The longer-term effects of afforestation with Pinus radiata and Eucalyptus grandis on streamflows were analyzed using data from two paired-catchment experiments in South Africa. The experiments are rare in that they have been maintained over longer periods than the typical rotation period for industrial timber plantations in the tropics or subtropics. In both experiments the planting treatments led to large reductions in streamflow, which increased with the age of the trees and were positively related to water availability. The pine plantation caused peak reductions in yield over a 5-year period of 44 mm a−1 or 7.7% a−1 for each 10% of catchment planted when the trees were between 10 and 20 years old. The eucalypt plantation caused peak reductions over a 3-year period of 48 mm a−1 and 10% a−1 for each 10% of catchment planted. However, as the plantations matured (over 30 years of age in the case of pines and over 15 years of age in the case of eucalypts) the flow reduction trend was reversed, and streamflow effects appear to be tending toward preafforestation levels. The longer-term effects of planted forests need not be as harmful on the water yield of catchments as has been predicted from shorter-term studies. The implication of these results is that if trees are grown on very long rotations, they may be used for restoring degraded catchments or as a means of storing carbon without completely denuding available water resources.
 Timber plantations are a rapidly expanding land use, especially in the subtropics and tropics. By a recent assessment there is an area of 271 million ha of planted forest globally, half of which is plantation and half seminatural, with a growth in forest plantation area of 35.8% between 1990 and 2005 [Food and Agriculture Organization of the United Nations (FAO), 2006]. In the tropics and warmer subtropics new plantings total around 2 million ha per annum [Evans, 1999]. Plantings of eucalypt plantations in particular have increased sharply, with roughly 50% of a total area of 17 million ha of eucalypt plantations having been established during the last decade [Turnbull, 1999].
 Of all their associated environmental issues, the hydrological effects of timber plantations have probably received the most attention. In some places, South Africa for example, the expansion of timber plantations is restricted because of their water consumption. Elsewhere, however, hope is still fostered that the planting of trees can assist in restoring favorable hydrological characteristics to degraded watersheds. In China alone there is a total of 21.8 million ha of planted forest that is designated as protection forest and a further 5.7 million ha of planted “social and cultural forest” [FAO, 2006].
 The United Nations Environmental Program (UNEP) continues to link numerous environmental ills to the loss of forest cover (Agenda 21). The mindset created by such policies and attitudes has caused, in the opinion of Calder [2005, p. 29], “governments, development agencies and UN organizations to commit funds to afforestation or reforestation programs in the mistaken belief that this is the best way to improve water resources.” In reviewing the evidence from scientific studies of forests and reforestation on streamflow, Calder concludes that forest will yield less water than shorter vegetation and that afforestation is likely to reduce dry season flows. Tree planting is also being promoted as a means of carbon sequestration, and a recent paper has pointed out the potentially high costs in terms of water resources of using timber plantations for these purposes [Jackson et al., 2005].
 In this paper we use two paired-catchment experiments from South Africa to investigate the longer-term effects of afforestation with Pinus radiata and Eucalyptus grandis on streamflows. The experiments are rare in that they have been maintained for longer periods than the typical rotation for industrial plantations of this sort and they afford us the opportunity to develop a broader picture of the hydrology of timber plantations. Both experiments have been analyzed previously [van Wyk, 1987; Bosch and Smith, 1989; Scott and Smith, 1997]. The planting of these fast-growing trees that are exotic to South Africa, led to large reductions in streamflow that increased with the age of the trees. Scott and Smith  used these general trends as a basis for empirical models that predict the percentage reductions in streamflow as a function of tree age and genus. For this paper, however, we have been able to analyze the longer-term effects of the treatments, revealing that water use diminishes as plantations age and streamflows are restored to some extent, in a manner similar to that shown in maturing native eucalypt forests in Victoria, Australia by Langford  and Kuczera .
2. Description of the Experimental Catchments
 The two experimental catchments are in very different climatic zones in South Africa. The general location of these research areas, Jonkershoek and Westfalia, is indicated in Figure 1.
 The Jonkershoek Forestry Research Station (latitude 33°57′S; longitude 18°55′E) was established in 1935 near Stellenbosch at the foot of rugged sandstone mountains on the Southwestern coast of South Africa. The characteristics of the two Jonkershoek research catchments are summarized in Table 1.
Table 1. Physical Features and Key Hydrological Characteristics of the Four Research Catchments Used in This Studya
Langrivier and Westfalia-B are the control catchments; amsl is above mean sea level.
