A modelling study of the effects of land management and climatic variations on groundwater inflow to Lake St Lucia, South Africa Lars Været & Bruce Kelbe & Sylvi Haldorsen & Richard H. Taylor Abstract Over the past few years groundwater has been recognized as an important contributor of freshwater to Lake St Lucia, South Africa during periods of prolonged drought. This has led to a management strategy aiming at increasing the groundwater recharge and minimizing groundwater use through active manipulation of the vegetation. For the Eastern Shores on the edge of Lake St Lucia, the replacement of vast areas of pine (Pinus elliottii) plantations with grassland over the past decade, combined with a strict burning regime, has led to a general rise of the water table, which has increased the groundwater seepage to Lake St Lucia. A numerical groundwater model has been applied to assess the effects of local management strategies on the mass balance of a shallow aquifer and these are compared to the effects of predicted climate and sea-level change for this area. The simulations indicate that local management actions that are being applied to the Eastern Shores have positive effects on the groundwater flux into Lake St Lucia and that they outweigh potential negative effects of future climate and sea-level change predicted for this area.
Received: 10 June 2008 / Accepted: 29 April 2009 Published online: 16 June 2009 * Springer-Verlag 2009 L. Været ()) : S. Haldorsen Department of Plant and Environmental Sciences, Norwegian University of Life Sciences, P.O. Box 5003 1430, Ås, Norway e-mail:
[email protected] B. Kelbe Department of Hydrology, University of Zululand, Private Bag X1001, Kwa-Dlangezwa, 3886, South Africa R. H. Taylor Ezemvelo KZN Wildlife, Private Bag X01, St Lucia Estuary, 3936, South Africa Present Address: L. Været, Markveien 23V, 1406, Ski, Norway Hydrogeology Journal (2009) 17: 1949–1967
Keywords Groundwater management . Climate change . Land use . Ecology . South Africa
Introduction Groundwater is important for many ecosystems and may for instance provide near-shore wetlands, estuarine and shallow marine ecosystems with freshwater. These systems can play a significant social and economic role, and may often be of great importance to the maintenance of biodiversity. Temporal and spatial changes of the related groundwater systems may influence these ecosystems in various ways (Murray et al. 2003). Cheng and Ouazar (2004) pointed out that anthropogenic disturbances and inappropriate management of coastal aquifers may cause irreversible damage and destroy these aquifers as freshwater resources. In addition to a local anthropogenic influence, climate variations can be of importance to coastal aquifers. There is today a particular concern that future climate change might cause changes in recharge and discharge of primary aquifers (Cheng and Ouazar 2004). Along with climate change there is also a concern that future sea-level rise near coastal aquifers may lead to elevated groundwater boundary conditions and cause an inland shift of the mixing zone between fresh and saline groundwater (Cheng and Ouazar 2004; IPCC 2007). Only in a few cases has there been a direct comparison between the impacts of management strategies versus the impact of predicted climatic variations on a coastal groundwater system. In this paper, the effects of land use change have been compared to the effects of a specific global change on the coastal groundwater system of Lake St Lucia (Fig. 1) in South Africa. This is an important, internationally recognized linked estuary-lake system which forms the core feature of the iSimangaliso Wetland Park in northeast South Africa (earlier named The Greater St Lucia Wetland Park). Studies of the groundwater systems in this area (Kelbe and Rawlins 1992a, b; Meyer et al. 1993; Kelbe and Rawlins 1995; Kelbe et al. 1995) have been focused on the possible consequences of dune mining and forestry on the DOI 10.1007/s10040-009-0476-5
1950
Fig. 1 The Lake St Lucia on the northeast coast of South Africa (the Maputaland coast). Contour lines (20 m) show the flat coastal plain between the western upland part of the catchment and the tall dune system along the Indian Ocean coast. A–A’: The location of the geological profile shown in Fig. 2
groundwater resources. The unique importance of groundwater in this area for the estuarine ecosystems during severe droughts has been pointed out by Taylor et al. (2006). These studies and subsequent changes in land use have led to management controls on groundwater recharge to protect the ecological resources of the Eastern Shores of Lake St Lucia (Fig. 1). However, the effects of sea-level rise and climate change on the Hydrogeology Journal (2009) 17: 1949–1967
groundwater system have previously not been analyzed. The iSimangaliso Wetland Park is a UNESCO World Heritage Site and an understanding of long-term changes of the aquifers that form an integral part of the area are needed in order to establish adequate management plans. In a review paper on climate and sea-level changes on the South African East Coast, Været et al. (2008a) predict DOI 10.1007/s10040-009-0476-5
1951
a warmer and wetter climate, accompanied by sea-level rise, towards 2100. However, Vaeret et al. point out that great uncertainty in the model predictions makes it crucial for different scenarios to be considered. The impact of climate change outlined by Været et al. (2008a) are incorporated into a calibrated three-dimensional numerical groundwater model, which is used to quantify the spatial and temporal variations in the groundwater system of the Eastern Shores. Both anthropologically induced and natural variations in climatic driving factors of groundwater are examined for different climate and management scenarios.
Location, geology and geomorphology The Lake St Lucia system has a north–south orientation which stretches about 70 km parallel to the southeast coast of Africa. It is located on the southern end of the Maputaland coastal plain that extends north into Mozambique (Fig. 1). High forested coastal barrier dunes, with elevations up to 180 m a.s.l., flank the Indian Ocean coastline and stretch all the way up to the Maputo Bay in Mozambique. The Eastern Shores dune barrier separates Lake St Lucia from the Indian Ocean (Fig. 1). In the St Lucia area, the high barrier dunes do not extend beyond ∼1 km inland from the coast. Earlier research has shown that marine, alluvial and aeolian sand deposits of the Eastern Shores (Fig. 1; see geological description below) form the main unconfined aquifer discharging into Lake St Lucia (Kelbe et al. 1995; Taylor et al. 2006). The present study was, therefore, carried out in this area. The Eastern Shores consist of the high coastal barrier dune along the east coast and the slightly lower Embomveni undulating dune mound to the west; the latter reaches up to an elevation of 70 m a.s.l. Between these high areas, is the interdunal depressions area that forms Lake Bhangazi, the Mfabeni Swamp and interdunal wetlands drained by the Nkazana and Tewate streams (Fig. 1). Consolidated Cretaceous siltstones, which were deposited on the sloping continental edge soon after the separation of the African continent from Gondwana (e.g. Hobday 1979; King 1972; Maud and Orr 1975; Maud 1980), underlie the whole area shown in Fig. 1 to form the basement of the unconfined aquifer (Fig. 2). These rocks are exposed in the False Bay area along the western shores of the lake. The consolidated Mesozoic and overlying thin Neogene sedimentary rocks generally dip eastwards and their surface is located 30–50 m below the sea level (b.s.l.) in the Eastern Shores (Fig. 2; Davies Lynn and Partners 1992). In the Eastern Shores, the Mesozoic and Neogene rocks are overlain by unconsolidated to semi-consolidated middle to late Pleistocene sediments. These consist of the thin estuarine clays of the Port Durnford Formation to the west and the thicker dune sands of the Kosi Bay and Isipingo Formations to the east (Fig. 2; Davies Lynn and Partners 1992). The weathered dune sands of the Kosi Bay Formation form the core of the Embomveni mound (Fig. 1) and extend up to ground surface with elevation Hydrogeology Journal (2009) 17: 1949–1967
up to 70 m a.s.l. The Kosi Bay Formation is partly clayenriched in its lower parts, and grades upwards into the well-drained red sands. The youngest stratigraphic unit of the Eastern Shores is the Holocene dune sands of the Sibayi Formation, which caps the older strata over most of the area (Fig. 2). The Isipingo Formation aeolionite formed the core of a coastal barrier island during the Last Interglacial marine transgression, when the St Lucia embayment had a direct marine connection and the sea-level maximum was about 4–5 m higher than today (Wright et al. 2000). Deposition of the Sibayi Formation parabolic dunes against the older Isipingo Formation barrier island core constricted the marine connection to the proto-St Lucia lagoonal system during the late Holocene (Wright et al. 2000). During the marine regression down to 130 m below the present sea level of the Last Glaciation, rivers incised narrow bedrock valleys, which were later filled in and buried by a complex sequence of marine, lagoonal and fluvial deposits during the Holocene marine transgression (Davies Lynn and Partners 1992). Davies Lynn and Partners (1992) identified a major buried river channel incised into the Cretaceous siltstone just west of Mission Rocks and stretching northwards to Lake Bhangazi (“Alluvium” in Fig. 2). These authors also identified a possible fossil river mouth in the Mission Rocks area, which was probably formed during or prior to the Last Interglacial.
