- Email: [email protected]
Evaluating the influence of forest roads on shallow landsliding
Ecological Modelling 187 (2005) 85–98
Evaluating the influence of forest roads on shallow landsliding Marco Borga a,∗ , Fabrizio Tonelli a , Giancarlo dalla Fontana a , Federico Cazorzi b a
Department of Land and AgroForest Environments, AGRIPOLIS, University of Padova, via dell’Universit`a, 16, Legnaro, IT-35020, Italy b Department of Agricultural and Environmental Science, Polo Scientifico Rizzi, University of Udine, IT-33100 Udine, Italy Available online 25 March 2005
Abstract This study investigates how subsurface flowpaths are altered by forest roads and how these changes influence shallow landsliding susceptibility in steep, forested landscape. A simple conceptual model of the effect of forest roads on hillslope subsurface flow is developed. The model is incorporated into a hydro-geomechanical, threshold-based model for slope instability. In the model, the occurrence of shallow landsliding is evaluated in terms of drainage areas, ground slope and soil properties (i.e., hydraulic conductivity, bulk density, and friction angle). Model results allow to quantify the influence of roads on shallow landsliding hazard across a landscape and to generate hypotheses about the broader geomorphic effect of roads. Modelling results are compared with field data collected in four sites located in north-eastern Italy. Observed landslide patterns are broadly consistent with model estimates, a finding that underscores the utility of this simple approach for predicting the geomorphic effects of forest roads constructed on steep slopes. The approach used in this study may be useful for defining criteria for road design that reduce the effects of roads on geomorphic processes. © 2005 Elsevier B.V. All rights reserved. Keywords: Anthropogenic effects; Runoff and streamflow; Forest roads; Shallow landsliding; Subsurface Flow; Hillslope hydrology
1. Introduction The interaction between forest roads and geomorphic processes lies at the heart of several key issues concerning the effects of roads on the environment. Geomorphic effects of forest roads range from chronic and long-term contributions of fine sediment into stream to catastrophic effects associated to shallow landsliding during large storms. In steep, forested terrain prone ∗ Corresponding author. Tel.: +39 049 8272681; fax: +39 049 8272686. E-mail address: [email protected] (M. Borga).
0304-3800/$ – see front matter © 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.ecolmodel.2005.01.055
to landsliding, the greatest effect of roads on erosion rates is from increased rates of mass soil movement after road building (Gucinski et al., 2001). Major issues motivating concern about road-related erosion include potential degradation of aquatic habitat and water quality (Harr and Nichols, 1993) and risks to public safety and structures downstream (Burroughs, 1985; Pozzatti and Cerato, 1984). The magnitude of road-related mass erosion differs with climate, geology, topography, road age, construction practices and storm history (Gucinski et al., 2001). Several inventories have been conducted to assess road effects on mass failures, with more specific fo-
86
M. Borga et al. / Ecological Modelling 187 (2005) 85–98
cus on U.S. Pacific Northwest and New Zealand (Sidle et al., 1985; Swanson and Dyrness, 1975; Reid, 1981; Mosley, 1980; Coker and Fahey, 1993), each documenting increased rates of landsliding in road areas relative to unmanaged forested areas. Sidle et al. (1985) documented accelerated erosion rates from roads because of debris slides ranging from 30 to 300 times the forest rate. Influence of forest roads on geomorphic processes generally results from concentration of both runoff generated as overland flow from compacted road surfaces and intercepted subsurface flow by road cutslopes (Megahan, 1972; Anderson, 1983; Reid and Dunne, 1984; Luce and Cundy, 1994; Montgomery, 1994; Ziegler and Giambelluca, 1997; Luce and Black, 1999; Wemple et al., 2001). Subsurface storm flow dominates runoff generation in steep soil-mantled terrain where precipitation infiltrates and flows laterally either through macropores or over a lower conductivity zone. In these environments, road interception occurs when a seasonally high water table flowing over on impermeable base (e.g. bedrock) becomes deep enough to intersect the road ditch. Thus the fraction of the permeable soil occupied by the road cut becomes a controlling factor in the amount of interception. Subsurface flow intercepted along the road cut may be diverted to surface runoff, and then redirected downslope, modifying pre-existing flow paths on the hillslope. In steep, soilmantled terrain, the combination of ubiquitous subsurface flow and of the interception by road cuts yields an influence on hydrogeomorphic processes that may be much greater than one might expect from the small fraction of the land area roads occupy (Luce and Wemple, 2001). Road-generated runoff is usually routed by roadside ditches and therefore concentrated in particular areas below the road. The effect of this concentration of flow will depend on the characteristics of the receiving areas. When the road drainage is not connected to the stream network (at least through surface flow), road runoff may either reinfiltrate into the unchanneled terrain, or reinfiltrate below a gully that does not extend to the stream network. In these cases, concentrated road runoff may affect shallow landsliding potential in the receiving areas and decrease the critical source area required to initiate headwater streams (Montgomery, 1994). In cases where the road network is connected to the stream network road, runoff may enter a stream di-
rectly at a stream crossing culvert, or enter a stream indirectly through the formation of a gully, extending to the river network, below a ditch relief culvert. Therefore, the collective contribution of intercepted hillslopes to the road and the road surface drainage features determine the road impact on mass wasting on hillsides downslope. This impact may be large when roads intercept large amounts of subsurface flow and redirect it to unchanneled terrain, conditionally unstable, below the road. The purpose of this work is to gain further insight on road interactions with hillslope flow paths and on how these interactions influence shallow landsliding of concerned hillslopes. This paper addresses these issues by developing a conceptual and quantitative framework for evaluating how roads in different landscape positions (valley bottom, midslope, ridgetop) affect subsurface flow paths and associated mass failures. A threshold-based model for slope instability (Dietrich et al., 1993; Montgomery and Dietrich, 1994; Borga et al., 1998, 2002a) is extended to the case of hillslopes interested by road networks. A range of scenarios is used to generate hypotheses about the effects of road on hillslope stability. Simulation results are also compared with available field data collected in north-eastern Italy in order to assess how well the model captures the processes of interest.
2. A conceptual model of cutslope interception, throughflow rerouting and slope instability The conceptual model of cutslope interception and throughflow rerouting used in this study is based on coupling of digital topography with a simple model of steady-state rainfall-runoff to calculate the saturation deficit at any point in the landscape. The following assumptions are used to model the subsurface flow propagation: • Shallow lateral subsurface flow follows topographic gradient. • The entire soil profile is initially wet to field capacity. • Lateral discharge at each point is in equilibrium with a steady-state recharge R [LT−1 ], i.e. the infiltrating rainfall which passes through (or bypasses) the unsaturated zone to reach the saturated zone and eventually becomes subsurface runoff. The steady-state
M. Borga et al. / Ecological Modelling 187 (2005) 85–98
assumption implies that the specific upslope area (measured along the horizontal plane) is a surrogate measure of the subsurface flow at any point in the landscape. • The recharge rate equals the rainfall rate, thereby neglecting the vertical transport processes taking place between rainfall reaching the ground surface and recharge occurring at the soil base. Rainfall infiltrates quickly into the highly permeable soils dominated by subsurface runoff and this approximation should be adequate. • The capacity for lateral flux at each point is T sin θ, where T is the soil transmissivity [L2 T−1 ] and θ the surface slope. Furthermore, the approach with the hydrological model is to interpret the soil thickness (h [L]) as specified perpendicular to the slope. Analogously, hw [L] is the thickness of the saturated region above the bedrock, and drc [L] is the depth of the road cut base above the bedrock, both measured perpendicular to the slope (Fig. 1). Relative road cut depth (rrc ) is computed as drc /h. Assumptions above imply that depth integrated subsurface flow per unit contour length q [L2 T−1 ] is q=R
A b
(1)
where A [L2 ] is the upslope area and b [L] the length across which flow is accounted for. Furthermore, the relative wetness (for a generic pixel not influenced by road-rerouted subsurface flow) w=
hw h
(2)
Fig. 1. Schematic diagram illustrating the interaction of the hillslope water table with the road cut and the variables used in the model description.
may be expressed as RA w = Min ,1 Tb sin θ
87
(3)
The relative wetness has an upper bound of 1 with any excess assumed to form overland flow. Implementation of road network effects makes the following assumptions: • the amount of subsurface flow intercepted by the road is a linear function of the elevation hi of the water table relative to the base of the road cut, calculated as hi = hw − drc ; • the intercepted flow is redirected by the road in-board ditch, based on the road slope, while non-intercepted flow follows natural flowpaths on hillsides downslope; • ditch discharge is diverted by cross-ditches and relief culverts toward areas below the road, where road runoff may either reinfiltrate (where the receiving terrain is unchanneled) or enter a stream (where the road drainage is directly linked to the river network). This model of cutslope interception and throughflow rerouting has been broadly applied to study the effect of forest roads on watershed hydrology (Wemple et al., 1996; Tague and Band, 2001; Wigmosta and Perkins, 2001; Wemple and Jones, 2003) and its application on subsurface-flow dominated watersheds is supported by a vast amount of observations (Megahan, 1972; Jones, 2000; Wemple and Jones, 2003). Note that this approach neglects the drawdown effects of the seepage face on the water table (Atkinson, 1978), which may lead to overestimates of water table elevation in these calculations. This effect is likely small relative to other potential sources of error in the approach used here. For each pixel influenced by the road network, three upslope contributing areas are considered: (1) the portion of the contributing area located upslope from the road, which contributes according to the road cut depth and the current saturation deficit (A1 ); (2) the contributing area related to the subsurface discharge intercepted by the road cut depth as a function of the current saturation deficit and rerouted to the pixel by the road (A2 ); (3) the portion of the contributing area located downslope the road (A3 ).
