Plant and Soil

, Volume 324, Issue 1, pp 1–30

Desirable plant root traits for protecting natural and engineered slopes against landslides


  • Claire Atger
    • Pousse Conseil, Domaine de Fitzgerald, Le Mas RougeChemin du Mas Rouge
  • Anthony Glyn Bengough
    • Scottish Crop Research Institute
  • Thierry Fourcaud
  • Roy C. Sidle
    • Appalachian State University, Department of Geology
Marschner Review

DOI: 10.1007/s11104-009-0159-y

Cite this article as:
Stokes, A., Atger, C., Bengough, A.G. et al. Plant Soil (2009) 324: 1. doi:10.1007/s11104-009-0159-y


Slope stability models traditionally use simple indicators of root system structure and strength when vegetation is included as a factor. However, additional root system traits should be considered when managing vegetated slopes to avoid shallow substrate mass movement. Traits including root distribution, length, orientation and diameter are recognized as influencing soil fixation, but do not consider the spatial and temporal dimensions of roots within a system. Thick roots act like soil nails on slopes and the spatial position of these thick roots determines the arrangement of the associated thin roots. Thin roots act in tension during failure on slopes and if they traverse the potential shear zone, provide a major contribution in protecting against landslides. We discuss how root traits change depending on ontogeny and climate, how traits are affected by the local soil environment and the types of plastic responses expressed by the plant. How a landslide engineer can use this information when considering slope stability and management strategies is discussed, along with perspectives for future research. This review encompasses many ideas, data and concepts presented at the Second International Conference ‘Ground Bio- and Eco-engineering: The Use of Vegetation to Improve Slope Stability—ICGBE2’ held at Beijing, China, 14–18 July 2008. Several papers from this conference are published in this edition of Plant and Soil.


Soil cohesionRoot architectureSlope stabilitySoil mass wastingSuction

What is a trait?

A trait is defined as a distinct, quantitative property of organisms, usually measured at the individual level and used comparatively across species. A functional trait is one that strongly influences organismal performance (McGill et al. 2006). Plant quantitative traits are extremely important for understanding the local ecology of any site. Plant height, architecture, root depth, wood density, leaf size and leaf nitrogen concentration control ecosystem processes and define habitat for other taxa (Westoby and Wright 2006). An engineer conjecturing as to how plant traits may directly influence physical processes occurring on sloping land needs to consider how, for example, canopy architecture and litter properties influence the partitioning of rainfall among interception loss, infiltration and runoff. Plant traits not only influence abiotic processes occurring at a site, but also the habitat for animals and invertebrates. Depending on the goal of a landslide engineer or restoration ecologist, the immediate and long-term effects of plant traits in an environment must be considered if a site is to remain viable and ecologically successful. To stabilise a slope against shallow landslides, vegetation can be used, as plant root systems fix soil against slippage (Schiechtl 1980; Coppin and Richards 1990; Norris et al. 2008). Although root systems are designed to anchor the plant to the soil and provide access to nutrients and water, an understanding of the interactions between root traits and soil physical processes would enable the landslide engineer to better manage an unstable slope.

What types of mass wasting processes can be stabilized by plant roots?

Woody vegetation, particularly trees, can help prevent shallow landslides in two ways: (1) modifying the soil moisture regime via evapotranspiration; and (2) providing root reinforcement within the soil mantle (Fig. 1). The first factor is generally not very important for shallow landslides and debris flows that occur during an extended rainy season, except possibly in the tropics and sub-tropics where evapotranspiration is high throughout the year; deeper-seated landslides, such as earthflows, may experience slightly prolonged activity due to higher moisture contents after vegetation is removed (Sidle and Ochiai 2006). In most temperate regions, soils are nearly saturated and evapotranspiration is low during autumn and winter rainstorms when shallow slope failures typically occur, thus, soil water content is only minimally affected by the small water losses attributed to evapotranspiration (Sidle et al. 1985). In Canada, reduced evapotranspiration following logging may increase pore water pressures during moderate-sized winter storms, but for the large storms in which landslides generally occur, differences in pore water pressure due to logging were difficult to detect (Dhakal and Sidle 2004a). However, evapotranspiration facilitated by trees could extend the “window of susceptibility” for shallow landslides and debris flows if a large storm occurred near the beginning or end of the rainy season (Megahan 1983; Sidle et al. 1985). Also, when large and high intensity storms occur during drier conditions, evapotranspiration may reduce the potential of shallow landslides (Sidle and Ochiai 2006).
Fig. 1

a Conceptual illustration of a potential slip plane at the soil—bedrock interface of a hillslope, dynamic shallow groundwater table (landslide triggering mechanism), and stabilizing components of root systems (lateral reinforcement within the upper soil and anchoring of shallow soil mantle to bedrock). b A herringbone root system with one main axis on which are borne first order lateral roots. c A dichotomous root system with two external root tips borne on every lateral. 2nd and 3rd order lateral roots are present. Magnitude is defined as the number of root tips (n = 16 in both cases), and altitude the longest path from root base to a root tip. Depending on the branching pattern, the same volume of soil is not exploited equally, even when root volume and magnitude is the same in both cases. Dashed lines indicate the potential shear zone, the thickness of which can be influenced by the number of roots crossing through it, root length and branching angle. The circled area illustrates how root proliferation in e.g. a nutrient rich patch can influence root system architecture near a slip surface. In b), initiation and emergence of a lateral root on the upper side of the mother root can occur if the latter encounters an obstacle

A much more significant contribution of woody vegetation to the stability of shallow soils on steep slopes is the additional soil strength or cohesion attributed to root systems. In shallow soils, tree roots may penetrate the entire soil mantle and anchor the soil into more stable substrate (e.g., Wu et al. 1979) (Fig. 1a). Dense lateral root systems in the upper soil horizons form a membrane that stabilizes the soil (e.g., Schmidt et al. 2001) (Fig. 1a), and larger tree roots can provide reinforcement across planes of weakness along the flanks of potential slope failures (Sidle et al. 1985). Past research suggests that while woody roots significantly reduce shallow (<1 to 2 m deep) landslide potential on steep slopes, deeper soil mantles (>5 m) benefit little from such reinforcement, as root density decreases dramatically with depth and few large roots are able to anchor across the basal failure plane. The only benefit of root strength to the stabilization of deep-seated landslides, e.g., earthflows and slumps, is when very large lateral woody roots cross planes of weakness along the flanks of potential failures (Sidle and Ochiai 2006). However, such reinforcement would only benefit small deep-seated landslides whereas lateral roots also offer protection against most shallow landslides (Swanston and Swanson 1976) (Fig. 1a). Basal woody vegetation cover, organic matter, and dense, shallow root mats also prevent surface mass wasting, e.g. dry gravel on steep slopes (Mersereau and Dyrness 1972).

What information is used by engineers when modelling slope stability?

Investigating the stability of vegetated slopes implies taking into consideration the interaction among three physical or biological systems, i.e. soil, water and plants, through a multi-disciplinary approach (Coppin and Richards 1990). However, there are very few models that implement all these aspects as a whole, and even fewer which consider the time evolution of these interactions.

Modelling the mechanical reinforcement of soil by roots

The effect of roots on soil fixation has been reported by several authors, but quantifying the gain in soil shear strength is difficult to achieve. Pioneering modelling work by Wu (1976) and Waldron (1977) have introduced the root mechanical contribution as additive soil cohesion in the Coulomb’s failure criteria using a simple mechanistic model. The additional cohesion at the slip surface was defined by two variables: the average root tensile strength and root area ratio (RAR, or the fraction of a plane of soil occupied by roots). It was assumed that roots are initially perpendicular to the slip surface and bend according to the relative displacement of soil on both sides of the shear zone (see Fan and Su 2009). The tangential component of root tensile force thus directly contributes to the increase in soil shear strength, whilst the normal component augments the confining pressure. These first models of soil reinforcement have been shown to overestimate the additional cohesion due to roots in tension, as all roots are assumed to break at the same time (Bischetti et al. 2009a; Loades et al. 2009). A significant improvement of this approach has been proposed by Pollen and Simon (2005) and Pollen-Bankhead and Simon (2009) who applied a Fibre Bundle Model to rooted soils. This model considers that the root network breaks progressively from the weakest to the strongest roots, and that stresses of broken roots are redistributed to the remaining elements. More recent papers also pointed out the limitation of assuming that roots are initially oriented perpendicular to the slip surface, and propose using root architectural models as an improvement in slope stability analyses (Reubens et al. 2007; Danjon et al. 2008).

In most models, roots are considered as very flexible elements, thus limiting their application to the finest roots. However, from a mechanical point of view, a distinction must be made between fine and thin roots, which behave like cable elements, i.e. with a very low bending stiffness, and structural roots that are similar to beams, developing longitudinal shear stresses (Fournier et al. 2006; Reubens et al. 2007). The orientation of structural roots near the slip surface cannot be estimated using a simple geometrical rule, as performed by Wu et al. (1979), due to flexural rigidity that limits their deformations. Woody roots play an important role in tree anchorage and their mechanical interactions with the soil medium have been previously investigated using finite element models (FEM). These models take into account root architecture and structural roots have been considered as individual beam elements (Dupuy et al. 2005a, b; Dupuy et al. 2007; Fourcaud et al. 2008). Few numerical analyses have been carried out at a local scale to quantify the effect of such root element inclusions on soil shear strength (Wu 2007). Nor has stiffness of structural roots been introduced in slope stability models so far, except for the notable exceptions of models developed by Nakamura et al. (2007) and Tosi (2007).

