Dry season water use patterns underGuiera senegalensisL. shrubs in a tropical savanna* 1

Journal of Arid Environments (1998) 40: 53]67 Article No. ae980435

Dry season water use patterns under Guiera senegalensis L. shrubs in a tropical savanna Stephen R. Gaze w †, Joost Brouwer‡§, Lester P. Simmonds¶ & John Bromley w w

Institute of Hydrology, Wallingford OX10 8BB, U.K.

‡Dept. of Agronomy, Wageningen Agricultural University, PO Box 341, 6700 AH Wageningen, The Netherlands

¶ Dept. of Soil Science, University of Reading, PO Box 233, Reading RG6 6DW, U.K. The extent of rooting and the water use of Guiera senegalensis bushes at a fallow savanna site in south-west Niger were investigated. Low root length densities were found up to the maximum sampling distance (8 m horizontally and 2 m vertically) from the base of the bushes. Changes in soil water content of the soil profile were monitored using a neutron probe at seven locations over a 20 m transect between two G. senegalensis bushes. During the 1993r94 dry season, a progressive drying front was observed moving away from a G. senegalensis bush both horizontally (up to 10 m) and vertically (up to 4 m). Mean water loss from the top 4 m of the soil profile over the 8-month dry season was 151 mm. This was attributed to water use by G. senegalensis, and comprised 28% of the 1993 total annual rainfall. Clearing fallow savanna land for millet production will result in increased deep drainage partly through reduced dry season water use by the deep rooted bushes. q 1998 Academic Press Keywords: Guiera senegalensis; HAPEX-Sahel; fallow; savanna; semi-arid; water use; Niger; roots

Introduction The more or less stable coexistence of perennial woody and annual herbaceous plants in African savannas has been attributed to a complementarity of soil water use resulting from different rooting habits (Walker et al., 1981; Hesla et al., 1985; Knoop & Walker, 1985). The herbaceous plants root only in the upper soil layer whilst the woody plants have roots which also extend deeper into the soil profile and therefore have access to soil water which is not available to the shallow-rooted herbs. There is, however, some competition for soil water in the surface layers between the woody and herbaceous †Address for correspondence: Cambridge University Farm, Huntingdon Road, Girton, Cambridge CB3 0LH, U.K. §Formerly at ICRISAT Sahelian Center, BP 12404, Niamey, Niger, on secondment from Dept. of Soil Science and Geology, Wageningen Agricultural University, The Netherlands. 0140]1963r98r010053 q 15 $30.00r0

q Academic Press

54

S. R. GAZE ET AL.

plants (Knoop & Walker, 1985) because both components of the vegetation have roots in this surface layer. Indeed, in some savannas, shallow wetting depths result in no complementarity of rooting habit, with woody and herbaceous plant roots confined exclusively to the surface layers (Seghieri, 1995). Complementarity of soil water use (and other plant requirements such as nutrients and light) is also an important feature in agroforestry systems (Anderson & Sinclair, 1993; Smith et al., 1997). Much agroforestry research is directed towards minimizing competition and maximizing complementarity between the component species. The savanna ecosystem can be viewed as a natural agroforestry system, with herbaceous plants for grazing interspersed with irregularly spaced woody plants. Much of the savanna of south-west Niger develops during the normal agricultural cycle when previously cultivated land is left fallow for several years to restore soil fertility. It is therefore sometimes referred to as fallow savanna (e.g. Gash et al., 1991) or simply fallow (e.g. Wallace et al., 1994), but is typical of the low tree and shrub savanna described by Cole (1986). This paper reports on measurements of soil water use by plants in a fallow savanna in south-west Niger, with particular interest in the spatial and temporal pattern of water use through the 8-month dry season.

Methods Study site The study site was located at 138 159 N and 208 159 E, about 4 km west of the ICRISAT Sahelian Center (ISC), and 40 km south of Niamey, Niger. It was the fallow sub-site of the Southern Super-Site of the multinational Hydrologic Atmospheric Pilot Experiment in the Sahel, HAPEX-Sahel, which took place in Niger from 1991]1993 (Goutorbe et al., 1994). The site had not been cropped since about 1986, allowing the natural vegetation to regenerate. In 1992]1995, vegetation at the site comprised Guiera senegalensis L. shrubs with an herbaceous understorey and a few scattered trees, generally Combretum species. Guiera senegalensis is a multi-stemmed drought-deciduous bush, growing 2]3 m tall. Levy et al. (1997) counted 220 G. senegalensis bushes in an area of approximately 6750 m2 at this site in February 1992, averaging 327 bushes per ha, or 1 bush per 30 m 2 . The more common herbs included Cassia mimosoides L., Cenchrus biflorus Roxb., Digitaria gayana Stapf ex A. Chev., Eragrostis tremula Hochst. ex Steud., Mitracarpus villosus (Sw.) DC. and Sida cordifolia L. A complete listing of species recorded at the site in 1992 is given by Wallace et al. (1994), and comprised both annuals and perennials. Only the perennial grass Andropogon gayanus Kunth. stayed green throughout the year. However, because of grazing pressure, it remained as short, closely grazed clumps. Andropogon gayanus was not common at the site and tended to favour the open areas. The rest of the herbs emerged in early June following the start of the rainy season, flowered in late August and set seed and senesced in September at the end of the rains. Soil at the site comprised on average 88% sand, 3% silt and 9% clay, and was classified as a Psammentic Paleustalf under the USDA system (Soil Survey Staff, 1992) and an Arenosol in the FAO system (FAO-ISRIC, 1990). This sandy soil was 2 to 5 m deep overlying a weathered laterite layer. Below this was a sequence of Continental Terminal Miocene deposits (siltstones and mudstones) with a ground-water table within an oolitic ironstone aquifer at about 32 m below the soil surface. The overall gradient of the site was 2]3%.