88
M. Borga et al. / Ecological Modelling 187 (2005) 85–98
In the next section it is shown how the model accounts for the road cut interception effect in the computation of the contributing area. 2.1. The threshold model for slope instability
cept of critical rainfall, introduced by Montgomery and Dietrich (1994), is worth using. The critical rainfall is the minimum steady-state rainfall predicted to cause instability. Coupling between the slope stability model and the steady-state hydrologic model illustrated above allows to derive the critical rainfall (Montgomery and Dietrich, 1994). Based on Eqs. (3) and (4), and for a terrain element not influenced by road runoff, critical rainfall is defined as Tb sin θ Cr + Cs ρs W R= + + A ρw gh cos θ tan φ ρw ρw gh tan θ × 1− (5) tan φ
Planar infinite-slope analysis has been widely applied to the determination of natural slope stability, particularly where the thickness of the soil mantle is small compared with the slope length and where landslides are due to the failure of a soil mantle that overlies a sloping drainage barrier. The drainage barrier may be bedrock or a denser soil mass. The infinite-slope stability model provides a onedimensional model for failure of shallow soils that neglects arching and lateral root reinforcement. Because The model application is based on the assumption of the geometry of an infinite slope, overall stability that differences in geology impose a broad control on can be determined by analysing the stability of a single, absolute rates of shallow landsliding upon which supervertical element in the slope. End effects in the sliding imposed topographic controls (contributing area and mass can be neglected, along with lateral forces on eislope) dominate the relative slide frequency. The reader ther side of the vertical element, which are assumed is referred to Borga et al. (2002b) for an analysis of the to be opposite and equal. Under these assumptions the critical rainfall concept and its relation with the steadyfactor of safety (FS) for a vegetated slope with slopestate assumption used in the model. parallel seepage, simplified for wet and dry soil density the same, is computed as follows: Cr + Cs + ρs g(h − hw ) cos θ + (ρs g − ρw g)hw cos θ + W cos θ tan φ FS = (4) hρs g sin θ + W sin θ where Cs and Cr are soil and root cohesion [ML−1 T−2 ], respectively, φ is the internal friction angle of the soil [degrees], ρs is wet soil density [ML−3 ], ρw is the density of water [ML−3 ], g is gravitational acceleration and W is the vegetation surcharge [ML−1 T−2 ]. Since steady-state flow conditions are assumed, effective stress parameters (effective friction angle and effective cohesion) must be used in the analysis. Inclusion of root cohesion and vegetation surcharge into stability analysis allows one to develop a more complete analysis of forest practices, combining road construction and logging activities. Indeed, in potentially unstable areas, shallow failures may increase after trees are cut, as their root systems progressively decay. In several cases, only two factor of safety classes (FS ≥ 1 and FS < 1) are deemed enough to estimate the susceptibility of a landscape to shallow landsliding, and analysis of stability in relation to a range of precipitation values is required. In these situations, the con-
The stability of terrain elements whose contributing area includes road rerouting is potentially affected by road runoff. For these elements, the computation of road cut interception effect (hence contributing area) is function of the unknown critical rainfall. Indeed, road cut interception is function of both relative road cut depth (rrc ) and of the local saturation deficit (i.e., of the critical rainfall, according to Eq. (3)). Owing to these reasons, a generic algorithm has been devised for the computation of the contributing areas. The algorithm requires estimation of relative road cut depth (rrc ) for each grid element associated to a road, and it is based on the computation of the interception effect for a number of discrete rainfall values. For each of these values, the contributing area for each element of the landscape is calculated. Then, Eq. (5) is solved iteratively for each element.
M. Borga et al. / Ecological Modelling 187 (2005) 85–98
The algorithm is based on the following steps: (i) for each discrete rainfall value and for each model grid associated to a road, the value of w is computed based on Eq. (3); (ii) for these grids, the λ1 and λ2 weighting coefficients are computed as follows: r
rc
,1 , w w − rrc λ2 = max , 0 = 1 − λ1 w
λ1 = min
(6)
where λ1 represents the portion of the contributing area corresponding to the fraction of subsurface flow which is not intercepted by the road cut and follows natural flowpaths on hillsides downslope, and λ2 represents the portion of the contributing area corresponding to the fraction of subsurface flow which is intercepted by the road cut and follows the artificial road drainage path. The contributing area for a grid element interested by a road is therefore made up of three parts: • the fraction of the contributing area located upslope from the road: Λ1 A 1 = λ1 a dl (7) Lr1
where Lr1 is the length of the road segment intersecting the contributing area of the given grid element; • the contributing area related to the subsurface discharge intercepted by the road cut depth and rerouted to the pixel by the road: Λ 2 A2 = λ2 a dl (8) Lr2
where Lr2 is the length of the road segment which contributes intercepted flow to the given grid element; and • A3 , which is the portion of the contributing area located downslope the road. At this point, for each discrete rainfall values and for each generic landscape pixel it is possible to compute the product R(Λ1 A1 + Λ2 A2 + A3 ). The value of the product which agrees with the following equality: R (Λ1 A1 + Λ2 A2 + A3 )
89
Cr + Cs + = Tb sin θ ρw gh cos θ tan φ tan θ × 1− tan φ
ρs W + ρw ρw gh
(9)
is identified iteratively. (Note that the term R(Λ1 A1 + Λ2 A2 + A3 ) is monotonic increasing with R, and this allows to find always a solution to the problem.) When roads cross repeatedly the same hillslope, the algorithm needs to be repeated once for each crossing. This allows the computation of critical rainfall for each landscape pixel. A more detailed derivation of the model and of the algorithms included is presented elsewhere (Borga et al., 2004). For a given threshold of critical rainfall, the area of the basin with critical rainfall less than the threshold will coincide with the area characterised by having a factor of safety (computed with reference to the threshold) less than unity. Based on the concept of critical rainfall, four stability classes can be defined: unconditionally unstable, unstable, stable and unconditionally stable (Montgomery and Dietrich, 1994). Ground is unconditionally unstable if it is unstable even when dry. The condition for unconditionally stable slopes is expressed by tan θ < +
1 1− W/ρw gh + ρs /ρw
tan φ
Cr + Cs ρs gh cos θ (1 + W/ρs gh)
(10)
Unconditionally stable elements are those predicted to be stable even when saturated. The condition for unconditionally unstable slopes is expressed by
tan θ ≥ tan φ +
Cr + Cs ρs gh cos θ (1 + W/ρs gh)
(11)
Unstable elements are those predicted to fail according to Eqs. (5) and (9). Only potentially unstable elements are affected by road-related drainage modification. The impact can be evaluated in relative terms by mapping the critical rainfall before and after drainage modification due to road
M. Borga et al. / Ecological Modelling 187 (2005) 85–98
90
Fig. 2. Location map for the four study areas.
network (Rh and Rr , respectively), and computing the relative impact statistic (RI) as follows: RI =
Rh − Rr Rh
surface flow and increase critical rainfall; in this case, RI < 0.
(12) 3. Study areas and model application
The RI statistic was selected because it is dimensionless and is easily interpreted. If the road network has no effects on subsurface runoff propagation (hence, on shallow landsliding), RI = 0. If the cross-road ditches concentrate and increase saturated subsurface throughflow, hence decreasing the critical rainfall, RI > 0. Alternatively, road drainage may reduce saturated sub-
The model was implemented and evaluated on four study areas in north-eastern Italy (Fig. 2): the Vendevolo hill, on the Euganean Hills close to Padova, and three sites on the upper Noce river valley (Fondo, Lavazz`e and Livo). Topographic and climatic characteristics of the study areas are reported in Table 1.
Table 1 Main morphometric and climatic characteristics of study areas
Elevation range of concerned hillslope (m a.s.l.) Position on slope Slope (◦ ) 10 year, 12 h rainfall (mm) Slope is reported as average value over each landslide scar.
Fondo
Lavazz`e
Livo
Vendevolo
1200–1300 Lower 1/3 25.0 61.0
1350–1450 Middle 1/3 40.0; 29.5 61.0
700–800 Lower 1/3 25.6; 27.0 63.0
300–400 Upper middle 1/3 39.2; 36.4 73.0
M. Borga et al. / Ecological Modelling 187 (2005) 85–98
Road-related mass failures occurred in each of these sites, thus providing a basis for model evaluation in landscapes characterised by different climate, vegetation, geology and topographic structures. Field work comprised: road alignments and drainage mapping; landslide mapping; measurement of the characteristics of the road segment, such as slope, width and cutbank depth, and of road runoff drainage structures; post-event analysis of occurrences of culvert filling by sediments eroded along the road surface during storm events. Road drainage patterns and the associated drainage diversions were mapped in the field. This mapping phase may be highly uncertain, due to the different topographic features influencing road drainage propagation. To reduce these sources of uncertainty, road surface drainage patterns were mapped during and after runoff-producing events. Road surface widths measured in the field range from 3 to 6 m; max road slopes range around 20%. As it is typical in these areas, road drainage structures are represented by in-board ditches (though rarely used) and cross-road ditches. Cross-road ditches are excavated across the road at an angle and at sufficient depth, with armouring as appropriate, to divert both road surface water and ditch water off or across the road. The roads are of cut-and-fill design, consisting of a cutbank, a crowned road surface, and a fill slope. Hillslopes contributing to road segments are generally steep, ranging from 25◦ (Fondo) to 40◦ (Lavazz`e). Soils are shallow, ranging in depth from 0.2 m (Vendevolo and Lavazz`e) to over 1.5 m (Fondo and Livo), while cutbank depths ranges from 0 (some road segments for the Vendevolo case) to several meters (road cuts on bedrock at Lavazz`e). A sequence of events typical of major flood events in north-eastern Italy (prolonged rainfall, rain-on-snow events), occurred on autumn 2000, triggered the shallow landslides at Fondo, Lavazz`e and Livo. A 200-year return time, flash-flood producing storm (occurred on May 31, 1995) triggered the mass failures at Vendevolo. Inventoried shallow failures were generally originated by excessive road runoff focused on previously unchanneled hillsides below the road and thereby reinfiltrated. Neither fillslope slides or cutslope slides were considered in this study. Excessive road runoff was in most of the cases associated to disruption of road drainage structures, which were plugged by sedi-
91
ments transported by road runoff. Volumes of mass failures (estimated from measurements of the length, width and depth of the scar features) ranged from 50–60 m3 (Lavazz`e) to 3000 m3 (Fondo). For the three cases of Livo, Lavazz`e and Vendevolo, forest vegetation tended to buffer the flow of debris and sediment so that it did not reach large streams. For the Fondo case study, the material from the slide entrained and deposited along the channel bed, causing a temporary diversion of the river. 3.1. Study area description Fondo, Lavazz`e and Livo are characterised by an alpine climate, with a mean annual rainfall ranging from 1000 to 1300 mm. Precipitation occurs mainly as snowfall from November to April. Runoff is usually dominated by snowmelt in May and June but summer and early autumn floods represent an important contribution to the flow regime. The climate of the Vendevolo hill area is subMediterranean, with two precipitation peaks in spring and autumn. Mean annual rainfall amounts to 800 mm. On Fondo and Lavazz`e sites vegetation cover consists mainly of forest stands made up by spruce and larch; on Livo, vegetation is dominated by conifers and contains some deciduous species. Chestnut stands and shrubs represent main vegetation on the Vendevolo. Even though Fondo, Lavazz`e and Livo sites are just 15–20 km away from each other, they show pronounced differences when considering their geologic structure. Indeed, this area is crossed by the periadriatic lineament, which divides the Austroalpine basement (Lavazz`e) from the Southalpine one (Fondo and Livo). The Austroalpine basement consists of several subunits which form a complex nappe pile. The individual units are characterised by differences in lithological composition, mineralogy, protolith age and geochemistry, the occurrence of meta-igneous rocks and by different tectonometamorphic evolutions. Paragneiss is the dominant lytotype in the Austroalpine area at Lavazz`e, while dolomites and limestones are present in the Fondo area and fluvioglacial sediments characterise the Livo area. At Vendevolo, bedrock geology includes mainly rhyolithe, while soils are volcanic sandy loams. Stones and cobbles are locally abundant. The soil cover is variable in depth, and there are localized bedrock outcrops.
