Scale Water Balance Model: a GIS-Approach for Southern Africa

B.F. Alemaw1 and T.R. Chaoka2
Geology Department, University of Botswana P. O. Box 0022, Gaborone, Botswana 1 2 E-mail: alemaw@mopipi.ub.bw chaokatr@mopipi.ub.bw
ABSTRACT: A distributed GIS-based hydrological model is developed using GIS and computational hydrology techniques. The model is based on water balance consideration of the surface and subsurface processes. The surface water balance processes include precipitation infiltration, overland runoff and canopy surface interception losses on daily time steps; The subsurface process considers soil moisture accounting on a monthly basis. The model was used to estimate generated runoff from matrix of specific geo-referenced grids representing southern Africa. All regional and seasonal dispensation of water balances have been based on standard GIS formats for storage, spatial display and interpretation of results. Considering the 1961-90 climatic period, we have mapped the regional variation of the mean annual soil moisture (SM), actual evapo-transpiration (AET), and generated runoff (ROF) across the SADC. The model estimates the mean SM of the region to be about 148 mm per year. There is a wide spatial range in the distribution of SM over the due to the fact that the absolute soil moisture is dependent on the water retention properties of the soils considered across the region. The model prediction of the mean annual AET in the region reaches a maximum of 1500 mm, with mean 420 mm. The mean annual generated runoff from the land catchment in the region is about 151 mm/year although there is a significant inter-regional variation among the SADC countries, which is a function of the variation in the vegetation cover, soil and climate variation. Lower runoff regimes are dominant in arid areas in Botswana, Namibia and south-western part of the Republic of South Africa. Higher runoff regimes are the northern and western Tanzania, along the east coastal portions of Mozambique, central Mozambique, western Zambia and Malawi. Key words: water balance model; GIS; regional hydrology; Southern Africa; runoff; soil moisture; evapotranspiration
1.
Introduction
Water balance models have been widely used to simulate regional water balances and to study hydrologic effect of climatic change. The Thornthwaite and Mather (1957) conceptualisation of the catchment water balance for long term monthly climatic condition has become one of the outstanding research areas of water balance modeling of a catchment, region or of a continent. Typical examples are (1) a grid-based model for Latin America of Vorosmarty et al. (1989); (2) the Rhine flow model of Van Deursch and Kwadijk. 1993); and (3) the large scale water balance model for the upper Blue Nile in Ethiopia of Conway (1997). With the advent of increasing computing power and GIS technique, physical based hydrologic modeling has become important in contemporary hydrology for assessing the impact of human intervention and/or possible climatic change on basin hydrology and water resources. The use of distributed physically based (conceptual) hydrological modelling with prudent simplification is appropriate for providing a reasonable solution to large scale hydrological problems associated with planning and optimal allocation of resources. The present work concerns the development of such a model for the region of Southern Africa. The model can enable one to estimate expected monthly soil moisture, actual evapotranspiration and runoff. Such a model may be used for planning and decision support system in the region. Generally distributed type models can be developed either on the basis of dividing the drainage basin into grid meshes of specified resolution or by identifying as many as possible recognisable upstream sub-catchments in a watershed. A water balance model is developed in this study based on grid cell size of 30 by 30 minutes resolution, covering the region of southern Africa in a view to document the seasonal regimes of soil moisture, actual evapotranspiration and runoff.
A Continental Scale Water Balance Model: A GIS-Approach for Southern Africa
2.
Methodology
Model formulation The distributed GIS-based hydrological model (DGHM) developed in this study is based on consideration of surface and subsurface processes. In the subsurface zone, soil moisture accounting technique of Thornthwaite and Mather (1957), modified as part of this study, is used in computing the water balance components like actual soil moisture (SM), evapo-transpiration (AET) and runoff (TRO). The surface abstractions in terms of overland runoff and interception are accounted for by DGHM by using the SCS curve number method (SCS, 1985). The model leads to the creation of high-resolution data sets of SM, AET and TRO over the several catchments in the southern African region, which are represented by geographically referenced grids of 0.5X0.5 degree resolution. The Distributed GIS-based Hydrological Model (DGHM) for the southern African region proposed in this study is based on the combination of: (1) daily overland runoff computation using the curve number (CN) method and daily canopy evaporation or interception losses estimated from normalised vegetation indices (NDVI), in the surface processes; and (2) monthly soil moisture accounting approach in the subsurface zone, The model handles the temporal and spatial variations of the water balance components by first solving the temporal variation of each hydrological component for a single grid cell spatially over the region. This eventually allows for hydrologic regime establishment for the region of southern Africa. When written in volumetric units, water balances can be given by:
SM (t ) = PPT TRO AET F
(1)
where PPT is volume (depth) of precipitation, TRO is volume of runoff, AET is amount actual Evapotranspiration, F is commutative infiltration. S with a subscript denotes component of storage with the subscript M for soil moisture storage. Equation (1) is in the form, which explains that the central idea of water balance modelling is monthly accounting of soil moisture. The water balance model given by Equation (1) can be rearranged to estimate the total run off as follows:
TRO = PPT (SM + AET )
(2)
In Equation (2) PPT is replaced with EPPT, which is the effective precipitation, the precipitation in excess of interception loss and that flows overland, thus infiltrates to the soil.
Description and Structure of the Model The model development procedure and algorithm of the water balance computation at a grid cell (using the gird runoff generation (RG) component) is schematically shown in Figure 3. The precipitation input is first separated into the portion that is runoff over land (DRO) and the portion that infiltrates into the soil and enters the water balance calculation called effective precipitation (EPPT). EPPT is obtained from grid precipitation (PPT) by subtracting daily evaporation as interception losses (ES) and daily overland runoff (DRO) computed by CN method of rainfall abstraction, and aggregating for month of simulation. The subsurface processes of soil moisture, actual evapotranspiration and generated runoff are presented in the subsequent sections. Interception loss The net precipitation that enters the soil, or infiltration, is the aggregate effect of surface processes namely, incoming precipitation, and losses through surface evaporation as interception and direct runoff. In DGHM, therefore, the daily precipitation (P) input is first separated into the following. (i) The portion that is runoff overland (Qo), (ii) The portion that infiltrates (I) and enters the soil, and (iii) The daily wet canopy surface evaporation or canopy interception losses (ewct ). Interception is estimated with the assumption of absence of drainage with the water balance at the canopy described by:
dC = (1 ) p e wc dt
0C S
(3)
where p is the precipitation rate, ewc is the wet canopy evaporation rate, S is the canopy water storage capacity, C is the actual amount of water intercepted by canopy and is the free throughfall coefficient (the proportion of rain which falls to the ground without striking the canopy). From measurements of and LAI of different landuses tabulated in Rutter et al. (1975), Biftu (1998) established regression equation between them, and this is adopted in this study.
= 0.997 0.139 LAI; R 2 = 0.997 as a
Regional information
(PET, PPT, SOIL, NDVI) Surface process
Eff. Precipitation
(EPPT)
Evapotranspiration
Soil Moisture SM Direct runoff (DRO)
Excess s Runoff storage (RO)
Runoff from a grid (TRO)

