Impacts of invading alien plant species on water flows at stand and catchment scales

The impacts of invasions by alien plant species on the quantity of rainwater that reaches rivers and streams have been studied in several countries. Some studies have found that there is a large impact and others have found little or no impact. These conflicting conclusions can be explained largely by differences in the structure (e.g. height, depth of root systems) and the physiology (e.g. evergreen, deciduous, water stress tolerance) between the alien and the indigenous plant species. The greater the differences between the two, the greater the impact is likely to be.


Introduction
There is a growing body of knowledge on the biophysical and socio-economic impacts of terrestrial invasions by alien (introduced) plant species (Levine et al. 2003;Ehrenfeld 2010;Pyšek and Richardson 2010;Vilà et al. 2011;Pyšek et al. 2012;Rejmá nek and Richardson 2013;Dickie et al. 2014;Funk et al. 2014). However, the hydrological impacts of these invasions at a stand or catchment scale-such as the impacts on surface runoff, groundwater recharge and evaporation losses-have received comparatively little attention. Exceptions include studies on the impacts of riparian invasions in the western USA (Doody et al. 2011;Hultine and Bush 2011), Salix invasions in Australia (Doody and Benyon 2011;Doody et al. 2014) and various species in South Africa. Although this review emphasizes research done in South Africa, we show that the findings are globally relevant because there are general hydrological principles that determine the direction and magnitude of the impacts as discussed by Le Maitre (2004). This paper explores those principles in greater detail and addresses concerns about the quantitative basis and generalizability of research on the hydrological impacts of invasions raised by Hulme et al. (2013).
Research on impacts of invasions in South Africa has had a strong hydrological emphasis largely because of the country's climate, lack of forests and historical background. It is a dry country which receives less than half the global mean annual rainfall [+490 mm/year (hydrologists often express water quantities in millimetres depth, the same as the unit for rainfall; to convert to volumes-1 mm/year is equivalent to 10 m 3 /ha/year)] and has ,0.5 % of its area under forest (Mucina and Rutherford 2006). Therefore, it requires plantations of alien tree species to meet its needs for timber and fibre (Richardson et al. 2003;van Wilgen and Richardson 2014). Systematic planting of alien tree species for timber production began in the late 1800s but, by the 1930s, growing concerns about the hydrological impacts of these plantations led to a government-funded research programme (Kruger and Bennett 2013). This research demonstrated that afforestation results in substantial decreases in river flows in the affected catchments (Bosch and von Gadow 1990;Dye 1996a;Scott and Smith 1997). The prominence of alien tree species as invaders raised concerns that their impacts on water flows could be similar (Wicht 1945;Kruger 1977). When renewed concerns about these impacts were raised in the 1990s (Le Maitre et al. 1996), the lack of data on the invasion impacts resulted in the use of the afforestation research data for estimating the impacts of alien tree invasions in South Africa on water resources, particularly the effect on river flows (Le Maitre et al. 1996Maitre et al. , 2000Calder and Dye 2001;Cullis et al. 2007;van Wilgen et al. 2008). The preliminary results of this research were a key motivation for the establishment of the governmentfunded Working for Water programme (van Wilgen et al. 1998) and have been used in setting priorities for investments in control measures so that resources are effectively deployed (Forsyth et al. 2012). The state of knowledge of the hydrological impacts of invasive alien plants in South Africa was last reviewed by Gö rgens and van Wilgen (2004). There have been several additional studies since 2004 that have advanced the understanding of invasive alien plant water-use (e.g. transpiration and interception losses) and how this affects hydrological processes and river flows.
We first review the current understanding of the key factors that limit the water-use of plants at stand and catchment scales to establish a basis for understanding the impacts of invading species. We then summarize and place the information on hydrological impacts into the context of those limiting factors and draw some generalizations. We finally discuss some fundamental challenges for research. All the catchment-level studies of the impacts of invasive species in South Africa have been based on managed plantations of these species rather than invasions, so information on the impacts of plantations has also been included. The emphasis of this review is on stand and catchment scale impacts because a thorough review of studies at the leaf, plant, stand and catchment scale, and the coupling and decoupling between the different levels is given by Asbjornsen et al. (2011) and for invading versus native species by Cavaleri and Sack (2010). We have also focussed on data that covers at least one year as that time frame is meaningful for informing management decisions such as prioritizing control.