92
M. Borga et al. / Ecological Modelling 187 (2005) 85–98
Topographic data were gridded to generate raster maps with 10 m grid size for the different study areas. Maps of specific upslope drainage area were computed by using the procedure introduced by Tarboton (1997). In order to reduce the parameter uncertainty likely affecting model results, the model used here does not consider the effects of root strength and vegetation surcharge on slope stability. A sensitivity analysis of the influence of vegetation-related parameters on model results is on-going and results will be reported soon. Field work provided the basis for estimating values for soil hydraulic and geotechnical parameters. The value of transmissivity and the ratio ρs /ρw were considered constant over the different sites (equal to 30 m2 day−1 and 1.8, respectively), while the angle of internal friction was assumed equal to 42◦ for Lavazz`e and Vendevolo, and to 38◦ for Fondo and Livo. The soils were assumed cohesionless; operating in this way, the stability analysis model does not require estimation of the highly uncertain parameter soil thickness h (even though computation of the contributing areas requires estimation of soil thickness and cut slope depth along the road segments). Model assessment was carried out in two steps. In the first step, the shallow landsliding model was evaluated on non-road-related shallow landsliding, to examine its suitability to describe mass wasting processes in the study areas. Indeed, only if the contributing area (per unit contour width) and hillslope form are the main topographic attributes defining critical conditions for landsliding, the road interception/rerouting and slope stability model may have potential for predicting road impact on landsliding susceptibility. In the second phase, observations collected for the roadrelated landslides were used to evaluate in each study site the coupled road interception/rerouting and slope stability model. 3.2. Testing the shallow landsliding model Whereas non-road-related shallow landslides are widespread on Lavazz`e and Livo sites, very few non-road-related failures can be found on Fondo and Vendevolo study areas. Therefore, the shallow landsliding model (excluding the road interception/rerouting model) was tested on non-road-related slides on the Lavazz`e and Livo sites, in order to assess
Table 2 Percent of catchment and landslide area in each critical rainfall range for the Rio Lavazz`e basin Critical rainfall (mm day−1 )
Catchment area (%)
Landslide area (%)
Unconditionally unstable 0–50 50–100 100–200 >200 Unconditionally stable
2.9 17.6 10.8 9.2 13.4 46.1
6.6 39.2 28.2 6.6 19.4 0.0
the representativeness of model and data at least for these study areas. Assessment of the shallow landsliding model is based on field mapping of all non-road-related shallow landslide scars in the study area. The scars are then overlain onto the critical rainfall map and for each scar a critical rainfall value is assigned. Histograms of landslide-associated critical rainfall values are then made. Better model performance would be reflected in a larger difference between fractions of catchment and observed landslide area corresponding to low values of critical rainfall. Fifty three field mapped landslide scars located on the Rio Lavazz`e basin, which includes both Lavazz`e and Livo sites, were used for testing purposes. The basin is 41 km2 wide, and ranges in altitude from 550 to 2672 m a.s.l. Since the model applies only to the sources of landslides, observed runout zones along landslide paths were excluded from the analysis. The comparison of observed landslide locations with model predictions is reported in Table 2. The comparison was designed to account for the uncertainty associated with field mapping. To account for this effect, landslide scars were associated to the lowest critical rainfall computed in a 3 × 3 grid cell window centred on the grid cell of interest. While this procedure has the potential to positively bias the statistics, it allows to account for the event that a landslide observed on a grid characterised by the model as relatively safe is originated by the failure of a close grid with low-critical rainfall. The proportion of landslide area that occurred within critical rainfall categories <100 mm day−1 or considered as unconditionally unstable is around 74%, while the corresponding percentages of basin area reduce to 31.3%. These results suggest that landslides
M. Borga et al. / Ecological Modelling 187 (2005) 85–98
occur disproportionately in areas of low-critical rainfall, implying that the model discriminates areas of greater landslide hazard. However, results reported in Table 2 indicate also that many more cells are predicted to be unconditionally unstable than are observed as scars. A partial explanation of this result can be provided by the limitations of any field mapping, which captures only the most recent failures. On this basis, unfailed slopes characterised by the model as unconditionally unstable or with low-critical rainfall should be interpreted as likely sites of failure in the future rather than areas presumed to be tested and proved stable during previous storms. This discrepancy is also likely due to uncertainty in input parameter values and model structure. For instance, spatial and vertical variability of soil physical and mechanical parameters is disregarded in the model, so that values input to the model can be not representative of soil strength near shear surface. An encouraging result is that no landslides were observed on the basin area predicted as unconditionally stable. Overall, these results indicate that the model yields an encouraging first answer (at least for the two study sites where testing was possible). The results suggest
93
that slope, hillslope form and contributing area play a major role in controlling the spatial distribution of landslides in the studied basin. 3.3. Analysis of results from the custlope interception and slope stability model A visual assessment of model results was carried out by following a three-step procedure: (i) first, based on data gathered from the post-event analysis and by using the model techniques we reproduced the likely functioning of road drainage structures during the landslide triggering storm events; (ii) by using the hydraulic and geomechanical soil parameters reported above for the four sites, a Relative Impact map was obtained for each landscape; (iii) then, the Relative Impact map was compared with the location of the road-related landslide scars caused by the storm events. During the post-event analysis (in the first few months after the storms), the entire road network of the study areas was surveyed. (For the Vendevolo site, we used reports made available from the civil lawsuit which focused on the case.) All erosional and depositional features within the road prism were identified. This provided data for the specification of
Fig. 3. Relative impact (RI) map for Fondo. The map shows also road drainage patterns, landsliding (contoured by thick lines) and road drainage directions (arrows). Contour interval is in meter, at 10 m intervals. Cross-road ditches indicated by dots.
94
M. Borga et al. / Ecological Modelling 187 (2005) 85–98
Fig. 4. Relative impact (RI) map for Lavazz`e. The map shows also road drainage patterns, landsliding (contoured by thick lines) and road drainage directions (arrows). Contour interval is in meter, at 10 m intervals. Cross-road ditches indicated by dots.
Fig. 5. Relative impact (RI) map for Livo. The map shows also road drainage patterns, landsliding (contoured by thick lines) and road drainage directions (arrows). Contour interval is in meter, at 10 m intervals. Cross-road ditches indicated by slashes.
M. Borga et al. / Ecological Modelling 187 (2005) 85–98
95
Fig. 6. Relative impact (RI) map for Vendevolo. The map shows also road drainage patterns, landsliding (contoured by thick lines) and road drainage directions (arrows). Contour interval is in meter, at 10 m intervals. Cross-road ditches indicated by slashes.
road drainage patterns and for the implementation of the model. Results are reported for the four sites in Figs. 3–6. Fig. 7a and b shows a picture of the landslide initiation site at Fondo. In these figures, the position of the eroded area corresponding to the landslides is reported. Several features are worth noting from the examination of the Relative Impact spatial patterns. First, unconditionally unstable areas are much more widespread in the Fondo and Lavazz`e cases than in Livo and Vendevolo. In the first two cases, unconditionally unstable areas consistently correspond to subvertical cliffs and areas with thin and discontinuous soil coverage with widespread rock outcrops. The model does not strictly applies to these areas. Bedrock outcrops would, nonetheless, be expected where landslide frequency exceeds the soil production rate, so this prediction is not inconsistent. In the last two cases, the model often produces isolated cells of predicted unconditional instability. Only in few cases these cells correspond to bedrock outcrops. Model and parameter uncertainty is likely at the origin of these predictions. The Relative Impact patterns show how road networks can alter pre-existing subsurface flow paths. Roads appear to influence potential shallow landsliding
far downstream of runoff interception sites. Absolute values of Relative Impact are highest immediately below the road but extend from the area below the road to a large portion of the hillside downslope (being most persistent on divergent areas and less on convergent terrain). In three of the four cases (Fondo, Lavazz`e and Livo) positive Relative Impact values extend to the stream channel, indicating the potential for debris flows entering an adjoining stream channel (as actually occurred in the Fondo site). Gradients of Relative Impact across adjacent terrain elements are higher for the Fondo and Vendevolo cases, where the impact of the rerouting effect due to cut slope interception and road drainage features was more important. Several hillslopes are characterised by negative Relative Impact values, indicating a decreasing shallow landslide hazard on these sites. The observed road-related landslide scars generally correspond well with positive values of the Relative Impact statistic. This is particularly evident for the Fondo and Vendevolo cases, where the impact of the rerouting effect due to cut slope interception and road drainage features is higher. In the Fondo case study, the landslide initiation site is captured well by the model, even though the eroded area extended also to terrain charac-
96
M. Borga et al. / Ecological Modelling 187 (2005) 85–98
Fig. 7. Road-related landslide at the Fondo study site. (a) The cross road ditch which caused the failure and the landslide source. (b) The picture shows the landslide initiation point.
terised by negative Relative Impact. This result illustrates clearly one of the limitations of this approach, which is based on a hillslope stability model which describes only the initial collapse of the soil. Model predictions are in any case largely affected by uncertainties in describing the processes of cut slope in-
terception and runoff rerouting, and in the structure and parametrization of the hillslope stability model. Given the assumptions and generalisation that are being made, the model predictions should not be expected to be right everywhere in a landscape. This is illustrated well by the Lavazz`e and Livo cases, where the mapped scars
M. Borga et al. / Ecological Modelling 187 (2005) 85–98
only marginally include terrain with positive Relative Impact. Particularly in the Lavazz`e case, the presence of subvertical cliffs and widespread bedrock outcrops in the upslope area renders very difficult the reliable application of the hydrological model (and related assumptions) incorporated in the hillslope stability approach. This visual assessment suggests that model predictions can be considered moderately accurate, even though the model provides only a rough representation of actual hydrological and geomechanical processes taking place in the study areas, and the limitations due to the assumptions incorporated in the model should be always borne in mind.