Fig. 1 Representation of Water balance at a grid cell in the distributed GIS-based water balance model (DGHM).
Modeling daily overland runoff in DGHM Before entering the estimation of the water balance components, the daily precipitation (P) is separated into the direct runoff (Qo) and the portion that infiltrates and so enters the water balance computation as the effective rainfall. The US Soil Conservation Service (SCS) has prepared a series of tables and nomographs by which the amount of daily overland runoff from intense precipitation can be estimated from information on slope, ground cover, soil type and antecedent precipitation. This method is usually called the SCS-CN curve method (CN formula). This method is usually recommended for daily overland runoff estimation. Estimation of the overland runoff using the CN formula for daily overland runoff is given by:
P 0.2 s P + 0.8s Qo = 0.0 Qo = 100 s = 254 1 CN
for for
P > 0.2 s P 0.2 s
(4)
where s is a parameter related to curve number (CN) [ SCS (1985) ,Chow et al. (1988)] as: (5)
The constant, 254, in Equation (5) gives s in mm. Thus, P and Qo are also expressed in mm. Thus, the monthly volume of overland runoff is the sum of the daily overland runoff, DRO; that is, the number of days in the month under consideration.
A Continental Scale Water Balance Model: A GIS-Approach for Southern Africa 3
DRO = Qo i , where, nd is
i =1
nd
The importance of this approach in developing the distributed GIS-based model (DGHM) is more explained in terms of the subsurface processes. It is based on the assumption that if one computes overland runoff from the total precipitation falling, then it forms the effective precipitation that goes to replenish the soil moisture. Modeling soil moisture Soil moisture is determined from the interaction between effective PPT and PET. During wet months (when effective PPT in excess of PET), soil moisture can increase up to a maximum of field capacity determined by soil texture and rooting depth. dSM (6) = EPPT - PET when EPPT > PET and SM < FC dt
dSM =0 dt
when SM = FC
(7)
dSM = a SM ( PET EPPT ) dt
when
EPPT < PET
(8)
where, FC is the soil moisture at field capacity of the soil millimetres, PPT is precipitation in mm/month; PET is potential evapo-transpiration in mm/month. During the dry periods where the EPPT < PET, the soil becomes increasingly dry, soil moisture becomes a function of potential soil loss. Thus, different authors assumed different relationship to find the soil moisture during the dry periods (Thornthwaite, 1948; Vorosmarty et al., 1989). For a particular soil, there is a linear relationship between Log (SM) and (EPPT-PPT) summed from the start of the dry season to the current month. Thus, to calculate the rate of change of soil moisture (S) through the dry season, the soil moisture is directly estimated from the empirical relation suggested by Vorosmarty et al. (1989). To calculate dSM/dt through Equation (8) for intermediate field capacities, the DGHM defines a slope a to the retention function as:
a = ln( FC ) /(1.1282 FC ) 1. 2756