DRY SEASON WATER USE BY SHRUBS

55

Climate The climate of the region is typical of the southern edge of the Sahelian zone with a single summer rainy season (June]September) and high temperatures throughout the year. Long-term average potential evaporation exceeds rainfall in all months except August, when potential evaporation more or less equals rainfall (Sivakumar et al., 1993). The long-term mean annual rainfall recorded at Niamey, 40 km north of the study site, is 545 mm. The corresponding mean annual potential evaporation is 2294 mm. In the south-west of Niger, there is a strong rainfall gradient from north to south of about 1 mm kmy1 (Sivakumar et al., 1993). The long-term average annual rainfall for the study site is therefore probably closer to 585 mm. Rainfall at the site was measured as part of the EPSAT-Niger (Estimation of Precipitation by Satellite}Niger) experiment (Lebel et al., 1992) using an automatically logged tipping-bucket (0.5 mm) raingauge installed 1.5 m above the soil surface. Details of the monitoring, maintenance, data collection and processing are given by Lebel et al. (1995).

Root sampling Samples for root length measurement were taken from the study site on 6 and 7 June 1994, at the beginning of the wet season when the G. senegalensis bushes were beginning to come into leaf. The herbs were just beginning to emerge but the most advanced herbs were still only at cotyledon stage. Samples were taken in two transects moving away from G. senegalensis bushes into an open area, about 20 m north of access Tube 5404 (see Fig. 1). One transect went from a bush into an open area several metres from the nearest A. gayanus. The other transect went from another bush into an open area with sparse, closely grazed A. gayanus. Samples were taken 1, 2, 4 and 8 m from the centre of the bushes. At each

Figure 1. Position of individual neutron probe access tubes in the transect between two Guiera senegalensis bushes at the savanna study site in south-west Niger.

S. R. GAZE ET AL.

56

position the depths sampled were 0]15, 15]25, 25]35, 35]50, 50]70, 70]90, 90]110, 110]150, 150]190 and 190]230 cm. A tube (49 mm internal diameter) was driven into the soil to obtain the samples. The wetting front at the time of sampling was at about 60 cm. Below this the soil was still quite dry and sampling was difficult. Consequently, at some positions it was not possible to sample to 230 cm depth. Following sampling, the soil was washed through a 0.5 mm sieve to separate the roots from the soil. The root samples were then counted using the line intersect method (Tennant, 1975). The G. senegalensis roots tended to have a darkish red colour, and an attempt was made to distinguish between G. senegalensis roots and herb roots at counting. However, with fine roots this was very difficult and some fine roots of G. senegalensis may have been counted as herb roots.

Soil water content monitoring A transect of seven neutron probe access tubes was installed between two G. senegalensis bushes towards the end of the 1993 wet season (Fig. 1). The bushes at each end of the transect were 20 m apart, with only herbaceous ground cover between them. The transect was oriented in a roughly west]east direction. There was a very gentle slope (f2%) running south-west across the transect. Tube 5104 was just upslope of the bush marking the western end of the transect. Tube 5704 was just downslope (and in the rainshadow) of the bush marking the other end of the transect. Distances between the tubes and the nearest G. senegalensis bush are given in Table 1. Tubes 5204, 5404 and 5604 were installed in early August 1993 using a trailer mounted drilling rig. Tubes 5104, 5304, 5504 and 5704 were installed on 9 September 1993 using a hand auger, following Bell (1987). All the tubes were installed with a good fit between the drilled hole and the access tube. Volumetric water content profiles were measured using a neutron probe (IH Mark II probe, Didcot Instrument Co., U.K.). Measurements were taken every 1]2 weeks at 10, 20, 25, 30, 40, 60, 80, 100, 120, 140, 160, 175 and 225 cm depths, and 50 cm increments thereafter to the bottom of the access tube. Maximum reading depths for each tube are given in Table 1. Data sets for the neutron probe calibration were obtained from the study site, a nearby millet field and the nearby ISC. There was no discernable difference between

Table 1. Distance from Guiera senegalensis bush and maximum reading depth for the transect of deep access tubes used to follow dry season water use patterns in a savanna in south-west Niger

Tube No. 5104 5204 5304 5404 5504 5604 5704 w

Distance from nearest bush (m)

Distance from bush at W end of transect (m)

Maximum reading depth (cm)

Maximum depth of wetting front in 1993 (cm)

1 2 5 10 5 2 1

1 2 5 10 15 18 19

375 475w 375 725 375 525 325

375 375 )375 375 375 275 275

Until late October 1993, then 375 cm.

DRY SEASON WATER USE BY SHRUBS

57

calibration data from these sites, so data were pooled. Separate calibration equations were derived for depths shallower than 30 cm. A single combined calibration equation was used for depths 30]100 cm. In the absence of calibration data for depths greater than 100 cm, the combined calibration determined for the 30]100 cm depth was also used for depths below 100 cm. This assumes that the soil at these depths is similar, or has little influence on the neutron probe calibration. However, the iron in the laterite layer renders this assumption invalid and so absolute water contents for depths greater than about 375 cm (approximate depth to laterite) may be erroneous. Consequently, the values at these depths indicate relative changes in soil water content. Full details of the calibration procedure and data sets are given by Gaze (1996).

Results Roots Figure 2 shows the mean root length densities recorded from the two transects moving away from a clump of G. senegalensis bushes into a clearing at the fallow site. Differences between the two transects were not significant. The G. senegalensis root length densities recorded were generally very low (-0.05 cm cmy3 ), though occasionally higher densities were measured near the soil surface (e.g. Fig. 2(c)). An attempt was made to distinguish between the red G. senegalensis roots and the pale to brown herb roots, but this was not easy for fine roots. It is therefore possible that some G. senegalensis roots were counted as herb roots, resulting in an underestimate of G. senegalensis root length density. Also, the size of the sample volume with respect to G. senegalensis root distribution was probably too small to obtain a representative sample of G. senegalensis roots. The following conclusions can nevertheless be drawn. Guiera senegalensis roots probably extend beyond the deepest sampling depth; a G. senegalensis root, 7.5 mm in diameter, was found in a sample taken from 170]210 cm depth, 2 m from the bush. It is therefore very likely that G. senegalensis bushes can extract water, if available, from greater than 2 m depth. Very low root length densities for G. senegalensis were found in some samples even at 8 m distance from the nearest bush (Fig. 2(d)). The potential soil volume that G. senegalensis bushes can exploit therefore extends far beyond the projected crown radius of about 1 m. The herb root length densities are high, considering that at the time of sampling (6]7 June 1994) herbs had not developed past the cotyledon stage. The root samples generally looked like older roots, and it is probable that the roots were predominantly herbaceous roots remaining from the previous season. There was no clear trend along the transect with increasing distance from the bush. At all distances, the majority of herb roots were found in the top 50 cm with relatively few roots below this depth, although in almost all cases some roots were recorded at the maximum depth sampled (2 m). Water availability in the top 50 cm will therefore be an important influence on herb growth. At all distances and depths, the root length densities measured for the herbs were greater than for the G. senegalensis.

Changes in water stored in the soil profile with time Figure 3 shows the evolution of the mean depth of water stored in different layers in the soil profile from the end of the wet season in 1993 to the early dry season at the end of 1994, measured using a neutron probe. Total annual rainfall at the site was 545 mm

58

S. R. GAZE ET AL.

Figure 2. Profiles of mean root length density for herbs (I) and Guiera senegalensis bushes (^) on a transect with increasing distance from G. senegalensis bushes in a savanna in south-west Niger. Distance from bushes: (a) 1 m; (b) 2 m; (c) 4 m; and (d) 8 m. Inset tables summarize the total root length per unit area for given depth intervals in the soil profile for Guiera senegalensis (G.s.) and herbs.

in 1993 and 650 mm in 1994. The 1993r94 dry season shows a rapid initial drying of the soil profile in the 0]1 m layer, with progressively slower rates of drying with increasing depth. There is no evidence that the wetting front exceeded 4 m depth in

DRY SEASON WATER USE BY SHRUBS

59

Figure 3. (a) Mean soil water storage for depth increments of 0]1 m (`), 1]2 m (v), 2]3 m (^), 3]4 m (') and 4]5 m (I), and (b) daily rainfall between September 1993 and November 1994 at a Guiera senegalensis savanna in south-west Niger.

1993. It might therefore reasonably be assumed that the drying of the soil profile in the 1993r94 dry season is attributable solely to evaporation from the soil surface and plant water use. The apparently high relative depth of water stored in the 4]5 m layer is probably an artefact of the laterite in the soil at this depth, which was not accounted for in the neutron probe calibration. The progressive wetting of the soil profile through the 1994 wet season can be clearly seen (Fig. 3). The wetting front reached the 4]5 m layer at the beginning of the 1994r95 dry season. By early to mid September 1994 (the start of the dry season), the water content at all layers was wetter than the same period in 1993. The subsequent drying of the soil profile was much more rapid than in the early 1993r94 dry season. This is partly attributable to both drainage and evaporation acting to dry out the soil profile in 1994, whereas drying in 1993 was only due to evaporation (including plant water use). The more rapid drying in 1994 may also be attributable to the maintenance of a green transpiring herbaceous ground cover for longer into the dry season than in 1993, due to several small rains in September and October 1994.

Patterns of dry season water use by G. senegalensis In 1993, the maximum depth of the wetting front exceeded the depth of monitoring only for Tube 5304 (Table 1). For those tubes deep enough to monitor changes in water content below 4 m, there was no evidence of the wetting front exceeding 4 m (see

S. R. GAZE ET AL.

60

Fig. 3). It is therefore assumed that during the 1993]94 dry season, any water lost from the profile was due to evaporation and not drainage. The two bushes at either end of the transect retained some green leaves throughout the dry season. The herbaceous understorey, with the exception of the perennial grass A. gayanus had senesced by 5 October 1993. Andropogon gayanus remained green throughout the dry season, persisting as closely grazed clumps in the vicinity of Tubes 5204, 5304 and 5404 (Fig. 1). It is therefore possible that some of the water loss around these tubes was due to water use by A. gayanus. Spatial variability in measured dry season water loss between tubes (see Table 2) makes it difficult to quantify water use by A. gayanus, but comparing total water loss for Tubes 5204 to 5404 with Tubes 5504 and 5604 (Table 2) suggests that it was at most 15]20 mm over the entire dry season. The much lower water use around Tube 5704 results from the reduced infiltration around this tube which was in the rainshadow of a G. senegalensis bush. Average daily water loss in a transect away from a G. senegalensis bush is shown in Fig. 4. The data presented are from Tubes 5104 to 5404; data are available for Tubes 5404 to 5704 but the trends are not as clear, largely because of lower infiltration around Tube 5704 and its effect on water availability. Water loss was calculated over the top 4 m of the soil profile, or the maximum depth of reading, which ever was less. At the start of the dry season, rates of water loss were much higher at 1]5 m from the bush (1.9]2.4 mm dayy1 ) than at 10 m from the bush (0.9 mm dayy1 ). Water loss from the profile was largely attributable to plant water use. Average evaporation rates from the soil surface during October were estimated to be about 0.2 mm dayy1 using the Pilbeam et al. (1995) model calibrated with microlysimeter data from the site (Gaze, 1996). As expected, the rates of water loss decreased as the dry season progressed. However, the decrease was more marked for tubes near the bushes than for those farther away. Water use at 1 and 2 m from the bush was negligible by the end of the dry season. Over the same period, water loss was about 0.5 mm dayy1 at 5 and 10 m from the bush. Despite the different patterns in water loss with time through the dry season, the overall losses during the dry season from each tube were broadly similar (Table 2). From 28 September 1993 to 11 May 1994, mean water loss from around Tubes 5104 to

Table 2. Proportion of total dry season water loss from around neutron probe access tubes which was from depths greater than 2 m in the soil profile. Water loss was measured between 28 September 1993 and 11 May 1994 in Guiera senegalensis savanna in south-west Niger

Access tube 5104 5204 5304 5404 5504 5604 5704 Mean all 5104 to 5404

Distance from nearest bush (m) 1 2 5 10 5 2 1

Water loss below 2 m depth (mm)

Total water loss (mm)

Proportion of total water loss from below 2 m depth (%)

73 96 116 71 94 84 36

131 168 165 141 159 137 93

56 57 70 51 59 61 39

82 89

142 151

56 59

DRY SEASON WATER USE BY SHRUBS

61

Figure 4. Temporal patterns of water loss during the dry season at different distances from the nearest Guiera senegalensis bush in a savanna in south-west Niger: (I) s 1 m; (^) s 2 m; (`) s 5 m; (v) s 10 m.

5404 was 151 mm, equivalent to 28% of the 1993 annual rainfall. Of this water loss, 59% (89 mm) was from depths greater than 2 m. In the absence of the deep-rooted G. senegalensis, this water would otherwise have contributed to ground-water recharge below the site.

Patterns of water use within profiles Analysis of the temporal sequence of measured soil water profiles showed that there was still some downward movement of water within the profile, at least early in the dry season. Klaij & Vachaud (1992) proposed a method for separating drainage and evaporation measurements based solely on neutron probe data. The technique could not be used in these profiles because the assumption of zero root extraction of water from the layer being considered was invalid because there were probably roots throughout the profile. An alternative simple method to separate evaporation from drainage was therefore used. The soil profile was divided into four layers: 0]1.1, 1.1]2, 2]3 and 3]4 m. If there was a net increase in water in a layer during the time interval considered, the water was assumed to have drained from the layer above. Any layer receiving water was assumed to have zero loss through upward movement, either via roots or the soil matrix. Upward loss from the layer above was calculated after the amount of water measured arriving in the receiving layer below had been subtracted. This method will tend to overestimate evaporation from the surface layers (by underestimating drainage to the layer below) and underestimate plant water uptake from the deeper layers. However, it is better than assuming zero internal redistribution within the profile. The mean soil water extraction (mm my1 dayy1 ), calculated in this way, for Tubes 5104 to 5404 as the dry season progressed is shown in Fig. 5. The results show that in October 1993 considerable soil water extraction occurred in the upper 2 m of the profile from around all tubes. Water extraction also occurred deeper in the profile for the three tubes nearest the G. senegalensis bush. However,

62

S. R. GAZE ET AL.

Figure 5. Temporal and spatial patterns of water extraction during the dry season at a savanna site in south-west Niger. Data are presented for measurements made at a number of distances from the nearest Guiera senegalensis bush. (a) October 1993; (b) Nov]Dec 1993; (c) Jan]Feb 1994; (d) Mar]May 1994. Distance from bush: (I) s 1 m; (^) s 2 m; (`) s 5 m; (v) s 10 m.

there was very little soil water extraction from below 2 m around the tube 10 m from the bush. As the season progressed the situation gradually changed. By Jan]Feb 1994, there was minimal water extraction from the top 2 m, and at greater depths rates of water extraction increased with increasing distance from the G. senegalensis bush. A similar pattern was observed in Mar]May 1994.

Discussion Root depth and water uptake The G. senegalensis bushes at the study site have a sparse but extensive root system enabling the bushes to access soil water several metres from the bush, both horizontally and vertically. This is consistent with the limited root data available in the literature for

DRY SEASON WATER USE BY SHRUBS

63

woody plants in semi-arid areas. Breman & Kessler (1995) cite data from Senegal where 5% of the G. senegalensis root biomass was reported to be at depths greater than 2 m for sandy soils receiving 300 mm rain annually. No data concerning the lateral spread of G. senegalensis roots are given, but Breman & Kessler (1995) do cite data on other semi-arid shrubs and trees. For example, lateral roots of several tree species have been reported to extend more than 40 m from the tree. Seghieri (1995) recorded lateral roots of woody plants extending more than 10 m from the trunk in a semi-arid savanna in north Cameroon. Whilst there were only very low root length densities of G. senegalensis recorded at 2 m depth in this study, roots at 2 m depth and greater were able to extract a considerable cumulative amount of water from the soil profile during the dry season (Table 2). In a recent comprehensive review of the literature, Jackson et al. (1996) reported that 57% of the root biomass of tropical grassland savanna was found in the upper 30 cm of the soil profile. It can be tempting to use such data to justify shallow rooting depths when modelling root water uptake. However, the reported average maximum rooting depth for the same biome was 15 m (Canadell et al., 1996), indicating at least the occasional presence of low, but functionally important root length densities at depth. The data presented for our fallow savanna site show that these sparse deep roots can play an important role in the water relations of tropical savannas and the annual water balance of these regions, as suggested by Jackson et al. (1996). Bate et al. (1982) described a two-dimensional model to simulate the competitive and complementary aspects of soil water uptake by woody and herbaceous plants based on an electrical resistance analogue of water uptake. Their model predicted that drying of the soil profile would start in the surface layers, with root water extraction by herbs and woody plants. The woody plant would initially obtain most of its water from the surface layer closest to its base. As this layer dried, competition between the herbs and woody plant for water would increase in the surface layer further from the shrub. Further drying would result in water extraction by shrub roots occurring progressively further horizontally and vertically from the shrub, with the herb roots confined to the surface layers only and unable to extract water from depth. The predictions from their model are consistent with the measured spatial and temporal pattern of water uptake reported in this study. The root distribution and water uptake data support the hypothesis that woody and herbaceous plants in savannas are to some extent complementary in their water use (Walker et al., 1981; Knoop & Walker, 1985). Root data were not obtained deeper than 2 m so conclusions cannot be drawn concerning the extent of complementarity in the spatial arrangement of rooting systems of the shrub and herb components. Indeed, during the wet season G. senegalensis and herbs will compete for water in the surface layers. However, the herbs senesce early in the dry season, allowing a temporal complementarity in water use with G. senegalensis using water progressively further horizontally and vertically away from the bush as soil water reserves close to the bush are exhausted.

Implications for annual water balance and ground-water recharge Water use of fallow savanna plants depends on water availability, the duration and extent of green leaf cover and the rooting characteristics of the plants. These all change through a season, from season to season, and from site to site. Rainfall in Niger, and in the Sahel generally, is highly variable both spatially and temporally (Lebel et al., 1992; Sivakumar et al., 1993). There can also be considerable redistribution of rainfall water at the soil surface with consequent highly spatial

64

S. R. GAZE ET AL.

variable infiltration over a field (Gaze et al., 1997). The combination of variable rainfall and infiltration results in variable amounts of water stored in the soil profile at the end of the wet season, both from one year to the next, and from one area to another, at small and large scale. This influences subsequent dry season plant water use. For example, Gash et al. (1991), working at a fallow savanna site at the ISC (4 km from our site), measured 135 mm total evaporation in 6 weeks from the end of September 1988. This is close to the cumulative average total evaporation of 142 mm recorded in our study over the whole dry season (Table 2). However, the 1988 wet season at the ISC had a relatively high rainfall with a total of 699 mm (ICRISAT, 1989). In contrast, Allen & Grime (1995) reported on evaporation measurements made from the same fallow savanna site as Gash et al. (1991) during the relatively dry 1990 wet season. They measured 459 mm total evaporation during a wet season of just 454 mm of rainfall. No measurements were made of dry season evaporation or plant water use, it being assumed that there was no soil water left in storage for plant water use. It would appear that as long as there is sufficient water within the G. senegalensis rooting zone, the bushes retain at least some green leaves and continue to extract water from the profile. Over the field of which the study site formed a part there was considerable variation in the time of leaf fall. For example, in 1995 part of the field was cleared and planted to millet by the local farmer. Stems which had regrown from the cut G. senegalensis bushes, and isolated bushes in this area showed no evidence of leaf fall several weeks into the dry season at the end of November 1995. At the same time bushes in other parts of the field which remained uncleared varied from leaf fall completed to some isolated bushes in open areas still retaining many of their leaves. Guiera senegalensis, therefore, appears able to adjust its green leaf area in the dry season according to the soil water status experienced by its roots, retaining green leaves when there is sufficient water and shedding them when the supply is exhausted. This is likely to moderate the effect of variable rainfall on the subsequent deep drainage to groundwater. Green leaves are likely to be retained longer into the dry season following a high rainfall wet season, resulting in higher dry season water use than following a low rainfall wet season. The ability of G. senegalensis to extract available water from the soil profile well into the dry season, provided it has sufficiently deep roots, has potentially important consequences for changes in the annual water balance as fallow savanna land is cleared for millet cultivation. With the increasing population and land use pressure in Niger, the duration that land is left fallow is decreasing (Ada & Rockstrom, ¨ 1993). Also, the area under fallow savanna is decreasing while the land area under permanent cultivation is increasing. In Western Australia, large-scale clearing of shrubs and trees for agricultural production has resulted in a doubling of drainage below the root zone (Nulsen et al., 1986). However, removal of deep rooted perennial plants from savanna does not necessarily result in increased deep drainage. Carlson et al. (1990) reported no change in water balance components in response to shrub removal from Texan rangeland. They attributed this to a compensatory increase in the herbaceous component of the savanna. Converting fallow savanna land in Niger to millet production involves total clearing of the land, followed by subsequent cultivation of a sparse millet crop, sometimes intercropped with cowpea. This results in less plant cover than the fallow savanna, both spatially and temporally, and therefore is likely to result in increased deep drainage. Gaze (1996), summarizing annual water balance studies on farmer-managed millet fields in the Sahel, reported a positive linear relationship ( r 2 s 0.98) between annual rainfall and drainage below the rooting zone. For the long-term annual mean rainfall at Niamey of 545 mm reported by Sivakumar et al. (1993), the predicted drainage beyond 3 m depth under traditionally-managed millet would be 242 mm. This compares with no drainage measured below 4 m at the fallow savanna site in this study in 1993 which also happened to receive 545 mm of rain that year.

DRY SEASON WATER USE BY SHRUBS

65

The fact that, in the environment encountered in this study, there is water available at depth in the soil profile at the end of the wet season for use by deep-rooted plants, suggests that there is scope for complementary water use between millet and trees in an agroforestry system. However, drying of the profile during the long dry season, whether by drainage or evaporation, means that, in the subsequent wet season, available water will initially be confined to the surface 1 to 2 m of the profile. Consequently, in the early wet season there would be competition for soil water between the tree and crop. From a water use perspective, the ideal agroforestry tree component would be one that had minimal water requirement during the early wet season, thus avoiding competition with the crop at this stage, and only started to use water once the wetting front was below crop rooting depth. Faidherbia albida (Del.) A. Chev., a common parkland agroforestry tree in the Sahel (Vandenbeldt, 1991), is ideally suited in this context, shedding its leaves at the beginning of the wet season and producing new green leaves at the start of the dry season. Another option could be to effect incomplete clearing of G. senegalensis bushes when converting fallow savanna for millet production (A. Wezel, pers. comm., 1994). This would minimize competition for water in the early wet season but leave some sparsely spaced bushes to use water deep in the soil profile through the dry season. The G. senegalensis bushes would have some value as a source of firewood (Louppe, 1991). Furthermore, the bushes would have a beneficial effect on the nutrient status of the soils by obtaining nutrients from deep in the soil profile and depositing them on the soil surface in leaf litter (Louppe, 1991). They may also have a role to play in reducing wind erosion, which can be a particular problem early in the wet season before millet crops are properly established (Sterk et al., 1996).

Summary Guiera senegalensis roots were found up to 8 m horizontally and 2 m vertically from the base of the bush. These distances were the maximum for which samples for root analysis were taken. From analysis of soil water content records at the study site it is likely that G. senegalensis roots extend at least 10 m horizontally and 4 m vertically. During the dry season of 1993r94, a progressive drying front was observed moving away from a G. senegalensis bush both horizontally (up 10 m) and vertically (up to 4 m). During the 8-month dry season, mean water loss from the top 4 m of the soil profile was 155 mm. This was attributed to water use by G. senegalensis and comprised 28% of the 1993 total annual rainfall. Of this water loss, 59% (89 mm) was from depths greater than 2 m. Dry season water use by deep rooted woody plants such as G. senegalensis can therefore be an important component of the annual water balance. Clearing fallow savanna land in Niger for millet production will result in increased deep drainage and ground-water recharge through reduced dry season water use by the deep-rooted G. senegalensis shrubs. Some of this deep drainage could be used by a suitable deep-rooted tree or shrub in an agroforestry system. S.R.G. received financial support from University of Reading Research Endowment Fund and the Institute of Hydrology. IH funding for this project was obtained from the European Union and the Natural Environment Research Council. The use of the ICRISAT Sahelian Center facilities, and the assistance of Ide ´ Sanda, Djibo Soumeila, Diafarou Amadou, Zourkafili Amadou, Alfari Sadou, Amadou Bonkano (dit Bohijo) and Sam Boyle (IH) in installing and maintaining the field equipment and collecting data is gratefully acknowledged. We thank ORSTOM, Niger, for the EPSAT-Niger rainfall data and the loan of the drilling rig for installing neutron probe access tubes.

66

S. R. GAZE ET AL.

References Ada, L. & Rockstrom, ¨ J. (1993). Diagnostic sur le systeme agraire du ‘‘Zarmaganda Central’’, Niger. Projet Energie II, BP 12860, Niamey, Niger. (Unpublished.) Allen, S.J. & Grime, V.L. (1995). Measurements of transpiration from savannah shrubs using sap flow gauges. Agricultural and Forest Meteorology, 75: 23]41. Anderson, L.S. & Sinclair, F.L. (1993). Ecological interactions in agroforestry systems. Agroforestry Abstracts, 6: 57]91. Bate, G.C., Furniss, P.R. & Pendle, B.G. (1982). Water relations of South African savannas. In: Huntley, B.J. & Walker, B.H. (Eds), Ecology of Tropical Savannas, pp. 336]368. BerlinHeidelberg, Germany: Springer-Verlag. Bell, J.P. (1987). Neutron Probe Practice. Report No. 19 (3rd Edn). Wallingford, U.K.: Institute of Hydrology. 51 pp. Breman, H. & Kessler, J.-J. (1995). Woody Plants in Agro-ecosystems of Semi-arid Regions. Advanced Series in Agricultural Sciences, Vol 23. London: Springer-Verlag. 340 pp. Canadell, J., Jackson, R.B., Ehleringer, J.R., Mooney, H.A., Sala, O.E. & Schulze, E.-D. (1996). Maximum rooting depth of vegetation types at the global scale. Oecologia, 108: 583]595. Carlson, D.H., Thurow, T.L., Knight, R.W. & Heitschmidt, R.K. (1990). Effect of honey mesquite on the water-balance of Texas rolling plains rangeland. Journal of Range Management, 43: 491]496. Cole, M.M. (1986). The Savannas: biogeography and geobotany. London: Academic Press. 438 pp. FAO-ISRIC (1990). Guidelines for Soil Description (3rd Edn). Rome: FAO. 70 pp. Gash, J.H.C., Wallace, J.S., Lloyd, C.R., Dolman, A.J., Sivakumar, M.V.K. & Renard, C. (1991). Measurements of evaporation from fallow Sahelian savannah at the start of the dry season. Quarterly Journal of the Royal Meteorological Society, 117: 749]760. Gaze, S.R. (1996). Water balance of farmer-managed millet and fallow savannah on sandy soils in south west Niger. PhD thesis, Department of Soil Science, University of Reading. 268 pp. Gaze, S.R., Simmonds, L.P., Brouwer, J. & Bouma, J. (1997). Measurement of surface redistribution of rainfall and modelling its effect on water balance calculations for a millet field on sandy soil in Niger. Journal of Hydrology, 188–189: 267]284. Goutorbe, J.-P., Lebel, T., Tinga, A. et al. (1994). HAPEX-Sahel: a large-scale study of land]atmosphere interactions in the semi-arid tropics. Annales Geophysicae, 12: 53]64. Hesla, B.I., Tieszen, H.L. & Boutton, T.W. (1985). Seasonal water relations of savanna shrubs and grasses in Kenya, East Africa. Journal of Arid Environments, 8: 15]31. ICRISAT (1989). ICRISAT West African Programs Annual Report, 1988. Hyderabad, India: ICRISAT. 125 pp. Jackson, R.B., Canadell, J., Ehleringer, J.R., Mooney, H.A., Sala, O.E. & Schulze, E.-D. (1996). A global analysis of root distributions for terrestrial biomes. Oecologia, 108: 389]411. Klaij, M.C. & Vachaud, G. (1992). Seasonal water balance of a sandy soil in Niger cropped with pearl millet, based on profile moisture measurements. Agricultural Water Management, 21: 313]330. Knoop, W.T. & Walker, B.H. (1985). Interactions of woody and herbaceous vegetation in a Southern African savanna. Journal of Ecology, 73: 235]253. Lebel, T., Sauvagoet, H., Hoepffner, M., Desbois, M., Guillot, B. & Hubert, P. (1992). Rainfall estimation in the Sahel: the EPSAT-NIGER experiment. Hydrological Sciences Journal, 37: 201]215. Lebel, T., Taupin, J.D. & Greard, M. (1995). Rainfall monitoring: The EPSAT-Niger setup and ´ its use for HAPEX-Sahel. In: Lebel, T. (Ed), Hydrologie et Meteorologie de Meso-echelle dans ´´ ´ ´ HAPEX-Sahel: dispositif de mesures au sol et premiers resultats, pp. 31]68. Paris: ORSTOM ´ Editions. 207 pp. Levy, P.E., Moncrieff, J.B., Massheder, J.M., Jarvis, P.G., Scott, S.L. & Brouwer, J. (1997). CO 2 fluxes at leaf and canopy scale in millet, fallow and tigerbush vegetation at the HAPEX-Sahel southern supersite. Journal of Hydrology, 188]189: 612]632. Louppe, D. (1991). Guiera Senegalensis: Espece ` agroforestiere? ` Micro-jachere ` derobee ´ ´ de saison seche et approvisionnement ´ energetique d’un village du centre-nord du bassin arachidier ` ´ senegalais. Revue Bois et Forets ´´ ˆ des Tropiques, 228: 41]47. Nulsen, R.A., Bligh, K.J., Baxter, I.N., Solin, E.J. & Imrie, D.H. (1986). The fate of rainfall in a mallee and heath vegetated catchment in southern Western Australia. Australian Journal of Ecology, 11: 361]371.

DRY SEASON WATER USE BY SHRUBS

67

Pilbeam, C.J., Daamen, C.C. & Simmonds, L.P. (1995). Analysis of water budgets in semi-arid lands from soil water records. Experimental Agriculture, 31: 131]149. Seghieri, J. (1995). The rooting patterns of woody and herbaceous plants in a savanna}are they complementary or in competition? African Journal of Ecology, 33: 358]365. Sivakumar, M.V.K., Maidoukia, A. & Stern, R.D. (1993). Agroclimatology of West Africa: Niger (2nd Edn). Patancheru, India: ICRISAT. 116 pp. Smith, D.M., Jarvis, P.G. & Odongo, J.C.W. (1997). Sources of water used by trees and millet in Sahelian windbreak systems. Journal of Hydrology, 198: 140]153. Soil Survey Staff (1992). Keys to Soil Taxonomy. SMSS Technical Monograph No 19 (5th Edn). Blacksburg, VA: Pocahontas Press, Inc. 556 pp. Sterk, G., Herrmann, L. & Bationo, A. (1996). Wind-blown nutrient transport and soil productivity changes in southwest Niger. Land Degradation and Development, 7: 325]335. Tennant, D. (1975). A test of a modified line intersect method of estimating root length. Journal of Ecology, 63: 995]1002. Vandenbeldt, R.J. (1991). Rooting systems of western and southern African Faidherbia albida (Del.) A. Chev. (syn. Acacia albida Del.) }a comparative analysis with biogeographic implications. Agroforestry Systems, 14: 233]244. Walker, B.H., Ludwig, D., Holling, C.S. & Peterman, R.M. (1981). Stability of semi-arid savanna grazing systems. Journal of Ecology, 69: 473]498. Wallace, J.S., Brouwer, J., Allen, S.J. et al. (1994). HAPEX-Sahel Southern Super-Site Report: an overview of the site and the experimental programme during the intensive observation period in 1992. Wallingford, U.K.: Institute of Hydrology. 55 pp.