Sediment Yield, Spatial Characteristics, and the Long-Term

Home , Earthflow

Sediment yield, spatial characteristics, and the long-term evolution of active earthﬂ ows determined from airborne LiDAR and historical aerial photographs, Eel River, California

Benjamin H. Mackey†, and Joshua J. Roering Department of Geological Sciences, 1272 University of Oregon, Eugene, Oregon 97403, USA

ABSTRACT period, equating to a regional yield of 1100 by sliding along transient shear surfaces with t km–2 yr–1 (~0.45 mm/yr) if distributed across a degree of internal deformation or fl ow. They In mountainous landscapes with weak, the study area. As such, a small fraction of span a range of landslide failure styles and have fi ne-grained rocks, earthfl ows can dominate the landscape can account for half of the re- been variably described as landslide complexes erosion and landscape evolution by supply- gional denudation rate estimated from sus- (Iverson, 1986a), landslides (Schulz et al., ing sediment to channels and controlling pended sediment records (2200 t km–2 yr–1 or 2009b), earth slides (Cruden and Varnes, 1996), hillslope morphology. To estimate the con- ~0.9 mm/yr). We propose a conceptual model and mudslides (Chandler and Brunsden, 1995; tribution of earthfl ows to regional sediment for long-term earthfl ow evolution wherein Glastonbury and Fell, 2008). Hereafter, we budgets and identify patterns of landslide ac- earthfl ows experience intermittent activity use the term earthfl ow as advocated by Hungr tivity, earthfl ow movement needs to be quan- and long periods of dormancy when limited et al. (2001) to describe large, slow-moving tifi ed over signifi cant spatial and temporal by the availability of readily mobilized sedi- landslides with macroscale, fl ow-like morphol- scales. Presently, there is a paucity of data ment on upper slopes. Ultimately, high-order ogy. These features morphologically resemble that can be used to predict earthfl ow behav- river channels and ephemeral gully networks “earth glaciers.” They are generally slow mov- ior beyond the seasonal scale or over spatially may serve to destabilize hillslopes, control- ing (<4 m/yr), large (>500 m long), deep seated extensive study areas. Across 226 km2 of rap- ling the evolution of earthfl ow-prone terrain. (>5 m thick), and mechanically dominated by idly eroding Franciscan Complex rocks of the fi ne-grained material, and they behave in a plas- Eel River catchment, northern California, INTRODUCTION tic or visco-plastic manner (e.g., Putnam and we used a combination of LiDAR (light de- Sharp, 1940; Kelsey, 1978; Iverson and Major, tection and ranging) and orthorectifi ed his- Earthfl ows 1987; Zhang et al., 1991b; Bovis and Jones, torical aerial photographs to objectively map 1992; Baum et al., 1993; Chandler and Brunsden, earthfl ow movement between 1944 and 2006. In mountainous landscapes without active 1995; Malet et al., 2002; Coe et al., 2003; By tracking the displacement of trees grow- glaciers, sediment production and hillslope Mackey et al., 2009). Earthfl ows classically ing on earthfl ow surfaces, we fi nd that 7.3% form are primarily controlled by mass-wasting have an hourglass planform shape, with an of the study area experienced movement over processes, including slow-moving, deep-seated amphitheater-like source zone, an elongate nar- this 62 yr interval, preferentially in sheared landslides known as earthfl ows. These large row transport zone, and a lobate compressional argillaceous lithology. This movement is dis- slope failures can establish or modify drainage toe or depositional area (Keefer and Johnson, tributed across 122 earthfl ow features that patterns, alter hillslope morphology, and impart 1983). Topographically, earthfl ows are com- have intricate, elongate planform shapes, a large perturbations on sediment budgets by reg- monly found on hillslopes with low planform preferred south-southwesterly aspect, and ulating the timing, magnitude, frequency, spatial curvature (Ohlmacher, 2007). a mean longitudinal slope of 31%. The dis- distribution, and grain size of sediment entering Active earthfl ows exhibit seasonal movement tribution of mapped earthfl ow areas is well- a channel or river network (Hovius et al., 1998; patterns primarily governed by precipitation and approximated by a lognormal distribution Korup, 2005b, 2006). This has direct implica- groundwater levels, and they can require several with a median size of 36,500 m2. Approxi- tions for fl ooding, channel network evolution, weeks of cumulative rainfall before the onset of mately 6% of the study area is composed sediment transport, aquatic habitat, and infra- movement (Kelsey, 1978; Iverson and Major, of earthfl ows that connect to major chan- structure. While the importance of landsliding 1987). Earthfl ows can potentially dominate nels; these fl ows generated an average sedi- in mountainous landscapes has been long recog- sediment delivery to channels in erosive land- ment yield of 19,000 t km–2 yr–1 (rock erosion nized, we are still challenged to understand and scapes (Putnam and Sharp, 1940; Swanson and rate of ~7.6 mm/yr) over the 62 yr study quantify the ways in which landslides, and in Swanston, 1977; Kelsey, 1978), and yet they particular earthfl ows, modulate the rate and lo- seldom fail catastrophically (Iverson, 2005). We cations of sediment fl ux, and control landscape specifi cally distinguish earthfl ows from large- † Present address: Division of Geological and form, over geomorphically signifi cant time displacement, catastrophic single-event failures, Planetary Sciences, California Institute of Technol- ogy, 1200 E. California Blvd., Pasadena, California scales (Dietrich et al., 2003; Korup et al., 2010). such as rockslides, translational bedrock slides, 91125, USA. Slow-moving earthfl ows are a class of land- debris fl ows, and rotational slumps (Cruden and E-mail: [email protected] slide (Cruden and Varnes, 1996) characterized Varnes, 1996).

GSA Bulletin; July/August 2011; v. 123; no. 7/8; p. 1560–1576; doi: 10.1130/B30306.1; 14 ﬁ gures; 1 table; Data Repository item 2011099.