accumulated potential water loss or APWL
(9)
where the numerator represents soil moisture (millimeters) with no net drying. The denominator is the model can calculate dSM/dt as a function of soil dryness and update SM.
( [PET PPT ]) in mm at SM=1 mm. With a determined, the
Calculations commence at the end of the wet season when it is assumed the soil is at the field capacity. Soil water stocks are then depleted during the dry season in accordance with the moisture retention function. For each wet month, soil moisture is determined by incrementing antecedent values by the excess of the available water over PET. This recharge may or may not be sufficient to bring the soil to field capacity at the end of subsequent wet season. Analysis of the ratio PPT/PET at the various grids in southern Africa has shown that the region can not be represented by one homogeneous climatic zone, rather it is a combination of climate influenced by the diversified hydro-climatic conditions in the region. In this analysis, it has been observed that the long-term monthly rainfall distribution varies greatly in the region. It is assumed that the ratio PPT/PET could be a good indicator of the seasonal distribution of monthly rainfall, and accordingly the wet season ends when the ratio PPT/PET just starts to fall below unity. In line with this, therefore, the soil moisture can be assumed to be at its field capacity. Consequently, DGHM fixes a starting month of any grid across the region according to the aforementioned criterion. The solution of the basic mass balance equation for the subsequent months can determined once the initial soil moisture state is obtained. Actual Evapotranspiration Once the actual soil moisture is determined, the corresponding actual evapotranspiration (AET) is calculated for the month. Following the Thornthwaite and Mather (1957) approach, AET is set equal to PET in wet months, when EPPT > PET. During this time it is assumed that EPPT is in sufficient abundance to satisfy all the potential water demands of the resident vegetation. During dry season when EPPT < PET, the monthly average AET is modified down wards from its potential value as shown below.
AET = PET
when
EPPT > PET
(10)
4
AET = EPPT -
dSM dt
when
EPPT < PET
(11)
where the soil moisture drops below the wilting point, the AET is equal to the EPPT but if there is no EPPT at all time, AET becomes zero. Total runoff (TRO) generated at a grid Most of the water does not leave the basin as soon as it becomes available as surplus. The portion that constitutes overland flow is assumed to flow out the watershed within the month it occurs but the portion that infiltrates may take a number of months to move slowly through upper layer of the soil column to emerge in the surface water courses as base flow. The lag or delay factor depends not only the size of the basin but also on the vegetation cover and the soil type, degree of slope, characteristics of the soil rock layers, etc. Thornthwaite suggested that 50 % of the water surplus could be assumed to runoff each month from large basins with the remainder being held over and added to the surplus of the next month. This factor is set according to Thornthwaite & Mather (1957) which is equal to 0.5.
RO = 0 .5 * (D + (EPPT - PET)) RO = 0.5 * D
when SM = FC & EPPT > PET
(12) (13)
when SM < FC & EPPT < PET
where, RO is rainfall-driven runoff or surplus runoff (mm/month) and D is the amount of detention storage in millimeters that is assumed to leave each grid cell next month. Finally, the total runoff from the grid is calculated by adding the DRO and the RO.
TRO
DRO + RO
(14)
Equation (14) gives the runoff from each grid that is ultimately routed as discharge for any catchment in the region.
3.
CREATION OF GEO-REFERENCED DISTRIBUTED INPUT DATABASE
Geo-referencing and GIS of the study area 0 The study region involves the southern African sub-continent that is encompassed between equator to 35 S 0 0 latitudes and 5 E - 55 E longitudes. It is divided into grids of 30 minutes spatial resolution representing the region by 7000 square grids (70 rows by 100 columns). Each grid point is assigned a unique ID number starting from 1 to 7000, each referenced to its latitudinal and longitudinal location. This forms the basic geo-referenced platform of the study region. Totally 3402 grids represent the land catchments of the southern African region Precipitation and Potential evapotranspiration Input Variables Rainfall is a climatic quantity showing greater variability within a short range of spatial variation. Despite the scarcity of data in the region, distributed 1961-90 average grid rainfall data base at 0.5X0.5 degree from Hulme (1996) is used in this study. The southern African regional precipitation has been based on the thin-plate splines approach developed by Mike Hutchinson from the Australian National University. There is no readily available potential evapotranspiration for the region of southern Africa. However, the southern African regional 1961-90 climatologies used in Hulme (1996) exist at 0.5 resolution. The region's monthly potential evapotranspiration was computed using the Penman's (1964) method and stored at 0.5 degrees grid resolution. GIS - based Soils and Vegetation Information The FAO/UNESO soil map of Africa, available at 10x10 minutes resolution, is used to derive the southern African soils data from which a 30x30 minutes resolution of soils database is created by pixel contraction (Eastman, 1997). The soils classification of the Africa (FAO, 1988) was based on agronomic characteristics and benefits it renders to agriculture with about 133 of such soil types. The rasterized soils map at grids of 30 X 30 minutes has been used to derive textural soil classes, which ultimately fall into six groups [Table 1].
Table 1. Distribution of textural soil classes in the southern African region.