In addition to root tensile strength, the pull-out resistance of roots (Table 1) has been incorporated as a criterion in slope stability models (Waldron 1977; van Beek et al. 2005; Pollen-Bankhead and Simon 2009). The maximum force required to pull-out a root is usually correlated with root diameter at the clamp position, root length and/or total root biomass including lateral ramification. However, pull-out tests do not allow the root-soil interface behaviour to be rigorously calibrated due to the difficulty in separating the effect of lateral root branching. Bond resistance of a single root is defined by the product between the root-soil interface shear strength and the total root surface (Ennos 1990). The interface shear strength can be approximated theoretically considering that shear resistance of the root-soil interface is proportional to soil shear strength. The influence of branching patterns on pull-out resistance in different soil types has been investigated using numerical (Dupuy et al. 2005a) and experimental (Stokes et al. 1996; Mickovski et al. 2007) approaches. These studies have shown that root topology, branching angle and branching density can significantly change the distribution of stresses and plastic strains within the soil medium, thus modifying resistance to pull-out, but these aspects are not yet directly considered in soil reinforcement models.
Table 1

Ranges (or means) of pull-out resistance for roots of different diameters and lengths


Soil type

Range (mean) of root diameter at clamp position or at breaking point (*) (mm)

Range (mean) of root length (mm)

Pull-out force range (mean) (N)

Range (mean) of equivalent failure stress at clamp position or at position of root failure (*) (MPa)


Picea sitchensis Bong. Carr.

Brown earth





Anderson et al. 1989

Picea sitchensis Bong. Carr.






Anderson et al. 1989

Casuarina glauca Sieb.

Brown loams and sandy loams

0.56–17.23 (4.49)



3.34–81.68 (22.43)

Docker and Hubble 2008

Eucalyptus amplifolia Naud.

Brown loams and sandy loams

0.45–16.43 (5.00)



8.88–130.65 (27.33)

Docker and Hubble 2008

Eucalyptus elata Dehn.

Brown loams and sandy loams

0.21–15.93 (4.85)



8.57–198.06 (31.49)

Docker and Hubble 2008

Acacia floribunda (Vent.) Willd.

Brown loams and sandy loams

0.31–13.33 (4.09)



11.19–217.89 (58.09)

Docker and Hubble 2008

Triticum aestivum L.

Sandy clay loam





Easson et al. 1995

Triticum aestivum L.

Sandy loam





Easson et al. 1995

Helianthus annuus L.

Sandy loam





Ennos 1989

Allium porrum L.

Dry sandy loam





Ennos 1990

Allium porrum L.

Wet sandy loam





Ennos 1990

Alnus incana L.

Sand and gravel

2.42–3.19 (2.76) *


250–357 (299)

23.5–39.2 (30.4)*

Karrenberg et al. 2003

Populus nigra L.

Sand and gravel

3.25–4.07 (3.64) *


436–517 (475)

26.4–37.3 (31.4)*

Karrenberg et al. 2003

Salix elaeagnos Scop.

Sand and gravel

3.33–3.97 (3.62)*


578–705 (638)

43.3–57.8 (50.1)*

Karrenberg et al. 2003

Vetiveria zizanioides L.


0.30–1.45 (1.02)

110–275 (219)a

190–620 (470)


Mickovski et al. 2005

Crataegus monogyna Jacq.

Brown sandy clay

7–62 (21.6)


300–12,000 (2,880)

3–15 (8)

Norris 2005

Quercus robur L.

Brown sandy clay

2–9 (5.4)


30–440 (150)

2–14 (7)

Norris 2005

Pinus halepensis



3–24 (9)

Van Beek et al. 2005

aRooting depth, as the whole plant was pulled-out

Integration of root reinforcement models in slope stability models

Root reinforcement models are used to determine the contribution of vegetation to the factor of safety (FOS) of a particular slope. The FOS is defined as the ratio of the actual soil strength to the minimum shear strength required for equilibrium. To calculate the FOS of a vegetated slope, 2D or 3D numerical techniques exist which couple soil mechanics, soil hydrology and plant growth (Stokes et al. 2008), e.g. limit equilibrium (Greenwood 2006), finite difference (Wilkinson et al. 2002; van Beek et al. 2005) and finite element (Kokutse et al. 2006) models. Vegetation must be taken into account as reinforcement elements and elements that modify the hydrological regime of the slope (Simon and Collison 2002; Wilkinson et al. 2002; Dhakal and Sidle 2004b).

What are the traits useful for reinforcing soil on slopes?

Although the plant root traits most commonly used by engineers when modelling slope stability are tensile strength and RAR, several other important traits (Burylo et al. 2009) should also be considered (Table 2). First however, the type of root must be defined. Therefore, to better understand the different processes governing soil fixation by roots, three classes of roots will be referred to in this paper, depending on their diameter (although these classes do not necessarily reflect the developmental stage of a root, as the diameter of roots at different ages of maturity will depend on species):
  • Fine roots (>0.0–2.0 mm). A major function of these roots is to take up water and nutrients. If roots are young and not yet woody, root hairs may be present which aid uptake. The turnover of fine roots is relatively rapid.

  • Thin roots (>2.0–10.0 mm). In woody species most roots in this class will be lignified. Turnover is not known because some roots will become thicker, depending on a certain number of external and internal processes.

  • Thick roots (>10.0 mm). These roots are important for anchoring the plant to the soil and preventing uprooting. The spatial position of these roots will determine the position of the fine and thin roots in the soil, and thus influence indirectly nutrient and water uptake.

Table 2

Definitions and desirability of root traits useful to a landslide engineer



How can the trait be desirable for landslide engineers?

Root area ratio (RAR)

The fraction of a plane of soil occupied by roots

The greater the RAR, the more soil shear strength will be increased (Fig. 5b)

Tensile strength

Resistance to breaking in tension

A higher resistance will enable a root to mobilise its full strength during pull-out and increase soil shear strength. Therefore, for a given RAR, many small roots are more desirable than a few thick roots

Root thickness

Diameter of root

Thin roots are relatively stronger in tension. Thickness is an indicator of root longevity (thicker roots live longer), bending stiffness and the ability to penetrate soil

Shallow/deep rooting

Inherent rooting depth

If deeper growing roots cross the potential soil shear surface, slope stability is enhanced. Anchorage of roots to bedrock will improve substrate fixation, but root growth into cracks may cause cracks to enlarge and result in substrate failure

Root length density (RLD)

Defined as either root length per unit volume of soil or root length per unit surface of a soil

Increasing root length augments the pull-out resistance up to a critical length, from which roots will break in tension instead of slipping out of the soil. The water uptake rate from a given horizon will increase with RLD. Along with evapotranspiration rate, slope stability is affected by soil water content, particularly in tropical regions, therefore higher RLD is desirable, especially in deeper horizons

Specific root length (SRL)

Root length per unit root dry mass

High SRL implies more numerous thinner roots and low SRL means less but thicker roots (see ‘Tensile strength’ above). Is a useful descriptor for herbs and perennials where a high SRL is desirable

Root angle

Angle between mother and daughter root in vertical or horizontal plane

Soils containing roots with a range of orientations develop wider shear zones and can slowly mobilise reinforcement from roots via their tensile strength even at large shear displacements

Production of adventitious roots

Roots initiated from the base of the stem, with later roots emerging higher up the stem

Adventitious roots can grow into soil and debris deposited on the upslope side of a plant, thus fixing shallow soil layers


Physical organisation of root branching

Root topology influences resource uptake and efficiency and can significantly change the distribution of stresses and plastic strains within the soil medium, thus modifying root resistance to pull-out. Dichotomous systems are better anchored than herringbone systems


Forks appear when the growth potential of one lateral root changes as a shift occurs in the mother root’s functioning

Production of forks will anchor more superficial roots to the soil, especially if forks grow vertically downwards

Root clustering

Proliferation of roots within a given volume of soil

If roots are clumped within cracks and biopores in the soil, the time required for total water extraction increases because of the greater distance that water must flow through the soil structural unit to the roots. But if clusters of roots occur in resource-rich patches, water uptake is faster around the cluster compared to a zone without roots, thus preferential water extraction occurs from specific parts of the soil. A homogenous distribution of roots within the soil is more desirable

Root response to soil stresses

Effect of adverse physical, chemical, or biological conditions on root growth

Physical—large variation in plant responses to soil compaction, drought and waterlogging

Chemical—certain genotypes may overcome nutrient deficiencies or grow in presence of toxic compounds

Biological—species dependent susceptibility and resistance to disease

Root decay rate

Rate of dry mass loss per unit dry mass initially present

Roots which decay slowly fix soil for a longer period. Root decay may create preferential flow pathways in soil. Root decay rates vary greatly according to root diameter, chemical composition and species

Resprouting ability

The ability of a plant to produce a new shoot from a root or stem if part of the plant’s biomass has been removed

Usually in perennials only. If a plant can resprout, from either the shoot or root system, it will be able to survive in a newly disturbed environment e.g. in a landslide deposition zone, even if above-ground biomass is destroyed

Mycorrhizal interaction

Symbiotic association between a fungus and plant roots. Arbuscular mycorrhizas hyphae enter into the root cells and ectomycorrhizas consist of a hyphal sheath covering the root tip and a hartig net of hyphae surrounding cells within the root cortex