Extensive work has been undertaken on the 1978; Keefer and Johnson, 1983; Bovis and we lack conceptual models describing the behavior of individual earthfl ows at the daily and Jones, 1992; Ohlmacher, 2007). Accurate inven- long-term evolution of an earthfl ow-prone seasonal scales (e.g., Iverson and Major, 1987; tories of regional earthfl ow activity will enable landscape, and even the long-term evolu- Malet et al., 2002; Coe et al., 2003; Schulz et al., us to quantify rates of earthfl ow-generated sedi- tion of an individual earthfl ow. For clay-rich, 2009a). Other research foci include specifi c earth- ment yield and identify spatial patterns of slide earthfl ow-prone hillslopes (Fig. 1), there is fl ow mechanics such as shear zone dilatancy and activity that infl uence long-term movement. nothing comparable to the colluvial hollow strengthening (Iverson, 2005; Schulz et al., 2009b) and debris-fl ow framework applicable to soil- and the evolution of earthfl ow material strength Long-Term Earthfl ow Evolution mantled uplands (Dietrich and Dunne, 1978; over time (Maquaire et al., 2003). Many attempts Lehre and Carver, 1985), or the competing have been made to model earthfl ow mechan- When observing an earthfl ow-prone land- soil creep and fl uvial erosion processes ap- ics and rheology (e.g., Brückl and Sehidegger , scape, especially on high-resolution digital plicable to gentle ridge and valley topography 1973; Craig, 1981; Savage and Chleborad, 1982; topography (Fig. 1), a striking feature is that (Perron et al., 2009). Iverson, 1986c; Vulliet and Hutter, 1988; Baum nearly all hillslopes appear to have been af- In an attempt to address this knowledge gap, et al., 1993; Angeli et al., 1996; Savage and fected by mass movement, and the morphology we here propose a largely unexplored alterna- Wasowski , 2006; van Asch et al., 2007), primarily of earthfl ow and landslide activity is widespread tive to theories of long-term external forcing. to describe or replicate the observed behavior of (Putnam and Sharp, 1940). Multiple generations Earthfl ow activity may be dependent on fac- specifi c earthfl ows. of dormant slope failures appear to surround the tors intrinsic to individual earthfl ows, namely, comparatively small fraction of active terrain. the long-term balance of soil and rock entering Earthfl ow Spatial Patterns and This observation has prompted speculation that and leaving the earthfl ow. The rate of material Sediment Yield longer-term external forcing may modulate leaving the earthfl ow via mass translation and earthfl ow activity, such that many of the dormant fl uvial erosion is unsustainably fast over long To assess the contribution of landsliding to slope failures we see in the landscape today are time periods (Mackey et al., 2009), whereas regional erosion rates, we must quantify the an artifact of past periods of activity. Sources of the supply of readily mobilized earthfl ow col- transfer of sediment from slope failures to ac- external forcing include glacial-interglacial or luvium is limited by rates of bedrock weather- tive channels and compare this sediment yield shorter-term variations in climate (e.g., Bovis ing and expansion of the earthfl ow source zone. to catchment-wide erosion data sets, such as and Jones, 1992; Fuller et al., 2009), episodes The imbalance of material entering and exiting suspended sediment records. Complicating of base-level fall (Palmquist and Bible, 1980), the earthfl ow can only be reconciled with brief this task, slope failures vary in timing, loca- earthquakes (Lawson, 1908; Keefer, 1984), or periods of earthfl ow activity, separated by long tion, size, mechanism, postfailure behavior, and land use (Kelsey, 1978). periods of dormancy, to allow time for mass to the effi cacy of sediment delivery to the chan- Although the seasonal behavior of indi- accumulate via weathering and recharge the nels (Benda and Dunne, 1997; Lave and Bur- vidual earthfl ows has been well documented, earthfl ow source zone. bank, 2004; Korup et al., 2010). Some types of landslides (e.g., debris slides and translational 400 500 landslides) can deliver all the failed mass to m 600 m + m 40.0794°°N channels shortly after failure, and such behavior 700 123.4811°°W is amenable to approximating landslide-derived m erosion rates by magnitude-frequency statistics of landslide areas and volumes (e.g., Hovius

et al., 1997; Lave and Burbank, 2004; Stark and 300 m Guzzetti, 2009; Larsen et al., 2010). In contrast, individual earthﬂ ows can continue moving and actively supplying sediment to a channel for

hundreds of years (Bovis and Jones, 1992), 200 so the rate of sediment fl ux at the earthfl ow- m channel interface is more pertinent to sediment yield than earthfl ow size or volume. To assess the contribution of earthfl ow movement to regional erosion rates, earthfl ow velocities and sediment yields must be accurately mapped and averaged over suffi ciently large temporal and spatial scales. This helps to ensure

that estimates of earthfl ow-generated sediment Eel yield are not biased by variations in decadal-scale R climate (Bovis and Jones, 1992; Mackey et al., 200 ive Boulder Creek m r 2009), or dominated by a peak in activity (e.g., 0 250 500 m a surge) from an individual earthfl ow (Kelsey, 1978). In sharp contrast to the many data sets Figure 1. Shaded relief image of light detection and ranging (LiDAR)–derived digital topog- of catastrophic landslides (e.g., Malamud et al., raphy of earthfl ow-prone hillslope along the Eel River, California. Trees have been fi ltered 2004a), long-term, regional-scale inventories of from the LiDAR data, leaving the bare earth. This hillslope has multiple generations of slow-moving landslides are scarce (cf. Kelsey, earthfl ow activity with a range of sizes and failure styles.