Sand No. of grids % coverage 831 24 Sandy loam 320 10 Clay 331 9 Clay loam 1432 42 Silt loam 365 11 Lithosol 122 4
To set broader rooting depths, two rooting depth categories in compliance with Vorosmarty et al. (1989) have been created. These are: (1) VEG1 - this includes farmlands (or extensive crop production area), savannah with scattered trees, shrub lands, etc., and percentage forest of 50% or less of NDVI. (2) VEG2 -dense forest, natural vegetation cover and deciduous forest cover, and percentage of forest above 50% of mean annual NDVI have used as a criterion. The nature of soil at a particular location and for a given vegetation cover affects the rooting depth. This has been noted in Vorosmarty et al. (1989) and is adopted. Based on the above vegetation grouping, the root depth for each soil is assigned based on the values. The SCS soil groupings for the overland flow module is assigned according to the definition given in Singh (1992) and is tabulated in Table 2. Table 2. Rooting depth assigned for various soil textures and SCS soil groupings.
Veg. Group NDVI Sand 2.5 1.0 A Assigned Root Depth (M) For Different Soil Sandy Loam Silty Clay Loam Loam 2.0 2.0 1.6 1.0 1.3 1.0 B D C Types Clay 1.2 0.7 C Lithosol 0.1 0.1 B
VEG2 >50% VEG1 50% SCS soil group (Singh, 1992)
Satellite data have been used extensively for mapping the land use and for monitoring the seasonal change of vegetation of river basins ( Nemani & Running, 1989). The Normalized Difference Vegetation Index (NDVI) is 2 commonly applied to derive the leaf area index (LAI) from channels 1 and 2 of NOAA-AVHRR data at 1-km resolution. Monthly twelve NDVI images, expressed in percentage, for 1987 are acquired from USGS web pages. Based on NDVI values the LAI for different land use groups can be estimated by empirical formulae. Derivation of soil retention parameters The ultimate importance of the textural reclassification of soils and vegetation is to derive retention parameters of the soils such as field capacity (FC) and wilting point (WP) values for each soil group to determine the soil moisture capacities. Based on field measurements, Saxton et al. (1986) have developed a technique to estimate the matrix potential of different soils by using multiple regression techniques. So at a matrix potential equivalent to field capacity (33 KPa) and wilting point (1500 KPa) for a unit meter depth of soil, approximate values of FC and WP for the various soil textures have been extracted [see Table 1]. Accordingly, the FC and WP values as per the rooting depth is then derived for the whole region as summarised in Table 3. Derived field capacity
-5 S -5 S -10 S -10 S
Available water content
-15 S
-15 S
-20 S
-20 S
-25 S Field Capacity (mm)