4. Conclusions The effect of forest roads on potential shallow landsliding in steep, forested terrain has been examined through a combination of field observations and predictive modelling, with focus on the interaction of subsurface flow with roads. A distributed slope stability model, designed to portray the effects of subsurface runoff interception and rerouting by roads on shallow mass wasting, was developed. The model includes a hydrological model for the computation of water table position and subsurface runoff interception and rerouting by the roads network, and a geomechanical model for the analysis of hillslope stability. The hydrological model is based on a steady-state model for shallow subsurface runoff, while the geomechanical model incorporates an infinite-slope Coulomb failure model. The model was applied in four sites in north-eastern Italy, characterised by fairly different climate, geology, geomorphology and road network structure. The model was shown, through independent testing on non-road-related and road-related shallow landslides, to be moderately accurate, even though the description of hydrological, landslide initiation and road drainage processes provided by the model is quite simplistic. Overall, the findings of this study are consistent with the hypothesis that some road segments intercept subsurface flow, route it to ditches and thence on previously unchanneled hillslopes, altering the hillslope flow paths and the potential for shallow landsliding. Despite its relative utility in predicting road impact on landsliding potential, the theoretical model devel-
97
oped here does not capture potentially important aspects of hillslope-scale behaviour. Bedrock topography that does not correspond to surface topography would be expected to produce saturation-deficit patterns that correspond poorly to predicted patterns based on this model. Furthermore, a complication which has been neglected in the model is the presence of natural pipes and possibly of other macropores in carrying significant downslope flows. Pipes act as bypasses to soil flow and therefore speed delivery to the slope base of whatever inflow they receive. Presence of pipes connected with the surface poses serious limitations to the representativeness of the coupled hydrological and slope stability model. However, owing to difficulties and uncertainties in pipeflow modelling, it is likely that this component remains an intractable part in the hydrological model. We propose that the simple theoretical model developed here is most useful for examining the coupled control of hillslope topography and road configuration on shallow landsliding on steep hillslopes. A central issue for further research concerns the interactions between the initial collapse and the resulting debris flow and fluvial sediment transport. Stability slope theory such as that incorporated in the model describes only the initial collapse of the soil. Understanding of the processes controlling the transformation of landslide to debris flow is critical for determining the distance of travel of the debris and the gradient on which it will come to rest. This is particularly important, since these properties affect both the extent of the hazard zone, as distinct from the extent of the unstable slope, and the effects of the road-related hillslope failure on stream channels. Finally, coupled models for automated mapping of spatial patterns of failure potential and downstream hazard may be extended to provide a basin-wide assessment of the nature of road-related sediment supply to channel networks in humid mountain basins.
Acknowledgments This research was supported through funding from Sixth Framework Program of the European Commission (FLOODsite project, EC Contract number: GOCE-CT-2004-505420) and the INRM (Istituto Nazionale Ricerca sulla Montagna) to DTeSAF (University of Padova). We thank Paolo Campedel and
98
M. Borga et al. / Ecological Modelling 187 (2005) 85–98
Francesca Crivellari for assistance in the geological and geotechnical surveys in the Lavazz`e and Livo areas; Vincenzo D’Agostino for the collaboration in the analysis of the Vendevolo case study. References Anderson, M.G., 1983. Road-cut slope topography and stability relationships in St. Lucia, West Indies. Appl. Geogr. 3, 104– 114. Atkinson, T.C., 1978. Techniques for measuring subsurface flow on hillslopes. In: Kirkby, M.J. (Ed.), Hillslope Hydrology. Wiley, Hoboken, NJ, pp. 73–120. Borga, M., Dalla Fontana, G., Da Ros, D., Marchi, L., 1998. Shallow landslide hazard assessment using a physically based model and digital elevation data. J. Environ. Geol. 35 (2–3), 81–88. Borga, M., Dalla Fontana, G., Gregoretti, C., Marchi, L., 2002a. Assessment of shallow landsliding by using a physically based model of hillslope stability. Hydrol. Process. 16, 2833– 2851. Borga, M., Dalla Fontana, G., Cazorzi, F., 2002b. Analysis of topographic and climatic control on rainfall-triggered shallow landsliding using a quasi-dynamic wetness index. J. Hydrol. 268 (1–4), 56–71. Borga, M., Tonelli, F., Selleroni, J., 2004. A physically-based model of the effects of forest roads on slope stability. Water Resour. Res. 40 (12), W12202 10.1029/2004WR003238. Burroughs, E.R., 1985. Survey of slope stability problems on forest lands in the West. In: Swanston, D. (Ed.), Proceedings of a Workshop on Slope Stability: Problems and Solutions in Forest Management, February 6–8, 1984, Seattle. Gen. Tech. Re PNW-180. U.S. Department of Agriculture, Forest Service, Pacific Northwest Forest and Range Experiment Station, Portland, OR, pp. 5–16. Coker, R.J., Fahey, B.D., 1993. Road related mass movement in weathered granite, Golden Downs and Motueka Forests, New Zealand. A note. J. Hydrol. (NZ) 31 (1), 65–69. Dietrich, W.E., Wilson, C.J., Montgomery, D.R., Mckean, J., 1993. Analysis of erosion threshold, channel networks and landscape morphology using digital terrain models. J. Geol., 259–278. Gucinski, H., Furniss, M.J., Ziemer, R.R., Brookes, M.H., 2001. Forest roads: a synthesis of scientific information. Gen. Tech. Re PNW-GTR-509. U.S. Department of Agriculture, Forest Service, Pacific Northwest Research Station, Portland, OR, 103 pp. Harr, R.D., Nichols, R.A., 1993. Stabilizing forest roads to help restore fish habitats: a northwest Washington example. Fisheries 18 (4), 18–22. Jones, J.A., 2000. Hydrologic processes and peak discharge response to forest removal, regrowth, and roads in 10 small experimental basins, western Cascades, Oregon. Water Resour. Res. 36, 2621–2642. Luce, C.H., Cundy, T.W., 1994. Parameter identification for a runoff model for forest roads. Water Resour. Res. 30 (4), 1057–1069.
Luce, C.H., Wemple, B., 2001. Introduction to the special issue on hydrologic and geomorphic effects of forest roads. Earth Surf. Process. Landforms 26 (2), 111–113. Luce, C.H., Black, T.A., 1999. Sediment production from forest roads in western Oregon. Water Resour. Res. 35 (8), 2561–2570. Megahan, W.F., 1972. Subsurface flow interception by a logging road in mountains of central Idaho. In: Scallany, S.C., McLaughlin, T.G., Striffler, W.D. (Eds.), Proceedings of the Watersheds in Transition Symposium. American Water Resources Association, Urbana, IL, pp. 350–356. Montgomery, D.R., Dietrich, W.E., 1994. A physically based model for the topographic control on shallow landsliding. Water Resour. Res. 30 (4), 1153–1171. Montgomery, D.R., 1994. Road surface drainage, channel initiation, and slope instability. Water Resour. Res. 30 (6), 1925–1932. Mosley, M.P., 1980. The impact of road erosion in the Dart Valley, Nelson. NZ J. Forestry 25 (2), 184–198. Pozzatti, A., Cerato, M., 1984. Note pratiche sulla progettazione delle strade forestali. L’Italia forestale e montana 5, 263–274 (in Italian). Reid, L.M., 1981. Sediment production from gravel-surfaced forest roads, Clearwater basin, Washington. Publication FRI-UW-8108. University of Washington, Fisheries Research Institute, Seattle. Reid, L.M., Dunne, T., 1984. Sediment production from forest road surfaces. Water Resour. Res. 20, 1753–1761. Sidle, R.C., Pearce, A.J., O’Loughlin, C.L., 1985. Hillslope stability and land use. Water Resource Monograph, vol. 11. American Geophysical Union, Washington, DC, 140 pp. Swanson, F.J., Dyrness, C.T., 1975. Impact of clearcutting and road construction on soil erosion by landslides in the western Cascade Range. Oregon Geol. 3 (7), 393–396. Tague, C., Band, L., 2001. Simulating the impact of road construction and forest harvesting on hydrologic response. Earth Surf. Process. Landforms 26, 135–151. Tarboton, D.G., 1997. A new method for the determination of flow directions and contributing areas in grid digital elevation models. Water Resour. Res. 33 (2), 309–319. Wemple, B.C., Hones, J.A., Grant, G.E., Selker, J.E., 1996. Runoff generation mechanisms in a steep, forested catchment: controls on flow contributions to a road network. EOS Trans. Am. Geophys. Union 77 (46), F188. Wemple, B.C., Swanson, F.J., Jones, J.A., 2001. Forest roads and geomorphic process interactions, Cascade Range, Oregon. Earth Surf. Process. Landforms 26, 191–204. Wemple, B.C., Jones, J.A., 2003. Runoff production on forest roads in a steep, mountain catchment. Water Resour. Res. 39 (8), 1220. Wigmosta, M.S., Perkins, W.A., 2001. Simulating the effects of forest roads on watershed hydrology. In: Wigmosta, M., Burgess, S. (Eds.), Land Use and Watersheds: Human influence on Hydrology and Geomorphology in Urban and Forested Areas. Water Science Application Series, vol. 2. AGU, Washington, DC, pp. 127–143. Ziegler, A.D., Giambelluca, T.W., 1997. Importance of rural roads as source areas for runoff in mountainous areas of northern Thailand. J. Hydrol. 196, 204–229.