Key Limiting Factors
Hydrologists have been constructing and refining conceptual approaches for understanding and predicting the relationships between rainfall, surface runoff, evaporation and vegetation at a range of scales for more than a century (Budyko 1974;Eagleson 1978;Reich et al. 1997;Rodriguez-Iturbe et al. 1999;André assian 2004;Cavaleri and Sack 2010;Asbjornsen et al. 2011;Moore and Heilman 2011;Moles et al. 2012;Van Bodegom et al. 2012;Porporato and Rodriguez-Iturbe 2013). Calder (2005) proposed a relatively simple and practical way of understanding the key factors controlling evaporation. It involves four sets of factors: three are physical and one involves plant traits. He also proposed that no more than two can be the primary controls in a given situation. The physical factors are: (i) energy availability from solar radiation or, in certain situations, advected energy; (ii) soil moisture availability, especially in strongly seasonal climates; and (iii) precipitation droplet size and its effect on interception. The plant traits are: (i) plant physiology, including whether it is evergreen or deciduous and its moisture stress tolerance; and (ii) plant size above ground (height, stem diameter, leaf area) and depth of the root system. Calder's limits concept therefore links hydrology and plant traits to explain how vegetation plays a critical role in regulating water fluxes in the terrestrial component of the water cycle. The principles proposed by Moore and Heilman (2011) are similar but include more detail on root systems and soil characteristics.

Climatic factors
Evaporation is the second largest component of the hydrological cycle after rainfall, so an understanding of the controls on evaporation is critical. The widely used and emulated model by Budyko (1974) for global and regional estimates of evaporation and, by deduction, surface runoff and other measures of liquid water fluxes, takes an important step in this direction. It is based on two fundamental relationships: These relationships are derived from the basic principle that actual evaporation in dryland situations is limited at the arid end of the climatic range by the supply of water (P) and at the humid end by the atmospheric evaporative demand (energy availability). The reason for the long-term measurements is that, in the short-term, inter-annual variations in rainfall, and thus soil moisture storage, and exploitation of soil moisture by deep-rooted plant species introduced to environments dominated by shallow-rooted species, can result in evaporation exceeding annual rainfall (Dye 1996b;Scott and Lesch 1997;Jarmain and Everson 2002;Farley et al. 2005;Clulow et al. 2011;Mendham et al. 2011). Plants growing in situations where additional water is available, such as on floodplains and over shallow aquifers, can also maintain high transpiration rates (Dye and Jarmain 2004;Engel et al. 2005;Farley et al. 2005;Doody et al. 2011;Moore and Heilman 2011;O'Grady et al. 2011). However, in terrestrial settings without supplementary sources of water, this imbalance cannot be maintained indefinitely, so the mean evaporation from vegetation will not exceed mean rainfall in dryland situations. In most cases, it will be less because of rainwater losses through evaporation from soil and litter, as well as runoff and water percolation beyond the reach of root systems to recharge groundwater. Using data on long-term rainfall and runoff from large catchments, Budyko (1974) found that the evaporative index can be estimated from the long-term climate dryness index using a curvilinear function 1 b ¼ (u tanh u 21 (1 2 cosh u + sinh u)) 0.5 with an error of 10 % [(Donohue et al. 2010), Fig. 1]. This model shows that (in large catchments) the evaporative index generally does not approach either the energy limit or the water limit (Donohue et al. 2010;Zhang and Chiew 2012). These two limits to evaporation apply to dryland situations or large catchments when environments with additional water amount to a small fraction of the total area: the water limit because long-term actual evaporation cannot exceed long-term rainfall; and the energy limit because long-term actual evaporation cannot exceed the energy available to drive it over these time spans.
The energy limit can also be exceeded in some situations where additional energy is available to drive evaporation, for example via advection (Everson 1999;Calder 2005). This is often the case for wetlands, oases or riparian woodlands in landscapes where prevailing winds introduce warm, dry air and increase evaporative demand (Everson 1999;Calder 2005). The relationships also predict that as rainfall increases, evaporation becomes a decreasing proportion of the rainfall as the energy limitation decreases both potential and actual evaporation ( Fig. 1) (Zhang et al. 2001).
There are deviations from the Budyko curve which are primarily due to two kinds of factors: (i) climatic-air temperatures which are directly related to evaporative demand (Thornthwaite 1948;Komatsu et al. 2012) and rainfall seasonality (Potter et al. 2005;Zhang et al. 2008); and (ii) vegetation-structure, eco-physiology and deciduousness (Bosch and Hewlett 1982;Brown et al. 2005;Calder 2005;Donohue et al. 2010Donohue et al. , 2012. Zhang et al. (1999Zhang et al. ( , 2001 analysed the effect of vegetation properties on the relationships between mean annual rainfall and mean annual evaporation for a dataset of more than 300 catchments. They derived general relationships between evaporation and rainfall for catchments either under: (i) grasslands (seasonal pastures or herbaceous vegetation) or (ii) Figure 1. The Budyko curve for the relationship between the longterm dryness index and the evaporative index and the energy and water limits to long-term evaporation.