Mycorrhizal infection usually enhances root branching density. Tree species predominately forming ectomycorrhiza can have a higher branching intensity than those forming arbuscular mycorrhiza. Ectomycorrhizal infected roots have been found to have a higher tensile strength. Mycorrhizal inoculation may be desirable on degraded soil

The first root (radical) which emerges from a seed grows downwards and in some species, becomes the taproot (Figs. 1b, c, 2) along which are borne lateral roots of different orders (Fig. 1b, c). Even if this taproot is removed through, e.g. injury, a new vertical root will emerge or an existing lateral root will grow downwards to replace the taproot, thus ensuring that the functional role of the taproot is maintained (Champagnat et al. 1974; Amin et al. 1987; Khuder et al. 2007). In other species, the radical dies shortly after germination and is never replaced. Roots are borne on the root collar, or stem base (Fig. 2). In this paper we shall refer to the first lateral roots emerging from the taproot or root collar as first order roots, lateral roots emerging from the first order roots as second order roots and so on (Fig. 1b, c).
Fig. 2

Depending on the strategy type, root systems develop differently for given species. In a short-lived and perennial lateral roots are borne on the taproot (black) and in b directly on the stem (shaded). Arrows show the direction in which the root system develops

With regard to soil fixation by plant roots, the most important trait to consider initially is rooting depth. To stabilize a slope against a shallow landslide, roots must cross the shear surface (Fig. 1) which can be up to 2 m deep in the middle of the slope (Cammeraat et al. 2005; van Beek et al. 2005; Norris et al. 2008). Rooting depth is species dependent when soil conditions are not limiting and root development changes significantly with depth, although in general, >80% of biomass is in the top 0.4–0.5 m of soil. The number and orientation of roots that the shear surface intersects will change significantly with rooting depth for the same plant, even for distances of only several cm (Fig. 1b, c). Roots vertically aligned to the potential slip surface will be mobilised throughout their length whereas those with a larger angle will be more likely to slip before failure in tension occurs. A greater number of roots and diversity of branching angles will increase the thickness of the shear zone (Abe and Ziemer 1991a; Mickovski et al. 2007). Although the true angle of roots in the shear zone is often ignored, and assumed to be between 40° and 70°, Greenway (1987) and Danjon et al. (2008) showed that such assumptions lead to an overestimation of the shear resistance contributed by roots. Even for the same root, the angle at which it grows with regard to the vertical can change several times along its length. Due to the physiology of roots, a root branch can be initiated at any point along a parent root, but not necessarily emerge fully from the parent root. Emergence is largely controlled by auxin fluxes and the bending of a root around an obstacle or as it grows through the soil can trigger the initiation and emergence of a lateral root (Fig. 1b, Lucas et al. 2008).

Thick roots act like soil nails on slopes, reinforcing soil in the same way that concrete is reinforced with steel rods (Fig. 1a). The spatial position of these thick roots also has an indirect effect on soil fixation in that the location of thin and fine roots will depend on the arrangement of thick roots. Thin and fine roots act in tension during failure on slopes and if they cross the slip surface, provide a major contribution to slope stability (Fig. 1, Waldron and Dakessian 1982). Root thickness is thus an important trait to consider, and is also an indicator of root longevity (thicker roots live longer), bending stiffness and the ability to penetrate soil (Clark et al. 2008) as well as to store and transport water (Roumet et al. 2006). For the landslide engineer, one of the first considerations with regard to root thickness will be how tensile strength changes with diameter. Although thicker roots require more force to be pulled out of the soil, when root strength is calculated (force/root cross-sectional area), thinner roots are significantly stronger than thick roots, due to changes in cellulose content (cellulose is highly resistant in tension) (Genet et al. 2005). Tensile strength values can vary from 20 to 730 MPa for roots in a diameter range of 0.15–4.5 mm (Bischetti et al. 2005), this surprisingly high latter value for Fagus sylvatica L. being stronger than that for steel (400–700 MPa). A list of tensile strength values for 67 species is available in Stokes et al. (2008), but in a model of slope stability, Danjon et al. (2008) used a generic power equation of root tensile strength versus diameter for several broadleaf and conifer species (equation from Genet et al. 2005):
$$ Tensile\,strength = 28.97{x^{ - 0.52}} $$

Genet et al. (2009) suggested that differences in tensile strength among tree species have little influence on the calculation of a slope’s FOS and that root distribution within the soil will affect slope stability much more.

Root length is one of the most studied traits in root systems and in herbaceous species is usually studied in conjunction with biomass, resulting in a parameter called specific root length (SRL). SRL is thus calculated as the root length per unit root dry mass. This ratio will provide information about the proportion of thin and thick roots within a root system (high SRL implies more numerous thinner roots and low SRL means fewer but thicker roots). A higher SRL would therefore be desirable, particularly at greater soil depths. Annual species usually have a higher SRL than perennials, possibly associated to a faster relative growth rate (RGR) and metabolic activity (Roumet et al. 2006). The RGR of a plant organ is calculated as the instantaneous rate of dry mass increase per unit dry mass already present (Wright and Westoby 1999). For a landslide engineer, how fast a plant grows will be of particular importance when a gap has occurred on a site after, e.g. tree fall, thinning of a plantation or clear-felling (Sidle 1992). The roots of plants remaining around the newly-created gap, or those of naturally generated or planted nursery seedlings, need to grow quickly to colonize the new patch of soil if soil slippage is to be prevented on a fragile slope. The inherent RGR of a species depends largely on its life form and ecological niche: in optimum conditions, plants from fertile habitats grow faster than plants from nutrient poor or stressful habitats and annuals grow faster than perennials. Although not a trait in itself, RGR can be correlated with various traits, e.g. SRL, although results from studies are conflicting. Wright and Westoby (1999) found that in seedlings of 33 species growing down a rainfall gradient in Australia, SRL decreased due mainly to an increase in root diameter (a larger proportion of thicker roots was present in lower rainfall areas) and was highly correlated with RGR. Such a strategy could indicate selection for increased efficiency of water uptake during seedling establishment in low rainfall habitat or possibly a greater diameter might confer an enhanced ability to penetrate dry soil (Wright and Westoby 1999). However, Poorter and Remkes (1990) found no relationship between SRL and RGR in 24 species. If the investment of resources in fast and slow-growing roots does depend on inherent RGR, one can suppose that these differences may be reflected in root tensile strength.

Although traits such as root diameter and SRL are useful indicators of plant structure, they are only simple descriptors of how a root system is configured. These traits will also change depending on the root order (Fig. 1), development stage and position within the branched hierarchy of the root system (Eissenstat et al. 2000). To better understand how root systems occupy soil over time and space, a landslide engineer needs to consider the 3D root architecture of any given species. This task is not easy, nevertheless, certain generalizations can be made, based on developmental observations and a knowledge of plant ontogenetic processes (the history of an organism from birth to death).

Root system architecture

The architecture of a plant depends on the nature and the relative arrangement of each of its parts (see Barthélémy and Caraglio 2007). From germination to death, various well-ordered and definite developmental stages can be characterized taking into account the developmental and functional features and the relative arrangement of the different parts. Until fairly recently, quantitative data on root architecture has been scanty not only due to the difficulty in accessing roots, but also because the morphological analysis of the root system is not as efficient as that of its aerial counterpart. Root systems do not bear morphological markers of growth rhythms e.g. leaf nodes, bud-scales or even well-defined annual growth rings, therefore temporal changes cannot be dated with hindsight.

The main differences between plant root architectural types are those which can be highlighted by comparing woody with non woody species. By assessing the initial rooting development, which is largely under genetic control in the first stages (Hermann 1977; Zobel 1996), two main rooting strategies can be seen depending on the development of the radical (Fig. 2):
  • In the first strategy type (Fig. 2) the whole root system is derived by the growth and lateral branching of the seedling radical borne opposite to the stem in the embryo. This primary root grows downwards and branches in a given orderly sequence, building up all the lateral woody root system of the plant (Kahn 1983; Bell 1991). During the branching process, the taproot is responsible for the gravitropic response and for the morphological and functional differentiation of lateral roots borne on the taproot (Dyanat Nejad and Neville 1972; Champagnat et al. 1974). The best illustration of such a rooting habit is shown in woody seedlings which possess a taproot, regardless of the taproot’s growth in the ensuing stages of plant development. Many dicotyledonous herbs and shrubs also exhibit such a rooting strategy (Kutschera and Lichtenegger 1997). In the juvenile developmental stages of plants exhibiting this strategy type, even if the taproot is removed through e.g. an injury, a new vertical root will emerge or an existing lateral root will grow downwards to replace the taproot, thus ensuring that the functional role of the taproot is maintained (Khuder et al. 2007).

  • In the second strategy type (Fig. 2), the radical stops growing or aborts soon after the seed germinates (Jourdan et al. 1995; Charles-Dominique et al. 2009). The adult root system is entirely adventitious1 and made up of many homologous roots initiated in an upward order at the base of the stem (Fig. 2). The best illustration of such a rooting habit is that present in cereals or bulbs, but tropical arborescent palms or ferns are also concerned (Hallé and Oldeman 1970; Jenik 1978; Bell 1991).