500 400 300 200 100 1
-25 S
-30 S
-30 S
Available water content (mm) = FC-WP

200 150 100 50 0
-35 S 5E
10 E
15 E
20 E
25 E
30 E
35 E
40 E
45 E
50 E
-35 S 5E
10 E
15 E
20 E
25 E
30 E
35 E
40 E
45 E
50 E
(a) field capacity (FC)
(b) available water content (AWC)
Fig. 2 Distribution of derived (a) field capacity (FC) and (b) available water content (AWC) in the SADC region used in water balance calculation.
The maximum retention capacity of soils depends on the type of soil and the vegetation cover of the area. It is obtained by multiplying the UFC and UWC values by the corresponding root depth. Derived field capacity (FC) distribution in the SADC region used in water balance calculation is presented in Figure 2a. On the other hand, Figure 2b shows the spatial variation of the derived available water content (AWC) distribution across the SADC region used in DGHM. Table 3. Relationship linking vegetation class, soil texture, rooting depth and moisture capacities of various soil groups in southern Africa.
Sand Sandy Loam Clay 1.2 0.7 48.5 35.8 582 153 339 89 Lithosol 0.1 0.1 27.3 13.2 27 14 27 14
Silt Loam Clay Loam The Root depth (m) 2.5 2.0 2.0 1.6 1.0 1.0 1.3 1.0 FC and AWC of soils as % of total volume of soils / m % (FC)/m 14.1 20.0 27.3 35.2 %(AWC)/m 6.3 9.1 13.2 35.8 Field Capacity and AWC per root depth of the plant GRP1 (FC) 353 400 546 563 (AWC) 196 218 282 243 GRP2 (FC) 141 200 355 352 (AWC) 78 109 183 152 Veg. Group GRP1 GRP2
(a) runoff
(b) evapotranspiration
10
Grid: at lat/long 10.25 S / 35.75 E PET
10
Grid: at lat/long 10.25 S / 35.75 E PET
PPT TRO
PPT AET
mm/day
mm/day
O N D J F M A M J J A S
0 O N D J F M A M J J A S
(c) Soil moisture

300
PPT & PET (mm/day)
8 6 4 2 0 1 2 3 4 5 6 7 8 9
PPT SM
250 200 150 100 50 0
10
11
12
Fig. 3 Seasonal variation of precipitation and PET versus model derived (a) runoff, (b) actual evapotranspiration, and (c) soil moisture for a selected grid cell the SADC region centred on Eastern Tanzania, which is dominated with soil of field capacity of 349 mm.
soil moisture, SM (mm)
10
PET
4.
MODEL APPLICATION AND DISCUSSION OF RESULTS
General principle For a single geographically referenced grid cell, the model is run independently for the 30-year period between 1961-90 to derive a regional map of hydrologic regimes. In each simulation year, the model in the subsurface zone, is run for a time loop of 12 months, by applying the PPT and PET. If a dynamic state is not achieved in soil moisture (SM), actual evapotranspiration (AET) and runoff (TRO) between subsequent simulations, simulation proceeds until a dynamic steady state is achieved with in 0.01 % error in all of these variables. When the steady state situation is arrived, the outputs SM, EAT and TRO are maintained, and the spatial control module is invoked and computation for another grid cell continues. The model is then applied for 7000 grids that represent the study area according to their spatial identifier. In order to establish the hydrologic regimes of the study region, the available 1961-1990 average hydroclimatological information, on monthly basis, has been used as input to the model DGHM. The seasonal variation of TRO, AET and SM for corresponding PET and PPT for a typical region in eastern Tanzania is shown in Figures 3a,b & c, respectively. Soil moisture The variation of 1961-90 mean annual soil moisture (MASM) in Southern African region, computed by the model, is shown in Figure 4. The MASM of the basin is estimated by the model is about 148 mm. The frequency distribution of MASM over the region portrays a significance spatial variation [Figure 4]. This is due to the fact that the absolute soil moisture depends on the water retention property of the soil, and hence on the field capacity of the soil. If the actual soil moisture is segregated by its field capacity value of the corresponding soil type, the actual regional variation can be observed clearly. It can be observed from Figure 4 that DGHM simulates the majority of the area northwestern Tanzania west of the Lake Victoria extending to west of Zaire and Central Mozambique to exhibit higher annual soil moisture regimes. The highland areas in Northern and eastern Tanzania have moderate soil moisture and most part of central and southern Africa from Zimbabwe to southern most of Republic of South Africa are dominated by low soil moisture regimes. Generally, the map of annual average soil moisture to some extent follows the pattern of precipitation. The variation of mean annual SM with respect to the mean annual PPT is depicted in Figure 5.
0S
Mean annual soil moisture - 1961-90