AoB PLANTS www.aobplants.oxfordjournals.org evergreen woodland or forest (Fig. 2). Other studies have confirmed that incorporation of vegetation features can improve the accuracy of such models (Donohue et al. 2010;Komatsu et al. 2012) in line with Calder's (2005) proposals on limiting factors. Research to date clearly shows that increases in woody plant density or replacement of grasslands by woody plants almost invariably increases the evaporation and decreases water availability by reducing surface runoff and groundwater recharge (Huxman et al. 2005;Wilcox and Thurow 2006).

Water availability
In dryland settings, water availability is limited by the proportion of rainfall captured and stored within the soils and underlying weathered material (regolith). However, there are situations where additional water is available from or via groundwater within the rooting zone such as in alluvial (riparian) or colluvial deposits or in deep soils and weathered profiles. Measurements in riparian invasions (Dye and Jarmain 2004), or after clearing riparian trees in natural settings, plantations or invaded catchments (Dye and Poulter 1993;Prinsloo and Scott 1999;Scott 1999;Everson et al. 2007;Salemi et al. 2012), show that water-use by the same species in the riparian zone is higher than in adjacent dryland situations. However, these studies also show that there is substantial spatial and temporal variability, both along and across the flood plain (Engel et al. 2005;Scott et al. 2006;Hultine and Bush 2011;Salemi et al. 2012). Some of this is due to vegetation characteristics but much of it is due to variations in the accessibility and volume of the additional water caused by, for example, variations in the depth to the water table and heterogeneities in the water storage capacity and transmissivity of the soils and the aquifer material (Scott et al. 2008;O'Grady et al. 2011;Funk 2013). In addition, the native vegetation in these habitats may have a similar structure and wateruse characteristics offsetting the gains from clearing and resulting in little or no net (incremental) change in wateruse (Scott 1999;Doody et al. 2011). Pertinent examples are the high evaporation rates reported for native riparian vegetation and riparian invasions by Acacia mearnsii (Dye and Jarmain 2004) or Salix babylonica invasions and native riparian eucalypt forest (Doody et al. 2011), or where deep soil moisture or groundwater is being exploited (Jobbá gy Engel et al. 2005;Farley et al. 2005;Benyon et al. 2006;Cleverly et al. 2006;Fritzsche et al. 2006;Kagawa et al. 2009;Clulow et al. 2011).

Plant traits
Vegetation structure and deciduousness affect interception and transpiration. Plant size was used by Le Maitre et al. (1996) as a key determinant of invasive plant wateruse. This was logical given the wide range of growth forms of invaders and the lack of data on the impacts on streamflow except for commercial plantation species (pines, eucalypts) and how they compared with native grass, woodlands and fynbos shrublands. However, the relationship between total biomass and the biomass of the leaves (and thus the transpiring and intercepting leaf area) varies between growth forms and over the lifespan of a plant, especially in trees. Although these changes follow allometric rules that also relate to plant water-use West et al. 1999;Niklas et al. 2003;Poorter et al. 2012;Zeppel 2013), they indicate that biomass per se is not a reliable indicator of transpiration (or interception).
At the plant level, a number of plant traits play key roles in regulating and limiting evaporation (i.e. both transpiration and interception losses) (Lavorel et al. 1997;Lavorel and Garnier 2002;Calder 2005;Moore and Heilman 2011;Van Bodegom et al. 2012). Key traits relating to water-use can be divided into two related groups: (i) the plant size (e.g. height, leaf area, root system depth); and (ii) physiology [e.g. hydraulic architecture (cavitation resistance, water flux rates, stomatal control), evergreeness]. One of the key relationships in woody plant species is the area of the sapwood and the leaves, which are related through what is known as the Huber value-the ratio of sapwood to leaf area (Carter and White 2009). The Huber value quantifies the ability of the plant xylem to conduct water and maintain transpiration. High ratios Figure 2. Generalized relationships between mean annual rainfall and mean annual evaporation (as a percentage) for catchments with two differing dominant vegetation types: grasslands, mainly seasonal pastures or trees (woodlands or forest). The evaporation data represent dryland settings as riparian zones comprise a small proportion of most catchments (1 -5 % and up to 10 % in some cases) or had vegetation which did not change when the rest of the catchment was converted from tree to grass cover. Derived from data in Zhang et al. (1999Zhang et al. ( , 2001 tend to occur in species which are conservative water users and tolerant of moisture stress, and low ratios in species which are not moisture stress tolerant (Tyree and Sperry 1988;McDowell et al. 2002). Unfortunately, data on sapwood areas and Huber values are available only for woody species with typical secondary growth, and so exclude many important invading species. A more widely available measure is the leaf-area index (LAI, m 2 of leaf/m 2 of ground within the canopy or at stand scale) which is a particularly important ecophysiological parameter because it is related to both the ability of the plant to absorb energy and to transpire water or intercept rainwater (Larcher 1975). So, it is directly linked to plant water-use and growth rate or productivity. Combined with leaf longevity or evergreeness and stomatal conductance, it provides a direct link to plant water-use and is, therefore, a potentially very useful factor for explaining why invasions by, or plantations of, certain species have more significant impacts on water resources than others (Le Maitre 2004;Everson et al. 2007Everson et al. , 2011Clulow et al. 2011;Fink and Wilson 2011).