These two main strategies define very early on the way in which the root system will colonize the soil and thus reinforce unstable soil, because they greatly influence root angle, relative horizontal versus vertical root spread and root density. In particular, in the first rooting strategy type, the spreading of horizontal roots away from the stem can result in large gaps in the soil with no roots present (Figs. 3, 4), whereas, in the second strategy type, many roots are continuously renewed from the base of the stem over a long period during plant development. On a slope where loose soil and debris can collect upslope of a plant, adventitious roots can therefore grow into this material, stabilising it better than deeper growing roots could.
Fig. 3

ad Development of a root system at different stages of ontogeny. Large central roots build up the framework of the root system (1, 2). The taproot (1) bears the woody laterals (2) which spread the thinner roots (3) away from the base of the trunk in order to explore the soil. These colonisation roots branch to develop thin woody ephemeral exploitation roots (4). All the woody roots branch into non woody laterals which are usually the fine absorbing roots (5)
Fig. 4

Forks occur in the lateral root branches of woody species. This branching pattern allows the root system to radiate widely along the soil surface

Roots of herbaceous plants

Root systems of herbaceous plants are usually more diffuse, or fibrous than those of woody plants, although bulbs, tuberous and storage roots are common. Fibrous root systems generally possess more fine and thin roots than those found in woody species. Tensile root strength can be similar to that of roots from woody species, whereas RAR can be higher (Mattia et al. 2005; De Baets et al. 2008; Loades et al. 2009), therefore, if shallow reinforcement of soil is required, fast-growing herbs can be a useful alternative to slower growing shrubs and trees. However, shallow mats of homogenously distributed fibrous roots can easily tear away from the subsoil in humid alpine conditions (Tasser et al. 2003). While such shallow root systems may provide protection against shallow surface erosion, they exert little to no protection against shallow or deep-seated landslides (Rice and Foggin 1971; Gabet and Dunne 2002).

At a more detailed level, topological indices have been used frequently to describe root systems of herbaceous species (Fitter et al. 1991) as well as young tree roots and the root systems of juvenile trees (Stokes et al. 1995; Tsakaldimi et al. 2009). The topology of a root system largely determines how roots occupy a given unit of soil and plays an important role in resource uptake and anchorage efficiency. Fitter (1985) classed herbaceous root branching patterns into three types, depending on magnitude (the number of branch tips), altitude (the number of links, or distance between two branches, in the longest path from the base to the tip) and the relationships between these two parameters. Herringbone root systems are composed of a single main root and one order of lateral roots (Fig. 1b) whereas in dichotomous systems, a mother root always possesses two daughter branches (Fig. 1c). Most branching patterns lie between these two extreme types and are called ‘random’ patterns. All three different types of topology can be found within the same root system, depending on root age and nutrient heterogeneity. In trees, shrubs and herbs, very young roots may need to explore nutrient rich patches, therefore send out ‘foraging’ herringbone roots which proliferate into random or dichotomous systems once a pocket of nutrients has been found. The branching pattern can affect the resistance to uprooting of a plant and for an equal root volume, dichotomous systems are significantly better anchored than herringbone systems if roots are rigid. This change in anchorage is because during uprooting, the total amount of soil mobilised during uprooting increases as it is carried upwards on the forks of lateral root branches (Stokes et al. 1996; Dupuy et al. 2005a, b). Similarly, in sand, the pullout resistance of model roots depends on the number and depth of lateral roots. The deeper the laterals, the greater the resistance to pullout due to the increase in stresses acting in the sand at depth, due to the overburden of sand above (Mickovski et al. 2007). The Coulomb friction law also predicts that if soil moves in a horizontal plane around the root, above a critical depth, the soil is lifted upwards. Below this critical depth, less energy is required to compress the soil and to move it horizontally around the soil (McKyes 1985).

Roots of woody plants

Woody plants generally develop two main classes of roots, long roots (called woody, coarse, skeleton, anchoring roots or macrorhizae) and short roots (called non woody, fine, feeding, absorbing roots or brachyrhizae). The long roots can become a permanent part of the root system acting as a framework and for anchorage. More ephemeral short roots emerge as lateral branches that do not undergo secondary thickening, and so are thinner than long roots. Discrimination between these two classes of root is usually on the basis of root diameter within the range 1 to 10 mm (Hermann, 1977), as defined previously in this paper.

Generally, trees develop large conical-shaped woody central roots which are perennial and build up the framework of the root system (Fig. 3). The central and vertical taproot ensures the anchorage of the plant and bears all the woody laterals (Atger and Edelin 1994a). Near the collar, the biggest woody laterals spread the thinner roots away from the base of the trunk in order to explore the soil (Figs. 3, 4). This framework develops laterally into two types of woody roots that are usually shorter and thinner (cylindrically shaped) and rarely live for a long time. These colonisation roots occupy briefly the soil around the skeleton roots as they are spread away from the tree stem, before branching to develop thin woody ephemeral exploitation roots (Fig. 3). Woody roots branch into non woody laterals which are usually the fine absorbing roots (Fig. 3) and which are more or less branched, depending on species. For a landslide engineer, it is important to remember that a slope is stabilised by the whole community of plants, and not individual root systems in isolation. The concept of root occupancy is useful here, representing the sum of the reinforcement due to new root material, plus the remaining root reinforcement of any residual decaying root systems (Watson et al. 1999). Watson et al. (1999) viewed the root system of a tree as an expanding circle, finding a rate of lateral root expansion of 0.44 m year−1 for radiata pine (Pinus radiata D. Don), as compared with 0.25 m year−1 for kanuka (Kunzea ericoides A. Rich), on clear-felled slopes in New Zealand. However, as the natural stand density of trees for kanuka (16,000 stems ha−1) is much greater than for radiata pine (typically 400 to 1,250 stems ha−1), the kanuka site reached its target root occupancy in only half the time taken by the densest stand of radiata pine. (In such studies however, it is always necessary to consider the plant’s developmental stage, as root extension rates of young trees are greater than those of mature trees).

Tree root systems have traditionally been classed into three types, depending on their overall shape: plate, heart and tap (Köstler et al. 1968). Plate root systems have large lateral roots and vertical sinker roots, heart systems possess many horizontal, oblique and vertical roots and tap systems one large central root and smaller lateral roots (species list available in Stokes et al. 2008). This classification is over-simplified however, as many species have a mixture of root system types. In particular, a ‘sinker’ root system type is frequently encountered and is composed of lateral roots from which descend vertical sinker roots (Danjon et al. 2005). With regard to tree overturning during storms, trees possessing heart and taproot systems have been classified as being the most resistant to uprooting and plate systems the least resistant (Dupuy et al. 2005b; Norris et al. 2008). For stabilizing a slope, the type of root system could play a role in species choice, with deeper taprooted systems planted in the middle of a slope and plate rooted systems at the top of the slope (Danjon et al. 2008).

Root traits and their interaction with soil properties

Roots reinforce slopes by the direct mechanical reinforcement of their roots, but also by drying the soil and so increasing the effective stress on potential shear surfaces (Fig. 5a). Root systems extract water from deeper in the soil than would occur for evaporation from the bare soil surface, and calculations based on the Penman-Monteith Equation for evapotranspiration suggest that this phenomenon may significantly strengthen slopes against shallow failures (Tarantino et al. 2002; Sidle and Ochiai 2006). However, evapotranspiration is often small during extended rainy seasons, especially in temperate regions and, as mentioned previously, slopes often fail when saturated by heavy rainfall. In Malaysia, where evapotranspiration is high throughout the year, stable slopes were associated with large root length densities (RLD) and relatively dry water contents, whereas unstable slopes tended to be relatively wet and unrooted (Osman and Barakbah 2006). RLD is a trait defined as either root length per unit volume of soil (i.e. root length distribution with depth) or root length per unit surface of a soil (i.e. the amount of roots per unit ground surface or crack surface) (Miyazaki 2006). However, simple water balance simulations in Peninsular Malaysia suggest that tropical forest removal would only moderately increase the period of movement of deep-seated failures and would have little effect on landslides in shallow soils (Sidle and Ochiai 2006).
Fig. 5

a Shear strength as a function of soil matric potential for three clay soils (data from Spoor and Godwin 1979) and b increase in shear strength as a function of root area ratio (RAR) for clay loam and chalk (data from Operstein and Frydman 2000), and from sandy loam (data from Mickovski et al. 2009)

Root traits and the soil water dynamics of slopes

Root extraction of soil water

The shear strength of a soil depends on the balance of forces operating on a potential shear plane. The matric suction associated with unsaturated soil tends to increase the resistance to shear as soil dries, by increasing the effective stress between particles in the soil matrix (Fig. 5). Plant roots act as small suction pumps, drawing water from the soil at matric potentials as negative as −1.5 MPa. Individual root tips may continue to grow at matric potentials as dry as −2.3 MPa, and even to −4 MPa if the rest of the plant is adequately supplied with water (Portas and Taylor 1976; Sharp et al. 2004).

The rate of water extraction achieved from a soil layer depends on the RLD in a particular zone of soil and the uptake rate per root length. If the root system is uniformly distributed in a volume of soil, a RLD as small as 0.1 cm cm−3 should be adequate to extract the water from that soil volume in a few days (Passioura 1991). However, if roots are clumped within cracks and biopores in the soil, the time required for water extraction may increase by an order of magnitude or more (Passioura 1991). This phenomenon is because of the greater distance that water must flow through the soil structural unit to the roots—several roots clustered within a single biopore may only be marginally more effective than an individual root in the same pore. Such clustering occurs particularly within the subsoil where the strength of the soil may be relatively large and restrict growth to continuous channels and planes of weakness or, e.g. if new roots grow preferentially along decomposing root channels. Little is known about the relative ability of roots to exploit biopores, though methods are being developed that might allow screening for this trait in crop plants (McKenzie et al. 2009). There are at least indications that species may differ substantially in their ability to extract water from the subsoil, even if the distribution of RLD is similar (Bremner et al. 1986; Passioura 1988).