Geological Society of America Bulletin, July/August 2011 1561 Mackey and Roering

Active Earthfl ow Characteristics, Sediment N Yield, and Landscape Evolution Eureka To assess the contribution of earthfl ows to regional sediment yield, and their control of landscape form, we investigated some variables that CA control sediment fl ux in earthfl ow terrain. These Sco Van variables provide a basis for quantifying regional CM Duzen earthfl ow erosion at geomorphically signifi cant time scales. We asked four basic questions: (1) how much of the terrain is active, (2) do ac- FS tive earthfl ows have characteristic spatial attri- AP butes such as slope, lithology, aspect, planform shape, probability distribution as a function of area, position on the hillslope, or connectivity LiDAR NF Eel to major channels, (3) what proportion of the coverage suspended sediment load in a channel network is attributable to earthfl ow mass movement, and Pacific MF Eel (4) can we combine remote-sensing data and fi eld Ocean DR observations to develop a conceptual framework SF Eel for the evolution of earthfl ow terrain? To address these questions, we focused on Eel the Eel River catchment in northern Califor- River nia (Fig. 2), where the combination of weak mélange rock, active tectonics, and high rates of seasonal rainfall leaves the watershed espe- 0 20 40 km 2 39.2560° N cially prone to slope instability. The 9450 km +123.2028° W Eel River has the highest sediment yield (2200 t km–2 yr–1) of any nonglacial river of its size in Figure 2. Eel River catchment and location of the study area along the contiguous United States (Brown and Ritter , main stem Eel River, northern California. Light detection and rang- 1971; Wheatcroft and Sommerfi eld, 2005) and ing (LiDAR) data were acquired over the 230 km2 region shaded in is an ideal location to study active earthfl ow white. Branches of the Eel River include Middle Fork (MF), North processes. We generated detailed, regional- Fork (NF), and South Fork (SF). Locations mentioned in text: scale maps of active earthfl ows underpinned by Alderpoint (AP), Dos Rios (DR), Fort Steward (FS), Scotia (Sco), airborne LiDAR, historical aerial photography, Cape Mendocino (CM). Note the strong northwest-trending struc- and fi eld inspection. We describe the challenges tural grain visible in the shaded relief background map. of accurately mapping active movement in in- herently unstable terrain, and outline methods we developed to objectively map active earth- and shale turbidite sequences (McLaughlin et al., north at ~5 cm/yr (Furlong and Govers, 1999) fl ows in this landscape. 2000). During oblique dextral translation, large and have a profound infl uence on the landscape, blocks of older, metasandstone, metabasalt, and causing river capture and drainage reversals REGIONAL STUDY AREA—EEL RIVER, blueschist-grade rocks were incorporated into the (Lock et al., 2006). A series of northwest-trend- NORTHERN CALIFORNIA penetratively sheared matrix of the Central belt ing emergent fault systems cut through the from the older eastern units. These more compe- northern California Coast Ranges associated Lithology and Tectonics tent blocks have a signifi cant local infl uence on with the advance of the Mendocino triple junc- the topography, persisting as erosion-resistant tion (Kelsey and Carver, 1988). The northern California Coast Ranges are com- topographic highs amid the mélange (Fig. 3). posed of the Franciscan Complex, a pene tratively The northwest-trending structural grain strongly Uplift Rates, Erosion, and Geomorphology sheared set of metasedimentary rocks that repre- infl uences the modern topog raphy, with major sents a Jurassic–Cretaceous accretionary prism. axial drainages and ridges trending northwest Due to active tectonics, weak rocks, and high The Franciscan Complex consists of three struc- (McLaughlin et al., 1982). rates of erosion, many geomorphic studies have turally separated belts, the Eastern, Central, and The postemplacement tectonic history of focused on the Eel River watershed. Localized Coastal belts, refl ecting the cumulative accretion northern California Coast Ranges has been rock uplift rates along the northern California of oceanic sediments to western North America dominated by the northerly migration of the coast approach 5 mm/yr near Cape Mendocino (Jayko et al., 1989; McLaughlin et al., 2000). Mendocino triple junction (MTJ) since the Mio- (Merritts and Bull, 1989) (Fig. 2), although The Central belt is especially prone to landslid- cene. The Pacifi c and Gorda plates are translat- these rapid rates appear to be related to uplift ing. It runs through much of the Eel River catch- ing north relative to North America along the of the coastal King Range (Dumitru, 1991) and ment and consists of an extensive Late Jurassic Cascadia megathrust, creating the San Andreas likely do not project inland. to Middle Cretaceous argillaceous mélange ma- fault to the south (Furlong and Schwartz, 2004). Several approaches have been used to constrain trix, encompassing blocks and slabs of sandstone This generates two zones of uplift that migrate background levels of rock uplift and erosion in the

1562 Geological Society of America Bulletin, July/August 2011 Earthﬂ ow activity and erosion, Eel River, California

+ 40.1496°°N N Hillslope geomorphology within the Eel 123.5818°°W River watershed is well described by Kelsey (1980) and Muhs et al. (1987), who highlighted the roles of contrasting geomorphic processes operating in different rock types. Topography is generally described as either “hard” or “soft,” Sp depending on its morphology and resistance to mass wasting. The harder, competent, steep sandstone rocks in the Franciscan Complex feature well-organized ridge and valley drainage networks, with erosion dominated by fl uvial and debris-fl ow incision (e.g., Stock and Dietrich, 2006), and local relief can exceed 1000 m. Toe- cm1 slope failures, known as debris slides, are also cm2 cm2 common in harder rocks (Kelsey, 1980; Kelsey et al., 1995). In contrast, the weaker, fi ne-grained mélange units (soft topography) have a dense but poorly developed drainage network, and longer low-gradient slopes (30–35%), and erosion is dominated by earthfl ows and ephemeral gullies Active (Fig. 1). The most comprehensive study of earth- earthflows fl ow processes in the Eel River catchment was undertaken by Kelsey (1977, 1978, 1980) in the Van Duzen river watershed, a large tributary of the Eel River at the northern end of the drainage basin (Fig. 2). Although comprising just 1% of the Van Duzen watershed area, earthfl ows contributed 10% of sediment to the channel.

cb1 Main Stem Eel River Study Site Legend Eel River This study focuses on a remote section of the Eel main stem Eel River between Dos Rios and Alder- Active earthflow Riv er point (Fig. 2). Brown and Ritter (1971) noted that cm1-Meta-argillite this site has one of the highest concentrations of cm2-Meta-argillite earthfl ow activity and sediment yield in the Eel River catchment and advocated further studies cb1-Sandstone of sediment production in this location. Lithol- ss-Sandstone Boundary of ogy in this area is predominantly the argillaceous study area mélange of the Central belt Franciscan Com- Sp-Serpentinite plex (Fig. 3), characterized by long, low-gradient 024km Resistant blocks slopes and extensive slope instability (Fig. 1). The argillaceous mélange is not conducive to conifer Figure 3. Simplifi ed lithological map modifi ed from Jayko et al. (1989) and McLaughlin growth and was open oak grassland at the time of et al. (2000) overlying a shaded relief map. See Table 1 for detailed descriptions. Active European settlement in the 1850s. This attracted earthfl ows are superimposed on the lithology. ranchers (Carranco and Beard, 1981), and the primary land use remains low-density cattle ranch- ing. Conifer growth and forestry are generally inland portion of the Eel River catchment. Fuller River catchment. This sediment yield equates to limited to isolated sandstone outcrops. Rainfall et al. (2009) used cosmogenic isotope measure- a catchment-averaged bedrock erosion rate of measured at Alderpoint averages 1.3 m/yr, and it ments and optically stimulated luminescence dat- ~0.9 mm/yr, and we adopt this as representative falls primarily between October and April. ing to calculate late Pleistocene–Holocene erosion of modern erosion rates. This period includes Slope failures in the study area range from rates for the upper South Fork Eel River (Fig. 2) of both anthropogenic effects (e.g., grazing, for- small slumps to huge earthfl ow complexes ~0.3 mm/yr (~750 t km–2 yr–1 assuming bedrock estry) and midcentury storms (Brown and Ritter, (Brown and Ritter, 1971; Mackey et al., 2009; density ρ = 2.5 g/cm3). Most relevant to the time 1971; Sloan et al., 2001; Sommerfi eld et al., 2002; Roering et al., 2009). On a single hillslope, earth- scale and focus of our study, Wheatcroft and Som- Sommerfi eld and Wheatcroft, 2007). Bed load is fl ows often exhibit complex crosscutting and merfi eld (2005) reanalyzed suspended sediment poorly constrained but decreases downstream as nested relationships and span a range of size, age, data (1950–2000) collected at Scotia (Fig. 2) and the weak rocks disaggregate, and it is signifi cantly activity state, and failure style (Fig. 1). Descrip- calculated an average suspended sediment yield less than suspended load at the Scotia gauging sta- tions of the large landslides and earthfl ows along of 2200 t km–2 yr–1 across 8063 km2 of the Eel tion (Brown and Ritter, 1971; Lisle, 1990). the Eel River canyon between Dos Rios and