Plants are also known to vary in their transpiration per unit leaf area (WU/LA). Thus there are species with higher WU/LA but a lower LAI than other species, and even variations within a species (Hatton et al. 1998;Carter and White 2009). These complications can be circumvented by the use of micro-meteorological and remote-sensing techniques for estimating evaporation at the canopy and stand scale via the energy balance. But they still require data about the canopy structure and the canopy or surface conductance at the sampling sites. Satellitebased remote sensing, as used by Meijninger and Jarmain (2014) and Jarmain and Meijninger (2012), can provide a basis for scaling-up micro-meteorological and stand-level measurements to landscape and catchment scales, but more work is needed before this approach can be applied across the rugged landscapes typical of high water yielding mountain catchments and for mixtures of species.
Nevertheless, the strong relationships involving LAI explain why it and other leaf-related indices of vegetation vigour or productivity (e.g. specific leaf area) are widely used in remote-sensing-based assessments of vegetation productivity and evaporation (Bastiaanssen et al. 1998;Gower et al. 1999;Asner et al. 2003;Cleugh et al. 2007;Glenn et al. 2011;Mu et al. 2011;Velpuri et al. 2013) and more widely in modelling vegetation dynamics (Running and Coughlan 1988;Landsberg and Waring 1997;Wright et al. 2001;Woodward and Lomas 2004;Dovey 2005). Other leaf traits also associated with competitiveness may account for greater water-use by invading species but the distinctions are not always clear cut (Grotkopp et al. 2002;Grotkopp and Rejmá nek 2007;Tecco et al. 2010;Drenovsky et al. 2012). Other comparisons among invasive species have found that they have, among others, higher leaf nutrient levels and specific leaf area and lower wood densities (Diaz et al. 2004;Leishman et al. 2007;Cavaleri and Sack 2010;Ordonez et al. 2010;Funk 2013;Tecco et al. 2013).
Major invading plant species whose stand and landscape-level water-use have been documented include a wide range of growth forms, height, evergreeness and root depth ( Table 1). Most of them are evergreen trees or shrubs, with the trees roughly divisible into the two groups used by Le Maitre et al. (1996), namely tall trees and medium trees. Deep root systems are also common among the trees with some taxa, such as Eucalyptus and Prosopis being well known for having root systems that reach depths of 10 -20 m or more Schenk and Jackson 2002a, b;Stromberg 2013). The effective depths of root systems are often misunderstood because the majority of the root mass is concentrated in the upper 0.5 m of the soil . However, deep-rooted species often are characterized by having a few roots, sometimes called sinker roots, which can reach great depths and are specialized for water transport (Pate et al. 1995;Dawson and Pate 1996;McElrone et al. 2004;Stromberg 2013) and may utilize hydraulic lift and redistribution (Oliveira et al. 2005). Deep root systems have been reported for pines in deep sands in Zululand, South Africa (Haigh 1966), and also for Eucalyptus in settings where the soils are considered shallow but there is deep weathering (Dye 1996b). It is likely that invading taxa with this trait will also be exploiting deep soil moisture or groundwater provided it is accessible.
Leaf-area index data were available for some taxa and they show quite a range of LAI values both between and within some taxa (Table 1). Some of the values are affected by the form of the leaves and how the LAI was calculated. Needle-leaved species generally have high LAIs while values for reed-like plants such as Arundo donax with photosynthetic stems or leaf sheaths depend on the interpretation of the leaf form and area. Many species can achieve an LAI of more than two, implying that they have the potential for high transpiration rates and potentially high interception losses. Shrub or scrambler species like Chromolaena have high LAI values (Table 1) and relatively high annual evaporation rates (Table 2), which suggests that their impacts will be closer to those of trees than their growth form indicates.