Evaluating the influence of forest roads on shallow landsliding Marco Borga a,∗ , Fabrizio Tonelli a , Giancarlo dalla Fontana a , Federico Cazorzi b a
Department of Land and AgroForest Environments, AGRIPOLIS, University of Padova, via dell’Universit`a, 16, Legnaro, IT-35020, Italy b Department of Agricultural and Environmental Science, Polo Scientifico Rizzi, University of Udine, IT-33100 Udine, Italy Available online 25 March 2005
Abstract This study investigates how subsurface flowpaths are altered by forest roads and how these changes influence shallow landsliding susceptibility in steep, forested landscape. A simple conceptual model of the effect of forest roads on hillslope subsurface flow is developed. The model is incorporated into a hydro-geomechanical, threshold-based model for slope instability. In the model, the occurrence of shallow landsliding is evaluated in terms of drainage areas, ground slope and soil properties (i.e., hydraulic conductivity, bulk density, and friction angle). Model results allow to quantify the influence of roads on shallow landsliding hazard across a landscape and to generate hypotheses about the broader geomorphic effect of roads. Modelling results are compared with field data collected in four sites located in north-eastern Italy. Observed landslide patterns are broadly consistent with model estimates, a finding that underscores the utility of this simple approach for predicting the geomorphic effects of forest roads constructed on steep slopes. The approach used in this study may be useful for defining criteria for road design that reduce the effects of roads on geomorphic processes. © 2005 Elsevier B.V. All rights reserved. Keywords: Anthropogenic effects; Runoff and streamflow; Forest roads; Shallow landsliding; Subsurface Flow; Hillslope hydrology
1. Introduction The interaction between forest roads and geomorphic processes lies at the heart of several key issues concerning the effects of roads on the environment. Geomorphic effects of forest roads range from chronic and long-term contributions of fine sediment into stream to catastrophic effects associated to shallow landsliding during large storms. In steep, forested terrain prone ∗ Corresponding author. Tel.: +39 049 8272681; fax: +39 049 8272686. E-mail address: [email protected] (M. Borga).
0304-3800/$ – see front matter © 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.ecolmodel.2005.01.055
to landsliding, the greatest effect of roads on erosion rates is from increased rates of mass soil movement after road building (Gucinski et al., 2001). Major issues motivating concern about road-related erosion include potential degradation of aquatic habitat and water quality (Harr and Nichols, 1993) and risks to public safety and structures downstream (Burroughs, 1985; Pozzatti and Cerato, 1984). The magnitude of road-related mass erosion differs with climate, geology, topography, road age, construction practices and storm history (Gucinski et al., 2001). Several inventories have been conducted to assess road effects on mass failures, with more specific fo-
86
M. Borga et al. / Ecological Modelling 187 (2005) 85–98
cus on U.S. Pacific Northwest and New Zealand (Sidle et al., 1985; Swanson and Dyrness, 1975; Reid, 1981; Mosley, 1980; Coker and Fahey, 1993), each documenting increased rates of landsliding in road areas relative to unmanaged forested areas. Sidle et al. (1985) documented accelerated erosion rates from roads because of debris slides ranging from 30 to 300 times the forest rate. Influence of forest roads on geomorphic processes generally results from concentration of both runoff generated as overland flow from compacted road surfaces and intercepted subsurface flow by road cutslopes (Megahan, 1972; Anderson, 1983; Reid and Dunne, 1984; Luce and Cundy, 1994; Montgomery, 1994; Ziegler and Giambelluca, 1997; Luce and Black, 1999; Wemple et al., 2001). Subsurface storm flow dominates runoff generation in steep soil-mantled terrain where precipitation infiltrates and flows laterally either through macropores or over a lower conductivity zone. In these environments, road interception occurs when a seasonally high water table flowing over on impermeable base (e.g. bedrock) becomes deep enough to intersect the road ditch. Thus the fraction of the permeable soil occupied by the road cut becomes a controlling factor in the amount of interception. Subsurface flow intercepted along the road cut may be diverted to surface runoff, and then redirected downslope, modifying pre-existing flow paths on the hillslope. In steep, soilmantled terrain, the combination of ubiquitous subsurface flow and of the interception by road cuts yields an influence on hydrogeomorphic processes that may be much greater than one might expect from the small fraction of the land area roads occupy (Luce and Wemple, 2001). Road-generated runoff is usually routed by roadside ditches and therefore concentrated in particular areas below the road. The effect of this concentration of flow will depend on the characteristics of the receiving areas. When the road drainage is not connected to the stream network (at least through surface flow), road runoff may either reinfiltrate into the unchanneled terrain, or reinfiltrate below a gully that does not extend to the stream network. In these cases, concentrated road runoff may affect shallow landsliding potential in the receiving areas and decrease the critical source area required to initiate headwater streams (Montgomery, 1994). In cases where the road network is connected to the stream network road, runoff may enter a stream di-
rectly at a stream crossing culvert, or enter a stream indirectly through the formation of a gully, extending to the river network, below a ditch relief culvert. Therefore, the collective contribution of intercepted hillslopes to the road and the road surface drainage features determine the road impact on mass wasting on hillsides downslope. This impact may be large when roads intercept large amounts of subsurface flow and redirect it to unchanneled terrain, conditionally unstable, below the road. The purpose of this work is to gain further insight on road interactions with hillslope flow paths and on how these interactions influence shallow landsliding of concerned hillslopes. This paper addresses these issues by developing a conceptual and quantitative framework for evaluating how roads in different landscape positions (valley bottom, midslope, ridgetop) affect subsurface flow paths and associated mass failures. A threshold-based model for slope instability (Dietrich et al., 1993; Montgomery and Dietrich, 1994; Borga et al., 1998, 2002a) is extended to the case of hillslopes interested by road networks. A range of scenarios is used to generate hypotheses about the effects of road on hillslope stability. Simulation results are also compared with available field data collected in north-eastern Italy in order to assess how well the model captures the processes of interest.
2. A conceptual model of cutslope interception, throughflow rerouting and slope instability The conceptual model of cutslope interception and throughflow rerouting used in this study is based on coupling of digital topography with a simple model of steady-state rainfall-runoff to calculate the saturation deficit at any point in the landscape. The following assumptions are used to model the subsurface flow propagation: • Shallow lateral subsurface flow follows topographic gradient. • The entire soil profile is initially wet to field capacity. • Lateral discharge at each point is in equilibrium with a steady-state recharge R [LT−1 ], i.e. the infiltrating rainfall which passes through (or bypasses) the unsaturated zone to reach the saturated zone and eventually becomes subsurface runoff. The steady-state
M. Borga et al. / Ecological Modelling 187 (2005) 85–98
assumption implies that the specific upslope area (measured along the horizontal plane) is a surrogate measure of the subsurface flow at any point in the landscape. • The recharge rate equals the rainfall rate, thereby neglecting the vertical transport processes taking place between rainfall reaching the ground surface and recharge occurring at the soil base. Rainfall infiltrates quickly into the highly permeable soils dominated by subsurface runoff and this approximation should be adequate. • The capacity for lateral flux at each point is T sin θ, where T is the soil transmissivity [L2 T−1 ] and θ the surface slope. Furthermore, the approach with the hydrological model is to interpret the soil thickness (h [L]) as specified perpendicular to the slope. Analogously, hw [L] is the thickness of the saturated region above the bedrock, and drc [L] is the depth of the road cut base above the bedrock, both measured perpendicular to the slope (Fig. 1). Relative road cut depth (rrc ) is computed as drc /h. Assumptions above imply that depth integrated subsurface flow per unit contour length q [L2 T−1 ] is q=R
A b
(1)
where A [L2 ] is the upslope area and b [L] the length across which flow is accounted for. Furthermore, the relative wetness (for a generic pixel not influenced by road-rerouted subsurface flow) w=
hw h
(2)
Fig. 1. Schematic diagram illustrating the interaction of the hillslope water table with the road cut and the variables used in the model description.
may be expressed as RA w = Min ,1 Tb sin θ
87
(3)
The relative wetness has an upper bound of 1 with any excess assumed to form overland flow. Implementation of road network effects makes the following assumptions: • the amount of subsurface flow intercepted by the road is a linear function of the elevation hi of the water table relative to the base of the road cut, calculated as hi = hw − drc ; • the intercepted flow is redirected by the road in-board ditch, based on the road slope, while non-intercepted flow follows natural flowpaths on hillsides downslope; • ditch discharge is diverted by cross-ditches and relief culverts toward areas below the road, where road runoff may either reinfiltrate (where the receiving terrain is unchanneled) or enter a stream (where the road drainage is directly linked to the river network). This model of cutslope interception and throughflow rerouting has been broadly applied to study the effect of forest roads on watershed hydrology (Wemple et al., 1996; Tague and Band, 2001; Wigmosta and Perkins, 2001; Wemple and Jones, 2003) and its application on subsurface-flow dominated watersheds is supported by a vast amount of observations (Megahan, 1972; Jones, 2000; Wemple and Jones, 2003). Note that this approach neglects the drawdown effects of the seepage face on the water table (Atkinson, 1978), which may lead to overestimates of water table elevation in these calculations. This effect is likely small relative to other potential sources of error in the approach used here. For each pixel influenced by the road network, three upslope contributing areas are considered: (1) the portion of the contributing area located upslope from the road, which contributes according to the road cut depth and the current saturation deficit (A1 ); (2) the contributing area related to the subsurface discharge intercepted by the road cut depth as a function of the current saturation deficit and rerouted to the pixel by the road (A2 ); (3) the portion of the contributing area located downslope the road (A3 ).
88
M. Borga et al. / Ecological Modelling 187 (2005) 85–98
In the next section it is shown how the model accounts for the road cut interception effect in the computation of the contributing area. 2.1. The threshold model for slope instability
cept of critical rainfall, introduced by Montgomery and Dietrich (1994), is worth using. The critical rainfall is the minimum steady-state rainfall predicted to cause instability. Coupling between the slope stability model and the steady-state hydrologic model illustrated above allows to derive the critical rainfall (Montgomery and Dietrich, 1994). Based on Eqs. (3) and (4), and for a terrain element not influenced by road runoff, critical rainfall is defined as Tb sin θ Cr + Cs ρs W R= + + A ρw gh cos θ tan φ ρw ρw gh tan θ × 1− (5) tan φ
Planar infinite-slope analysis has been widely applied to the determination of natural slope stability, particularly where the thickness of the soil mantle is small compared with the slope length and where landslides are due to the failure of a soil mantle that overlies a sloping drainage barrier. The drainage barrier may be bedrock or a denser soil mass. The infinite-slope stability model provides a onedimensional model for failure of shallow soils that neglects arching and lateral root reinforcement. Because The model application is based on the assumption of the geometry of an infinite slope, overall stability that differences in geology impose a broad control on can be determined by analysing the stability of a single, absolute rates of shallow landsliding upon which supervertical element in the slope. End effects in the sliding imposed topographic controls (contributing area and mass can be neglected, along with lateral forces on eislope) dominate the relative slide frequency. The reader ther side of the vertical element, which are assumed is referred to Borga et al. (2002b) for an analysis of the to be opposite and equal. Under these assumptions the critical rainfall concept and its relation with the steadyfactor of safety (FS) for a vegetated slope with slopestate assumption used in the model. parallel seepage, simplified for wet and dry soil density the same, is computed as follows: Cr + Cs + ρs g(h − hw ) cos θ + (ρs g − ρw g)hw cos θ + W cos θ tan φ FS = (4) hρs g sin θ + W sin θ where Cs and Cr are soil and root cohesion [ML−1 T−2 ], respectively, φ is the internal friction angle of the soil [degrees], ρs is wet soil density [ML−3 ], ρw is the density of water [ML−3 ], g is gravitational acceleration and W is the vegetation surcharge [ML−1 T−2 ]. Since steady-state flow conditions are assumed, effective stress parameters (effective friction angle and effective cohesion) must be used in the analysis. Inclusion of root cohesion and vegetation surcharge into stability analysis allows one to develop a more complete analysis of forest practices, combining road construction and logging activities. Indeed, in potentially unstable areas, shallow failures may increase after trees are cut, as their root systems progressively decay. In several cases, only two factor of safety classes (FS ≥ 1 and FS < 1) are deemed enough to estimate the susceptibility of a landscape to shallow landsliding, and analysis of stability in relation to a range of precipitation values is required. In these situations, the con-
The stability of terrain elements whose contributing area includes road rerouting is potentially affected by road runoff. For these elements, the computation of road cut interception effect (hence contributing area) is function of the unknown critical rainfall. Indeed, road cut interception is function of both relative road cut depth (rrc ) and of the local saturation deficit (i.e., of the critical rainfall, according to Eq. (3)). Owing to these reasons, a generic algorithm has been devised for the computation of the contributing areas. The algorithm requires estimation of relative road cut depth (rrc ) for each grid element associated to a road, and it is based on the computation of the interception effect for a number of discrete rainfall values. For each of these values, the contributing area for each element of the landscape is calculated. Then, Eq. (5) is solved iteratively for each element.
M. Borga et al. / Ecological Modelling 187 (2005) 85–98
The algorithm is based on the following steps: (i) for each discrete rainfall value and for each model grid associated to a road, the value of w is computed based on Eq. (3); (ii) for these grids, the λ1 and λ2 weighting coefficients are computed as follows: r
rc
,1 , w w − rrc λ2 = max , 0 = 1 − λ1 w
λ1 = min
(6)
where λ1 represents the portion of the contributing area corresponding to the fraction of subsurface flow which is not intercepted by the road cut and follows natural flowpaths on hillsides downslope, and λ2 represents the portion of the contributing area corresponding to the fraction of subsurface flow which is intercepted by the road cut and follows the artificial road drainage path. The contributing area for a grid element interested by a road is therefore made up of three parts: • the fraction of the contributing area located upslope from the road: Λ1 A 1 = λ1 a dl (7) Lr1
where Lr1 is the length of the road segment intersecting the contributing area of the given grid element; • the contributing area related to the subsurface discharge intercepted by the road cut depth and rerouted to the pixel by the road: Λ 2 A2 = λ2 a dl (8) Lr2
where Lr2 is the length of the road segment which contributes intercepted flow to the given grid element; and • A3 , which is the portion of the contributing area located downslope the road. At this point, for each discrete rainfall values and for each generic landscape pixel it is possible to compute the product R(Λ1 A1 + Λ2 A2 + A3 ). The value of the product which agrees with the following equality: R (Λ1 A1 + Λ2 A2 + A3 )
89
Cr + Cs + = Tb sin θ ρw gh cos θ tan φ tan θ × 1− tan φ
ρs W + ρw ρw gh
(9)
is identified iteratively. (Note that the term R(Λ1 A1 + Λ2 A2 + A3 ) is monotonic increasing with R, and this allows to find always a solution to the problem.) When roads cross repeatedly the same hillslope, the algorithm needs to be repeated once for each crossing. This allows the computation of critical rainfall for each landscape pixel. A more detailed derivation of the model and of the algorithms included is presented elsewhere (Borga et al., 2004). For a given threshold of critical rainfall, the area of the basin with critical rainfall less than the threshold will coincide with the area characterised by having a factor of safety (computed with reference to the threshold) less than unity. Based on the concept of critical rainfall, four stability classes can be defined: unconditionally unstable, unstable, stable and unconditionally stable (Montgomery and Dietrich, 1994). Ground is unconditionally unstable if it is unstable even when dry. The condition for unconditionally stable slopes is expressed by tan θ < +
1 1− W/ρw gh + ρs /ρw
tan φ
Cr + Cs ρs gh cos θ (1 + W/ρs gh)
(10)
Unconditionally stable elements are those predicted to be stable even when saturated. The condition for unconditionally unstable slopes is expressed by
tan θ ≥ tan φ +
Cr + Cs ρs gh cos θ (1 + W/ρs gh)
(11)
Unstable elements are those predicted to fail according to Eqs. (5) and (9). Only potentially unstable elements are affected by road-related drainage modification. The impact can be evaluated in relative terms by mapping the critical rainfall before and after drainage modification due to road
M. Borga et al. / Ecological Modelling 187 (2005) 85–98
90
Fig. 2. Location map for the four study areas.
network (Rh and Rr , respectively), and computing the relative impact statistic (RI) as follows: RI =
Rh − Rr Rh
surface flow and increase critical rainfall; in this case, RI < 0.
(12) 3. Study areas and model application
The RI statistic was selected because it is dimensionless and is easily interpreted. If the road network has no effects on subsurface runoff propagation (hence, on shallow landsliding), RI = 0. If the cross-road ditches concentrate and increase saturated subsurface throughflow, hence decreasing the critical rainfall, RI > 0. Alternatively, road drainage may reduce saturated sub-
The model was implemented and evaluated on four study areas in north-eastern Italy (Fig. 2): the Vendevolo hill, on the Euganean Hills close to Padova, and three sites on the upper Noce river valley (Fondo, Lavazz`e and Livo). Topographic and climatic characteristics of the study areas are reported in Table 1.
Table 1 Main morphometric and climatic characteristics of study areas
Elevation range of concerned hillslope (m a.s.l.) Position on slope Slope (◦ ) 10 year, 12 h rainfall (mm) Slope is reported as average value over each landslide scar.
Fondo
Lavazz`e
Livo
Vendevolo
1200–1300 Lower 1/3 25.0 61.0
1350–1450 Middle 1/3 40.0; 29.5 61.0
700–800 Lower 1/3 25.6; 27.0 63.0
300–400 Upper middle 1/3 39.2; 36.4 73.0
M. Borga et al. / Ecological Modelling 187 (2005) 85–98
Road-related mass failures occurred in each of these sites, thus providing a basis for model evaluation in landscapes characterised by different climate, vegetation, geology and topographic structures. Field work comprised: road alignments and drainage mapping; landslide mapping; measurement of the characteristics of the road segment, such as slope, width and cutbank depth, and of road runoff drainage structures; post-event analysis of occurrences of culvert filling by sediments eroded along the road surface during storm events. Road drainage patterns and the associated drainage diversions were mapped in the field. This mapping phase may be highly uncertain, due to the different topographic features influencing road drainage propagation. To reduce these sources of uncertainty, road surface drainage patterns were mapped during and after runoff-producing events. Road surface widths measured in the field range from 3 to 6 m; max road slopes range around 20%. As it is typical in these areas, road drainage structures are represented by in-board ditches (though rarely used) and cross-road ditches. Cross-road ditches are excavated across the road at an angle and at sufficient depth, with armouring as appropriate, to divert both road surface water and ditch water off or across the road. The roads are of cut-and-fill design, consisting of a cutbank, a crowned road surface, and a fill slope. Hillslopes contributing to road segments are generally steep, ranging from 25◦ (Fondo) to 40◦ (Lavazz`e). Soils are shallow, ranging in depth from 0.2 m (Vendevolo and Lavazz`e) to over 1.5 m (Fondo and Livo), while cutbank depths ranges from 0 (some road segments for the Vendevolo case) to several meters (road cuts on bedrock at Lavazz`e). A sequence of events typical of major flood events in north-eastern Italy (prolonged rainfall, rain-on-snow events), occurred on autumn 2000, triggered the shallow landslides at Fondo, Lavazz`e and Livo. A 200-year return time, flash-flood producing storm (occurred on May 31, 1995) triggered the mass failures at Vendevolo. Inventoried shallow failures were generally originated by excessive road runoff focused on previously unchanneled hillsides below the road and thereby reinfiltrated. Neither fillslope slides or cutslope slides were considered in this study. Excessive road runoff was in most of the cases associated to disruption of road drainage structures, which were plugged by sedi-
91
ments transported by road runoff. Volumes of mass failures (estimated from measurements of the length, width and depth of the scar features) ranged from 50–60 m3 (Lavazz`e) to 3000 m3 (Fondo). For the three cases of Livo, Lavazz`e and Vendevolo, forest vegetation tended to buffer the flow of debris and sediment so that it did not reach large streams. For the Fondo case study, the material from the slide entrained and deposited along the channel bed, causing a temporary diversion of the river. 3.1. Study area description Fondo, Lavazz`e and Livo are characterised by an alpine climate, with a mean annual rainfall ranging from 1000 to 1300 mm. Precipitation occurs mainly as snowfall from November to April. Runoff is usually dominated by snowmelt in May and June but summer and early autumn floods represent an important contribution to the flow regime. The climate of the Vendevolo hill area is subMediterranean, with two precipitation peaks in spring and autumn. Mean annual rainfall amounts to 800 mm. On Fondo and Lavazz`e sites vegetation cover consists mainly of forest stands made up by spruce and larch; on Livo, vegetation is dominated by conifers and contains some deciduous species. Chestnut stands and shrubs represent main vegetation on the Vendevolo. Even though Fondo, Lavazz`e and Livo sites are just 15–20 km away from each other, they show pronounced differences when considering their geologic structure. Indeed, this area is crossed by the periadriatic lineament, which divides the Austroalpine basement (Lavazz`e) from the Southalpine one (Fondo and Livo). The Austroalpine basement consists of several subunits which form a complex nappe pile. The individual units are characterised by differences in lithological composition, mineralogy, protolith age and geochemistry, the occurrence of meta-igneous rocks and by different tectonometamorphic evolutions. Paragneiss is the dominant lytotype in the Austroalpine area at Lavazz`e, while dolomites and limestones are present in the Fondo area and fluvioglacial sediments characterise the Livo area. At Vendevolo, bedrock geology includes mainly rhyolithe, while soils are volcanic sandy loams. Stones and cobbles are locally abundant. The soil cover is variable in depth, and there are localized bedrock outcrops.
92
M. Borga et al. / Ecological Modelling 187 (2005) 85–98
Topographic data were gridded to generate raster maps with 10 m grid size for the different study areas. Maps of specific upslope drainage area were computed by using the procedure introduced by Tarboton (1997). In order to reduce the parameter uncertainty likely affecting model results, the model used here does not consider the effects of root strength and vegetation surcharge on slope stability. A sensitivity analysis of the influence of vegetation-related parameters on model results is on-going and results will be reported soon. Field work provided the basis for estimating values for soil hydraulic and geotechnical parameters. The value of transmissivity and the ratio ρs /ρw were considered constant over the different sites (equal to 30 m2 day−1 and 1.8, respectively), while the angle of internal friction was assumed equal to 42◦ for Lavazz`e and Vendevolo, and to 38◦ for Fondo and Livo. The soils were assumed cohesionless; operating in this way, the stability analysis model does not require estimation of the highly uncertain parameter soil thickness h (even though computation of the contributing areas requires estimation of soil thickness and cut slope depth along the road segments). Model assessment was carried out in two steps. In the first step, the shallow landsliding model was evaluated on non-road-related shallow landsliding, to examine its suitability to describe mass wasting processes in the study areas. Indeed, only if the contributing area (per unit contour width) and hillslope form are the main topographic attributes defining critical conditions for landsliding, the road interception/rerouting and slope stability model may have potential for predicting road impact on landsliding susceptibility. In the second phase, observations collected for the roadrelated landslides were used to evaluate in each study site the coupled road interception/rerouting and slope stability model. 3.2. Testing the shallow landsliding model Whereas non-road-related shallow landslides are widespread on Lavazz`e and Livo sites, very few non-road-related failures can be found on Fondo and Vendevolo study areas. Therefore, the shallow landsliding model (excluding the road interception/rerouting model) was tested on non-road-related slides on the Lavazz`e and Livo sites, in order to assess
Table 2 Percent of catchment and landslide area in each critical rainfall range for the Rio Lavazz`e basin Critical rainfall (mm day−1 )
Catchment area (%)
Landslide area (%)
Unconditionally unstable 0–50 50–100 100–200 >200 Unconditionally stable
2.9 17.6 10.8 9.2 13.4 46.1
6.6 39.2 28.2 6.6 19.4 0.0
the representativeness of model and data at least for these study areas. Assessment of the shallow landsliding model is based on field mapping of all non-road-related shallow landslide scars in the study area. The scars are then overlain onto the critical rainfall map and for each scar a critical rainfall value is assigned. Histograms of landslide-associated critical rainfall values are then made. Better model performance would be reflected in a larger difference between fractions of catchment and observed landslide area corresponding to low values of critical rainfall. Fifty three field mapped landslide scars located on the Rio Lavazz`e basin, which includes both Lavazz`e and Livo sites, were used for testing purposes. The basin is 41 km2 wide, and ranges in altitude from 550 to 2672 m a.s.l. Since the model applies only to the sources of landslides, observed runout zones along landslide paths were excluded from the analysis. The comparison of observed landslide locations with model predictions is reported in Table 2. The comparison was designed to account for the uncertainty associated with field mapping. To account for this effect, landslide scars were associated to the lowest critical rainfall computed in a 3 × 3 grid cell window centred on the grid cell of interest. While this procedure has the potential to positively bias the statistics, it allows to account for the event that a landslide observed on a grid characterised by the model as relatively safe is originated by the failure of a close grid with low-critical rainfall. The proportion of landslide area that occurred within critical rainfall categories <100 mm day−1 or considered as unconditionally unstable is around 74%, while the corresponding percentages of basin area reduce to 31.3%. These results suggest that landslides
M. Borga et al. / Ecological Modelling 187 (2005) 85–98
occur disproportionately in areas of low-critical rainfall, implying that the model discriminates areas of greater landslide hazard. However, results reported in Table 2 indicate also that many more cells are predicted to be unconditionally unstable than are observed as scars. A partial explanation of this result can be provided by the limitations of any field mapping, which captures only the most recent failures. On this basis, unfailed slopes characterised by the model as unconditionally unstable or with low-critical rainfall should be interpreted as likely sites of failure in the future rather than areas presumed to be tested and proved stable during previous storms. This discrepancy is also likely due to uncertainty in input parameter values and model structure. For instance, spatial and vertical variability of soil physical and mechanical parameters is disregarded in the model, so that values input to the model can be not representative of soil strength near shear surface. An encouraging result is that no landslides were observed on the basin area predicted as unconditionally stable. Overall, these results indicate that the model yields an encouraging first answer (at least for the two study sites where testing was possible). The results suggest
93
that slope, hillslope form and contributing area play a major role in controlling the spatial distribution of landslides in the studied basin. 3.3. Analysis of results from the custlope interception and slope stability model A visual assessment of model results was carried out by following a three-step procedure: (i) first, based on data gathered from the post-event analysis and by using the model techniques we reproduced the likely functioning of road drainage structures during the landslide triggering storm events; (ii) by using the hydraulic and geomechanical soil parameters reported above for the four sites, a Relative Impact map was obtained for each landscape; (iii) then, the Relative Impact map was compared with the location of the road-related landslide scars caused by the storm events. During the post-event analysis (in the first few months after the storms), the entire road network of the study areas was surveyed. (For the Vendevolo site, we used reports made available from the civil lawsuit which focused on the case.) All erosional and depositional features within the road prism were identified. This provided data for the specification of
Fig. 3. Relative impact (RI) map for Fondo. The map shows also road drainage patterns, landsliding (contoured by thick lines) and road drainage directions (arrows). Contour interval is in meter, at 10 m intervals. Cross-road ditches indicated by dots.
94
M. Borga et al. / Ecological Modelling 187 (2005) 85–98
Fig. 4. Relative impact (RI) map for Lavazz`e. The map shows also road drainage patterns, landsliding (contoured by thick lines) and road drainage directions (arrows). Contour interval is in meter, at 10 m intervals. Cross-road ditches indicated by dots.
Fig. 5. Relative impact (RI) map for Livo. The map shows also road drainage patterns, landsliding (contoured by thick lines) and road drainage directions (arrows). Contour interval is in meter, at 10 m intervals. Cross-road ditches indicated by slashes.
M. Borga et al. / Ecological Modelling 187 (2005) 85–98
95
Fig. 6. Relative impact (RI) map for Vendevolo. The map shows also road drainage patterns, landsliding (contoured by thick lines) and road drainage directions (arrows). Contour interval is in meter, at 10 m intervals. Cross-road ditches indicated by slashes.
road drainage patterns and for the implementation of the model. Results are reported for the four sites in Figs. 3–6. Fig. 7a and b shows a picture of the landslide initiation site at Fondo. In these figures, the position of the eroded area corresponding to the landslides is reported. Several features are worth noting from the examination of the Relative Impact spatial patterns. First, unconditionally unstable areas are much more widespread in the Fondo and Lavazz`e cases than in Livo and Vendevolo. In the first two cases, unconditionally unstable areas consistently correspond to subvertical cliffs and areas with thin and discontinuous soil coverage with widespread rock outcrops. The model does not strictly applies to these areas. Bedrock outcrops would, nonetheless, be expected where landslide frequency exceeds the soil production rate, so this prediction is not inconsistent. In the last two cases, the model often produces isolated cells of predicted unconditional instability. Only in few cases these cells correspond to bedrock outcrops. Model and parameter uncertainty is likely at the origin of these predictions. The Relative Impact patterns show how road networks can alter pre-existing subsurface flow paths. Roads appear to influence potential shallow landsliding
far downstream of runoff interception sites. Absolute values of Relative Impact are highest immediately below the road but extend from the area below the road to a large portion of the hillside downslope (being most persistent on divergent areas and less on convergent terrain). In three of the four cases (Fondo, Lavazz`e and Livo) positive Relative Impact values extend to the stream channel, indicating the potential for debris flows entering an adjoining stream channel (as actually occurred in the Fondo site). Gradients of Relative Impact across adjacent terrain elements are higher for the Fondo and Vendevolo cases, where the impact of the rerouting effect due to cut slope interception and road drainage features was more important. Several hillslopes are characterised by negative Relative Impact values, indicating a decreasing shallow landslide hazard on these sites. The observed road-related landslide scars generally correspond well with positive values of the Relative Impact statistic. This is particularly evident for the Fondo and Vendevolo cases, where the impact of the rerouting effect due to cut slope interception and road drainage features is higher. In the Fondo case study, the landslide initiation site is captured well by the model, even though the eroded area extended also to terrain charac-
96
M. Borga et al. / Ecological Modelling 187 (2005) 85–98
Fig. 7. Road-related landslide at the Fondo study site. (a) The cross road ditch which caused the failure and the landslide source. (b) The picture shows the landslide initiation point.
terised by negative Relative Impact. This result illustrates clearly one of the limitations of this approach, which is based on a hillslope stability model which describes only the initial collapse of the soil. Model predictions are in any case largely affected by uncertainties in describing the processes of cut slope in-
terception and runoff rerouting, and in the structure and parametrization of the hillslope stability model. Given the assumptions and generalisation that are being made, the model predictions should not be expected to be right everywhere in a landscape. This is illustrated well by the Lavazz`e and Livo cases, where the mapped scars
M. Borga et al. / Ecological Modelling 187 (2005) 85–98
only marginally include terrain with positive Relative Impact. Particularly in the Lavazz`e case, the presence of subvertical cliffs and widespread bedrock outcrops in the upslope area renders very difficult the reliable application of the hydrological model (and related assumptions) incorporated in the hillslope stability approach. This visual assessment suggests that model predictions can be considered moderately accurate, even though the model provides only a rough representation of actual hydrological and geomechanical processes taking place in the study areas, and the limitations due to the assumptions incorporated in the model should be always borne in mind.