Geological Society of America Bulletin, July/August 2011 1563 Mackey and Roering

Alderpoint were initially confi ned to engineering the ability of researchers and practitioners to Acquisition of Airborne LiDAR reports by the California Department of Water map mass-movement features across a broad Resources (Dwyer et al., 1971; Scott, 1973; area. LiDAR-based mapping allows greater ac- The National Center for Airborne Laser Map- Smith et al., 1974). More recently, Mackey et al. curacy in both correct feature identifi cation and ping (NCALM) acquired a high-resolution (1 m) (2009) studied an earthfl ow on Kekawaka Creek, location than is available with traditional aerial LiDAR data set of the 230 km2 study area in a tributary to the Eel River. They highlighted photo and fi eld mapping approaches. September 2006 (Fig. 2). The raw data were the steady decline in earthfl ow velocity since Although LiDAR maps have been a great processed by NCALM and converted to digital the 1970s and determined long-term movement improvement on previous approaches (Schulz, elevation models (DEMs). The elevation grids rates by measuring meteoric 10Be accumulated 2007), confi dently mapping active movement included both unfi ltered elevation (data include in the earthfl ow soil. Roering et al. (2009) used in terrain with multiple episodes of deep-seated tree and building elevations, Fig. 5) and fi ltered satellite-based Interferometric Synthetic Aperture failure can be a subjective exercise (Van Den or “bare earth” elevation models from which Radar (InSAR) to monitor seasonal movement Eeckhaut et al., 2005, 2007). In landscapes vegetation and structures had been removed of the Boulder Creek landslide, and estimated prone to pervasive slope instability with mul- (Fig. 1) (Carter et al., 2007; Slatton et al., 2007). a minimum erosion rate of 1.5 mm/yr (~3400 tiple generations of landsliding, distinguishing From the bare earth LiDAR data, we extracted t km–2 yr–1) across the earthfl ow source area. between active and dormant mass movement is statistics of regional slope and aspect. Very little data exist on the depth of large land- diffi cult without additional information on fea- slides and earthfl ows along the Eel River. The ex- ture activity (e.g., Fig. 1). Historical Aerial Photo Orthorectifi cation ceptions are two landslides in the lower reaches Some qualitative estimates of earthfl ow ac- of the Middle Fork Eel River (California Depart- tivity state or age can be made from landslide The earliest extensive aerial photo set cover- ment of Water Resources, 1970), ~34 km south- morphology (McKean and Roering, 2004; ing our fi eld area in the Eel River catchment was east of our study area (Fig. 2), but in com parable Booth et al., 2009). Over time, morphological fl own in 1944 (U.S. Forest Service DDD series). Central belt Franciscan rock. Boreholes in the features attributable to landslide or earthfl ow These 1:24,000 scale photos are of high quality; 1.6-km-long, 0.45 km2 Salt Creek landslide, movement (e.g., sharp headscarps, tension trees, buildings, and subtle topographic features 3.5 km upstream of Dos Rios, revealed a landslide cracks, compressional folding, and lateral are readily identifi able. We acquired medium thickness of 33–35 m toward the toe. The larger, margins) become smoothed out and attenuated resolution (30 μm) scans of 18 photos to pro- 1-km-long, 0.64 km2 Salmon Creek landslide is by small-scale surfi cial or diffusive processes vide coverage of the study area. 16 km upstream from Dos Rios, and a drill hole (e.g., soil creep, bioturbation) when a landslide For quantitative analysis, aerial photographs in this landslide was sheared off at 34 m. stops moving (Wieczorek, 1984; González- must be referenced to the ground and orthorecti- Diez et al., 1999). Figure 4 shows earthfl ows fi ed to remove lens distortions and topographic METHODS—OBJECTIVELY MAPPING adjacent to the Eel River that qualitatively effects (Wolf and Dewitt, 2000). This requires EARTHFLOW ACTIVITY WITH appear to span a range of activity states. Two the calibrated focal length of the lens, the co- HIGH-RESOLUTION TOPOGRAPHIC currently active features with fresh kinematic ordinates of fi ducial markers on the edge of the ANALYSIS AND PHOTOGRAMMETRY structures (such as lateral levees and advanc- photo, a digital elevation model of the terrain, ing lobes) adjoin a dormant landslide, where and ground-control points (GCPs) to colocate Background—Landslide Mapping much of the small-scale morphology has been features on the photo with features on the erased, leaving subtle head scarps and lateral ground. For GCPs, we identifi ed features such Landslide mapping has traditionally been margins. The older landslide has been deeply as small trees, buildings, or isolated rock out- undertaken by a combination of stereo-pair aerial dissected by an axial gully, which is typical of crops on unequivocally stable terrain (such as photo and topographic map analysis, in concert many dormant earthfl ows. ridges and terraces) and colocated the identical with fi eld verifi cation. Aerial photo analysis can effi ciently cover a large area, but at the expense of accuracy, especially in forested terrain, and problems arise in accurately relocating features on a photograph to the base topographic map (Malamud et al., 2004a). Field mapping yields Figure 4. Active versus dormant Eel Rive greater accuracy (especially with modern global earthfl ow morphology. Two ac- r positioning system [GPS] technology), but at tive earthfl ows show sharp mor- the expense of high labor costs and the limited phological features, whereas extent of terrain that can be covered (Wills and Active a large dormant earthfl ow to McCrink, 2002; Haneberg et al., 2009). In re- earthflows the south has had much of the cent years, satellite-based InSAR mapping has fi ner-scale morphology attenu- proven to be effective in locating slow, sus- Old “dormant” ated by surfi cial processes. The earthflow tained mass movement (e.g., Hilley et al., 2004). northern active earthfl ow prop- Roering et al. (2009) successfully identifi ed ac- agated downslope from 1952 to tive earthfl ows in the Eel River catchment using 1966. The abandoned Northern data from the L-band PALSAR ALOS satellite, Pacifi c Railroad runs along the although this record only extends back to 2006. southwest side of the Eel River. The increasing availability of high-resolution + topography acquired through airborne laser 39.9757°°N 500 m swath mapping (LiDAR) has greatly enhanced 123.4503°°W

1564 Geological Society of America Bulletin, July/August 2011 Earthﬂ ow activity and erosion, Eel River, California

Unfiltered 40.0837°°N vegetation 123.4491°°W+

Active flow boundary

Figure 5. Unfi ltered shaded relief map of the Boulder Creek earthfl ow showing tree displacement vectors depicting earthflow movement. Margins of active earthfl ows and individual tree displacements are shown in black. Inset shows displace- Eel River ment at the midsection of the fl ow, where it crosses a structural barrier running obliquely across the transport zone. Note the large toe advancing from the Displaced northeast. Eel River runs south trees to north along the toe of the Boulder Creek earthfl ow.