These rainfall figures are from a single, lower-elevation gauge and underestimate the actual areal catch of the gauged catchment, which is probably in the order of 1.5 times this size.
Percentage afforestation (%)
Minimum elevation (m amsl)
Maximum elevation (m amsl)
Record length (years)
Mean annual precipitation and standard deviation (mm)
Mean annual runoff and standard deviation: pretreatment (mm)
1076 ± 368
1566 ± 457
649 ± 315
 The climate is humid, mesothermal, Mediterranean type with warm, dry summers and cool wet winters. The mean daily maximum for February (the hottest month) is 27.9°C and the mean daily minimum for July is 5.9°C [Versfeld and Donald, 1991]. The annual rainfall of the general research area averages 1390 mm with 83% of this falling in the 7 months of April to October.
2.1.1. Geology and Soils
 The geology of the research catchments is mixed. The lower elevations comprise Cape Granites of Cambrian age that are very coarse and decomposed to a depth of up to 10 m. The upper half to two thirds of the catchments is formed of quartzitic Table Mountain Sandstones of Ordovician age topped by Lower Paleozoic sandstones with intermittent shale bands that form the cliffs and rocky upper slopes of the catchments [Scott et al., 2000].
 The soils of the catchments are complex and of mixed origin, derived primarily of in situ decomposed granites and mixed colluvial material of mainly sandstone origin. The soils generally have a sandy loam texture, rich red-brown and yellow-brown colors, mostly low in organic matter but with a low bulk density (1 to 1.1 Mg m−3 in the surface horizon) and high infiltration capacity. Soil depths range from roughly one to 2 m, but are underlain by unconsolidated or decomposed material that allows free drainage of water as well as exploration by tree roots [Scott et al., 2000].
 The native vegetation of this area is mountain fynbos, a sclerophyllous shrubland between 2 and 3 m tall, that is dominated by Protea nerifolia, P. repens, Brunia nodiflora and Widdringtonia nodiflora [van Wilgen, 1982]. Evergreen forest with a height of 10 m or more develops along the stream banks and is dominated by Ilex mitis and Cunonia capensis. The vegetation in the upper elevations is shorter and is dominated by ericoid scrub.
2.1.3. Tree Planting Treatment
 The treated catchment, Tierkloof, was afforested to Pinus radiata in 1956 after 17 years of preplanting calibration with the neighboring Langrivier, which remained as the long-term control catchment under native fynbos vegetation. Pinus radiata is a native of Monterey, California, but is now the single most widely planted tree across the globe, 90% of which is in the southern hemisphere where it accounts for large areas of the coniferous plantations in New Zealand, Chile and Australia [Lavery and Mead, 1998]. Only the lower part (92 ha; 36%) of the Tierkloof catchment was planted. The trees were thinned in four stages from an initial planting of 1370 stems per hectare (sph) to a final stand density of 200 sph, and pruned to a final height of 10 m for the production of high quality sawlogs. All silvicultural treatments were standard operating procedure for sawlog production at the time, and are similar to those followed today.
 At 19 years of age the plantation had a mean height of 19.6 m [Le Maitre and Versfeld, 1997] and the mean annual increment of the plantation was around 15 m3 ha−1 a−1. Clear-felling of the Tierkloof plantation commenced in 1999 when the trees were 43 years old, but progress was slow initially and felling was only completed in 2004.
 The second experimental site is Westfalia, a private forest and agricultural estate on the Northeast mountain escarpment of South Africa (latitude 23°44′S; longitude 30°04′E). The experimental catchments were established in the 1970s by the Hans Merensky Foundation, with the assistance of the South African Forestry Research Institute [Smith and Bosch, 1989]. The physical characteristics of the two Westfalia catchments are given in Table 1.
 The climate at Westfalia is subtropical with a summer rainfall season. Monthly mean daily maximum temperature is between 21 and 30°C. Monthly mean daily minimum temperatures vary between 2 and 10°C [Bosch and Smith, 1989]. The mean annual precipitation of the Westfalia catchments is 1368 mm (Table 1). Almost 84% of rain falls during the summer months (October–March). The rainfall is mainly orographic, with the typical shower being a soft drizzle. Convective thunderstorms are common early in the rainy season.