-5S -10S
-15S
-20S
-25S
-30S 500
Soil moisture (mm)

400 300 200 100
5E
10E
15E
20E
25E
30E
35E
40E
45E
50E
Fig. 4 Mean annual soil moisture variation for the baseline 1961-90 over southern Africa It can also be inferred from Figure 5 that high precipitation grids experience high SM, explained by relatively high values of the ratio mean annual SM/FC. However, because PPT alone is a poor indicator of SM in absolute terms, and the variation of SM with PPT is not distinctively related, either linearly, or not otherwise. This can also be attributed because of the high variation of FC in the region shown in Figure 4, which ranges from 27 to 563 mm that can affect the ratio of actual SM to FC. But the dependence of SM on PPT becomes more vivid if it is segregated by the field capacity.
Mean annual SM vs PPT
Precipitation (mm/day)
4 3
0 0 100 200 300 400 500 600
Soil moisture (mm) Fig. 5 The relationship between precipitation and model derived mean annual soil moisture
Actual evapotranspiration The variation of mean annual actual evapotranspiration (MAET) in the Southern African region, simulated by the model DGHM, is shown in Figure 6. The model prediction of the mean annual (AET) in the region reaches a maximum of 1500 mm, with mean 420 mm. Mean annual AET vs PPT
7 6
Precipitation (mm/day)
0 0 1 2 3 4 5 6 7
Evapotranspiration (mm/day)
Fig. 6 The relationship between 1961-90 mean annual precipitation and DGHM model derived mean annual AET over the Southern African region. Considering the whole region, the actual mean annual AET accounts for 41 % of the mean annual PET over the region. The spatial variation of the computed annual AET over the Southern African region is shown in Figure 7. It can be seen that, PET is high in the area near the shore areas surrounded by the oceans and decreases towards hinterland. Nearly northern half of southern African region experiences higher evaporation than the remaining southern half of the region which experiences lower evaporation. The variation of the ratio of annual AET to PET is shown in same figure [Figure 7]. The regional variation of this ratio follows a similar tendency of the actual evaporation. It is higher in the areas surrounding the oceans and
tends to decrease hinterland as one goes from the periphery of the ocean shore areas, which reaches nearer to a value of 1.0. Since the mean annual AET values depend both on the distribution pattern of the rainfall as well as PET values, no general conclusion can be derived from PET alone with out involving the role of PPT & SM.
0S 0S
Mean annual AET - 1961-90

-5S -5S -10S -10S
Mean annual AET/PET ratio - 1961-90
-15S
-15S
-20S
-20S
-25S
-25S
-30S
Actual ET (mm/day)
5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5
-30S
AET/PET ratio
1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1
5E
10E
15E
20E
25E
30E
35E
40E
45E
50E
5E
10E
15E
20E
25E
30E
35E
40E
45E
50E
Fig. 7 Mean annual actual evapotranspiration and relation to Penman (AET/PET) variation over southern African region.
Runoff The mean annual runoff and the actual runoff coefficient variation for each grid (as computed from the ratio of the total runoff (TRO) to (PPT) is given in Figure 8a and 8b, respectively. The runoff coefficient map shown here is the ratio of the actual runoff generated from precipitation within each grid cell before starting to form a stream flow. (a) (b) Mean annual runoff - 1961-90 Mean annual runoff coefficient - 1961-90 0S 0S
-5S -5S
-10S
-10S
-15S
-15S
-20S
-20S
-25S
-25S
-30S
Grid runoff (mm/day)

5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5
-30S
mean annual runoff coefficient