features on both the unfi ltered LiDAR shaded trees often continue to grow on the landslide tion between active and stable trees. In some relief image and the photo. mass as it translates downhill and are readily areas with few or no trees suitable for quantify- We were not able to locate camera informa- identifiable in photographs, on unfiltered ing displacement, we were able to detect move- tion or calibration reports for the 1944 series of LiDAR maps, and in the fi eld. By comparing ment based on changing earthfl ow morphology. photos, but an estimated focal length (210 mm) differences between the 1944 photos and the un- For example, advancing toes or headscarp retro- from a database of typical camera specifi cations fi ltered LiDAR data acquired in 2006, we were gression can be readily identifi ed by comparing (Slama, 1980) achieved a median rectifi cation able to track the locations of individual trees the photos with the LiDAR maps. We took a con- error of ~2.3 m. We used image-processing soft- and construct slope-corrected displacement servative approach to mapping: if we could not ware to measure the coordinates of fi ducial marks vectors for the 62 yr time interval. We used this confi dently map either vegetation displacement around the margins of the photos on true-scale approach to objectively map historically active or morphological change, we did not include the scanned photos and assumed a perfect principal earthfl ows and discriminate between stable and terrain as an active earthfl ow. Some earthfl ows, point of focus (x,y = 0,0) to construct the photo moving terrain. such as the Mile 201 slide, underwent extensive coordinate system. We used the 1 m2 LiDAR bare The distribution of trees varies across the landscaping work to protect a railway that runs earth elevation as the elevation model and the un- study area and between different earthfl ows. adjacent the river. As trees and features on these fi ltered 1 m2 LiDAR slope and shaded relief maps Some earthfl ows have many long-lived trees that earthfl ows had been altered by heavy machinery , as reference images to rectify the photographs can be used to construct a detailed vector fi eld we relied on engineering reports (e.g., Scott, within ERDAS Imagine 9.3 software. of displacement (Fig. 5). On other earthfl ows, 1973) to estimate total displacement. Where we Through this process, we could rectify the trees can be sparse, and the extent of movement mapped isolated displacements in different parts photos with a high degree of accuracy, and relo- is more diffi cult to objectively discern with tree of an earthfl ow, we amalgamated these into one cate features on the photos as they were placed displacement vectors alone. To delineate the larger contiguous feature when justifi ed by mor- on the ground in 1944. We compared the posi- margins of earthfl ows, we used an iterative ap- phology. Additionally, we distinguished stable tion of stable features on the LiDAR with the proach, carefully comparing the rectifi ed 1944 patches within a larger earthfl ow from the neigh- rectifi ed photos across the study site in order to photos and the LiDAR imagery. The primary boring mobile terrain. estimate orthorectifi cation error. guide to mapping an earthfl ow margin was the We visited earthfl ow features in the fi eld over boundary between stable and moving trees. The four fi eld seasons to confi rm the reliability of Estimating Earthfl ow Movement margins of active earthfl ows typically feature a our technique. On the ground, active earthfl ows subtle morphological structure, such as a head- exhibit fresh headscarps, exposed “mole-track” The earthfl ows of the Eel River catchment scarp, lateral levee or pressure ridge, or a toe- lateral margins with slickensides (Fig. 6A), commonly have isolated oak trees and bushes lobe thrust, and we used these morphological disturbed, densely cracked hummocky terrain growing on the earthfl ow surface (Fig. 5). These features to guide the fi ne-scale margin delinea- (Kelsey, 1978; Keefer and Johnson, 1983), fresh

Geological Society of America Bulletin, July/August 2011 1565 Mackey and Roering extensional cracks, and occasionally trees that Department of Water Resources, 1970) to guide a sediment yield (metric t km–2 yr–1), we assume a have been stressed, toppled, or killed by intense depth estimates of the larger features in our density of earthfl ow colluvium of 2.1 ± 0.1 g/cm3 ground deformation (Fig. 6B). Conversely, study area. Because velocities can vary across (Kelsey, 1978; Mackey et al., 2009). stable or dormant earthfl ows, while retaining an earthfl ow toe due to differential movement, One limitation with this approach is the poten- much of the hummocky terrain and character- where possible we averaged multiple measure- tial for surface velocities to represent a maximum istic macroscale earthfl ow morphology, do not ments of tree displacement to obtain represen- rate, rather than a depth-averaged fl ow velocity, exhibit the same degree of fresh ground distur- tative earthfl ow toe velocities. To calculate the owing to deformation within the earthfl ow mass. bance (Fig. 4). annual sediment delivery attributable to earth- Although the term “earthfl ow” implies signifi - We mapped earthfl ow features as detailed fl ow movement into a channel, we multiplied cant internal deformation, fi eld data for earth- polygons using ESRI ArcMap 9.2 software and the average annual toe velocity (from the 62 yr fl ows similar to those in our study site emphasize constructed a statistical database of the indi- photo-LiDAR–derived vectors) by the width and that the predominant mechanism is sliding or vidual earthfl ows. By combining LiDAR, recti- depth of the earthfl ow toe (Fig. 7). Based on re- plug fl ow, and most deformation (>75%) is ac- fi ed photos, and fi eld inspection, we objectively peat measurements, we assigned uncertainties of commodated in a narrow basal layer (Keefer and mapped landslide displacement (≥5 m) over the 25% for the depth, 10% for the width, and 10% Johnson, 1983; Vulliet and Hutter, 1988; Zhang 62 yr interval across the 226 km2 study area. for the velocity when calculating sediment fl ux et al., 1991a; Swanston et al., 1995; Savage and from each individual earthfl ow. To convert to Wasowski, 2006; Glastonbury and Fell, 2008). Earthfl ow Sediment Production

During winter, the earthfl ows push out into the channels, and erosion of the toe occurs by A slumping from the toe face. Fine sediment (clay, silt, and fi ne sand) is carried off in suspension during high fl ows (Brown and Ritter, 1971), and coarse sand and gravel accumulates as bed load in the channel. Earthfl ows can also transport large boulders (up to 15 m diameter), which are Figure 6. Field evidence for pushed into the channel, and which can armor earthfl ow activity. (A) Lateral the bed while they slowly weather and disag- margin of an active earthfl ow gregate in place, frequently creating knickpoints showing slickenlines. (B) View on the stream profi le (Kelsey, 1978). These up the active Penstock earthfl ow large boulders are estimated to contribute less showing the disturbed ground B than 2% of the earthfl ow volume, with 90% and distressed trees. Shrubs in of the earthfl ow colluvium fi ner than 76 mm foreground are ~1 m tall. in diameter (Smith et al., 1974). Much of the argillaceous rock rapidly disaggregates when subjected to fl uvial erosion and wetting and dry- ing processes in the channel and becomes suspended load. In calculating sediment production from earthfl ows, we distinguished earthfl ows that discharge sediment directly into a channel or major gully from fl ows that are disconnected from the channel network and do not represent active sediment sources. We only included the earthfl ows connected to channels when calculat- Penstock ing earthfl ow sediment production. earthflow