4. Conclusions The effect of forest roads on potential shallow landsliding in steep, forested terrain has been examined through a combination of field observations and predictive modelling, with focus on the interaction of subsurface flow with roads. A distributed slope stability model, designed to portray the effects of subsurface runoff interception and rerouting by roads on shallow mass wasting, was developed. The model includes a hydrological model for the computation of water table position and subsurface runoff interception and rerouting by the roads network, and a geomechanical model for the analysis of hillslope stability. The hydrological model is based on a steady-state model for shallow subsurface runoff, while the geomechanical model incorporates an infinite-slope Coulomb failure model. The model was applied in four sites in north-eastern Italy, characterised by fairly different climate, geology, geomorphology and road network structure. The model was shown, through independent testing on non-road-related and road-related shallow landslides, to be moderately accurate, even though the description of hydrological, landslide initiation and road drainage processes provided by the model is quite simplistic. Overall, the findings of this study are consistent with the hypothesis that some road segments intercept subsurface flow, route it to ditches and thence on previously unchanneled hillslopes, altering the hillslope flow paths and the potential for shallow landsliding. Despite its relative utility in predicting road impact on landsliding potential, the theoretical model devel-
97
oped here does not capture potentially important aspects of hillslope-scale behaviour. Bedrock topography that does not correspond to surface topography would be expected to produce saturation-deficit patterns that correspond poorly to predicted patterns based on this model. Furthermore, a complication which has been neglected in the model is the presence of natural pipes and possibly of other macropores in carrying significant downslope flows. Pipes act as bypasses to soil flow and therefore speed delivery to the slope base of whatever inflow they receive. Presence of pipes connected with the surface poses serious limitations to the representativeness of the coupled hydrological and slope stability model. However, owing to difficulties and uncertainties in pipeflow modelling, it is likely that this component remains an intractable part in the hydrological model. We propose that the simple theoretical model developed here is most useful for examining the coupled control of hillslope topography and road configuration on shallow landsliding on steep hillslopes. A central issue for further research concerns the interactions between the initial collapse and the resulting debris flow and fluvial sediment transport. Stability slope theory such as that incorporated in the model describes only the initial collapse of the soil. Understanding of the processes controlling the transformation of landslide to debris flow is critical for determining the distance of travel of the debris and the gradient on which it will come to rest. This is particularly important, since these properties affect both the extent of the hazard zone, as distinct from the extent of the unstable slope, and the effects of the road-related hillslope failure on stream channels. Finally, coupled models for automated mapping of spatial patterns of failure potential and downstream hazard may be extended to provide a basin-wide assessment of the nature of road-related sediment supply to channel networks in humid mountain basins.
Acknowledgments This research was supported through funding from Sixth Framework Program of the European Commission (FLOODsite project, EC Contract number: GOCE-CT-2004-505420) and the INRM (Istituto Nazionale Ricerca sulla Montagna) to DTeSAF (University of Padova). We thank Paolo Campedel and
98
M. Borga et al. / Ecological Modelling 187 (2005) 85–98
Francesca Crivellari for assistance in the geological and geotechnical surveys in the Lavazz`e and Livo areas; Vincenzo D’Agostino for the collaboration in the analysis of the Vendevolo case study. References Anderson, M.G., 1983. Road-cut slope topography and stability relationships in St. Lucia, West Indies. Appl. Geogr. 3, 104– 114. Atkinson, T.C., 1978. Techniques for measuring subsurface flow on hillslopes. In: Kirkby, M.J. (Ed.), Hillslope Hydrology. Wiley, Hoboken, NJ, pp. 73–120. Borga, M., Dalla Fontana, G., Da Ros, D., Marchi, L., 1998. Shallow landslide hazard assessment using a physically based model and digital elevation data. J. Environ. Geol. 35 (2–3), 81–88. Borga, M., Dalla Fontana, G., Gregoretti, C., Marchi, L., 2002a. Assessment of shallow landsliding by using a physically based model of hillslope stability. Hydrol. Process. 16, 2833– 2851. Borga, M., Dalla Fontana, G., Cazorzi, F., 2002b. Analysis of topographic and climatic control on rainfall-triggered shallow landsliding using a quasi-dynamic wetness index. J. Hydrol. 268 (1–4), 56–71. Borga, M., Tonelli, F., Selleroni, J., 2004. A physically-based model of the effects of forest roads on slope stability. Water Resour. Res. 40 (12), W12202 10.1029/2004WR003238. Burroughs, E.R., 1985. Survey of slope stability problems on forest lands in the West. In: Swanston, D. (Ed.), Proceedings of a Workshop on Slope Stability: Problems and Solutions in Forest Management, February 6–8, 1984, Seattle. Gen. Tech. Re PNW-180. U.S. Department of Agriculture, Forest Service, Pacific Northwest Forest and Range Experiment Station, Portland, OR, pp. 5–16. Coker, R.J., Fahey, B.D., 1993. Road related mass movement in weathered granite, Golden Downs and Motueka Forests, New Zealand. A note. J. Hydrol. (NZ) 31 (1), 65–69. Dietrich, W.E., Wilson, C.J., Montgomery, D.R., Mckean, J., 1993. Analysis of erosion threshold, channel networks and landscape morphology using digital terrain models. J. Geol., 259–278. Gucinski, H., Furniss, M.J., Ziemer, R.R., Brookes, M.H., 2001. Forest roads: a synthesis of scientific information. Gen. Tech. Re PNW-GTR-509. U.S. Department of Agriculture, Forest Service, Pacific Northwest Research Station, Portland, OR, 103 pp. Harr, R.D., Nichols, R.A., 1993. Stabilizing forest roads to help restore fish habitats: a northwest Washington example. Fisheries 18 (4), 18–22. Jones, J.A., 2000. Hydrologic processes and peak discharge response to forest removal, regrowth, and roads in 10 small experimental basins, western Cascades, Oregon. Water Resour. Res. 36, 2621–2642. Luce, C.H., Cundy, T.W., 1994. Parameter identification for a runoff model for forest roads. Water Resour. Res. 30 (4), 1057–1069.
Luce, C.H., Wemple, B., 2001. Introduction to the special issue on hydrologic and geomorphic effects of forest roads. Earth Surf. Process. Landforms 26 (2), 111–113. Luce, C.H., Black, T.A., 1999. Sediment production from forest roads in western Oregon. Water Resour. Res. 35 (8), 2561–2570. Megahan, W.F., 1972. Subsurface flow interception by a logging road in mountains of central Idaho. In: Scallany, S.C., McLaughlin, T.G., Striffler, W.D. (Eds.), Proceedings of the Watersheds in Transition Symposium. American Water Resources Association, Urbana, IL, pp. 350–356. Montgomery, D.R., Dietrich, W.E., 1994. A physically based model for the topographic control on shallow landsliding. Water Resour. Res. 30 (4), 1153–1171. Montgomery, D.R., 1994. Road surface drainage, channel initiation, and slope instability. Water Resour. Res. 30 (6), 1925–1932. Mosley, M.P., 1980. The impact of road erosion in the Dart Valley, Nelson. NZ J. Forestry 25 (2), 184–198. Pozzatti, A., Cerato, M., 1984. Note pratiche sulla progettazione delle strade forestali. L’Italia forestale e montana 5, 263–274 (in Italian). Reid, L.M., 1981. Sediment production from gravel-surfaced forest roads, Clearwater basin, Washington. Publication FRI-UW-8108. University of Washington, Fisheries Research Institute, Seattle. Reid, L.M., Dunne, T., 1984. Sediment production from forest road surfaces. Water Resour. Res. 20, 1753–1761. Sidle, R.C., Pearce, A.J., O’Loughlin, C.L., 1985. Hillslope stability and land use. Water Resource Monograph, vol. 11. American Geophysical Union, Washington, DC, 140 pp. Swanson, F.J., Dyrness, C.T., 1975. Impact of clearcutting and road construction on soil erosion by landslides in the western Cascade Range. Oregon Geol. 3 (7), 393–396. Tague, C., Band, L., 2001. Simulating the impact of road construction and forest harvesting on hydrologic response. Earth Surf. Process. Landforms 26, 135–151. Tarboton, D.G., 1997. A new method for the determination of flow directions and contributing areas in grid digital elevation models. Water Resour. Res. 33 (2), 309–319. Wemple, B.C., Hones, J.A., Grant, G.E., Selker, J.E., 1996. Runoff generation mechanisms in a steep, forested catchment: controls on flow contributions to a road network. EOS Trans. Am. Geophys. Union 77 (46), F188. Wemple, B.C., Swanson, F.J., Jones, J.A., 2001. Forest roads and geomorphic process interactions, Cascade Range, Oregon. Earth Surf. Process. Landforms 26, 191–204. Wemple, B.C., Jones, J.A., 2003. Runoff production on forest roads in a steep, mountain catchment. Water Resour. Res. 39 (8), 1220. Wigmosta, M.S., Perkins, W.A., 2001. Simulating the effects of forest roads on watershed hydrology. In: Wigmosta, M., Burgess, S. (Eds.), Land Use and Watersheds: Human influence on Hydrology and Geomorphology in Urban and Forested Areas. Water Science Application Series, vol. 2. AGU, Washington, DC, pp. 127–143. Ziegler, A.D., Giambelluca, T.W., 1997. Importance of rural roads as source areas for runoff in mountainous areas of northern Thailand. J. Hydrol. 196, 204–229.