0.40 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00 -0.05
5E
10E
15E
20E
25E
30E
35E
40E
45E
50E
5E
10E
15E
20E
25E
30E
35E
40E
45E
50E
Fig. 8 Spatial variation of mean annual generated runoff and runoff coefficient over southern African region derived from the baseline 1961-90 climatology.
The mean annual generated runoff from the land catchment in the region is about 151 mm/year although there is a significant inter-regional variation among the SADC countries. Nevertheless, the variation of the runoff generation for each grid varies greatly. Lower runoff regimes are dominant in arid areas in Botswana, Namibia and south-western part of the Republic of South Africa. Higher runoff regimes are the northern and western Tanzania, along the east coastal portions of Mozambique, central Mozambique, western Zambia and Malawi.
10
mean=151 mm, st. dev. 246, size=3401 grids
No of grids in SADC
1500
1000
500
0 0 100 200 300 400 500 600 700 800 900 1000
Mean annual runoff (mm)

Fig. 9 Frequency distribution of mean 1961-90 annual runoff of the SADC region simulated by the developed model DGHM.
Runoff (mm)
0 0 1 2 3 4 5 6
Rainfall (mm)
Fig. 10 Mean annual generated runoff in relation to the rainfall of the baseline 1961-90 climatology at the grids of southern African region.
Figure 9 shows the frequency distribution of the mean 1961-90 annual runoff of the SADC region simulated by the developed model DGHM. It demonstrates the mean annual runoff in the region of southern Africa is 151 mm with majority dominated with low runoff regime below 500 mm per year.
It has been observed that the runoff generation is highly influenced by the distribution pattern of precipitation in the region. The mean annual variation of runoff generated is linear at high and low precipitation regimes in the region. In the medium precipitation regimes, there is a wide variation in runoff generation as depicted in Figure 10. The annual variation of runoff as a function of the annual precipitation in mm/day at grids across southern Africa is presented in Figure 10. It offers the relative runoff volume generated at each grid. However, a clear conclusion can not be drawn for the region, as to the inter-grid and regional variation of input information responsible for the runoff generation dictated by the processes in surface as well as in subsurface soil moisture variation.
5.
CONCLUSION
A distributed GIS-based hydrological model (DGHM) is developed using GIS and computational hydrology techniques. The model is based on water balance consideration of the surface and subsurface processes. The model has been used to create a GIS-based high resolution data sets of monthly soil moisture, evapotranspiration and grid runoff in the region. Considering the 1961-90 climatic period, we have mapped the regional variation of the mean annual soil moisture (SM), actual evapo-transpiration (AET), and generated runoff (ROF) across the SADC. The model estimates the mean SM of the region to be about 148 mm per year. There is a wide spatial range in the distribution of SM in the region due to the fact that the absolute soil moisture is dependent on the water retention properties of the soils considered across the region. The model prediction of the mean annual AET in the region reaches a maximum of 1500 mm, with mean 420 mm. The mean annual generated runoff from the land catchment in the region is about 151 mm/year although there is a significant inter-regional variation among the SADC countries, which is a function of the variation in the vegetation cover, soil and climate variation. The model enables one first to establish the hydrologic regimes of the region, and finally to deal with the prediction of discharge from the various drainage basins. It ultimately permits a through assessment of the region's water resources availability. Such models can be externally coupled with Atmospheric Global Circulation Models (AGCM) and can be used to study the impact of possible climatic change conditions on the regions water resources. Acknowledgments: The authors thank DAAD for the research grant to fund part of the study, and Dr. Mike Hulme at University of East Anglia for providing climatic data of Southern Africa.
REFERENCES Biftu, G.F. (1998) Semi-distributed hydrological modelling using remoteley sensed data and GIS.Unpublished PhD thesis. University of Alberta. Conway, D. (1997) A Water Balance Model of the Upper Blue Nile in Ethiopia, Hydrological Science Journal, 42(2). Eastman, J.R. (1997) IDRISI for windows users guide. Clark University, Graduate School of Geography, Worcester, Massachusetts 06610, USA. FAO (of United Nations) (1988) World Soil Resources, An explanatory Notes on the FAO world soil reference map, World Soils Resources Report, No. 66. Hulme, M., Arntzen, J., Downing, T., Leemans, R., Malcolm, J., Reynard, N., Ringrose, and Rogers, D. (1996) Climate Change and Southern Africa: An exploration of some potential impacts and implications in the SADC region, University of East Anglia, Norwich, UK. Nemani, R. R., and Running, S. W. (1989). Testing a theoretical climate-soil-leaf area hydrologic equilibrium of forests using satellite data and ecosystem simulation. Agricultural and Forest meteorology, 44: 245-260. Rutter, A. J., Morton, A. J., and Robins, P. C. (1975) A Predictive model of rain interception in forests, 1. Generalization of the model and comparison with observations in some coniferous and hardwood stands. J. Appl. Ecol, 12:364-380. Saxton,K.E., Rawls, W.J., Romberger, J.S. and Papendick, R.I. (1986). Estimating generalized Soil-Water Characterstics from Textre. Soil SCI. SOC. AM. J., Vol. 50
Singh, V. J. (1992) Elementary hydrology. Prentice hall, Inc., Englewood Cliffs, pp. 973 Thornthwaite, C.W. and Mather, J.R. (1957) Instructions and Tables for computing potential evapotranspiration and the water balance, Drexel Inst. Techno. Publ. Clim., X(3). Thornthwaite, C.W. (1948) An approach toward a rational classification of climate. Geographical review, 38(1): 55-94. Van Deursch, W.P.A., and Kwadijk, J.C.J.(1993) An Integrated GIS Water Balance Model for the River Rhine, Proceeding of Viena Conference, IAHS Publ., No.211. Vorosmarty, C.J., More, B., Grace, A.L., Gildea, M.P., Melillo, J.M., Peterson, B.J., Rasteller, E.B., and Steudler, P.A. (989). A Continental Scale Models of Water Balance and Fluvial Transport: An Application to South America, Global BioGeochemical Cycles, 3(3) 241-265.
13

Continental Scale Water Balance Model

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Continental Scale Water Balance Model

Uploaded by

Copyright:

Available Formats

3rd WaterNet/Warfsa Symposium 'Water Demand Management for Sustainable Development', Dar es Salaam, 30-31 October 2002

A Continental Scale Water Balance Model: a GIS-Approach for Southern Africa

A Continental Scale Water Balance Model: A GIS-Approach for Southern Africa

SM (t ) = PPT TRO AET F

TRO = PPT (SM + AET )

A Continental Scale Water Balance Model: A GIS-Approach for Southern Africa

= 0.997 0.139 LAI; R 2 = 0.997 as a

Soil Moisture SM Direct runoff (DRO)

Excess s Runoff storage (RO)

Runoff from a grid (TRO)

P 0.2 s P + 0.8s Qo = 0.0 Qo = 100 s = 254 1 CN

P > 0.2 s P 0.2 s

dSM = a SM ( PET EPPT ) dt

EPPT < PET

a = ln( FC ) /(1.1282 FC ) 1. 2756

( [PET PPT ]) in mm at SM=1 mm. With a determined, the

EPPT > PET

A Continental Scale Water Balance Model: A GIS-Approach for Southern Africa

EPPT < PET

RO = 0 .5 * (D + (EPPT - PET)) RO = 0.5 * D

when SM = FC & EPPT > PET

when SM < FC & EPPT < PET

CREATION OF GEO-REFERENCED DISTRIBUTED INPUT DATABASE

A Continental Scale Water Balance Model: A GIS-Approach for Southern Africa

Table 1. Distribution of textural soil classes in the southern African region.

VEG2 >50% VEG1 50% SCS soil group (Singh, 1992)

Available water content

-25 S Field Capacity (mm)

Available water content (mm) = FC-WP

(a) field capacity (FC)

(b) available water content (AWC)

Grid: at lat/long 10.25 S / 35.75 E PET

Grid: at lat/long 10.25 S / 35.75 E PET

(c) Soil moisture

250 200 150 100 50 0

A Continental Scale Water Balance Model: A GIS-Approach for Southern Africa

soil moisture, SM (mm)

MODEL APPLICATION AND DISCUSSION OF RESULTS

Mean annual soil moisture - 1961-90

Soil moisture (mm)

A Continental Scale Water Balance Model: A GIS-Approach for Southern Africa

Mean annual SM vs PPT

0 0 100 200 300 400 500 600

Mean annual AET - 1961-90

Mean annual AET/PET ratio - 1961-90

Grid runoff (mm/day)

mean annual runoff coefficient

A Continental Scale Water Balance Model: A GIS-Approach for Southern Africa

mean=151 mm, st. dev. 246, size=3401 grids

Mean annual runoff (mm)

A Continental Scale Water Balance Model: A GIS-Approach for Southern Africa

You might also like