Dryland invasions
There is a large body of information globally on the effects of different vegetation types and changes in vegetation  Table 1. A summary of information on key traits which are known to affect transpiration and interception rates of invading plant taxa including plantations of these species. Information on typical plant height was taken from Henderson (2001) or based on personal observations ( # ). Root depths from Canadell et al. (1996) and Schenk and Jackson (2002a, b) Table 2. A summary of data on evaporation from stands dominated by invading plant species, including selected information from commercial forest plantations of those species. Mean annual rainfall has been adjusted to the mean for the catchment rather than a particular rain gauge. Remote sensing-based estimates of annual evaporation from stands with a canopy cover of at least 35 % in KwaZulu-Natal and the Western Cape in South Africa were taken from Jarmain and Meijninger (2012) and Meijninger and Jarmain (2014 (2009) types on runoff and other components of the hydrological cycle. Most of it is on the effects of changes in the natural vegetation, or from natural vegetation to cultivated land, but there is information on differences between, or changes from, non-woody to woody vegetation or changes in the structure of woody vegetation (e.g. Bosch and Hewlett 1982; Zhang et al. 2001). Most of the studies of hydrological impacts have focused on dryland invasions and most of these were on the short-term effects on soil moisture balance or evaporation and not on runoff (Levine et al. 2003;Cavaleri and Sack 2010).
Estimates of evaporation are also available for invasions or plantations of some of the taxa whose traits were summarized in the previous section (  (Table 2). Data for plantations of the major species were also given by Meijninger and Jarmain (2009) and Jarmain and Meijninger (2012). They were lower than those reported for catchment studies: pines in the Western Cape 735 + 215 mm/year, eucalypts and wattles in KwaZulu-Natal were 690 + 190 and 615 + 140 mm/year, respectively, but still higher than the natural vegetation they had replaced. Early studies of interception in eucalyptus plantations in Mpumalanga and pines in Jonkershoek (Western Cape) found that they were typically low for plantation tree species because of the high intensity of rainfall and temporal pattern of events (,10 % of total evaporation) (Dye 1996a). However, more recent studies by Everson et al. (2007) and Bulcock and Jewitt (2012) in the mist-belt region of KwaZulu-Natal, which is characterized by lowintensity rainfall events, recorded high interception losses in A. mearnsii (+30 %), E. grandis (+15 %) and P. patula (+21.4 %). These differences are consistent with a limiting factor of raindrop size (Calder 2005) in addition to rainfall amount and intensity per event. More research is needed to determine how representative the high and low values are of other areas and under different rainfall intensity regimes and for a range of stand-ages, densities, leaf-area indexes and site conditions. Studies of plantation species suggest that water-use efficiency is also an important factor, with the key difference between native and introduced tree species being the slow growth rates of the native tree species rather than differences in water-use (transpiration rates) (Wise et al. 2011).
A couple of studies were of herbaceous weeds and found substantial changes in water-use relative to natural vegetation which are likely to result in changes in water flows. A decrease of 56 % in runoff after simulated rainfall was observed in an area of perennial grassland invaded by the thistle Centaurea maculosa (Lacey et al. 1989). Centaurea solstitialis invasions in annual grasslands resulted in a reduction in soil moisture equivalent to 1050 -1200 m 3 /ha/year in one study (Gerlach 2004), while ET increased by 40 mm/year (23 %) compared with native perennial grasslands in another (Enloe et al. 2004). Centaurea maculosa is short-lived (,10 years) but is deep-rooted and forms multi-aged stands, whereas C. solstitialis is an annual but has deep roots and continues growing after the grasses have senesced; both maintain a high canopy cover (additional information from USDA NRCS 2013).
In summary, the differences in evaporation, and thus in water discharges, between native vegetation and matched invasions (or tree plantations) show that invasions typically have a higher water-use than native vegetation. The differences are consistent with expectations given the changes in vegetation structure (e.g. height, root depths, LAI) and physiology (e.g. deciduousness) (Calder 2005;Moore and Heilman 2011;Funk 2013). Thus, the hydrological impacts of invasive alien plant species are not special or exceptional, although differences in their physical and physiological traits may allow them to maintain greater water-use than the native species they replace. Information about these traits can be used to provide more robust estimates of the water-use of species whose water-use has not yet been measured.

Riparian invasions
In riparian or floodplain settings (or areas with aquifers accessible by plant roots), theory predicts and data show that evaporation will be greater than the adjacent dryland areas because water availability is no longer the primary limiting factor (Scott 1999;Dye and Jarmain 2004;Calder 2005;Hultine and Bush 2011;Moore and Heilman 2011;Salemi et al. 2012). Evaporation from riparian invasions by A. mearnsii exceeded that for the native vegetation in the Western Cape and in the KwaZulu-Natal midlands by 171 and 424 mm/year, respectively (Table 3) (Dye and Jarmain 2004). The greater annual evaporation in the Western Cape was attributed largely to high daily transpiration rates during the dry, hot summer. The greater difference between invaded and natural sites in KwaZulu-Natal was primarily due to seasonal (winter) dormancy in the riparian grassland. In both cases, the evaporation in the adjacent dryland communities was lower than that for the riparian communities. Pinus species growing in a riparian zone were found to use 200 mm/year more water than pines in the adjacent dryland fynbos (Table 3) (Dzikiti et al. 2013a). Short-term increases in low flows of 9 -31 m 3 /ha/day have been reported after clearing riparian invasions (Dye and Poulter 1993;Prinsloo and Scott 1999;Rowntree and Beyers 1999;Everson et al. 2001). The relative gains in streamflow from riparian versus dryland clearing of plantation trees range from 3.35 times at Biesievlei to 2.39 times at Two Streams (Table 3). These are substantial gains but they are also short-term and will decrease in the long-term as the native vegetation re-establishes itself (Scott 1999), as found in the Two Streams study (Everson et al. 2007;Clulow et al. 2011). The extent of the decrease will depend on the ability of the native species to access the same sources of additional water. Where the native species root systems are as deep as those of the invaders, there will be little or no gain in the long-term, but where they are much shallower the long-term gains could be large.