2.2.1. Geology and Soils
 The bedrock underlying these catchments is a biotite-bearing Archean granite gneiss (Nelspruit type) with diabase dykes and sills crisscrossing the area with some intrusions of Turfloop granite (A. Döhne, personal communication, 1984). The soils are deep, well drained and red, with a kaolinitic clay content of between 20 and 60% (Döhne, personal communication, 1984). The soils are stable with low-erodibility characteristics, and a high permeability.
 The native vegetation (that occupies catchment B and also covered catchment D before the treatment) is an evergreen and broad-leafed scrub forest (10 m tall), grading to tall forest (20 m) along the riparian areas. In terms of species the forest has affinities to tropical rain forest. Dominant species are Syzygium cordatum, Nuxia floribunda, Rapanea melanophloeos and Trimeria grandiflora [Bosch and Versfeld, 1984].
2.3. Tree Planting Treatment
 In February 1981, the riparian zone of Westfalia catchment D, occupying roughly 10% of the catchment area, was cleared of its cover of indigenous forest and the cut material left on the ground. This was part of an experiment to measure the influence of riparian vegetation, and only the data prior to March 1981 has been included in the calibration period data set. In December 1982 all the forest and regrowth in catchment D were cut and the material stacked and burned. In March 1983, the cleared area was afforested with Eucalyptus grandis up to the stream edge, [Smith and Bosch, 1989] leaving the catchment 100% afforested. Eucalyptus grandis is a native of New South Wales in Australia, but is the most widely planted eucalypt in South Africa. The largest plantings of this popular species and its hybrids are in industrial plantations in Brazil and other central and South American countries [Commonwealth Agricultural Bureau, 2000]. The trees were planted at a density of 1370 sph and thinned at 3, 5 and 8 years down to 600, 400 and 305 sph, respectively. They were pruned in their ninth year to 10 m. When 9.5 years old the mean height of the eucalypts was 33.6 m and the mean growth rate of the plantation is 29.1 m3 ha−1 a−1 (A. Mostert, Hans Merensky Foundation, personal communication, 2007).
 In 1990, which was the beginning of a dry 5-year period, the stream in Catchment D started to dry up, and from the dry season in 1991 it ceased flowing altogether. In 1995 the riparian zone in Catchment D (about 12% of the catchment area) was clear-felled and, since then, the area has been kept clear of eucalyptus regrowth. After exceptionally high rainfall in the summer of 1995/96, the stream in Catchment D started flowing again (February 1996) and it continues to flow.
3.1. Data Collection
 All four of the study catchments are gauged with sharp-crested 90°V notch weirs surmounted with 1.83 m (6 ft) wide rectangular notches. The height of the rectangular notch varies from catchment to catchment. Stage heights are monitored continuously at these gauging stations using Belfort recording float gauges. Rainfall is also measured using recording gauges. Digitized streamflow and rainfall records were summed by month for the analysis.
3.2. Data Analysis
 The effects on streamflow of replacing native vegetation with timber plantations were determined using the paired-catchment method. The method is based on the assumption that the relationship between the streamflow of two physiographically similar catchments will remain the same provided that the vegetation of these catchments remains the same or changes in a similar fashion. Streamflow in each of the treatment catchments was calibrated against their control catchments over a pretreatment period when the vegetation in each was mature and did not change. The expected streamflows were estimated using regression models developed from the long calibration periods. The deviations from regression during the treatment period were taken as treatment effects.
 Several regression models were tested to best express the relationship between flows in the treatment and control catchments during the calibration periods. Model selection was done on the basis of adjusted R2 (proportion of variance explained) and an inspection of residuals. Additional criteria used to select between apparently suitable models were general simplicity and minimizing prediction error sum of squares.
 For the analysis of streamflow we used monthly streamflow volumes as the computational unit. Compared to using streamflow volumes summed over a shorter period, of a day or a week, for example, monthly volumes greatly reduce the autocorrelation between adjacent residuals from the regression model. Compared to using annual streamflow totals, monthly flows provide for more robust statistics because of the larger number of observations. The general results of the analysis remain the same regardless of which summation period is used.
 The selected calibration models are shown in Table 2 indicating the predictor terms in the model, the values of the regression coefficients, the fit of the models (adjusted R2) and error degrees of freedom. Log linear models were preferred to simple linear models because the residuals from the log linear regressions are randomly distributed whereas the residuals from the simple linear models are positively related to the size of flows.
Table 2. Details of the Pretreatment Calibration Models Against Which the Effects of Afforestation in Tierkloof and Westfalia-D Catchments Were Tested and Measureda
Regression Coefficient b
Error Degrees of Freedom
Here ln represents the natural log, and T and C represent the monthly flow in the treated and control catchments, respectively.