Landslide volumes and their contribution to Road sediment yield can potentially be estimated from Active flow Road the planform area (Hovius et al., 1997; Lave and margin Figure 7. Oblique view of the Burbank, 2004; Malamud et al., 2004a). Given Penstock earthflow looking Mean 35 m sparse constraints on earthfl ow depths, and the displacement southeast across Kekawaka small fraction of total earthfl ow volume that (1944–2006) Creek, illustrating the measure- enters the channel network in any given year, we 160 m wide ments taken to estimate annual focused on the depth and velocity of material at sediment fl ux from each earth- the earthfl ow toe. Kekawaka Creekek fl ow. Center of earthfl ow toe is We have no direct data as to the depth to the 40.1032° N, 123.4940° W. failure surface in our study area (e.g., from drill 8 m toe holes), but approximations can be made by calculating the toe height at the channel interface (Kelsey, 1978). We used the depth data from the N 1 m contours Middle Fork Eel River earthfl ows (California

1566 Geological Society of America Bulletin, July/August 2011 Earthﬂ ow activity and erosion, Eel River, California

If, for simplicity, we assume a Newtonian defor- was 0.4 (0.2–0.7) m/yr (median and interquar- River or major tributary), whereas 60 are located mation profi le, and that sliding accounts for 75% tile range). By comparing the locations of stable higher on hillslopes and do not discharge into of movement (leaving 25% for internal defor- features both on the photos and in the LiDAR large channels (Table DR1 [see footnote 1]). mation), then we are potentially overestimating data, we estimated the error associated with This characterization is distinct from our pre- mass fl ux by 10% when not accounting for in- photo rectifi cation. Rectifi cation error was small vious method of documenting earthfl ows that ternal deformation. Since the depth behavior of in comparison to the tree displacement, with a constitute active sediment sources (described earthfl ows in our study site is unconstrained, we median error of 2.3 (1.3–3.2) m (Fig. 9B). in section 3.5, “Earthfl ow Sediment Produc- do not explicitly include this in our calculations tion”). In that classifi cation, some earthfl ows of sediment delivery to channels. Rather, we Earthfl ow Shape and Area positioned high on slopes are bisected by or note that we are potentially overestimating sedi- feed into large gullies and thus indirectly supply ment delivery by ~10% if internal deformation is The planform shapes of individual earth- sediment to the primary fl uvial network, but do 25% of the total movement. Drag from earthfl ow fl ows vary greatly and can diverge signifi cantly not span ridgetop to channels. margins does not appear to have a major effect from the classical hourglass planform. Many on the planform velocity profi le. Lateral strain earthfl ows have complex and intricate margins, Slope and Lithology is primarily taken up by fault-like longitudinal whereby multiple small tributary earthfl ows structures on the margins or within the earthfl ow feed into a centralized transport zone (Figs. 5 We calculated the longitudinal slope of each body (Fig. 5) (Schulz et al., 2009b). By taking and 8). Earthfl ows can bifurcate around resis- earthfl ow (headscarp-to-toe elevation range/ the average measurements of multiple trees on tant topography, lower ridges, and capture the centerline length), in addition to the average the toe, we minimize errors from variations in drainage area of adjacent terrain. In several slope of all 4 m LiDAR DEM pixels (16 m2) planform velocity. cases, stable regions persisted within an earth- within each earthfl ow polygon (Fig. 12; Table fl ow complex (Fig. 8). The 122 earthfl ows have DR1 [see footnote 1]). To compare earth- RESULTS a median area of 36,500 m2 and interquartile fl ow topog raphy to the distribution of regional range of 12,500–117,000 m2. The largest active slopes, we calculated the slope of all 4 m pixels Spatial Distribution of Earthfl ows earthfl ow in the study site, the Boulder Creek in the study area. earthfl ow (3.1 km2; Fig. 5), is over 3 times the The longitudinal earthfl ow slope distribution Across our study area along the Eel River, area of the next largest feature (0.94 km2), (Fig. 12A) has an approximate normal distribu- we identifi ed 122 earthfl ow features that moved the Island Mountain earthfl ow (Fig. 8). The tion with a mean value of 31% ± 7% (mean ± during the interval 1944–2006 (Fig. 8). Over larger earthfl ows dominate small (<3 km2), standard deviation). Larger earthfl ows are gen- the study area of 226 km2, we mapped active scallop-like tributary catchments, which typi- erally less steep (Fig. 12B); specifi cally, no earthfl ows covering 16.5 km2, indicating 7.3% cally have low planform curvature, and could earthfl ows with an area over 250,000 m2 have of terrain moved between 1944 and 2006. We be described as “earthfl ow basins.” The Boulder longitudinal slopes exceeding 30%. note this to be a minimum estimate, since it is Creek earthfl ow is the dominant feature in the We compared normalized distributions of possible for earthfl ows with less than 5 m of ~15 km2 Boulder Creek catchment. 4 m slope pixels for the entire study area, and cumulative displacement to be active, but below The probability distribution of earthfl ow the 16.5 km2 of active earthfl ows, respectively our level of detection, or for earthfl ows without areas (Fig. 10) has a highly positive skew and is (Fig. 12C). The distributions are statistically dif- trees or notable surface deformation to have well described by a lognormal distribution with ferent based on an F-test (α = 0.05), but mean gone undetected. Table DR1 documents the spa- parameter values of μ = 10.65 ± 0.27 and σ = values are very close at 34% for earthfl ows and tial attributes of the individual earthfl ows.1 1.51 ± 0.17 (95% confi dence bounds). The tails 36% for all terrain. The study area has a greater of many landslide area distributions are recog- fraction of steeper terrain, as shown by a more Earthfl ow Displacements nized to show power-law behavior, with fre- positively skewed distribution than the earth- quency decreasing as the inverse power of area. fl ows (Fig. 12C). In total, we tracked displacements of 998 fea- The tail of the earthfl ow distribution, where area Of the 122 mapped active earthfl ows, 98 tures (trees, shrubs, or rocks) distributed across exceeds 80,000 m2, can be approximated by a (82% of the active earthflow area) occur in the 122 earthfl ows (with an average density of power law with an exponent of –1.06 (Fig. 10). the penetratively sheared mélange unit (cm1) one displaced feature in every 17,000 m2 of earth- There is a statistically different change in (Fig. 3; Table 1). fl ow terrain), shown on Figure 8. The distribu- the aspect ratio of earthfl ows as they increase tion of displacements is positively skewed, with in area. Earthfl ows with planform areas less Earthfl ow Aspect a median displacement of 23.9 m, and an inter- than 80,000 m2 have an average aspect ratio quartile range (middle 50%) of 14.6–42.3 m (averaged length/width) of 4.5 ± 3.1 (mean ± Kelsey (1978) noted that earthfl ows along the (Fig. 9A). The maximum displacement was standard deviation) (Fig. 11A). In comparison, Van Duzen River tend to have a southerly aspect. 175 m, recorded on the Boulder Creek earth- when earthfl ow area exceeds 80,000 m2, the as- To test whether active earthfl ows along our Eel fl ow (Fig. 5). Over the 62 yr interval, the median pect ratio is signifi cantly different at 7.2 ± 3.9 River study site have a preferential aspect, we earthfl ow velocity across all moving features (Fig. 11B), indicating that larger earthfl ows tend calculated the median aspect of 4 m DEM pixels to be more elongate (Fig. 11C). within each earthfl ow polygon (Fig. 13A). The 1GSA Data Repository item 2011099, Table DR1, Of the larger earthfl ows, 25 extend from the mean earthfl ow aspect is 210°, with a mean re- a .xls fi le that documents the spatial characteris- channel up to a ridgeline or major break in slope sultant vector of 0.42. Applying Rayleigh’s test tics and movement of the earthfl ows, and Figures at the scale of the hillslope lengths (1–3 km). for signifi cance of a mean direction, for this data DR1–DR2, which illustrate the stages of earthfl ow evolution described in Figure 14 with fi eld exam- Among the 97 earthfl ows that do not span from set, the critical resultant vector (α = 0.05, n > ples, is available at http://www.geosociety.org/pubs/ channel to ridge top at this scale, 37 have a 100) is 0.17, indicating that the 122 earthfl ows ft2011.htm or by request to [email protected]. prominent interface with a major channel (Eel have a signifi cant mean southwesterly aspect.