Hydrology of the Eastern Shores The hydrology of the study area is controlled by the St Lucia lake water level on the western boundary, the Indian Ocean on the eastern boundary and several internal drainage systems (e.g. Bhangazi, Nkazana and Tewate, see Fig. 1). The regional and temporal variations in rainfall (Fig. 3) are the driving factors behind the water balance and salinity variations of Lake St Lucia. Annual rainfall during 1951–2006 for seven weather stations around Lake St Lucia is presented in Fig. 3. There is a noticeable variation between the stations, showing an east–west decrease in rainfall, which was also pointed out by Taylor et al. (2006). Studies by Stassen and Mynhardt (1992) implied that the river runoff from the west is the main source of water for the lake under normal climatic conditions and that groundwater has an insignificant role, contributing only ∼5% (Kelbe et al. 1995) of the average total freshwater input. The groundwater fluctuations on the Eastern Shores are driven by variability in rainfall at different time scales. Figure 3 shows a near decadal cyclicity in rainfall that is also reflected in the groundwater hydrographs (Fig. 3), while Været and Sokolic (2008b) also show how responses in the water table are related to seasonal and event-related variations in rainfall. Groundwater levels respond rapidly to rainfall and drop more slowly and gradually during droughts. In agreement with Kelbe et al. (1995), the groundwater contribution does not form an important component of the DOI 10.1007/s10040-009-0476-5
1952
Fig. 2 Geological cross section A–A’ (location: see Fig. 1) showing the Cretaceous rocks and the Tertiary Port Durnford Formation underlying the Pleistocene and Holocene aquifer of the Eastern Shores. Based on Davies Lynne and Partners (1992). A saltwater/freshwater interface calculated from simulated water-table levels using the Ghyben-Herzberg relation (Todd 1980) illustrates the theoretical influence of saltwater above the base of the Eastern Shores aquifer represented by the Cretaceous siltstone
freshwater budget of Lake St Lucia in periods of average and above-average rainfall. During extreme drought, however, the rivers become dry and provide almost no inflow into the lake. Even though groundwater levels on the Eastern Shores drop considerably under such dry conditions (Fig. 3), a freshwater seepage persists into the Nkazana, Tewate and other main streams and along the lake shoreline, becoming the most important freshwater contribution to the lake (Taylor et al. 2006). Salinity measurements at several places in the lake show near decadal cyclicity caused by the variation between dry and
wet weather conditions (Taylor et al. 2006). The importance of the groundwater seepage for the freshwater inflow into St Lucia consequently follows a similar decadal trend (Taylor et al. 2006).
Vegetation Lubke et al. (1992) mapped the vegetation cover of the Eastern Shores south of Lake Bhangazi. Ten plant communities were recognized, including open-water communities. Open-water surfaces related to these latter
a Annual rainfall (mm)
2500 2000 1500 1000 500
2005
2003
2001
1999
1997
1995
1993
1991
1989
1987
1985
1983
1981
1979
1977
1975
1973
1971
1969
1967
1965
1963
1961
1959
1957
1955
1953
1951
0
Year
b
13 B4
Groundwater head (m a.s.l.)
12 11 10 D5
9 8 A3
7 6 5 4
Fig. 3 a Average annual rainfall for seven weather stations around Lake St Lucia (names and locations are shown in Fig. 1) during 1951– 2006. The vertical lines indicate the variance within the seven stations. Data provided by the South African Weather Service. b Groundwater levels for boreholes A3, B4 and C5 on the Eastern Shores from 1973 to 2006 (for borehole locations, see Fig. 4). Data provided by Ezemvelo KZN Wildlife Hydrogeology Journal (2009) 17: 1949–1967
DOI 10.1007/s10040-009-0476-5
1953
communities have been mapped by Sokolic (2006) and Været and Sokolic (2008b), and their distribution changes dramatically with variations in the groundwater elevation. A simplified vegetation map for the Eastern Shores is presented in Fig. 4, based on the work by Lubke et al. (1992), aerial photographs and digitized data from the St Lucia Wetland Authority. For modelling purposes, the vegetation cover has been simplified into five zones based on their estimated potential evapotranspiration rate (ET) and rooting depth. Commercial pine (Pinus elliottii) plantations were first established on the Eastern Shores in 1961 with new areas being added in 1981 covering a total area of ∼53 km2 or about one third of the area south of Lake Bhangazi (Fig. 4). The plantations were usually established on areas which had previously been grassland. Earlier studies in this area (Rawlins and Kelbe 1991a, b) and other areas (Salama et al. 2002) indicate that deep-rooted pine
plantations have a significant effect on the recharge conditions through extensive interception, and through evapotranspiration rates from the unsaturated soil and water-table fringe (capillary zone). Re-establishment of the natural vegetation has therefore been important in the recent management policy of the Eastern Shores, and, since the early 1990s, the pine plantations have been phased out by clear felling and most of these areas have been restored to grassland. The clear felling was completed in 2006 and an intense burning policy has been initiated to keep woody plants in the reclaimed grassland to a minimum. The area north of Lake Bhangazi (Fig. 1) has been managed as a wilderness area with minimal influence since the 1950s and has therefore not gone through extensive changes in land use in comparison to the central and southern parts of the Eastern Shores. The gradual planting and removal of pine has been included in the
Fig. 4 Vegetation map for Eastern Shores with pine plantation at its greatest extent. The boreholes labelled A–G have been monitored regularly since 1973; the rest are boreholes where groundwater depths were measured 20–22 April 1993 Hydrogeology Journal (2009) 17: 1949–1967
DOI 10.1007/s10040-009-0476-5
1954
Climate change and sea-level rise
Materials and methods The goal of this study was to compare the effects on the groundwater system of local management practices (restoration of natural vegetation through the removal of pine plantations) coupled with climate change. Numerical steady state (SS) simulations were conducted for two
2003/04
97/98
94/95
91/92
88/89
85/86
82/83
79/80
2000/01
DJF season DSF - linear: y = -0.0126x + 6.9617 R2 = 0.0194
76/77
24 73/74
0 70/71
22 67/68
20
50 64/65
18
100
61/62
16
150
58/59
200
55/56
14
52/53
12
250
49/50
10
300
46/47
8
350
43/44
6
400
40/41
4
450
37/38
2
500
34/35
0
550
31/32
600
28/29
DJF rainfall (mm)
Baroclinic wave disturbances in the southern oceans create weather perturbation with a period of about 6 days around the east coast of southern Africa (Preston-Whyte and Tyson 1973) suggesting a 5–10 day analytical period for rainfall analysis that has been used by Rawlins and Kelbe (1998) in their model of recharge to the groundwater on the Eastern Shores. Usman and Reason (2004) studied the relationship between dry spell frequency (DSF) and interannual drought occurrences in southern Africa. They used DSF between 1979–1980 and 2001– 2002 to assess spatial and temporal patterns in the consistency of rainfall during the mid-summer (December–January–February, DJF) season. A dry spell was defined as a pentad (a period of 5 days) with less than 5-mm rainfall. A similar DSF approach have been applied for St Lucia based on rainfall recorded at the Cape St Lucia weather station (South African Weather Service (SAWS) weather station 0339720 3) for the years 1928– 2006. There is a decreasing trend in DSF for 1928/1929– 2004/2005 (Fig. 5), and the total DJF rainfall increases over the same period (Fig. 5). This analysis suggests that the St Lucia area has become wetter rather than drier over the past 80 years. No conclusive predictions for future climate change have been presented for the Maputaland area. Based on existing models and analysis of the climate variability in the past Været et al. (2008a) concluded that the most likely scenario for Lake St Lucia and its catchments is a
warmer (∼2–3°C) and wetter (∼5–10% increase in rainfall) climate, associated with a sea-level rise of ∼0.4 m, towards 2100. The results based on the local weather data from Cape St Lucia (Fig. 5) therefore accord with the conclusions by Været et al. (2008a). Usman and Reason (2004) used the Nino 3.4 Indices to calculate correlation between El Niño Southern Oscillation (ENSO) and DSF for DJF, and found a very high degree of correlation (0.8) for a spatial average over southern Africa for the period from 1979–1980 to 2001– 2002. The same method applied to the Cape St Lucia record shows a very low correlation factor (0.2) for the short period analyzed by Usman and Reason (2004), and even lower correlation for the longer period 1928/29– 2004/05. Therefore, variations on a decadal scale (Fig. 5) seem more important for this area than the ENSO. Over the same period 1928/1929–2004/2005 there has been a rise (∼1.1°C) in annual average minimum temperatures, while maximum temperatures have decreased (∼1.3°C). A reduced difference between minimum and maximum temperatures is also in accordance with the results from regional climate models for this region (Hudson 2002; Hudson and Jones 2002). This change may be related to increased cloudiness (Hudson 2002).
DSF
model calibration process through the evapotranspiration and recharge parameters in the model.
DJF rainfall - linear: y = 0.2933x + 121.18 R2 = 0.0113
Fig. 5 1928–2005 December–January–February (DJF) rainfall and Dry Spell Frequency (DSF), their 5-year running means and linear trends. El Niño and La Niña events over the same period are indicated with black and grey arrows respectively. The slope of the linear trend lines is given by y=0.2933x+121.18 and y=–0.0126x+6.9617 for DJF rainfall and DSF respectively, while R2 is 0.0113 and 0.0194 Hydrogeology Journal (2009) 17: 1949–1967
DOI 10.1007/s10040-009-0476-5
1955
management strategies (presence of commercial pine or restoration of grasslands) under different climate and sealevel scenarios proposed by Været et al. (2008a). Even though Været et al. (2008a) found a wetter than present climate to be most likely in this area, no conclusive models have been presented. Therefore, a drier than present scenario has also been included in the analysis. This scenario can generally be applied to evaluate the response of the groundwater system to prolonged drought. The effects caused by a rise in temperature have not been modelled explicitly. The proposed scenarios used in the prediction models are summarized in Table 1. During the peak of the latest drought (2002–2007), plants growing on the margins of the lake that have low salinity tolerances were used as indicators of zones with sustained groundwater seepage (Taylor et al. 2006). Based on this vegetation study, four zones within the model domain were found to be of particular importance during prolonged droughts. These are the shorelines of (1) Catalina Bay south and Brodies Crossing, (2) Catalina Bay north, (3) Dead Tree Bay, and (4) Tewate Bay (Fig. 1).
Conceptual hydrogeological model Kelbe and Rawlins (1992a, b) and Kelbe et al. (1995) identified the (1) Isipingo, Sibayi, and Kosi Bay Formations, and (2) the Port Durnford Formation as the two main conceptual hydrostratigraphic units of the regional aquifer in their simulation studies. Idealized hydraulic values for the two main units have been estimated by Kelbe and Rawlins (1992a) using numerical models calibrated against the borehole measurements, but constrained within the range of laboratory analyses and field tests presented by Davies Lynn and Partners (1992). Since the hydraulic properties of the two hydrostratigraphic units are not well established, and appear to be indistinguishable in many areas, and the water table is generally near the interface between the hydrostratigraphic units, they have been lumped together in this study as a single vertical layer forming an unconfined aquifer with spatial variability presenting average hydraulic conditions in the vertical. The assumption of a single layer model is
unlikely to have a great influence on the model simulations and is more consistent with the spatial resolution of the model grid. The spatial extent of the units is derived from detailed calibration of the model. The model domain configures the study area as a small island that is almost completely surrounded by the Indian Ocean and Lake St Lucia that form the main drainage boundaries. The sloping Cretaceous siltstone forms the basement of the aquifer (Fig. 2). The spatial variability in the hydrostratigraphic units are identified as zones with vertically homogeneous hydraulic properties that are aligned with the stratigraphic (Fig. 2), geomorphic features (Fig. 6) and identified palaeochannels. The frontal dune and underlying formations form a hydrostratigraphic unit that is assumed to run the full length of the study area along the coastal margin that was divided into north–south components. It is assumed that a hydrostratigraphic unit coincides with the palaeochannel that runs through Lake Bhangazi from Tewate Bay, which is part of the ancient drainage channel identified by Davies Lynn and Partners (1992) stretching northwards from Mission Rocks. The assumed extension is supported by the existence of the only persistent water hole in the otherwise dry lake at Tewate Bay during the severe drought from 2002 to 2007. This water hole is sustained by significant groundwater seepage. The swamp area to the south of Lake Bhangazi is part of an extensive peat land that is assumed to extend from Lake Bhangazi to Catalina Bay (Fig. 1). This peat land overlies the main aquifer and is therefore designated as a separate hydrostratigraphic unit in this model. To the west of the deep alluvial channel, the extensive Embomveni mound comprises the western hydrostratigraphic unit (zone 3 in Fig. 6). Anomalies in groundwater levels in the southern portion of this unit are described in the model.
Numerical modelling The groundwater dynamics of the Eastern Shores of Lake St Lucia were simulated using the three-dimensional finite difference MODFLOW groundwater model developed by the United States Geological Survey (USGS; Harbaugh et al. 2000; Hill et al. 2000; McDonald and Harbaugh 1988;
Table 1 Model scenarios described as different combinations of sea level, land use, and rainfall scenarios Sea level
Land use
Rainfall
Scenario no.
Present
Pine plantation
–10% Present +10% –10% Present +10% –10% Present +10% –10% Present +10%
i ii iii iv v vi vii viii ix x xi xii
Grassland +0.4 m
Pine plantation Grassland
Scenario v represents the present conditions. The “present condition” is defined as the average annual precipitation based on the weather record from 1961 to 1990 Hydrogeology Journal (2009) 17: 1949–1967
DOI 10.1007/s10040-009-0476-5
1956
Fig. 6 Model domain and boundaries. The main hydrostratigraphic units (HRUs) are aligned with the geomorphic features and palaeochannels identified by Davies Lynn and Partners (1992). Calibrated values for hydraulic conductivity (K) are given in relation to x, y and z orientations, along with specific yield (Sy). Zone numbers in the table correspond to the numbers given to each HRU in the figure
McDonald and Harbaugh 1996) that was configured, run and analyzed in Groundwater Vistas (GWV; Rumbaugh and Rumbaugh 2004). The model domain is shown in Fig. 6. The model comprised a rectangular grid with node resolution of 100 × 100 m covering an area of 496 km2. The aquifer forms an area in the model domain of about 266 km2 (Fig. 6) with Lake St Lucia and the Indian Ocean boundaries covering the remainder of the model domain. Marine tides have been ignored because of the time resolution of the model (> day). Both the Indian Ocean and the Lake St Lucia system are configured in the Hydrogeology Journal (2009) 17: 1949–1967
groundwater model as specified (constant) head boundaries. The head values were changed to simulate future scenarios of sea-level rise. The height and the steep nature of the shoreline make inundation caused by a 0.4 m sea-level rise possible only along a very few limited parts of the shoreline. The effect of inundation on groundwater flux to the lake has therefore not been considered in the model. Neither have the potential effects of shoreline erosion caused by sea-level rise. A simple calculation based on the Ghyben–Herzberg principle (Todd 1980) shows that the theoretical salt– fresh interface is situated above the base of the aquifer in a zone 200–300 m inland from the east coast. For the remainder of the model domain the salt–fresh interface is situated well below the base of the aquifer and is not likely to affect groundwater flux into Lake St Lucia along the western margin. Density differences have therefore been ignored. The groundwater flow at the extreme northern boundary was assumed to be perpendicular to the shorelines across the very narrow section of the land and consequently this boundary could be adequately represented as a specified (no) flow boundary (Anderson and Woessner 1992). The internal drainage lines (Fig. 1) were treated as river boundaries in GWV using the river function of Prudic (1989). The river function treats the interaction between the aquifer and the drainage line as a head-dependent boundary condition. The flow of water into or out of the aquifer across the riverbeds with specified hydraulic properties is dependent on the head difference between the water level in the stream and the groundwater head in the underlying cell. Lake Bhangazi was configured as a head-dependent boundary using the lake function (LAK3) of Merritt and Konikow (2000). The model was configured in steady-state mode using average hydrometeorological conditions and then calibrated against the average measured head values in the central Eastern Shores (1973–2006) and the occasional flow measurement in the Nkazana Stream (Fig. 1) to establish initial values of the hydraulic conductivity and recharge and evapotranspiration boundary condition parameters. In addition, Meyer et al. (1993) and Davies Lynn and Partners (1992) presented values of depths to groundwater under the coastal dunes measured during various geological investigations in the region. These have been included in the model calibration of the water table along the Eastern Shores region. The model was then converted to transient mode and calibrated against the long record of borehole measurements. The period of transient simulation (1973–2006) was split into 70 time steps (stress periods) that were derived from both rainfall and borehole records reflecting periods of relatively uniform recharge conditions (Fig. 7). The longest stress period is close to 700 days while the shortest is 25 days, with an average stress-period length of 170 days. Temporal variations within the span of one stress period will be lost in the simulations that will induce some error (residual) in the comparison between the model and observed head and flux values. Similarly, daily rainfall smoothed over the DOI 10.1007/s10040-009-0476-5
1957 14 Smoothed rainfall
SP rainfall
12
mm/day
10 8 6 4 2
1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005
0
year
Fig. 7 Stress period (SP) rainfall and daily rainfall smoothed over the average stress period (time step) length (170 days)
average stress period length (Fig. 7) will not fully reflect the very short and very long stress periods. The groundwater recharge and discharge are specified through boundary conditions that simulate vertical flow paths, which represent percolation and evapotranspiration processes. These are balanced by the lateral flow through the specified boundaries under steady-state conditions. The Eastern Shores study area is almost entirely covered by well drained, highly permeable sediments without any visible signs of surface runoff except for the two streams mentioned. However, there is considerable variation in the vegetation cover as defined by the leaf area index (proportion of land cover by leaves in the vertical plain) for the different land use types (from forest to grasslands to swamp lands, Fig. 4). This will cause considerable differences between the rainfall and the recharge lost to interception and unsaturated soil moisture depletion (assuming no surface runoff for this area). Similarly, these vegetation types have very different rooting depths and rooting densities which can cause considerable variations in the extraction of soil water from the unsaturated zone and from the capillary (groundwater) zone when roots are in hydraulic contact with the phreatic surface. Since the one important objective of the study is to simulate the impact of land use, these factors are included in the groundwater model through the calibrated recharge rate for each land use and a conceptualized evapotranspirative loss function for the vegetation with roots close to the water table. This is simulated in MODFLOW using the EVT package described by McDonald and Harbaugh (1988). The EVT package calculates the evapotranspiration loss from the groundwater based on the simulated head, an ET surface, an extinction (rooting) depth where no ET occurs and a maximum evapotranspirative flux for the Hydrogeology Journal (2009) 17: 1949–1967
vegetation type. The EVT package uses a linear function between the maximum and zero evapotranspiration rates. This package has been upgraded to a curvilinear function by Banta (2000), but was not used in this study because the vertical rooting density of the land use types was not sufficiently well known to justify its application. The changes in land use through planting and clear felling of pine forests have been mapped by local management authorities. The transient data were incorporated into the groundwater model through the recharge and evapotranspiration packages in MODFLOW. The Eastern Shores was divided into recharge and evapotranspiration zones based on vegetation type (Fig. 4). Meyer and Godfrey (1995) calculated general (net) groundwater recharge rates in northern Maputaland ranging from 18% of mean annual precipitation (MAP) near the coast to 5% some 50 km inland. However, Bredenkamp et al. (1993) derived values that were approximately twice this rate. These estimates are net recharge rates that assume no further loss through evapotranspiration from the groundwater. In this study, the recharge rate must reflect the gross recharge rate if the evapotranspirative loss from the groundwater is simulated separately. In this study, the average (net) recharge was initially calibrated to be between 20 and 40% of mean annual precipitation (MAP), depending on land use type and the average depth to water table, to incorporate the simulation of groundwater evapotranspiration and concur with Bredenkamp et al. (1993). It was also calibrated against the sparse stream flow measurements in conjunction with the calibration of the hydraulic conductivities to simulate the average water-table profile. In the calibration process, it was assumed that the recharge varied from 100% of the period rainfall for open water bodies to 35% for full canopy forest based on the analysis of the rainfall– DOI 10.1007/s10040-009-0476-5
1958
Fig. 8 Daily potential evaporation and rainfall measured at St Lucia weather station (WS in Fig. 4, right panel) in 1989
groundwater response by Rawlins (1991) for the Eastern Shores. Analyses of a short record of daily potential evaporation (EP) from the Penman Model and the precipitation for St Lucia show that EP (atmospheric demand) varies seasonally throughout the year. The highest EP occurs during the summer and the lowest during the winter months (Fig. 8). From this relationship, monthly averages of EP for a season were developed to impose an upper limit to the combined evapotranspiration loss from the rainfall loss and groundwater evapotranspiration. It was established from the daily weather data (Fig. 8) that there is an inverse relationship between EP and rainfall. Short wet periods are associated with low EP while longer dry periods are associated with high EP. The EP value for each stress period was derived from the seasonal values and adjusted for periods with high or low rainfall. The maximum evapotranspirative loss from the groundwater (EPGmax), when the roots are in full hydraulic contact with the phreatic surface for each land use type, was estimated from the average difference between EP and rainfall loss to recharge. Potential groundwater evapotranspiration estimates (EPG) based on the above relationships were combined with estimates from previous studies (Kelbe and Rawlins 1992a, b; Kelbe et al. 1995; Salama et al. 2002; Wejden 2003), to derive estimates of the EPGmax values for each land use type (Table 2).
The evapotranspiration model in MODFLOW also requires estimates of the extinction (maximum rooting) depth where it is assumed no evapotranspiration will occur if the water table is below this elevation. The simulated water table profile is generally independent of the topographical surface profile, except in areas of inundation. Consequently, the inclusion of the evapotranspiration from the groundwater requires the specification of the surface topography to simulate the depth to the water table. A topographical surface in the groundwater model was derived from interpolation of the 5-m contours from the 1996 orthophotos (Surveyor General Job number 985) by Armstrong B, Sokolic F and Taylor R (Ezemvelo KZN Wildlife, South Africa, A digital elevation model (DEM) for Lake St Lucia and surrounding areas, unpublished report, 2002). The assumed rooting depth of the different land use types has been estimated from the studies of Canadell et al. (1996), Crowe (2005), Groom (2004), Guswa (2008), Murphy and Lodge (2006), and discussion with ecologists in the region. The assumed rooting depth for each land use type in the model is given in Table 2. Analysis of the borehole samples drilled on the Mfabeni plain (MN Mkhwanazi, University of Zululand, unpublished data, 2008) showed that the organic content of the soil profile extended to below 12 m which is inferred as an indicator of the extent of the roots below the recently felled commercial forest.
Table 2 Parameter values applied to the model for evapotranspiration (ET) and recharge for different land uses under average hydrometeorological conditions Land use
Rooting depth (m)
ET rate (m/day)
Recharge (m/day)
Grassland Swamp forest Dune forest Swamp Pine forest
2 1 4 1 10
0.0024 0.0030 0.0024 0.0033 0.0036
0.0018 0.0016 0.0012 0.0020 0.0035
ET rate refers to maximum evapotranspirative loss for each land use. The spatial distribution of different land uses are given in Fig. 4 Hydrogeology Journal (2009) 17: 1949–1967
DOI 10.1007/s10040-009-0476-5
1959
sufficiently adequate for an evaluation of the impact of land use on the groundwater regime during climate change studies with longer time frames if the predicted impacts are greater than the indicated error range. The model simulates similar variability within the range of values measured, and the parameters used for evaluating the impact are all sensitive to the expected change in conditions.
Results Fig. 9 The sensitivity of the model (expressed as a sum of squares of the residual, SSR, for the comparative difference (residual) between the simulated heads and observed heads when the model parameters (maximum evapotranspirative loss and rooting depth) were varied over a range of values (parameter multiplication factor) for the indicated land use types
Sensitivity and calibration The groundwater model showed considerable sensitivity to changes in both the recharge and evapotranspiration parameters for specific hydrostratigraphic units or zones (Fig. 9). Similar sensitivity for some units was identified for some of the hydraulic parameters (e.g. hydraulic conductivity). The hydrostratigraphic unit parameters were subjectively adjusted to calibrate the model using stream flow and water-table profiles to establish the parameter sensitivity and obtain the best estimate of the hydraulic and recharge parameter values. The probability of the residual error for all the comparisons with borehole measurements derived following the calibration of the model parameters is presented in Fig. 10. The plot shows that the differences between the predicted and measured groundwater heads were within 2 m of each other for about 45% of the observations; the plot also shows that the differences between the predicted and measured groundwater heads were within 3 m of each other for 60% of the time. An example of the model fit for transient conditions in selected boreholes is shown in Fig. 11. The model calibrations indicate that the simulated heads in the one-layer model have an overall fit to the observed water levels that is within 2(3) m for 45(60)% of the time. The discrepancies between observed and predicted series can be attributed to several factors that include: the temporal resolution of the recharge (stress) periods, inaccuracies in the exact location and elevation of the boreholes, inconsistencies in the borehole series measurements and spatial inaccuracies in the selected rainfall series. The measured variability in water level can exceed 4 m in some areas (see Fig. 11); this may occur at an intermediate time during a stress period which would induce large errors due to temporal smoothing by the model. However, despite these uncertainties, given comparative fit between the transient series of predicted and observed heads within 2 (3) m for 45(60)% of the time for all the boreholes, it is assumed that the model is Hydrogeology Journal (2009) 17: 1949–1967
Effects on groundwater caused by the predicted rise in sea level are compared to those effects caused by changes in land use under three different rainfall regimes. The results are presented as a series of steady-state model simulations for twelve different hydrological states (scenarios) that are categorized in Table 1 and described below. The simulation results are summarized in Fig. 12 (groundwater contours), Fig. 13 (mass balance), and Fig. 14 (fluxes to discharge boundaries).
Effects of land use changes Management practice for nearly 40 years since the late 1950s was to maintain pine plantations over vast areas of the central and southern Eastern Shores (Fig. 4). Pine plantations are shown as scenarios i–iii and vii–ix in Table 1. To bring the vegetation back to what is assumed to be the natural state, the pine has been replaced by grassland through clear felling and clearing of woody plants followed by intense burning over the past decade. This is referred to as “reintroduction of grassland” and applies to scenarios iv–vi and x–xii. The reintroduction of grassland affects the groundwater mass balance (Fig. 13) and results in a rise in the regional water table (Fig. 12) as well as increased outflow across the discharge boundaries (Fig. 14).
Effects of sea-level rise The simulations of the effects of sea-level rise gave only small differences in the total mass-balance components for the whole model domain when comparing the different estimates published by the Intergovernmental Panel on Climate Change (IPCC 2007) for future sea-level rise: namely the median (+0.4 m), the minimum (+0.18 m) and the maximum (+0.59 m). Therefore, only the results for the median estimate of 0.4 m sea-level rise have been included (scenarios vii–xii) in the comparison with the other scenarios. The groundwater contour plots in Fig. 12 show the range from lowest to highest average simulated watertable elevation, illustrated by a selection of five scenarios (i, ii, v, vi and xii). The other seven scenarios that are not illustrated (iii, iv, vii, viii, ix, x and xi in Table 1) show variations in water-table elevation within DOI 10.1007/s10040-009-0476-5
1960
Fig. 10 a The plot of observed heads against the residual difference for all boreholes. b The residual probability plot shows that about 45% of all simulated heads are within 2 m of the observed heads, and 60% are within 3 m of the observed heads
this range. A rise in the discharge boundary of this magnitude does not affect the regional water table, except in low-lying areas. The simulations show only small changes in the mass balance for a +0.4 m rise in sea level (Fig. 13), with decreased outflow across the discharge boundaries (Fig. 14). The decreased outflow is accommodated by increased loss through ET and increased discharge to the streams. ET increases in lowlying areas close to the western discharge boundary (Lake St Lucia) where a greater proportion of the roots Hydrogeology Journal (2009) 17: 1949–1967
come in contact with groundwater due to a rise in the water table.
Effects of changes in rainfall An increase or decrease in rainfall refers to a ±10% change. A change in rainfall affects the mass budget (Fig. 13), and thereby water table elevation (Fig. 12) and outflow across discharge boundaries (Fig. 14). DOI 10.1007/s10040-009-0476-5
1961
Fig. 11 Observed and simulated heads for a selection of boreholes (see Fig. 4 for locations)
The combined effects of changes in land use, sea level, and rainfall The combined effects of the above changes have been simulated and are referenced to the present sea level, present rainfall and previous pine conditions shown in Fig. 4 and labelled as scenario ii in Table 1. The simulated effects of the recent management policy where pine plantations (Fig. 13, scenario ii) have been replaced with grassland (Fig. 13, scenario v), shows a considerable rise in the water table, with more than 4-m difference under the central Embomveni mound. Adding 10% rainfall (Fig. 13, scenario vi) to the change in land use generates a general rise in water table, also under Embomveni, but of a considerably less magnitude than the change in land use alone. An additional 0.4 m sea-level rise (Fig. 13, scenario xii) only affects groundwater levels in the lower lying areas. The lowest average water table is illustrated in Fig. 13, scenario i for present sea level, pine plantations at their greatest extent, and a decrease in rainfall. Total mass balance has been calculated for all scenarios (Fig. 13). The most important factors are the rainfall and land use; both have a direct influence on the variations in total recharge. The greatest increase (∼14%) takes place with a 10% increase in rainfall in combination with restored grassland (from scenario ii to scenario vi), where Hydrogeology Journal (2009) 17: 1949–1967
both components contribute equally to the overall increase in recharge. The lowest mass budget is seen for a rainfall 10% lower than the present and the same distribution of pine plantation as shown in Fig. 4 (scenario i). Rainfall and land use significantly affect water loss through alteration of evapotranspiration (ET). The mass balance summary for the model domain shows that an increase in rainfall leads to increased ET because there is generally an increase in water level that penetrates further into the rooting zone of the vegetation. The restoration of grassland resulted in a reduction in rooting depth and hence the potential ET that resulted in an overall decrease in actual ET. The lowest ET values are found for a combination of grassland and a precipitation of 10% lower than that of today (scenarios iv and x). A reduction of ∼20% in ET loss takes place with a restoration of grassland and reduced rainfall (scenarios iv and x). A sea-level rise has a more moderate influence on the ET, due to the small rise in the water-table profile. The greatest increase in the overall ET loss (14%) can be expected for a rise in sea level, combined with pine plantations and increased rainfall (scenario ix). The simulations (Fig. 13) indicate that groundwater runoff to streams increases in nearly all situations where there is a rise in sea level, an increase in rainfall and a DOI 10.1007/s10040-009-0476-5
1962
Fig. 12 Contour plots for five selected scenarios. Contour lines are given at 2 m intervals. a pine, –10% rainfall, present sea level; b pine, present rainfall, present sea level; c grassland, present rainfall, present sea level; d grassland, +10% rainfall, present sea level; and e grassland, +10% rainfall, +0.4 m sea level (see Table 1)
ET, out
Rivers, out
Constant head, out
Recharge, in
m 3/day
900000
100000
800000
200000
700000
300000
600000
400000
500000
500000
400000
600000
300000
700000
200000
800000
100000
900000 1000000
0 Scenario # Rainfall Land use
m 3/day
0
1000000
i -10 %
ii
iii
Present +10 %
iv -10 %
Pine
Sea level
v
Present +10 % Grassland
Present
vi
vii -10 %
viii
ix
x
Present +10 %
-10 %
Pine
xi
xii
Present +10 % Grassland
+0,4 m
Fig. 13 Model summary mass balance given as absolute values (m3/day) for 12 different scenarios Hydrogeology Journal (2009) 17: 1949–1967
DOI 10.1007/s10040-009-0476-5
1963
Fig. 14 a Outflow (m3 per day per km of coastline) across specified constant head boundaries for different combinations of sea level, land use, and rainfall. b Outflow to Lake St Lucia for different combinations of land use and rainfall. Sea level is the same (present) for all combinations
change from pine plantations to grassland. An increase in sea level alone (scenario viii) has only a minor effect, while removal of pine plantations (scenario v) increases the flow into rivers by ∼50%. The combination of sealevel rise, the restoration of grassland, and increased rainfall (scenario xii), will increase water flow into the rivers by more than 60%. With the restoration of grassland, almost all transmission of water from the rivers into the aquifer stops (scenarios iv–vi and x–xii). Simulated flux through the shoreline into the Indian Ocean and into Lake St Lucia increases with increased rainfall (∼4%) and with grass replacing pine plantations (∼12%). The flux for pine plantation coupled with a +10% rainfall (scenario iii) is almost the same as the flux for grassland coupled with a –10% rainfall (scenario vii). The Hydrogeology Journal (2009) 17: 1949–1967
greatest increase in flux (∼16%) is the result of a combination of increased rainfall and a reintroduction of grassland (scenarios vi and xii). A sea-level rise moderates the increase by ∼4%. The greatest decrease in flux takes place with increased sea level, and reestablishment of pine plantations in combination with decreased rainfall (–11%; scenario vii). Figure 14 shows the calculated groundwater fluxes (m3/day per km coastline) into the lake shoreline and the Indian Ocean coastline. As the groundwater flow into Lake St Lucia is very important for the estuary and for the lake margin vegetation (Taylor et al. 2006), the flux along four important seepage zones (Catalina Bay south and Brodies Crossing, Catalina Bay north, Dead Tree Bay, and Tewate Bay, for location see Fig. 1) have been simulated. DOI 10.1007/s10040-009-0476-5
1964
The model simulations show that two thirds of the groundwater flows to the Indian Ocean while the rest flows into Lake St Lucia as diffuse seepage or via specified drainage lines (e.g. Nkazana and Tewate streams). However, the seepage flux varies considerable along these shorelines. The flux along the specified seepage zones is much higher than the average for the Lake St Lucia shoreline as a whole (Fig. 14). The fluxes from Brodies/Catalina Bay south, Central Dead Tree Bay and Tewate Bay are closer to the flux into the Indian Ocean than the average flux into Lake St Lucia. Generally both an increase in rainfall and a removal of pine plantations result in increased outflow to Lake St Lucia. A change in land use from pine plantations to grassland causes a greater increase in total outflow from the aquifer compared to the effect of increased rainfall alone (Fig. 14). The effect of land use is negligible for Tewate Bay since there has never been pine plantation in this area. Sea level rise generally results in a small decrease in outflow with the most noticeable effect seen for Tewate Bay (Fig. 14). The northern part of Catalina Bay (west of the Nkazana Stream outlet) represents an exception from the trends described above. Although fed by groundwater from the southern parts of the extensive Embomveni Mound, and identified by the ecologists as an important groundwater seepage zone, the simulated flux per kilometre of coastline is much lower than for the other three seepage zones, and even lower than the average simulated flux into the lake in total. However, this is the area where the model calibration suggests that there is some uncertainty in the hydraulic conditions.
Discussion Concern has been expressed regarding the effects of future changes in climate and sea level on groundwater resources (Cheng and Ouazar 2004; IPCC 2007). Aquifers are already under great stress to meet increasing freshwater demands, not only through direct consumption, but also in the production of ecosystem goods and services. South Africa’s demand for wood is predominantly met by industrial forest plantations, and the demand will continue to increase with the expected economic growth (DWAF 1996). The current study and other studies (e.g. Salama et al. 2002) have suggested that a change in land use, resulting in a reduction in pine plantations, may mitigate, and even surpass potential negative effects of future global change, while an expansion of plantations are likely to reduce the groundwater recharge and discharge drastically. Local differences apply, and such changes in land use must be guided by local management plans where and when appropriate. There is a growing awareness internationally that more trees are not always good and deforestation is not only bad (FOU 2007). South Africa has applied this in their national management strategies, and despite the increased demand for wood, South Africa has granted fewer permits for new afforestation in later Hydrogeology Journal (2009) 17: 1949–1967
years because of a country-wide water scarcity (DWAF 1996; FOU 2007), and the deforestation of the Eastern Shores is a result of this overall way of thinking. Groundwater seepage from the Eastern Shores has been recognized as an important component of the average total freshwater budget to Lake St Lucia in dry periods when surface flow declines (Kelbe and Rawlins 1992a). Taylor et al. (2006) pointed out that protection of the primary aquifers of the Eastern Shores is of great importance for the resilience of ecosystems in Lake St Lucia. These authors highlighted the seepage zones along the shorelines of the Eastern Shores-Lake St Lucia interface as an area of particular importance as refugia for the survival of various organisms during droughts and high salinities. Along with the removal of more than 5,000 ha of pine trees during the last decade, the current management policy is to burn all areas of the Eastern shore below 30 m a.s.l. frequently and severely to reduce the spread of indigenous woody plant components. The short-term effect of deforestation on groundwater level has been confirmed by the observations of inundated pine stumps by rising groundwater soon after pine felling during dry periods. This effect is also reflected in the groundwater level records (Fig. 3), and through the simulations in this study (Fig. 12). The transient simulation for the calibrated model indicates that the model, for these impact assessments, adequately reflects the variations in groundwater levels and discharge rates for the recharge and ET conditions since 1973 (Fig. 11). The model reproduces the observed responses in water levels over most of the Eastern Shores study area to within 2–3 m of the observed levels. However, while the model adequately simulates the head variations for the south-western part of Embomveni, it underestimates the head elevation for this area. This probably leads to lower simulated seepage rates than would be observed along the shoreline in Catalina Bay north but still within 10–20% of the expected. It is suspected that this is due to the perching or damming of the groundwater in this region. Unfortunately, this could not be further investigated using a one layer model. The overall loss of groundwater to ET is a function of root depth and its relation to the groundwater depth. Any process that causes a change in the water-table elevation will affect how much of the root system is located in contact with the groundwater zone, and thereby also the actual evapotranspiration. Thus, a shift in land use from deep-rooted pine plantations (root depth 5–10 m) to grassland with shallower roots (1–2 m) would have decreased the overall loss of groundwater contribution through ET. The linear model of ET from the groundwater is a simple approach and a more physically based function for this relationship may have improved the model. A recent development through MODFLOW-2005, which includes the simulation of the Unsaturated Zone Flow (UZF1) Package for modelling unsaturated flow between the land surface and the water table, has just been released by Niswonger et al. (2006) and would reduce some of the uncertainty in the predictions. Accuracy of the digital elevation model (DEM) used in the groundwater model is DOI 10.1007/s10040-009-0476-5
1965
important for the ET calculation, and the DEM (based on 5-m contours) applied is considered adequate for this study because of the smooth undulating nature of the surface topography. The importance of topography and the depth to water table is evident in areas like Embomveni where depth to water table is such that pine roots may reach down to the groundwater while grass roots cannot. The overall simulated result from a change in land use in the Embomveni area has predicted a decrease in the groundwater contribution to ET, a subsequent elevation of the water table, which produced a steeper hydraulic gradient to the drainage boundaries, creating a greater flux of water into Dead Tree Bay. In areas with no change in land use such as east of Tewate Bay, a change in rainfall and/or sea-level rise will have a greater effect on water-table elevation and ET loss from the groundwater. The impact of change in land use is insignificant under the high dunes because the depth to water table is too deep for roots (mainly indigenous dune forest) to dip into the water table, i.e. there is no extraction from the aquifer in this area from evapotranspiration. The lowering of groundwater levels through evapotranspiration from the groundwater by deep-rooted trees such as pine, has been identified in several studies (Kelbe and Rawlins 1992a, b; Kelbe et al. 1995; Le Maitre et al. 1999; Salama et al. 2002; Sun et al. 2000). Salama et al. (2002) simulated a continued drop in water level for Gnangara Mound, Australia, with the persistence of pine plantation, even if pumping for human utilization was stopped. Salama et al. (2002) further simulated a recovery of the water table after pine removal, even with continued pumping. This is in accord with the results reported here and shows clearly that a consideration of the dominant vegetation types is probably one of the most important factors in the management of groundwater aquifers. For the Eastern Shores, a rise in sea level decreases the predicted overall water-table gradient because of internal drainage boundary changes and thereby reduces the flux into the Indian Ocean and Lake St Lucia (Fig. 14). Areas presently covered by wetlands along the lake margin (see Fig. 4) may become further inundated, but these areas are so small compared to the total size of the aquifer that the effect on overall outflow is most likely to be negligible. The flooding of low-lying areas due to sea-level rise has therefore not been incorporated into the model predictions. A 10% decrease in rainfall should be sufficient to include the possible effects of temperature rise on evaporation rates, although this has not been investigated explicitly in this study. The combined negative effects of sea-level rise and decreased rainfall on groundwater outflow to Lake St Lucia are nearly counterbalanced by the already implemented change in land use. Without a rise in sea level the simulations indicate that the removal of the pine forest has actually increased the outflow to Lake St Lucia so much that this would overshadow a lower rainfall regime. This suggests that the sustained outflow to Lake St Lucia during extreme and prolonged droughts, as presented by Taylor et al. (2006), would Hydrogeology Journal (2009) 17: 1949–1967
continue in the future even with a worst-case scenario of higher sea level and decreased rainfall. Groundwater seepage from the Eastern Shores is therefore believed to remain an important factor to retain the resilience of Lake St Lucia in the future. This will, in particular, apply if future rainfall over the St Lucia catchment areas decreases such that the river flows into the lake decrease. Scholes and Archer (1997) found that the expansion of woody plants was related to interactions between browsers, grazers, climate change, and fires. On the Eastern Shores, man and fire have influenced the landscape for a long time—long enough for fire to be regarded as one of the natural processes that have shaped the vegetation types as they are currently known—resulting in a fire subclimax state of the vegetation. Local management authorities seek to retain this by mimicking the processes that occurred prior to the time when this became a protected area, i.e. what was happening more than a century ago in order to manage the groundwater through land use. Bond and Midgley (2000) proposed that elevated atmospheric CO2 levels, as seen over the past century and predicted for the future, increases the photosynthetic capabilities of woody plants. This promotes the growth of certain woody plants that are able to store additional carbohydrates produced in underground roots. These plants become more resilient to the effects of fires. The stored carbohydrates allow them to sprout much sooner after being burnt to the ground than those plants that are unable to store carbohydrates. This effect promotes the wooding up of the grassland. Bond and Midgley (2000) also suggest that the same effect occurred in palaeoecosystems resulting in woody plant expansions during periods of high CO2 concentrations, and grassland expansion during periods of low CO2 concentrations. If this hypothesis of Bond and Midgley (2000) is correct, then burning may not be the tool to control the wooding up of grasslands over a longer period of time. Based on the model results, the management strategy of removing the pine plantations over the past decade or so must be deemed as a success in reducing groundwater losses. There is substantial knowledge to be gained from further development of the Eastern Shores groundwater model where the effect of the successions natural vegetation could be included. For transient models a refinement of the stress periods may improve how the model simulates short-term changes. The understanding of the mechanisms of how land use changes affect recharge and groundwater ET losses can be further improved through the incorporation in the groundwater models of emerging methods that include the unsaturated zone processes. Further investigation is necessary to explain the lack of model fit for the south-western part of Embomveni and additional layers in the model may be necessary to reflect complexities in the geology/lithology of this area. The major changes in water-table elevation seen for different scenarios will have implications for the wetland distribution on the Eastern Shores in the future. Wetlands are of major importance to the biodiversity on the Eastern Shores, and wetland dynamics and possible DOI 10.1007/s10040-009-0476-5
1966
consequences of future climate change are included in current studies of the Eastern Shores groundwater system.
Conclusions Steady-state groundwater simulations for the Eastern Shores under different climate and sea-level scenarios suggest that local management practice may have a more substantial effect on the groundwater regime than effects of the predicted future climate and sea-level changes. The current management strategy has positive effects on the groundwater budget, with rising water tables and increased seepage to Lake St Lucia resulting from land-use changes. The simulations show that groundwater seepage from the Eastern Shores will continue to play an important role for the resilience of Lake St Lucia in the future. Analyses of the impact of climate change and land use may make managers of drought-exposed systems better prepared to cope with global change. It is believed that similar studies also will prove to be important in many other areas in the world. Acknowledgements We want to thank the following institutions and persons: Financial funding was provided by the Norwegian Council for Higher Education’s Programme for Development Research and Education, and by the Norwegian University of Life Sciences. C. Fox and the late A. Myeza at Ezemvelo KZN Wildlife kindly provided groundwater monitoring data and assisted during field work
References Anderson MP, Woessner WW (1992) Applied groundwater modelling: simulation of flow and advective transport. Academic Press, New York Banta ER (2000) MODFLOW-2000, The US Geological Survey modular ground-water model: documentation of packages for simulating evapotranspiration with a segmented function (ETS1) and drains with return flow (DRT1). US Geological Survey Open-File Report 00-466 Bond WJ, Midgley GF (2000) A proposed CO2-controlled mechanisms of woody plant invasion in grasslands and savannas. Glob Change Biol 6:101–109 Bredenkamp DB, Botha LJ, Venter D et al (1993) St Lucia Lake: Geohydrology and related aspects. In: Taylor RH (ed) Proceedings of the workshop on water requirements for Lake St Lucia. Fanies Island, 12–13 May 1992, Department of Environment Affairs, Pretoria, South Africa Canadell J, Jackson RB, Ehleringer JR, Mooney HA, Sala OE, Schulze ED et al (1996) Maximum rooting depth of vegetation at the global scale. Oecologia 108(4):583–595 Cheng AHD, Ouazar D (2004) Coastal aquifer management monitoring, modelling, and case studies. Lewis, Boca Raton, FL Crowe P (2005) The influence of soils and species on tree root depth. Forestry Commission, Information Note Nv 2005, Edinburgh, UK Davies Lynn and Partners (1992) Landform geomorphology and geology, chapter 2. In: Coastal and environmental services, Environmental Impact Assessment—Eastern Shores of Lake St Lucia (Kingsa/Tojan) Lease Area. Specialist reports, vol 1, Biophysical Environment, Rhodes University, Grahamstown, South Africa DWAF (1996) Sustainable forest development in South Africa: the policy of the government of national unity. White paper, Hydrogeology Journal (2009) 17: 1949–1967
Ministry of Water Affairs and Forestry, Pretoria, South Africa. http://dwafapp4.dwaf.gov.za/webapp/?page=documents&sub Page=legislation. Cited 7 September 2007 FOU (2007) State of the world’s forests 2007. Food and Agricultural Organization of the United Nations, Rome. http:// www.fao.org/docrep/009/a0773e/a0773e00.htm. Cited 7 September 2007 Groom PK (2004) Rooting depth and plant water relations explain species distribution patterns within a sandplain landscape. Func Plant Biol 31(5):423–428 Guswa AJ (2008) The influence of climate on root depth: a carbon cost-benefit analysis. Water Resour Res 44, W02427. doi:10.1029/2007WR006384 Harbaugh AW, Banta ERHMC, McDonald MG et al (2000) MODFLOW-2000, the U.S. Geological Survey modular ground-water model: user guide to modularization concepts and the ground-water flow process. US Geol Surv Open-File Rep 00-92 Hill MC, Banta ER, Harbaugh AW, Anderman AE et al (2000) MODFLOW-2000, the U.S. Geological Survey modular ground-water model: user guide to the observation, sensitivity, and parameter-estimation processes and three post-processing programs. US Geol Surv Open-File Rep 00-184 Hobday DK (1979) Geological evolution and geomorphology of the Zululand coastal plain. In: Allanson BR (ed) Lake Sibaya, chapter 1. Monographiae Biologicae Series, vol 36, Springer, New York Hudson DA (2002) Future changes in temperature and precipitation extremes over southern Africa. DEFRA report 2/2/02, Hadley Centre for Climate Prediction and Research, Met Office, Bracknell, UK Hudson DA, Jones RG (2002) Regional climate model simulations of present-day and future climates of southern Africa. Hadley Centre technical note 39, Hadley Centre for Climate Prediction and Research, Met Office, Brucknell, UK IPCC (2007) Climate Change 2007: the physical science basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, Cambridge University Press, Cambridge, UK Kelbe BE, Rawlins BK (1992a) An evaluation of the geohydrology of the Eastern Shores catchment of Lake St Lucia. Final report to Department of Water Affairs and Forestry, Pretoria, South Africa Kelbe BE, Rawlins BK (1992b) Hydrology: environmental impact assessment for eastern shores of Lake St Lucia, chapter 3. In: Coastal and environmental services: environmental impact assessment—Eastern Shores of Lake St Lucia (Kingsa/Tojan Lease Area). Specialist reports, vol 1: Biophysical Environment, Rhodes University, Grahamstown, South Africa Kelbe BE, Rawlins BK (1995) Geohydrology of the Eastern Shores of St Lucia. In: Taylor RH (ed) Proceedings from the workshop on water requirements for Lake St Lucia. Fanies Island, 12–13 May 1992, Department of Environment Affairs, Pretoria, South Africa Kelbe BE, Rawlins BK, Nomquphu W et al (1995) Geohydrological modelling of Lake St Lucia. Hydrological Research Unit report, Department of Water Affairs and Forestry, University of Zululand, KwaDlangezwa, South Africa King LC (1972) The natal monocline, 1st edn. University of Natal Press, Pietermaritzburg, South Africa Le Maitre DC, Scott DF, Colvin C et al (1999) A review of information on interactions between vegetation and groundwater. Water SA 25(2):137–152 Lubke RA, Avis AM, Phillipson PB et al (1992) Vegetation and Floristics, chapter 7. In: Coastal and environmental services, Environmental Impact Assessment—Eastern Shores of Lake St Lucia (Kingsa/Tojan Lease Area). Specialist reports, vol 1, Biophysical environment, Rhodes University, Grahamstown, South Africa Maud RR (1980) The climate and geology of Maputaland. In: Bruton MN, Cooper KH (eds) Studies on the ecology of Maputaland, Rhodes University and The Natal Branch of the DOI 10.1007/s10040-009-0476-5
1967 Wildlife Society of Southern Africa, Grahamsown and Durban, South Africa Maud RR, Orr WN (1975) Aspects of post Karoo geology in the Richards bay Area. Trans Geol Soc S Afr 78(11):101–109 McDonald MG, Harbaugh AW (1988) A modular three-dimensional finite-difference ground-water flow model. Techniques of Water-Resources Investigations, book 6, chapter A1, US Geological Survey, Reston, VA, 586 pp McDonald MG, Harbaugh AW (1996) User's documentation for MODFLOW-96, an update to the U.S. Geolgogical Survey modular finite-difference ground-water flow model. US Geol Surv Open-File Rep 96-486, 220 pp Merritt ML, Konikow LF (2000) Documentation of a computer program to simulate lake-aquifer interaction using the MODFLOW ground-water flow model and the MOC3D solute-transport model. US Geol Surv Water Resour Invest Rep 00-4167, 146 pp Meyer R, Godfrey L (1995) Characterization and Mapping of the Groundwater Resources, Kwazulu Natal Province. Mapping Unit 7, Directorate of Geohydrology, Department of Water Affairs and Forestry, Pretoria, South Africa Meyer R, McCarthy MJ, Eglington BM (1993) Further investigation of the geology and geohydrology of the dune cordon on the Eastern Shores of Lake St Lucia. EMAP-C-93032, Division of Earth, Marine and Atmospheric Science and Technology, CSIR, Pretoria, South Africa Murphy SR, Lodge GM (2006) Root depth of native and sown perennial grass-based pastures, North-West Slopes, New South Wales. 2. Estimates from changes in soil water content. Aust J Exp Agr 46(3):347–359 Murray BR, Zeppel MJB, Hose GC, Eamus D et al (2003) Groundwater-dependent ecosystems in Australia: it's more than just water for rivers. Ecol Manage Restor 4(2):110–113 Niswonger RG, Prudic DE and Regan RS et al (2006) Documentation of the Unsaturated Zone Flow (UZF1) Package for modeling unsaturated flow between the land surface and the water table with MODFLOW-2005, section A, chapter 19, book 6. Modeling Techniques, Techniques and Methods 6-A19, US Geological Survey, Reston, VA Preston-Whyte RA, Tyson PD (1973) Note on pressure oscillation over South Africa. Mon Weather Rev 101:650–653 Prudic DE (1989) Documentation of a computer program to simulate stream-aquifer relations using a modular, finite difference, groundwater flow model: US Geol Surv Open-File Rep 88-729 Rawlins BK (1991) A Geohydrological assessment of the behaviour and response of the Zululand coastal plain to both environmental influences and human activity. MSc Thesis, University of Zululand, South Africa Rawlins BK, Kelbe BE (1991a) The influence of land use on the hydrology of the dune systems of the Natal north coast. In: Everard DA, von Maltitz GP (eds) Dune forest dynamics in relation to land-use practices. Foundation for Research and Development, Pretoria, South Africa Rawlins BK, Kelbe BE (1991b) Case study of the hydrological response of a shallow coastal aquifer to afforestation. In:
Hydrogeology Journal (2009) 17: 1949–1967
Nachtnebel HP, Kovar K (eds) Hydrological basis of ecological sound management of soil and groundwater. IAHS Publication No 202. IAHS, Wallingford, UK Rawlins BK, Kelbe BE (1998) Groundwater modelling of the impact of commercial forestry on an ecologically sensitive coastal lake hydrology in a changing environment. International Conference of the British Hydrological Society, Exeter, UK, July 1998 Rumbaugh JO, Rumbaugh DB (2004) Guide to using groundwater vistas. Environmental Simulations, Reinholds, PA, USA Salama RB, Vermooten S, Silberstein RP, Hatton TJ, Hodgson GA, Pollock D et al (2002) Capture of more recharge and discharge is the only means of increasing the sustainable yield of Gnangara Mound. Paper presented at United Nations Environment Program (UNEP) International Conference on Sustainability of Water Resources, Murdoch University, Perth, Australia, 13–14 November 2002 Scholes RJ, Archer SR (1997) Tree-grass interactions in savannas. Annu Rev Ecol Syst 28:517–544 Sokolic F (2006) The use of satellite remote sensing to determine the spatial and temporal distribution of surface water on the Eastern Shores of Lake St Lucia. MSc Thesis, School of Environmental Sciences, University of KwaZulu-Natal, South Africa Stassen R, Mynhardt V (1992) A hydrological study of the Lake St Lucia catchment, part I. Department of Water Affairs and Forestry, Directorate of Hydrology, Pretoria, South Africa Sun G, Riekerk H, Kornhak LV et al (2000) Ground-water-table rise after forest harvesting on cypress-pine flatwoods in Florida. Wetlands 20(1):101–112 Taylor R, Kelbe BE, Haldorsen S, Botha GA, Wejden B, Været L et al (2006) Groundwater-dependent ecology of the shoreline of the subtropical Lake St Lucia estuary. Environ Geol 49(4):586– 600 Todd DK (1980) Groundwater hydrology. Wiley, New York Usman MT, Reason CJC (2004) Dry spell frequencies and their variability over southern Africa. Clim Res 26:199–211 Været L, Sokolic F (2008a) Methods for studying the distribution of groundwater-dependent wetlands: a case study from Eastern Shores, St Lucia, South Africa. In: Været L (2008) Responses to global change and management actions in coastal groundwater resources, Maputaland, southeast Africa. PhD Thesis, Norwegian University of Life Sciences, Norway Været L, Haldorsen H, Kelbe BE, Botha GA et al (2008b) Changes in climate and sea level in Maputaland, south-eastern Africa: from palaeodata to near future model scenarios. In: Været, L (2008) Responses to global change and management actions in coastal groundwater resources, Maputaland, southeast Africa. PhD thesis, Norwegian University of Life Sciences, Norway Wejden B (2003) Groundwater discharge along Eastern Shores of Lake St Lucia, South Africa: a basis for ecological refuges. MSc Thesis, Agricultural University of Norway, Norway Wright CI, Miller WR, Cooper JAG (2000) The late Cenozoic evolution of coastal water bodies in Northern Kwazulu-Natal, South Africa. Mar Geol 167(3–4):207–229
DOI 10.1007/s10040-009-0476-5