Tierkloof streamflow (monthly totals)
lnT = a + blnC
lnT = a + blnC
 The Jonkershoek control catchment, Langrivier, does not correlate particularly well with the treated catchment (Table 2) apparently because the upper reaches are rugged offering a mix of aspects, cliffs and steep rocky slopes in the lee of high peaks that cause singularly high but largely ungauged precipitation during large frontal rain events. This weaker correlation causes a weaker calibration model which, when combined with the smaller planted area in Tierkloof (36%), limits the ability to detect significant treatment effects in this experiment.
 The significance of treatment effects was assessed by the dummy variable method of multiple regression analysis. The t test for entry of a dummy variable into the regression models can be shown to be the equivalent of an F test for the extra sum of squares because of the entry of an additional term into the model [Kleinbaum and Kupper, 1978]. The application of this multiple regression technique for the analysis of paired-catchment experiments is described fully by Hewlett and Bosch  and Scott and van Wyk .
 In both catchments the effects of timber plantations were highly significant and marked (Figures 2 and 3and Tables 3 and 4). In the case of pines (Tierkloof), significant effects of tree planting were only detected 6 years after planting, and annual flow reductions peaked at 53 mm for each 10% planting of the catchment at 17 years after planting, while the peak reductions over any 5-year period were 44 mm a−1 for each 10% of catchment planted, centered on this same age, or 7.7% a−1 per 10% of catchment planted, centered on the age of 16 years. During the later part of the rotation, as the trees aged from 30 to 45 years, flow reductions became progressively smaller declining to half of what they had been at their peak (Figure 2). In the last 10 years of the rotation, only three individual years had streamflow that was significantly different (p < 0.05) from that predicted by the calibration equation.
Table 3. Summary Table of the Effects of Pine Plantations on Streamflow in the Tierkloof Catchmenta
Period, Hydrological Years
Age Midpoint (years)
Tierkloof, Mean Annual Runoff (mm)
Mean Annual Precipitation (mm)
Flow Reductions (mm/10% Planted)
Flow Reductions (%/10% Planted)
The reductions in streamflow were summed by hydrological year and were averaged over (mostly) 5-year periods after planting and are expressed as absolute (mm) and relative (%) values per 10% of the catchment planted.
Flow reductions were not statistically significant. All other values for flow reduction are significant at p < 0.05.
Table 4. Summary of the Effects of Eucalypt Planting on Streamflow in the Westfalia-D Catchmenta
Period, Hydrological Years
Tree Age Midpoint (years)
Mean Annual Precipitation (mm)
Flow Reductions (mm/10% Planted)
Flow Reductions (%/10% Planted)
The precipitation and reductions in streamflow were summed by hydrological year and were averaged over 3-year periods after planting and are expressed as absolute (mm) and relative (%) values per 10% of the catchment planted. All flow reductions are significant (p < 0.05).
 The eucalypt plantation (Westfalia) had a larger and earlier effect on streamflow (Figure 3). Significant effects of eucalypt planting were measured from the first full year after planting and flows ceased for 4 years when the trees were 9–12 years old. There were two peaks of absolute flow reductions averaged over 3-year periods; 42 and 48 mm a−1 for each 10% planted when the trees were 5–7 and 12–14 years old, respectively. When the trees were between the ages of 9–12 years, in a period of below-average rainfall, the stream dried up giving maximum relative flow reductions of 10% per 10% planted. The riparian zone (12% of the catchment area) was cleared when the plantation was 12 years old, and following a wetter year when the trees were 13 years of age streamflows returned and the calculated flow reductions became smaller as the eucalypts matured from 17 to 21 years (Figure 3). The clearing of the riparian zone and the regrowth of vegetation in that area would have played some part in the observed streamflow responses at Westfalia, contributing to the fluctuating pattern in flow reductions in the later parts of the rotation (Figure 3).
Van Wyk  analyzed streamflow from Tierkloof over the first 25 years of afforestation using a slightly different methodology, namely the adjusted values method described by Brakensiek and Amerman , using both precipitation and control catchment streamflow as independent predictor variables. He concluded that after 25 years it was not clear whether streamflow reductions had peaked, but that annual flow reductions had leveled off at 47.5 mm per 10% planted from 16 years after planting. This result is very similar to that of our analysis. Our results and those of van Wyk, show that the pine plantation in Tierkloof has a large effect on streamflow compared to studies from other parts of the world. The generalization from a review by Bosch and Hewlett  would predict a slightly smaller maximum flow reduction of 40 mm per 10% of catchment planted, while the generalization from the review of Sahin and Hall  would predict an average reduction over 5 years of 33 mm per 10% planted. However, larger changes in streamflow have been recorded for coniferous forests in similar and higher rainfall zones [Bosch and Hewlett, 1982; Sahin and Hall, 1996].
 The Westfalia experiment was analyzed by Bosch and Smith  using the same method as used in this paper, but only including the first three hydrological years after the planting of eucalypts. They estimated a flow reduction of 20 mm (4.8%) per 10% of catchment planted in the third full hydrological year after the planting of eucalypts. They concluded that eucalypts had an earlier influence on streamflow than pine plantations, and recognized that peak reductions had probably not yet been reached. The first 10 years of the Westfalia-D data, plus a second experiment, were used by Scott and Smith  to build a general flow reduction model for eucalypts on productive sites. This model (shown in Figure 3) predicts streams to dry up and remain dry while trees are mature.
 Our results show very high flow reductions (3-year averages of 42 and 48 mm a−1 for each 10% planted) for a young eucalypt plantation during high rainfall years, i.e., when there is good water availability. During drier years absolute reductions were lower but all the available water was used by the plantation, resulting in peak reductions in relative flow reductions, namely a dry stream or 10% reduction per 10% of catchment planted. The peak reductions are only slightly higher than the peaks predicted for eucalypts from Bosch and Hewlett's  review (40 mm per 10% planted) but are considerably higher than the 18 mm per 10% planted predicted from the generalizations generated by Sahin and Hall , though again not as high as peak reductions recorded for elsewhere. The peak reductions recorded in Westfalia-D, though, are remarkably high when one considers that the effects of the eucalypts are being measured against a preafforestation baseline provided by native evergreen scrub forest. According to the general curves for afforestation effects of Zhanget al. , generated from a range of international experiments, little difference in flow would be expected from such a conversion of forest to forest. These contrasts provide clear grounds for considering productive commercial plantations as a separate type from slow-growing native forests.
 This study has shown that in these South African timber plantations streamflow reductions do not reach an asymptote, as predicted by the general flow reduction curves of Scott and Smith [1997, Figures 2 and 3], but diminish over time as the plantations mature. The general trend in streamflow reductions in response to planting resembles the pattern, though with a different time scale, of the measured transpiration cycle of young Eucalyptus grandis stands in Mpumalanga, South Africa [Olbrich, 1994]. The young E. grandis stands had peak transpiration rates at around 3 years of age and the stand-level transpiration declined after this peak in association with reductions in leaf area index and water use per unit of leaf area.
 In these fast-growing plantations it is estimated that greater transpiration is the main cause of the large flow decreases [Scott and Lesch, 1997]. As transpiration rates decline with advancing age of the trees so the total evaporation is reduced. It is thought therefore that interception changes play a lesser role in the diminished impacts of mature plantations. For similar eucalypt plantations, Dye [1996a, 1996b] measured leaf area index (LAI) values of 3 to 4 at tree age of 4 years but declining soon thereafter (LAI of 1.7 by 10 years of age). Reduced LAI would lead to reductions in interception, but the bulk of the observed water use reductions are thought to be attributable to a reduction in transpiration with age.
 The pattern in streamflow over time after afforestation also resembles the response in streamflow to fire in, and regrowth of, mountain ash (Eucalyptus regnans) forest in Victoria, Australia [Langford, 1976; Kuczera, 1987] and overmature fynbos [Scott, 1993]. Wildfire in the mountain ash forests near Melbourne led to brief increases in streamflow (for 2–3 years), followed by large decreases in flow while the young ash regrowth developed and matured. Beyond an age of roughly 30 years, streamflow from these forests increased gradually toward prefire levels [Langford, 1976; Kuczera, 1987]. Further work on these Eucalyptus regnans forests has related the increasing streamflow in maturing forest (over 30 years of age) to lower transpiration rates associated with reductions in leaf area index and a smaller sapwood area [Haydon et al., 1997]. In a similar way, clear-felling of mature indigenous evergreen forest on the Westfalia estate did not lead to substantial or sustained streamflow increases, but rather to a flow reduction once the vegetation had recovered a full canopy and was regrowing vigorously (within a year), even though the biomass was much smaller than that of the forest that had been cleared [Scott and Lesch, 1996]. Andréassian  cites additional examples of declining water use over time from Australia, USA and Europe.
 We speculate that these catchments represent large reservoirs formed of deeply weathered soil and that these soils act as buffers for short-term changes in components of the hydrological cycle. Large increases in evaporative losses, such as occur after the afforestation of shorter vegetation, draw on the reservoir of stored water in the catchment as well as the rainfall in the current year. The effects of afforestation on streamflow may therefore be lagged because of the buffer provided by the reservoir of stored water. Over longer rotations, there is a negative relationship between plantation age and streamflow reductions. This means that during the later stages of a long sawlog rotation some degree of replenishment of soil water stores occurs, to counterbalance net losses from storage in the early part of the rotation. The implication of this finding is that the hydrological effects of long-rotation crops has probably been overestimated in the past, such as by the models of Scott and Smith , while the hydrological effects of short-rotation crops, such as eucalypts grown for pulp and mining timber, may have been underestimated (because of a lag between evaporation and streamflow response). Regardless of overestimates of the effects of long-rotation tree crops, though, the timber plantations would still have used much more water than the shorter or less productive vegetation they replaced.
6. Implications for Watershed Restoration and Carbon Sequestration
 The results of this study show that the hydrological effects of timber plantations is less simple than was previously believed. To the extent that one can extrapolate from industrial plantations to plantings for amenity and other nontimber purposes, it appears that, in the long term, such plantations may have smaller water resource effects than previously estimated. At the very least, the water use of tree crops ought to be evaluated over the full length of the rotation. Provided that trees are not felled at close to their peak growth rates (to maximize volume production, as is typical for commercial pulp crops) but are allowed to mature for an extended period, then the hydrological effects may be reduced. One may also speculate that trees that are not particularly fast growing or dominant, might be suitable for planting for catchment restoration purposes.
 The use of trees for watershed restoration or carbon sequestration may be feasible, therefore, provided the following factors are considered.
 1. Long rotations would be needed in order to benefit from the reduced water use of mature forest cover (implying that the plantations would not be managed primarily for optimal timber yields or even commercial returns);
 2. The plantation cover would need to be managed to maintain a normal age class distribution (roughly equal areas of each age class) in order to spread the burden of vigorous young growth (and consequent high water use) in time and, similarly,
 3. A normal age class distribution would be desirable within each major drainage so as to disperse, on a spatial basis, the impact of the high water use phase of the trees.
 Results from two long-term afforestation experiments in South Africa show that the sharp decreases in streamflow that result from rapid establishment and growth of trees does not continue indefinitely. The Pinus radiata took 6 years to significantly reduce streamflows but reductions were around 42 mm a−1 per 10% of catchment planted or 6%/10% planted between the tenth and twentieth years after planting. Once the pine plantation was 30 years old it was clear that streamflow reductions were becoming smaller and by 40 years the streamflows were no longer statistically significantly different from preafforestation levels. The faster-growing Eucalyptus grandis plantation reached peak water use between 6 and 14 years after planting, reducing water yields over this period by around 40 mm per 10% of catchment planted or all available water, causing the stream to stop running for 4 years. Streamflows returned strongly once the eucalypts were older than 15 years and following the clearing of eucalypts from the riparian zone, and flows were restored to roughly half of expected flows over the period of 15–21 years after planting.
 The fast-growing timber plantations have an early and marked affect on streamflows from the afforested watersheds, that appears to reflect the growth rate of the trees. But the high levels of streamflow reduction are not sustained after the trees reach maturity and streamflows recovered to a substantial degree by the end of long sawlog rotations (20 and 40 years for eucalypts and pines, respectively). The results of this study imply that trees may have a useful role in catchment restoration provided they are managed on long rotations that aim primarily at the restoration of positive watershed characteristics rather than commercial timber production. Similarly, with the right management, trees may be grown in plantations for carbon sequestration without necessarily having an unacceptable effect on water resources. Overall, timber plantations have a large negative effect on streamflows, but the reductions in water resources can be mitigated to some extent by increasing rotation lengths.
 We acknowledge the foresight of the pioneers of forest hydrology in the Department of Water Affairs and Forestry (South Africa) and of the Hans Merensky Foundation for establishment and maintenance of these experiments, the technicians who maintained the experiments and handled data reduction and Godfrey Moses, Alfred Moshole, and Adrian Simmers, in particular; and the Water Research Commission (Pretoria) and the CSIR for funding contributions. We thank three anonymous reviewers of this paper for useful suggestions toward its improvement.