Geological Society of America Bulletin, July/August 2011 1567 Mackey and Roering

448000 452000 456000 460000 464000

Eel River Canyon Neafus 40.1495°N Earthflow activity 1944–2006 Peak 123.5818°W 4444000 (1213 m) 4444000

Alderpoint 7.5 km downstream KW 201 Kekawaka Eel Rive Creek

4440000 r 4440000

PS 4436000 4436000

Pipe Creek

4432000 IM 4432000

Chamise Creek

Eel River 4428000 4428000 Legend

Eel River

Tributary drainages 190 Active earthflow 1944–2006 displacement 1250 4424000 Elevation 4424000 (m) 95 Study 4 km North Fork area Dos Rios 35 km Eel River upstream 448000 452000 456000 460000 464000 Figure 8. Map of the 226 km2 study area, active earthfl ows, and tree displacements, from 1944 to 2006. Extent of light detection and ranging (LiDAR) coverage is shown as shaded relief, colored by elevation. Specifi c earthfl ows: Boulder Creek (BC), Island Mountain (IM), Mile 201 (201), Kekawaka (KW), Penstock (PS), Mile 190.5 (190). Background image is 30 m grid shaded relief. Coordinates are UTM Zone 10N.

1568 Geological Society of America Bulletin, July/August 2011 Earthﬂ ow activity and erosion, Eel River, California

A Length of earthflow displacement vectors parent preferred orientation of the terrain is not signifi cant at α = 0.05. The amalgamated earth- 80 fl ow terrain (Fig. 13B) has a pronounced bias to 60 n = 998 the southwest, with a mean direction vector of Median = 23.9 m 218°, and resultant vector of 0.31, and is there- 40 IQR = 14.6–42.3 m fore signifi cant at α = 0.05. Count 20 Figure 9. (A) Histogram of 998 Earthfl ow Sediment Flux tree displacements mapped 0 across the fi eld area. (B) Histo- 40 80 120 160 We estimated an average sediment fl ux of gram of rectifi cation errors, cal- 120,000 ± 10,000 m3/yr (errors summed in culated from the offset between Tree displacement (m) quadrature) for 62 earthfl ows that discharge di- stable features on the photo rectly into a channel or creek (see Table DR1 and light detection and rang- B Rectification error of stable features [see footnote 1]) and thus can be considered ing (LiDAR) data. Interquar- tightly connected to the fl uvial system. These 25 tile range (IQR) is middle 50% earthfl ows have a combined area of 13.6 km2 n = 220 of data. Note the scale change 20 (6.0% of the study area). This equates to a Median = 2.3 m from parts A to B. 15 sediment yield for the 62 earthfl ow features of IQR = 1.3–3.2 m 19,000 ± 2000, t km–2 yr–1, or a rock erosion

Count 10 rate of ~7.6 mm/yr (bedrock density 2.5 g/cm3). 5 When distributed over the study area of 226 km2, 0 the mass translation of active earthﬂ ows gen- 02468erates an average regional sediment yield of 1100 t km–2 yr–1 (~0.45 mm/yr). This rate is Rectification error (m) half of the 2200 t km–2 yr–1suspended sediment yield calculated for the Eel River (1950–2000) by Wheatcroft and Sommerﬁ eld (2005). Hence, 100 across our study area from 1944 to 2006, 6% of the terrain supplies approximately half of the Onset of power law sediment to channels. (80,000 m2) Median earthflow size DISCUSSION (36,500 m2) Erosion and Sediment Delivery

–1 10 α Our results show that from 1944 to 2006, Slope ( ) = –1.06 large, slow-moving earthfl ows were the pri- 2 (R = 0.98) mary process of erosion along the main stem Eel River, preferentially occurring in weak mélange lithology. Mass translation by earth- Earthflow data fl ows contributed ~120,000 m3/yr (~250,000 Lognormal fit (all data) t/yr) to the channel network in our study area. 2 Power-law fit (x > 80,000) The sediment yield from the 13.6 km of active Cumulative probability of slide number 10–2 earthfl ows in Eel Canyon (19,000 t km–2 yr–1) is similar to a comparable photogrammetric study 104 105 106 (1941–1975) by Kelsey (1978) of Van Duzen Earthflow area (m2) earthfl ows (24,900 t km–2 yr–1). When averaged over the 226 km2 study area, sediment delivery Figure 10. Distribution of earthfl ow areas. Cumulative probability plot of earthfl ow area, via active earthfl ows (1100 t km–2 yr–1) accounts showing the lognormal fi t over the whole data set, and a power-law fi t to earthfl ow areas for approximately half the average sediment exceeding 80,000 m2. yield of the Eel River catchment during the sec- ond half of the twentieth century (2200 t km–2 yr–1), despite earthfl ows encompassing just 6% Given the strong northwest-trending structural discriminate by individual earthfl ow as in Fig- of the study area terrain. control on topography, we plotted the aspect of ure 13A. Figure 13B highlights the asymmetry We attribute the additional suspended sedi- all 4 m pixels within the study area to determine of regional aspect; the oblate circular rose dia- ment yield (~50%) to both fl uvial erosion of whether this biased the aspect of earthfl ows. To gram shows a bimodal distribution with slight the earthfl ow surfaces, and erosion from the properly compare the topographic trend with preference for cells facing southwest. The mean nonearthfl ow portion (~94%) of our study area earthfl ow terrain, we identifi ed the aspect of all direction of all topography is 246°, with a small in the form of gullies, streambank erosion, soil 4 m pixels within active earthfl ows, but did not mean resultant vector of 0.04, such that this ap- creep, and isolated shallow landslides. In hard