There are large species and varietal differences in the distribution of root length with depth (Jackson et al. 1996), and so it should be possible to choose vegetation for slopes that extract water from greater depths. In terms of the total amount of water extracted from a slope, generally the larger the biomass the larger the potential evapotranspiration rate. Water extraction by the root system must be viewed in the wider context of the water balance of the whole slope.

Effects of roots on preferential flow pathways in the soil

Preferential flow paths are regions of the soil where relatively rapid transport of water and solutes may occur under heavy rain or flooding (Perillo et al. 1999). Rapid macropore flow paths may be difficult to locate or predict using core-sampling techniques, and may only become apparent at large lysimeter or hillslope scales (Deeks et al. 2008). Macropores formed by plant root channels comprised at least 55% of all macropores in a Japanese forest soil (Noguchi et al. 1997), whilst 60% of all preferential flow paths under recently tilled alfalfa (Medicago sativa L.) plots were associated with roots or decomposed root channels (Perillo et al. 1999). The length of these preferential flow paths increases with rainfall intensity and soil water content (Tsuboyama et al. 1994; Perillo et al. 1999), and may result in relatively rapid transport of water through the surface horizons to less permeable horizons or to bedrock (Fig. 6). In gently sloping terrain preferential flow paths generally serve to facilitate vertical transport of water and solutes; however, in steep topography, especially during high rainfall, these preferred flow networks can rapidly transport water downslope potentially increasing pore water pressure at the soil-bedrock interface or within the soil when hydrologic discontinuities are encountered (Sidle and Ochiai 2006).
Fig. 6

Illustration of how a decayed root channel can initiate a preferential flow path (PFP) in soil. The dark coloured area in the Bt horizon indicates where dye has flowed preferentially along the channel formed by the decomposing lignified tissue of a root, stained black. (Reprinted from Perillo et al. (1999), with permission from Elsevier)

It is likely that differences exist in the preferential flow pathways associated with root systems of different species, although this is not well documented. Root systems with relatively rapid root turnover rates will presumably result in more decaying root channels than plants with slow turnover. Whilst the thick structural roots of trees may persist for many years, finer roots belonging to the same root system e.g. <2 mm tree roots, may be born and die within a period of weeks or months (Eissenstat et al. 2000; Block et al. 2006). Even within a single root system, there can be very large differences in root lifespan related to root diameter—for example, fine tree roots turnover relatively rapidly (e.g. 30 to 300 d for fine poplar roots; Block et al. 2006), as compared with much thicker structural tree roots that can persist for many years. This temporal nature of fine root dynamics needs to be reflected in models (Abe and Ziemer 1991b).

Roots play a major role in the formation of soil structure by providing a source of carbon, via root turnover and exudation, and by subjecting the rhizosphere to alternating wet-dry cycles that bind rhizosphere particles together with mucilage (Watt et al. 1993). The interaction of the root system with soil type influences the nature of structure formation. For example, grass root systems often have RLD of 20 cm cm−3 or more in the surface 0.1 m of soil, with relatively rapid turnover rates. This gives rise to the typical crumb-like soil structure found under grass swards. The presence of the grass roots and their associated fungal and bacterial communities rapidly binds the soil together and decreases erosion from establishing newly engineered slopes. The average depth of most grass root systems is usually <1.0 m and so the effect is confined to the surface horizons and has little affect on landslides as previously noted.

Soil constraints on root system morphology

Root system form varies not only between species but also within a species. The first stages of root development are genetically driven and shape is not highly variable. As a plant matures, the morphology of its root system and the distribution of roots in the soil are greatly affected by the immediate soil environment as well as aboveground influences. However, for a given soil constraint, e.g. mineral deficiency, plant responses can be significantly different depending on species and the part of the root system considered. Even the response of a single root to any one stress will vary according to the root type and its branching order (Drew 1975; Thaler and Pagès 2000).

When analysing the consequences of environmental constraints on root development, roots which undergo many of the common stresses often respond by first reducing and then by arresting lateral root growth. Therefore, a mother root inhibits lateral root emergence before restricting its own growth (May et al. 1965, 1967; Hackett 1972; Drew et al. 1973). In heterogeneous soil, when a stressed root reaches a zone with conditions ideal for growth, any inhibition of branching can be followed by a proportional stimulation of lateral root development. Therefore, when one part of a root or a root system has been suppressed, the remainder can react and exhibit what has been called compensatory growth (Drew et al. 1973). Studying the competitive effects of neighbours on lateral root spread in a Creosote bush (Larrea tridentata Sesse and Moc. ex DC.) population, Brisson and Reynolds (1994) demonstrated that the occurrence of such compensatory growth induced a sharply asymmetrical distribution of horizontal roots. As the asymmetrical formation of root systems is frequently encountered on mountain slopes (see below), studies of plant growth on slopes must be analyzed carefully, taking into account this potentially compensatory root growth. Nevertheless, not all species can exhibit compensatory growth to avoid neighbours e.g. Platanus sp. and the west African species Aucoumea klaineana Pierre. Such species have roots which can fuse and graft with each other (both within the same tree and between individuals), thus allowing sap exchange between individuals (Norris et al. 2008). The fused root systems can act as a network, holding soil and stones in place and interlocking with rocks and coarse fragments in the soil, enhancing the lateral membrane strength of the soil-root complex (Fig. 1). Nevertheless, many tree species demonstrate compensatory growth which results in the whole form of the root system changing so that, e.g. a taprooted or heart rooted system can be transformed into a plate root system when influenced by local soil conditions, such as the presence of a hard pan or a seasonal water table (Fig. 1a, Danjon et al. 2005; Norris et al. 2008). The expression of these different forms is known as plasticity.

Shallow and deep root endemics

Schenk and Jackson (2002) collated a database of rooting depth for 1,300 plants from water-limited (arid to sub-humid) environments. These authors found that plant rooting depths in water-limited environments are highly correlated with infiltration depth. Perennial herbs and forbs had average rooting depths of approximately 0.75 m whereas the rooting depth of annuals was usually <0.5 m, although these depths could reach several metres. Shrubs and trees on deep soils have a mean rooting depth of 2.2 m whereas shrubs and trees growing on shallow soils over bedrock have a mean rooting depth of 7.9 m. What is happening in this latter case is that roots of woody plants on shallow soils tend to grow along fractures deep into the underlying bedrock (Fig. 1a). Schenk (2008) therefore asks the question—how do roots at the boundary between soil and bedrock locate rock fractures that are large enough for them to grow in? Poot and Lambers (2008) studied six species of Hakea, two of which are endemic to shallow soils over ironstone and four are commonly found on deeper soils in the same region. When grown in containers, the shallow soil endemics explored the bottom of the containers more efficiently than the deep soil endemics. This superior foraging capacity will allow roots to locate cracks in the underlying bedrock, leading in turn to possible supplies of nutrient and water rich pockets. Therefore, the shallow soil endemic will be able to better tolerate water and nutrient stress, but the disadvantages of occupying this particular ecological niche include higher energy costs for constructing deeper roots, the possibility of oxygen deficiency and lower water and nutrient availability (Schenk 2008). For the landslide engineer, the ability of vertical roots to grow into cracks in the bedrock is of extreme but conflicting interest. If plants on shallow soils are able to grow through the soil and anchor the plant to the underlying bedrock, soil on a slope would be better fixed (Fig. 1a). However, the roots of many species, e.g. Pinus pinaster Ait., are not able to penetrate the bedrock and grow along the surface of the rock, resulting in an unstable shallow root plate (Fig. 1a; Danjon et al. 2005). In some species, e.g. eucalyptus (see Stone and Kalisz 1991), roots can grow as deep as 60 m, often being found in subterranean or aquatic caves (Hubble et al. 2009). However, in plants where roots are anchored to the bedrock, root growth through cracks could enlarge them over time, thus destabilising the rock and causing it to fracture, leading in turn to slope failure (Sati and Sundriyal 2007).

Plasticity in root traits

The ability of a genotype to change its phenotype was originally thought to be an unstable process. However, it is now recognized that this ability to change, termed phenotypic plasticity, is central to plant ecological development (Bradshaw 2006). Under genetic control, plasticity enables sessile plants to adjust to spatial and temporal heterogeneity, thus minimizing stress (Fourcaud et al. 2008). Plasticity has been found in relation to nutrient supply (Drew 1975), water availability, and soil strength (Bingham and Bengough 2003).

Heterogeneity in resource supply

When plant fine roots encounter a nutrient rich patch, they usually proliferate, or cluster, within it (Fig. 1b), where proliferation is the increase in initiation of new laterals (Hodge 2004). Lateral root length can also increase (Drew 1975), but by producing more branches, a given unit volume of soil will be exploited more efficiently. Total root system biomass does not necessarily change because an increase in root mass within a nutrient patch can occur to the detriment of the root system growth outside the patch (Drew 1975). Within a patch, roots can be longer and thinner (Farley and Fitter 1999) and the trait root tissue density (RTD) may differ in the proliferated roots (Eissenstat 1991). Lignified roots are not sensitive to changes in nutrient supply whereas in fine roots, changes in physiological plasticity can also occur so that ion uptake is increased, which might in turn act as a feedback mechanism, triggering the production of new lateral roots (Hodge 2004). A trait describing lateral root spread may be useful for estimates of the area over which plants anchor the surface soil (Fig. 1). In shrubs and trees, lateral spread was found to be typically between 2 and 16 m in radius (Stone and Kalisz 1991; Schenk and Jackson 2002) but only 0.10–0.60 m for herbaceous plants, with semi-shrubs and stem succulents intermediate. Most (95%) herbaceous perennials have lateral root spreads of <1.5 m, and 95% of all shrubs have been found to possess lateral root spreads of <15 m (Schenk and Jackson 2002).

Added to the complexity of understanding root plastic responses to soil nutrient status is the consideration of microbial communities. Mycorrhizal fungi colonise roots of a host plant, and benefit from this mutual symbiosis by having direct access to carbohydrates produced by the plant. In return, the plant can access a larger surface area of soil via the mycelial network of the mycorrhiza and take advantage of the mycorrhiza’s ability to ‘scavenge’ or ‘mine’ phosphate in the soil (Lambers et al. 2008). Mycorrhizas can also proliferate in nutrient rich patches, and increase root production and alter RLD (Cui and Caldwell 1996; Hodge et al. 2000). In 25 North American tree species, species predominately forming ectomycorrhiza were found to have a higher branching intensity than those forming arbuscular mycorrhiza (Comas and Eissenstat 2009). Root tensile strength increases in ectomycorrhizal-infected roots (Ba 2008) but this response appears to be species and site dependent. In a growing literature on the subject, mycorrhizas are seen as useful tools in the restoration of degraded soil, but again, species and site are important factors to consider before undertaking any restoration project using mycorrhizas (e.g. Palenzuela et al. 2002; Estaun et al. 2007). In certain conditions, the introduction of mycorrhizas at a site may even be detrimental to plant establishment (see Walker et al. 2009).

Root response to zones of increased water availability are less clear. Lateral soil moisture heterogeneity can result from variations in topography, water infiltration, differential root competition and agricultural irrigation whereas vertical heterogeneity, e.g. waterlogging, is often seasonal and predictable (Bauerle et al. 2008). Root growth is usually reduced in dry soils and within dry patches; resource allocation is also reduced so that outside the dry patch, preferential root growth occurs (Coutts 1982). Different species have different strategies to water supply; cacti may produce low-cost, short-lived fine roots in response to periods of precipitation (Schenk and Jackson 2002; Snyman 2006), whereas Eissenstat (1991) showed that Citrus sp maintain roots in dry soil and increase root growth only slightly in wetter soil. Bauerle et al. (2008) demonstrated that morphological plasticity in root systems of fast growing vines (Vitis vinifera cv. Merlot) was greater than that in slow-growing vines, but that the tolerance to moisture stress was similar in both types. Plastic changes will influence how the plant extracts water from the soil, with some root morphologies being more efficient than others (Tsutsumi et al. 2003). When clusters of roots occur in resource-rich patches, water uptake is faster around the cluster compared to non-rooted zones, thus preferential water extraction occurs from specific regions of soil (Kazda and Schmid 2008).

Soil texture and compaction

Soil texture and compaction have major implications for the physical stresses that roots will experience (da Silva et al. 1994; Bengough et al. 2006). Sandy soils will store less plant-available water (water held between field capacity and wilting point) than clays, and so will be more susceptible to drought. Clay soils hold more available water, but they may have insufficient air-filled porosity, limiting root growth by hypoxia under wet conditions.

Mechanical impedance is often the major limitation associated with soil compaction. The elongation rates of both main axes and lateral roots may be decreased, unless small pores in the soil allow fine lateral roots to proliferate (Goss 1977). Due to the slower elongation rate of the main root axes, the number of lateral roots per unit length of the main axis will tend to increase in compacted soils. Roots thicken and follow a more tortuous path in more compacted soils (Konopka et al. 2009). Root depth is often restricted on former-mining sites, and operations such as deep ripping may be required to enable tree roots to penetrate deeper into dry profiles to obtain sufficient water (Szota et al. 2007). Species differences exist in the extent to which trees can root into compacted soils, with Corsican Pine (Pinus nigra Arnold.) showing superior rooting than Italian alder (Alnus cordata Loisel.), Japanese larch (Larix kaempferi Lindl.) and birch (Betula pendula Roth.; Sinnett et al. 2008). Plants may exhibit compensatory growth as they respond plastically to localised areas of compaction, this response varying between species (Bingham and Bengough 2003). Physical limitations to root growth can be thought of in terms of the Least Limiting Water Range (Bengough et al. 2006; da Silva et al. 1994), which defines critical values whereby root growth will be significantly limited if exceeded. For matric potential, soils drier than −1.5 MPa will severely limit root growth and in such dry soils mechanical impedance will also limit root growth. The limit chosen for mechanical impedance is a penetrometer resistance of 2 MPa whilst aeration is likely to be limiting if there is less than 0.1 cm3 cm−3 air filled porosity in a region of soil.

Root growth on hillslopes

Landslide engineers are particularly interested in how plants grow on slopes. Several studies have been carried out on the direction of root growth of plants growing on sloping terrain. Chiatante et al. (2003) argue that in Quercus cerris L., preferential root growth occurs either up- or downslope, thus enhancing anchorage along the axis of static mechanical loading. Mullen et al. (2005) also showed that in Arabidopsis seedlings inclined at different angles, lateral root growth was strongly controlled by gravity; although roots emerged on the upper side of the taproot, their direction of growth was quickly altered and they curved downwards, towards the lower side of the taproot, resulting in a significantly higher number of lateral roots downslope. However, in other species, it has been found that preferential growth of lateral roots occurred upslope or even perpendicular to the slope direction (Nicoll et al. 2006; Khuder et al. 2006; Khuder 2007). Nevertheless, although lateral roots of seedlings might grow preferentially in a given direction, as they elongate they become less sensitive to gravity (Kiss et al. 2002). At the tissue level, differences have also been observed. In 8 month old Spartium junceum L., a significantly higher lignin content was found in root systems growing on slopes compared to those growing on flat ground (Scippa et al. 2006). As lignin and cellulose content are usually inversely proportional, more lignin means less cellulose. Lignin is resistant in compression, and it is not known why it should be present in higher quantities in roots growing on slopes. Differences in root tissue structure will influence tensile strength (Genet et al. 2005) and in trees growing on slopes, tensile strength was found to be greater in upslope roots compared with downslope and horizontal lateral roots (Schiechtl 1980), although similar studies have shown no differences between roots from young Robinia pseudoacacia L. grown on a slope in controlled glasshouse conditions (Khuder 2007).

The diversity of root responses to slope angle observed thus probably depends on several factors including species, experimental conditions and plant age. For example, Di Iorio et al. (2008) found that in S. junceum, soil type had a greater influence than slope angle on the distribution of surface roots. Surface roots were more numerous in clay than in loam soil, independent of the slope angle, and grew upslope in clay soil only. Stone and Kalisz (1991) suggest that upslope laterals grow obliquely or horizontally into weathered joints of bedrock and can attain great lengths. Nutrient and water availability also play a major role: Pierret et al. (2007) found that in fields on sloping terrain in Laos, both crop (rice—Oryza sativa L. and Job’s tear—Coix lacryma-jobi L.) and fallow vegetation (15 native herbaceous species) developed more shallow root length per unit soil volume when growing in gentle than steep slopes. In conclusion, it is not possible to say that roots grow preferentially up- or downslope; the combined effects of all interacting factors must be taken into account.

Few studies have been carried out on how root traits change with altitude, and this topic is of much interest to landslide engineers. In the few studies existing, high altitude grasses in New Zealand were found to have thicker roots and smaller SRL than plants at low altitude (Craine and Lee 2003). However, Körner and Renhardt (1987) found that SRL was greater in herbaceous dicots at higher elevations. An investigation into the spatial arrangement of roots within the root system, and how their distribution affects resource uptake, may help explain these differences in results. Soethe et al. (2006) examined root architecture of several tree species growing at two different altitudes in Ecuador. At the higher altitude, soil depth was restricted and root extension significantly greater, resulting in a wide-spreading but shallow root-soil plate. The increase in root plate width would help compensate for the shallowness of the root system, thereby maintaining anchorage ability. Genet et al. (2006) found that root tensile strength decreased at very high altitudes in Abies georgii var. smithii., although the functional significance of this decrease is not yet clear. It could be that there is less investment in roots at high altitudes, which have been found to have a faster turnover rate (Graefe et al. 2008).

Plasticity in root traits can therefore be highly diverse, and to add to their complexity, responses may be quite different depending on whether a plant is grown individually, as a monoculture, or with a mixture of species (Huber-Sannwald et al. 1996, 1997). Not only do traits change over time because of their inherent genetic expression (Barthélémy and Caraglio 2007), but depending on the frequency and intensity of temporal signals, e.g. nutrient availability and rainfall events, morphological and physiological responses will differ within a plant root system (Hodge 2004).

How do plant root traits change with time?

At the beginning of this paper, we said that a trait is defined as a distinct, quantitative property of organisms. However, it must not be forgotten that traits are not static properties and that they evolve over time as the plant matures (Barthélémy and Caraglio 2007). The nature of a trait will thus depend on endogenous growth processes and be influenced by exogenous constraints.

Developmental stages in woody root systems

From germination to death, successive developmental stages can be defined according to the nature and topology of the different root categories developed in the branched system (Fig. 3). Tree species have complex rooting forms and can undergo reiterative branching processes which influence root distribution and interactions with the local environment. One such process leads to the formation of forks (Fig. 4) which pin more superficial roots to the underlying soil substrate. Forks appear when the growth potential of one lateral root changes as a shift occurs in the mother root’s functioning. Horizontal perennial roots can build up two kinds of forks; those near the taproot are often made up of one or two homologous horizontal, perennial roots and one vertical tap root which branches from the lower surface of the fork, away from the trunk. Forks give rise only to horizontal roots, no vertical members can develop. This branching pattern allows the root system to radiate widely along the soil surface (Fig. 4, Atger and Edelin 1994a, b; Edelin and Atger 1994). A second type of reiterative branching pattern can occur, which involves the outgrowth of a new lateral root (neoformation) or the activation of a dormant lateral root on an old part of the root system, e.g. after wounding or pruning.

Fine root dynamics

Although many data are available on the subject, engineers rarely consider the dynamics of fine root production in models of landslide risk. Fine root production reflects the annual changes that can occur belowground and how a system responds to disturbance (Vogt et al. 1996). Usually, only roots <3 mm in diameter are considered, which are those having the highest tensile strength (Bischetti et al. 2005, 2009a; Genet et al. 2005), and thus of great interest to landslide engineers. The biomass of thick roots is less affected by environmental change and so is fairly static over the short-term.

Hendricks et al. (2006) showed that the method by which fine root production is estimated is of upmost importance. Fine root biomass was measured using nitrogen and carbon budget techniques, standard soil cores, ingrowth cores (soil cores are removed and the soil replaced with root-free soil into which new roots can grow) and minirhizotrons (a transparent cylindrical tube inserted into the soil and down which a camera is inserted). By comparing the different methods in three forest types, Hendricks et al. (2006) demonstrated that annual fine root biomass values ranged from 0 to 4,618 kg ha−1 year−1 when production was measured from soil cores and by using minirhizotrons, respectively. The minirhizotron technique provided the most reliable data and this study underlines the shortcomings in the various root sampling techniques. As discussed earlier, root architecture is highly variable and plastic responses are many. Soil cores therefore need to be large and numerous enough to capture the heterogeneity in fine root production across a slope (Genet et al. 2008). Nevertheless, temporal changes in dynamics throughout a year will only be achieved by regular sampling or through using minirhizotrons or large scale rhizotrons. The latter consist of panes of glass pressed against the soil surface in a trench dug into the slope (e.g. Thongo M’bou et al. 2008). Root initiation, elongation rates and mortality can thus be observed as frequently as desired. In a study of Eucalyptus clones in Congo, seasonal dynamics were observed in such a rhizotron and it was found that fine roots grew continuously throughout the year, but that a significant decrease in elongation rates occurred at the end of the dry season (Thongo M’bou et al. 2008). These authors could also follow root growth dynamics with regard to depth and correlate elongation rates with water variation in the soil profile.

Integrating information for the management of mountain slopes

Natural regeneration or planting mixtures?

Surprisingly, the distribution of trees, and hence their root systems, in forests prone to landslides has been studied little (Schmidt et al. 2001; Sakals and Sidle 2004; Kokutse et al. 2006; Danjon et al. 2008; Genet et al. 2008, 2009) and yet is of major importance as landslides tend to occur in areas of reduced root distribution (Roering et al. 2003), where gaps occur between trees. Genet et al. (2008) studied root reinforcement in plantations of Cryptomeria japonica D. Don. of three different ages and spatial densities. It was found that in older plantations, where trees had been thinned and large gaps occurred between trees, the greater the number of trees in given transects and the smaller the average distance between groups of trees, the higher the FOS. No relationships between FOS and stand structure were observed in the youngest plantation, due to the high stand density and low spacing between trees, resulting in a more homogeneous structure. Sakals and Sidle (2004) proposed that model simulations of root cohesion over time and space would allow foresters to determine how gap formation between trees would affect slope stability. Similarly, Danjon et al. (2008) suggested using maps of root cohesion in forest stands to establish appropriate types of planting, e.g. rows or staggered lines of trees, depending on the FOS of the slope. Simulating changes in root architecture using plant growth models (Jourdan et al. 1995; Dupuy et al. 2005c; Drouet and Pagès 2007) could in the future be combined with slope stability models (Kokutse et al. 2006) and is an area where much promising research could be carried out.

Once e.g. a clear-felling or a fire has occurred on a slope, the site will be relatively fragile until new vegetation colonizes the soil. Although the roots remaining in the soil provide a certain amount of strength for the early years (e.g. Watson et al. 1999; Sidle et al. 2006; Jackson and Roering 2009), a forester has to decide how best to manage the site depending on the long-term goal. If not under the obligation to plant fast-growing monocultures of timber species, natural regeneration of the local vegetation may be a solution. Very little information exists on the slope failure risk associated with natural regeneration, although higher species diversity has been shown to improve soil aggregate stability (Pohl et al. 2009). Genet et al. (2009) showed that in the first phase of succession in naturally regenerated slopes in the Sichuan, China, when big node bamboo (Phyllostachys nidularia Munro.) was the dominant species, there was no contribution of vegetation to the FOS due to the shallow rhizomatous rooting nature of big node bamboo. The FOS did increase over time, as deeper rooting deciduous tree species colonised the sites studied. Similarly, Roering et al. (2003) observed that where gaps occurred in a naturally regenerated, temperate forest, sword fern (Polystichum munitum Kaulf.) was present and always found in the vicinity of the landslide scarp. Schmidt et al. (2001) discuss the possibility of planting conifer seedlings immediately after clear-felling, to reduce the window of landslide hazard which occurs due to subsequent root decay and hydrologic response. The duration of this window varies from 15 to 25 years and is highly site and species specific (Ziemer 1981; Watson et al. 1999). For example, Cammeraat et al. (2005) showed that in abandoned bench terraced slopes at different phases of vegetation succession, mass wasting processes increase after abandonment, due to the limited contribution of anchorage by roots at potential slip planes and the fast transfer of rainfall to the potential slip plane. It took up to 40 years for mass wasting process to decline. If natural regeneration is the choice of the landslide engineer, rather than restoration or prevention of landslides through plantations, then promotion of the recovery of self-sustaining communities is feasible by stabilization with native ground cover, applications of nutrient amendments and facilitation of dispersal to overcome establishment bottlenecks (Walker et al. 2009).

Regeneration strategies

The ability of a plant to resprout after a disturbance is an important consideration to be made in the management or restoration process. Resprouting is the ability of a plant to produce a new shoot from a root or stem if part of the plant’s biomass has been removed through a disturbance event such as grazing, fire, windstorms, avalanche etc (Bond and Midgeley 2001). Not all species possess this ability. Annual species usually cannot resprout as resources need to be allocated and stocked in storage organs. As construction costs are higher for such organs, these costs need to be weighed against the benefit that a plant will gain after a disturbance. If a plant can resprout, it will be able to survive in the newly disturbed environment, even when above-ground biomass is destroyed (Stokes et al. 2007). Very few studies have been carried out on the influence of edaphic factors on resprouting ability (Sakai et al. 1997). Guerrero-Campo et al. (2006) carried out a study of 123 commonly occurring plant species found on eroded land in north-east Spain. These authors found that species with coarse and deep tap-roots tended to be root-sprouting and those with fine, fasciculate and long main roots (which generally spread laterally), tended to be shoot-rooting. Landslide engineers can find information about such traits useful when deciding on management strategies with regard to coppicing. However, Sati and Sundriyal (2007) suggested that Alnus nepalensis D. Don can promote landslide activity after resprouting. If the roots of this species are exposed through soil loss, resprouting occurs which initially reduces soil erosion. The roots of several individuals then form an interlocking matrix which acts as a single unit. As the plants mature, the weight of this root-soil matrix increases and can result in a downward movement of the integrated soil mass. A similar phenomenon has also been observed in the rhizomatous big node bamboo, which resprouts from the running rhizomes located just underneath the soil surface (Stokes et al. 2007). Vogt et al. (2006) showed that abandoned stands of coppiced Sweet chestnut (Castanea sativa Mill.) became more likely to uproot as trees became taller and more mechanically unstable, particularly when growing in pits or steep depressions. On steep slopes, uprooted trees are likely to fall downslope, transporting sediment attached to the root plate and even resulting in soil mass movement (Norman et al. 1995). After clear-felling of a coppiced stand, it is not known if root decay occurs, thus reducing soil reinforcement, and should be further investigated. A careful management of coppiced stands is therefore required to meet the balance between the protection of surface soil through erosion and a deeper reinforcement of the slope.

Choice of planting stock

If seeding or natural regeneration of vegetation at a site is not an option and a site is to be planted with container seedlings, the landslide engineer should use the best quality planting stock available, to avoid seedling mortality and poor growth when plants are planted out. To promote lateral root growth in container seedlings, different types of pots exist, e.g. slitwall containers have open slits down the sides of the container so that roots can grow through the slits (Rune 2003). Ribwall containers have solid ribbed walls along which roots grow, thus avoiding root spiralling, but root area is less in these pots and spiralling more frequent than in slitwall containers (Rune 2003). Root spiralling occurs when lateral root growth is restricted by the pot size and roots grow around the pot walls, forming a root ball. A symmetrical root system shape with unrestricted lateral and vertical roots would fix soil earlier and more efficiently than seedlings with root balls. Container-grown trees can also have a significantly different root architecture to naturally regenerated seedlings, due largely to the availability of nutrient and water resources in nursery conditions, e.g. naturally regenerated Quercus ilex were found to possess herringbone-like root systems whereas container seedlings had more randomly branched root systems (Tsakaldimi et al. 2009). Many different techniques exist to manipulate root systems during the juvenile stage of growth, e.g. by root pruning before transplanting, thus changing the shape of the existing root system and affecting the subsequent branching pattern when roots regenerate from the pruned areas (Girouard 1995). Cuttings can also have a significantly different root architecture to seedlings. Seedlings generally possess their original taproot whereas cuttings do not necessarily, and only adventitious roots can be present, depending on the parent material of the cutting (Charrier 1969: Sasse and Sands 1997). Live pole cuttings are often used to stabilize slopes and river banks (Norris et al. 2008) and also develop an adventitious root system with the mechanical function of the pole analogous to a large taproot. It was found that not only does root system development depend on soil type, but the anchorage of adult Populus sp grown from live poles depends on whether soil is frictional (sandy) or cohesive (clayey), with anchorage improved in the latter (Dupuy et al. 2007).

The engineer must therefore carefully select planting stock, as seedlings grown in ideal nursery conditions and a particular soil type may not be adapted to stressful field conditions. In such cases, root growth of healthy seedlings can be arrested when substrate conditions change drastically. Before planting out therefore, it would thus be wise to determine how the most limiting factors to growth at the given site affect root traits. A list of desirable root traits for the landslide engineer is given in Table 2.

Perspectives for future research

One of the current areas where more information is needed is the development of a simple yet elegant slope stability model which takes into account the impact of vegetation. User-friendly softwares, e.g. Slip4ex (Greenwood 2006), exist which calculate the FOS of a slope, but rely on data from Wu’s (1976) and Waldron’s (1977) models of additional cohesion, which overestimate the cohesion provided by roots. The FBM (Pollen and Simon 2005; Pollen 2007; Pollen-Bankhead and Simon 2009) is highly promising but requires further development with regard to the spatial arrangement of roots in soil along with their mechanical properties. It is unlikely that a simple model can consider 3D root architecture, therefore parallel studies using techniques like FEM need to be carried out to determine how the position of roots in a block of soil influence its shear strength, taking into account root size, mechanical properties and soil physical properties.

Most of the existing slope stability models which consider vegetation are of a static nature (but see Pollen 2007), i.e. they just consider the distribution of roots at a given time, nevertheless, such models are still useful for estimating landslide risk for a given slope configuration. Exploring the effect of reforestation scenarios or more generally the impact of slope management can be performed making assumptions on the expected rooting patterns and using the previously mentioned static models (see Dhakal and Sidle 2003 as an example of considering root decay and regrowth after harvesting). However, coupling growth models, in particular based on structural-functional notions (Fourcaud et al. 2008), with soil, climate and hydrological models is an exciting challenge to answer the question of the consequences of climate change on substrate mass movement. Such models can also take into account the temporal nature of root dynamics and trait expression.

How root systems interlock with rocks and coarse fragments in the soil is not quantified, but is a character which could be manipulated by the landslide engineer to enhance the lateral membrane strength of the soil matrix component (Fig. 1a). We also need more information on the properties of the interface zone between roots and soil and how this changes friction along the root surface. Roots have rough surfaces due to the presence of bark, roots hairs and exudates (Mickovski et al. 2007). The latter can influence soil aggregate and rhizosheath formation which is also an important aspect for the landslide engineer to consider (Pohl et al. 2009). How root tensile strength changes along a root, within a root system, with plant and root age and how it is influenced by the soil environment, needs to be elucidated along with its interaction with other traits and plant form and history.

It is particularly important that we develop better ways of measuring root distribution with depth. If root traits could be predicted from aboveground measurements or characteristics, it would be possible to map the contribution of vegetation to slope stability and thus highlight zones of potential failure (Roering et al. 2003; Preti et al. 2009). This type of mapping could also consider species diversity and the effect of different species mixtures on slope stability. How root systems grow when planted individually or in competition with other species differs, and soil exploration and exploitation can be highly influenced by a diversity of root forms (Fig. 7). More work needs carrying out therefore on plant diversity with regard to slope stability, as well as how to plant and manage along a slope, taking into account local variations in topography and ecology (Genet et al. 2009). To this end, the development of expert systems and species databases would be very useful, especially open source systems where data can be added freely by end-users and researchers (Norris et al. 2008).
Fig. 7

Illustration of rooting types and depths for seven herbaceous species growing at 470 m altitude, Klagenfurt, Austria. From left to right, Dactylis glomerata L., Knautia arvensis L., Arrhenatherum elatius L., Pastinaca sativa L., Bromus hordeaceus L., Carum carvi L., Holcus lanatus L. and Crepis biennis L.. Where species’ mixtures are diverse, plant root systems occupy space differently than if grown individually or as a monoculture, e.g. the root system of K. arvensis exploits soil horizontally beneath the root systems of its neighbours. By increasing the diversity of root forms, the occupation of soil will be increased, and hence the overall fixation of soil along a slope (reprinted with permission from the publisher; see Kutschera and Lichtenegger 1997)

With regard to slope stability in a changing climate, the impacts of potential climate change scenarios outlined in the recent Fourth Assessment Report by the Intergovernmental Panel on Climate Change (Cruz et al. 2007) on landslide and severe surface erosion are mixed. For slow, deep-seated landslides, the likelihood that future climate change will decrease susceptibility or movement rates is actually greater than cases where rates would increase (Sidle and Burt 2009). Shallow landslides are likely to increase in a warmer climate if the local frequency of intense storms increases. Since such landslides are more influenced by rooting strength, modification of vegetation cover and harvesting methods in steep, managed forests may be required to ensure stable slopes in a changing climate. However, before such expensive measures are implemented on a widespread basis, better predictions of the spatial changes in rainfall response are needed than are currently available from climate change models. For landslides and severe surface erosion, more focus needs to be placed on sustainable anthropogenic practices, which are much more manageable in the context of current knowledge and practical policy-making decisions, than the uncertain impacts of global warming.

Climate change can also affect the expression of root traits; the increased carbon gain under elevated CO2 may augment RLD and mycorrhizal colonization, whilst at the same time decreasing tissue nitrogen concentrations (Eissenstat et al. 2000). Tree wood strength may increase but genetic variation is great (Beismann et al. 2002). Increases in soil temperature and changes in precipitation patterns will affect root growth dynamics. Therefore, we need more understanding of how the expression of traits and their interactions are affected by the local environment, in particular soil temperature and precipitation. At the moment it is not possible to say with any certainty how traits will be affected because responses of soil to changes in climate will be localized. One thing is certain though, once a landslide has occurred through either land use mismanagement or the consequences of climate change, the replacement of soil on the denuded slope can take up to thousands of years through natural processes. In a world where the population is expected to reach 9 billion by 2040 (, agricultural soil is precious and hillslope stability should be a priority for governments needing to feed rapidly increasing populations.

In conclusion, it is clear that biologists and landslide engineers need to work together to develop long-term strategies for the ecological management of steep slopes (Norris et al. 2008). We have included many ideas, data and concepts presented at the Second International Conference ‘Ground Bio- and Eco-engineering: The Use of Vegetation to Improve Slope Stability—ICGBE2’ held at Beijing, China, 14–18 July 2008. Several papers from this conference are published in this edition of Plant and Soil (Bischetti et al. 2009a; Fan and Su 2009; Mickovski and van Beek 2009; Pohl et al. 2009; Reubens et al. 2009; Stangl et al. 2009; Tsakaldimi et al. 2009; Walker et al. 2009) and several more are published in a special edition of Ecological Engineering (Bathurst et al. 2009; Bischetti et al. 2009b; Genet et al. 2009; Hubble et al. 2009; Loades et al. 2009; Preti et al. 2009; Schwarz et al. 2009; Stokes et al. 2009 submitted; Wang et al. 2009; Zhang and Dong 2009). The third conference in this series will be held in Vancouver, Canada, 2012 and the first author of this paper can be contacted for further details.


Adventitious root development can be encountered in the first rooting type. Many tropical trees exhibit a very large cone of adventitious roots at the base of the trunk which reinforces the original primary root system. This type can be seen as a mixed type. In temperate species the adventitious development of root is often less spectacular because only the few first centimetres of the trunk base are concerned.



Funding for AS, CA and TF was received from INRA (Jeune Equipe), CNRS (EcoPente project) and Agropolis Fondation, Montpellier. AMAP (Botany and Computational Plant Architecture) is a joint research unit which associates CIRAD (UMR51), CNRS (UMR5120), INRA (UMR931), IRD (2M123), and Montpellier 2 University (UM27). The Scottish Crop Research Institute receives grant-in-aid support from the Scottish Government Rural and Environment Research and Analysis Directorate. Thanks are due to the Chinese Academy of Sciences and LIAMA, in particular B. Hong and X. Zhang, for their help in the organisation of the ICGBE2 conference.

Copyright information

© Springer Science+Business Media B.V. 2009