When the Two Streams study began, the plantation was already established and there was almost no streamflow although it was the middle of the rainy season (Everson et al. 2007). The runoff to rainfall ratio was only 2.18 % from 1 January 2000 to 30 April 2004 but increased to 7.2 % in the period after clearfelling (December 2003-November 2008 despite re-afforestation in 2006. Little lateral water flow reached the riparian zone while the dryland trees were present although the deeper soils remained moist (Everson et al. 2007;Clulow et al. 2011). The sap flow data (larger diameter trees have greater sap flow rates), and greater diameters of riparian zone trees (1.53 times the dryland trees), provide strong evidence that transpiration was greater in riparian than non-riparian trees.
An important finding was made by the Two Streams study, namely that the evaporation from the A. mearnsii stand (largely dryland) exceeded the annual rainfall by a substantial margin (Table 1) (Clulow et al. 2011). This supported the findings of previous studies (Scott and Lesch 1997;Jarmain and Everson 2002). The ability to maintain such high transpiration rates seems to be mainly due to the trees developing deep root systems (.4.8 m deep) which exploited the moisture stored in the sub-soil and regolith (Clulow et al. 2011). These findings support the conclusions of other studies which have suggested that plantation trees are able to deplete soil and regolith water stores (Dye 1996b;Clulow et al. 2011). The observations also explain how afforestation of some catchments has dried up the streams completely, in some cases resulting in lags of a few years between clearing and streamflow recovery to pre-afforestation conditions (Scott and Lesch 1997;Scott et al. 2000a). Unsustainable soil moisture exploitation could also be happening at other sites where there are dryland invasions by deeprooted species on deep soils and weathered material. Where deep-rooted trees have been present for some years, it may require some time to replenish the soil moisture storage and restore the normal water balance, in some cases more than a year (Scott and Lesch 1997;Scott 1999;Everson et al. 2007).
Two assessments of the water-use of riparian tree invasions in arid environments have been done in South Africa. The one estimated a groundwater loss of 50.4 m 3 / month during the growing season (October-February), or 251.9 m 3 /ha/year (25 mm/year) by Prosopis species hybrids (Fourie et al. 2007). However, the canopy cover of the Prosopis stand in that study was only 21 % so the equivalent for a closed plant canopy was roughly 120 mm/year. The other recorded peak transpiration by Prosopis species stands of 80 m 3 /ha/month and a total annual use of 345 m 3 /ha/year (35 mm) (Dzikiti et al. 2013b). The canopy cover was 31 % so the equivalent for a closed plant canopy would be 111 mm/year. These estimates assume that all this groundwater could be saved by clearing Prosopis but the long-term saving would depend on the water-use of the native tree species that replace the Prosopis. The Rugseer River, where these studies were done, is ephemeral with extended dry periods (it is only estimated to flow for 36 % of the time) and its catchment gets 150 -250 mm/year of rainfall so groundwater availability is very limited. Plant moisture stress measurements showed high pre-dawn stress levels (values lower than 23.0 MPa, Dzikiti et al. 2013b), which explains the low transpiration rates.  Table 3. Observed and modelled evaporation and impacts on streamflow for native and invaded riparian settings, including afforested riparian zones in plantations. MAP, mean annual precipitation; Et, evaporation. a Calculated using the results of the break point modelling in the report. These estimates are low compared with those for native Prosopis woodlands in the southern USA of 350 -1100 mm/year, with most studies giving estimates between 350 and 750 mm/year (Le Maitre 1999;Scott et al. 2000bScott et al. , 2006Scott et al. , 2008Nagler et al. 2005a, b) but these were all on floodplains of perennial rivers. The evaporation from the native woody vegetation found along South African river systems in arid environments has only been briefly assessed. Everson et al. (2009) reported daily evaporation on a few summer days from riparian vegetation in the non-perennial Seekoei River as varying from 2.8 to 3.3 mm. This is potentially lower than for Prosopis invasions given that the native species seldom approach the density and canopy cover of Prosopis invasions (Le Maitre 1999;Wise et al. 2012). Prosopis trees also tend to produce new leaves earlier in the spring than native riparian species (Van den Berg 2010). Wise et al. (2012) used an estimated difference of +600 m 3 / ha/year (60 mm/year) for dense floodplain (riparian) invasions based on the available literature. Some studies have found large reductions similar to those reports for South Africa. The indications are that Prosopis species are likely to use more groundwater than the equivalent native species in South Africa and in similar settings in other countries where they are invasive (e.g. India, Australia, Ethiopia, Kenya).