Geological Society of America Bulletin, July/August 2011 1569 Mackey and Roering

of inactive earthfl ow features in the landscape. 100 y = 0.5x 0.21 Earthfl ow movement may be the dominant local 0.3 r 2 = 0.26 A erosional process for only a brief time. n = 122 Our 62 yr study period averages across 10 decadal-scale variation in earthfl ow movement 0.2 and does not capture potential fl uctuations in earthfl ow activity over longer time scales. Bovis Aspect ratio (L/W) 1 and Jones (1992) used tree ring data and tephro- chronology to correlate earthfl ow movement 0.1 103 104 105 106 107 with long-term (102–104 yr) changes in climate, C Area of all flows (m2) and it is possible that earthfl ows in the Eel River were more active during the cooler and wet- ter climatic conditions during the Last Glacial 0 Maximum (e.g., Adam and West, 1983; Barron 0 481216 20 24 et al., 2003). This does not require that inactive Aspect ratio of flows <80,000 m2 earthfl ow terrain be a relic of past climatic conditions, however, because other mechanisms, B 0.2 such as internal forcing within individual earthflows, can account for the distribution A of earthfl ow activity we observe today (see sec- mean 4.54, st. dev 3.10 n 77 tion “Long-Term Earthfl ow Evolution” below). B An important fi nding is that 1.3% of the study 0.1 mean 7.17, st. dev 3.88, n 45 area (17% of the total active earthfl ow area) moved during the study period but did not reach

Probability Probability F = 16.9, P = 0.000 active channels. Earthfl ows that redistribute material on hillslopes but do not deliver sediment directly to channels are an important consideration when calculating sediment budgets. These un- 0 connected earthfl ows may predispose the terrain 0 4 8 12162024to erosion by other processes, such as localized gullying (Schwab et al., 2008), or ultimately lead Aspect ratio of flows >80,000 m2 to reactivation of larger earthfl ows by changing Figure 11. Earthfl ow shape aspect ratio. (A) Histogram of earthfl ow aspects with area the distribution of stress on the hillslopes. <80,000 m2. (B) Histogram of earthfl ow aspects with area >80,000 m2. In A and B, horizontal Brown and Ritter (1971) noted that the sec- bar shows mean ±1 standard deviation. (C) Log-log plot showing dependence of earthfl ow tion of the Eel River watershed between Dos aspect ratio on earthfl ow area. Rios and Scotia has greater sediment yield than the rest of the drainage due to the extensive mélange lithology and the high number topography, steep ridge and valley morphology fl ow terrain could exceed 30,000 t km–2 yr–1 of active earthfl ows, meaning our study area suggests that debris fl ows are a primary ero- (~12 mm/yr) if fl uvial erosion and gullying from along the main stem of the Eel River has sional process, although such terrain was sparse the earthfl ow surface is taken into account. a higher sediment yield than the Eel River in our fi eld area (Fig. 3; Table 1). There is po- We found that 6% of the landscape covered catchment average. The proportion of regional tential for fl uvial erosion of the surface of active by active earthfl ows contributed at least 50% erosion attributable to earthfl ow processes at earthfl ows through localized gullying (Kelsey, of the basinwide averaged suspended sediment the catchment scale may not be as large as 1978; Schwab et al., 2008; Roering et al., 2009), yield. The ratio of fractional earthfl ow area our results suggest, but it cannot be quantifi ed although we did not account for this in our es- to the fraction of earthfl ow-derived suspended with available data. timates of earthfl ow mass movement. In other sediment for the Eel River study site (6:50) ap- studies of northern California earthfl ows, esti- proximates that found by Kelsey (1978) in the Spatial Characteristics mates of fl uvial erosion of the earthfl ow surface Van Duzen watershed (1:10). Both studies illus- ranged from 10% (Nolan and Janda, 1995) to trate that the surface of active earthfl ows in Fran- Our results show that earthfl ows have a dis- 50% (Kelsey, 1978) of the total earthfl ow sedi- ciscan mélange erode an order of magnitude tinctive position in the landscape and character- ment fl ux. We predict our main stem Eel River more rapidly than catchment-averaged values, istic spatial attributes, including planform shape study site would fall somewhere between these highlighting the highly erosive nature of earth- and size distribution. The larger earthfl ows span values—the earthfl ows along the Eel River are fl ows. This rapid rate of erosion indicates that ridge-to-channel length scales and are signifi - not as active or as incised as those along the Van earthfl ows evolve at unsustainably fast rates over cantly more elongate than smaller fl ows. Earth- Duzen River during the time of Kelsey’s (1978) periods beyond several hundred years (Kelsey, fl ows that only partially span the length of the study, but they have higher rates of activity than 1978; Mackey et al., 2009) and so risk exhaust- hillslope are more common higher on the hill- those in Redwood Creek as documented by ing their supply of source material. The legacy slope toward the ridges, as opposed to the lower Nolan and Janda (1995). Therefore, the actual of cycles of rapid erosion followed by lengthy sections of hillslopes, where they are can inter- sediment yield from the 6% of active earth- periods of dormancy is refl ected in the ubiquity act with major gullies or creeks.

1570 Geological Society of America Bulletin, July/August 2011 Earthﬂ ow activity and erosion, Eel River, California

A Longitudinal slope of earthflow B Longitudinal slope versus earthflow area et al., 2002; Malamud et al., 2004a, 2004b; Korup , 2005a; Stark and Guzzetti, 2009). To our 60 12 Mean = 31 knowledge, the probability distribution of area St. dev = 7 has not been previously documented for large, 8 40 30% slope slow-moving earthﬂ ows. Frequently, a rollover or departure from 4 20

Frequency power-law decay occurs for small landslide areas, which is attributed to either a change in 0 0

Longitudinal slope (%) process, or sampling artifact due to mapping 0 10 20 30 40 50 103 104 105 106 107 resolution (Malamud et al., 2004a). The devia- Longitudinal slope (%) Earthflow area (m2) tion from power-law behavior in our earthfl ow data set (~80,000 m2) is well above the detec- 2 Normalized earthflow and study area slope histogram tion level of our methodology (<~1000 m ), (4 m grid – 1% bins) suggesting a process-based explanation rather 3 than mapping bias. The power-law exponent 1) Earthflows (α value) for large earthfl ows in our data set 2 Mean St. dev (1.06) is very low compared to other studies, 1) 34 17 which typically fi nd α values of ~1.5 (Stark and 2) 36 20 Guzzetti, 2009). Our data set also shows that the 1 2) Study area onset of power-law behavior occurs at a much Frequency (%) higher value of earthfl ow area (~80,000 m2) than 0 most landslide inventories, where the rollover 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 occurs at 700–5000 m2 (Hovius et al., 1997; Slope (%) Stark and Hovius, 2001; Malamud et al., 2004a; