«HOW WIDE IS A ROAD? THE ASSOCIATION OF ROADS AND MASS- WASTING IN A FORESTED MONTANE ENVIRONMENT MATTHEW C. LARSEN* AND JOHN E. PARKS US Geological ...»
EARTH SURFACE PROCESSES AND LANDFORMS, VOL 22, 835–848 (1997)
HOW WIDE IS A ROAD? THE ASSOCIATION OF ROADS AND MASS-
WASTING IN A FORESTED MONTANE ENVIRONMENT
MATTHEW C. LARSEN* AND JOHN E. PARKS
US Geological Survey, GSA Center, 651 Federal Blvd, Guaynabo, Puerto Rico 00965, USA
Received 11 March 1996; Revised 27 July 1996; Accepted 6 August 1996
Earth surf. process. landforms, 22, 835–848 (1997) No. of ﬁgures: 9 No. of tables: 2 No. of refs: 41 KEY WORDS: mass-wasting; landslides; disturbance; forests; GIS; highways; humid topics
INTRODUCTIONMass-wasting has confounded road builders for as long as humans have constructed transportation routes through mountainous terrain. This is particularly true in mountainous humid-tropical settings where frequent high-intensity rainfall often results in widespread mass-wasting. The term ‘mass-wasting’, also referred to here as landsliding, refers to the downward and outward movement of hillslope-forming materials – natural rock, soils, artiﬁcial ﬁlls or combinations of these materials (Schuster, 1978). Mass-wasting can include falls, topples, slides, spreads and ﬂows. These phenomena are part of the process of hillslope erosion that is responsible for introduction of sediment into streams, rivers, lakes, reservoirs and ﬁnally the oceans. When highways are constructed in mountainous environments, the frequency of mass-wasting commonly increases (Varnes, 1978).
However, in most settings the actual mass-wasting zone of disturbance associated with highway construction is not well known.
Determination of the mass-wasting zone of disturbance is important for land-use managers and highway engineering who must deal with the costly and sometimes life-threatening problems caused by road-related landslides. Realistic assessment of the impact of proposed highway construction must be quantiﬁable as road costs may be signiﬁcantly increased by landsliding during construction (Sowers, 1971). In addition, the environmental impact of new road construction may be greater than anticipated if the effect of landslides associated with highways is only considered immediately proximal to the road. Finally, understanding of forest disturbance regimes is one of the dominant thrusts in modern forest ecology. Improved cognizance of the combined effects of anthropogenic and mass-wasting disturbance may be essential to the characterization of disturbance regimes in montane, landslide-prone environments. It is therefore important to determine: (1) to what degree the presence of roads in mountainous terrain is associated with landslide frequence; and (2) the width of the zone of mass-wasting disturbance identiﬁed with highways.
* Correspondence to: M. C. Larsen © 1997 by John Wiley & Sons, Ltd.
CCC 0197-9337/97/090835–14 $17.50 836 M. C. LARSEN AND J. E. PARKS A 201km2 area of Puerto Rico that includes secondary forest and relatively undisturbed forest, known as the Luquillo Experimental Forest (LEF), was selected to address these questions. This study area encompasses the topographic and climatic conditions that typify much of rural Puerto Rico as well as many other humid-tropical regions. The objectives of this study were to determine: (1) if a zone of higher landslide frequency could be recognized adjacent to roads when compared to landslide frequency in areas distant from roads; (2) the width of that disturbance zone; and (3) the rate of mass-wasting disturbance in that zone. This work was accomplished using a digitized spatial data base that included 1609 landslide locations and dimensions mapped from aerial photographs, topography, and a road network from 1:20 000 scale topographic maps (Larsen and TorresSánchez, 1996). The data were analysed with vector-based geographic information system (GIS) software, ARC/INFO (ESRI, 1993).
PREVIOUS WORKThe principal causes of anthropogenically induced instability result from increased weight on the hillslope from ﬁll, hillslope oversteepening, removal of slope support in roadcuts, alteration of surface runoff paths, and enhanced runoff rates (Sidle et al., 1985). The effect of mass-wasting on roads has been evaluated from the perspective of numerous disciplines. Geomorphologists, engineers, geologists and geographers, among others, have assessed landslide frequency and magnitude in a variety of environments. Fredricksen (1970), Eckhardt (1976), Beschta (1978) and Duncan et al. (1987) studied the problem in humid-temperate settings. Much of their work focused on areas where logging was active and mass-wasting had caused extensive damage to access roads. Mass-wasting in these areas degraded ﬁsh habitat through the introduction of large amounts of sediment into rivers and streams. Wolfe and Williams (1986) determined that landslide frequency in areas impacted by logging and associated road building was increased by three to 26 times in comparison with nearby undisturbed forested areas. An extensive summary of landslide problems resulting from road building at sites mainly in the Paciﬁc Northwest is included in Sidle et al. (1985). They report rates of soil mass movement associated with roads that are as much as 300 times greater than the rates in undisturbed forest.
Studies of road-related mass-wasting in the humid tropics have been limited in part by the level of economic development of most nations in the tropics. The humid tropics have been deﬁned as those areas with consistently high receipt of solar radiation, heat and moisture (Reading et al., 1995). Anderson (1983) examined the inﬂuence of soil pore-water pressure on road cuts and showed that elevated pore pressure resulting from heavy rainfall caused slope failures along road cuts on the island of St Lucia, West Indies. In addition, he noted that angle and slope plan curvature were signiﬁcant factors in determing whether a road cut was likely to fail.
Maharaj (1993), working in a 15 km2 watershed in the Port Royal Mountains, Jamaica, documented 866 mostly rainfall-triggered landslides. The greatest frequencies of these landslides were associated with several bedrock types and slopes in excess of 20°. Fifty-four per cent of the landslides were mapped along highways, indicating a strong association of mass-wasting with anthropogenic landscape disturbance.
Haigh et al. (1988, 1993) related landslide frequency along Himalayan highways to such factors as the presence or absence of forest, hillslope angle, and rock or soil type. In Thailand, numerous landslides were triggered by a major storm in 1988 in the steeply sloping Khao Luang mountains (DeGraff, 1990). Although road-associated mass-wasting was not discussed by the author, he noted that landslides were larger and most abundant in the anthropogenically perturbed regions of the mountains where rubber was cultivated on steep slopes. Anthropogenic alteration of slopes has resulted in increased frequency of rainfall-induced slope failures in Singapore as well (Chatterjea, 1994). At two urban sites with grass-covered slopes, Chatterjea documented 103 failures after rainstorms. The failures were more abundant and of larger magnitude than in nearby forested areas. Pitts (1992) discusses a variety of slope stability problems that have occurred in Singapore, noting that slopes composed of ﬁll have been the sites of large landslides. He states that lithology is of only marginal importance because of the geotechnical similarities among weathered bedrock types in this humid-tropical setting. The more important factor is structure as shear strengths along structures (i.e. joint planes) are lower than shear strength in the intact material.
Road widening in the Sungei Gombak catchment, Malaysia, resulted in landsliding associated with intense rainfall (Douglas, 1967). Douglas (1968) describes an additional process in the Sungei Gombak catchment, 837
LANDSLIDE FREQUENCYidentical to that observed in parts of the LEF and elsewhere, in which large core stones that have been exposed by accelerated erosion of surrounding soil are destabilized. The core stone may then move, allowing soil and saprolite resting above it to slip downslope, often onto road surfaces. In much of Peninsular Malaysia, landslides are common in deep cuts for road and building excavations, mining and quarry sites, and where housing is constructed on steep terrain (Tan, 1984). Tan notes that inﬁltration from heavy rains on ﬁll and cut slopes is a particular problem for highway construction in this and other tropical settings.
Another body of literature documents various methods of evaluating landslide hazard through regional mapping and use of a GIS. Brabb (1995) demonstrated the utility of a GIS for assessment of geologic (earthquake, landslide and ﬂood) hazards in San Mateo County, California. This approach included the distribution of landslides in relation to bedrock dip, hillslope angle, soil inﬁltration and other factors. Also working in San Mateo County, California, Wentworth et al. (1987) described a similar approach where several soil engineering characteristics were mapped using a GIS. Wagner et al. (1988) used computerized spatial mapping along 400 km of roads in Nepal to evaluate the inﬂuence of slope and soil type in relation to landslide occurrence. They demonstrated a strong correlation between landsliding and bedrock type along roads.
Hydrologic factors and degree of rock weathering were deemed important as well. Mehrota et al. (1991), working in the Himalayas, examined landslide locations in relation to slope, lithology, land use, drainage and structure. These factors were used to derive a numerical weighting from which landslide susceptibility maps were developed. The distribution of mass movements in the Philippines was evaluated by Moore et al. (1991).
They attributed landslide occurrence to seismic activity and tropical storms and found that deforestation and construction of logging roads dramatically increased the extent and impact of landslides. At a site in Italy, Carrara et al. (1991) related the distribution of landslides to land use practices, particularly highway construction and maintenance. Kingsbury et al. (1991) used GIS software to evaluate landslide frequency in relation to hillslope angle using triangular irregular network (TIN) modelling to derive 5° slope angle categories for an area near Wellington, New Zealand. The TIN-generated slope characteristics were investigated with a landslide data base derived from aerial photographs. Using landslide frequency, three landslide-hazard classes were determined according to slope: 0–4° (low hazard), 5–15° (moderate hazard), and greater than 15° (high hazard).
In Puerto Rico, Molinelli (1984) documented the association of landsliding with the construction of a major trans-island highway constructed in the 1970s. Landslide problems associated with highway construction in the LEF were studied by Sowers (1971) after construction of a short stretch of mountain-top highway triggered dozens of small slumps and greatly increased the cost and time needed for completion of the road. Scatena and Larsen (1991) and Larsen and Torres-Sánchez (1991) described rainfall-triggered landslides in the LEF triggered by 200–300 mm of precipitation associated with Hurricane Hugo which struck the island in 1989.
These landslides were abundant on hillslopes that faced the prevailing wind-driven rain during the storm. The landslides were relatively small and shallow because of the short duration (6 h) of the rainfall. Larsen and Torres-Sánchez (1996) used a GIS to analyse 4000 landslides in a 900 km2 area of Puerto Rico. They determined that landslides were most abundant on east- and northeast-facing hillslopes that were anthropogenically modiﬁed, with gradients greater than 12° and elevations greater than 350 m.
The studies described above provide important insights into landslide frequency and magnitude in a number of mainly tropical environments. The characterization of hillslope angle over a region is common to many of the investigations as a method for evaluating landslide probability. In addition, many of the studies relate high landslide frequency to the alteration of slope and soil stability associated with construction of highways. None of the analyses, however, attempt to determine how far the zone of landslide disturbance extends from roads.
SETTING Puerto Rico is the smallest island of the Greater Antilles, located about 1700 km southeast of Miami, Florida (Figure 1). The island is in the trade-wind belt at the boundary between the Caribbean Sea and the Atlantic Ocean at 18°N, 66°W. Because of tectonically controlled geologic complexity and strong orographic control on island rainfall distribution, a variety of land-use, topographic and soil characteristics exist in the relatively small (9000 km2) area of Puerto Rico. Several major bedrock types, typical of island arc systems throughout the 838 M. C. LARSEN AND J. E. PARKS Figure 1. Location of the Luquillo Mountains study area and annual average rainfall isohyets (in mm). Inset shows position of Puerto Rico in the Caribbean basin. Rainfall data from Calvesbert (1970).