Although there are studies of the water-use of poplars grown for biomass, the growing conditions and silvicultural treatments (typically intensive management) are so different from those in invasions and native forests that their findings are not applicable. Two studies have examined poplar species (Populus deltoides) water-use in South Africa (Dye et al. 2008). Poplar plantations are typically located on alluvial soils but are not usually planted up to the actual river banks. The modelled annual transpiration was 818 mm with peak values of 6 -8 mm/day during the period from October to December. The trees were very susceptible to a fungal disease which meant that transpiration began declining in early January rather than in April when leaf-shedding should begin. The annual estimate was adjusted to compensate for this to some extent but it is probably still conservative. However, studies of riparian poplar forests in the western USA provide similar estimates of evaporation (Table 3).
An assessment of the impacts of willow (S. babylonica) invasions on river systems in Australia found that evaporation differed substantially between trees growing in the active (flowing) river channel and trees on the banks (Doody and Benyon 2011;Doody et al. 2011). In active river channels, the total annual evaporation ranged from 1755 to 2410 mm/year (transpiration plus shaded water evaporation) compared with 563 mm/year for river bank invasions and open water evaporation of 1396 -1604 mm/year). The mean annual rainfall in the study area was 404 mm suggesting there were markedly different moisture regimes between banks and active channels because there would be little or no lateral groundwater inflow to the floodplain. Although S. babylonica is deciduous, its annual evaporation rates may exceed those reported for evergreen species (e.g. Table 1) provided there is sufficient water available.
Early reports on Tamarix invasions in the USA estimated a daily water-use reaching 200 m 3 /ha/day (20 mm/day , resulting in an estimated total flow reduction of 1.4 -3.0 billion m 3 /year for the larger rivers of the western USA (Zavaleta 2000). But more recent estimates suggest that tamarisk water-use is about the same volume of water as, or even less than, the native riparian forest species they replace (Table 1) (Scott et al. 2008;Nagler et al. 2009;Doody et al. 2011;Hultine and Bush 2011;Moore and Owens 2012).
The same may be true of A. donax invasions in California where there are native reeds and other species (Watts and Moore 2011). However, Arundo invasions in South Africa are often more extensive than those of Phragmites, and also occur in situations where Phragmites and Typha are absent (D. C. Le Maitre, pers. obs.), which could increase its impacts compared with native riparian species. Arundo also tends to remain evergreen while Phragmites and Typha die back in the winter which may affect the relative water-use.
In summary, studies of invasive species in riparian settings have confirmed that their annual transpiration or evaporative water-use can exceed that of the native riparian vegetation, provided that there are changes in vegetation structure, phenology or other traits. This means that the impacts of riparian invasions can be much greater than that is indicated by their extent or the proportion of the landscape that can be categorized as riparian. The differences seem to be less in environments where the native riparian vegetation is evergreen (e.g. fynbos) than where the native vegetation is deciduous (e.g. grasslands, savanna) and the invaders are evergreen (Dye and Jarmain 2004).

Challenges for Research
The variety of species The 28 invasive taxa in South Africa mapped by Kotzé et al. (2010) cover a wide range of growth forms with varying physiology, phenology, rooting depths, LAIs, specific leaf areas and other key traits that affect their potential water-use (Table 1). Even so, they are a subset of the major and emerging invading species in South Africa and elsewhere in the world. However, this review has found that the impacts of the different species do vary AoB PLANTS www.aobplants.oxfordjournals.org in ways that are consistent with their key traits (Fig. 3, Calder 1991, 2005Le Maitre 2004) so this information can be used for extrapolating the results for known species to unknown species in the interim and to prioritize measurements on other species.

Stand density and age
The hydrological impacts vary depending on the size (age) (Fig. 3) and density of the invasions with the effects of increasing size and density (canopy cover) being seen in the typical sigmoidal change in streamflow associated Figure 3. Effects of combinations of major plant traits that have been found to influence the impacts of plant invasions on water resources relative to natural vegetation (after Calder 1991(after Calder , 2005Le Maitre 2004). Plant traits are inter-related but can be grouped into those related to size and those related to physiology. High impacts on water resources will occur where there are marked contrasts in these traits (e.g. evergreen versus deciduous, deep versus shallow roots), and the more contrasting these are, the greater the difference is likely to be. In some cases contrasts may compensate for each other (e.g. evergreen trees with low stomatal conductances versus deciduous trees with high conductances) (Doody et al. 2011). In South Africa the most marked contrasts are where short, fairly shallow-rooted, winter-deciduous grasslands are replaced by tall, deep-rooted, evergreen trees . In contrast, invasions by tamarisks in North America have had little impact because they are similar to poplars in their growth form, rooting depth and leaf seasonality (Doody et al. 2011;Hultine and Bush 2011). with increasing tree or stand age (Bosch and von Gadow 1990;Dye 1996a;Le Maitre and Versfeld 1997;Scott et al. 2000a;Zhao et al. 2012). Initially a plantation comprises small saplings, with a low (,1 %) canopy cover, and streamflow reductions are not detectable but, as the trees grow, the reductions become evident with the maximum reductions being reached and sustained following canopy closure, which is well before biomass peaks (Le Maitre and Versfeld 1997). The same structural changes occur as invasions progress (Moody and Mack 1988), so these relationships need to be investigated more fully to allow the effects of size, density and canopy cover to be explicitly included in estimates of the impacts. The same changes occur when native woody plants encroach (i.e. density and canopy cover increases) and insights can be gained from research into the hydrological impacts and controlling factors (e.g. Huxman et al. 2005;Wilcox et al. 2008).

Water availability
Invasions occur in two different situations: (i) upland or dryland areas where the available moisture is limited to the rainfall which is retained in the rooting profile and (ii) riparian zones, floodplains or other areas where groundwater is available within the rooting depth. In this second situation, the potential and actual water-use generally is not primarily limited by water availability but by the climatic conditions and the growth form, root depth, phenology and physiology of the plants. More work is needed on riparian invasions both to quantify the impacts of invaders that have not been studied yet and to obtain data on the water-use of the native communities they replace (Salemi et al. 2012), or that replace them following clearing.

Range of climates and invaded vegetation types
Invasions occur across a very wide range of natural vegetation types and climatic conditions so robust guiding principles or rules need to be established for scaling-up from existing measurements to areas where there are no data, especially in semi-arid and arid areas. The ideas discussed by Porporato et al. (2004) on seasonality and temporal patterns in rainfall events could be useful in this context. The same considerations apply to situations where more water is available than the soil moisture derived from local rainfall infiltration and percolation. For example, an ephemeral river will have less groundwater available for invaders to use in the long-term than a perennial river system. Remote sensing-based estimates of evaporation provide a tool for addressing many of these issues and ensuring that data from sitespecific studies can be scaled up to landscapes and catchments.

Conclusions
Invasive species do not differ fundamentally from native plant species in their growth forms or physiology. Nevertheless, there are a number of factors that contribute to their greater water-use compared with similar native species, including: † Plant traits, notably their size, root depths, leaf area or leaf-area index, specific leaf area and transpiration rates. The greatest impacts are found where the invaders are evergreen trees, and the dominant native species are seasonally dormant grasses, but there may be little or no impact where they have similar growth forms and canopy structure (e.g. invasive, deciduous Tamarix versus native Populus or Salix on rivers in the USA). Interestingly, invasive, deciduous Salix in Australia can match and, in some settings exceed, the annual water-use of evergreen, native Eucalyptus species which are known to have high water-use rates. † The ability of some invaders to form dense stands compared with co-occurring native species also contributes to disproportionately high stand level water-use by the invasions although the water-use rates by individual species of similar transpiring leaf area maybe similar (S. Dzikiti et al., unpubl.). † The role of the deep root systems of many species which allow them to access soil moisture and groundwater in deep soils and weathered material and in floodplain alluvium where there is additional soil moisture and groundwater (Moore and Heilman 2011). † This review did not explore the implications of the efficiency with which many of the key invading species produce woody tissues (see Cavaleri and Sack 2010;Funk 2013) and become tall compared with similar native woody species, notably the major genera used in plantations (Pinus, Eucalyptus, Acacia) (Gush and Dye 2009;Wise et al. 2011). These are aspects that require further study.
As expected, invaders in dryland settings have a lower water-use than those in floodplains and unconsolidated aquifers and, thus, have less impact on surface runoff per unit area or groundwater. A number of issues require further investigation and addressing them should be a key priority for research on the impacts on invading alien plant species on river flows and groundwater resources.

Sources of Funding
This work was funded by the Working for Water programme of the Natural Resource Management programmes, Department of Environment Affairs, South Africa.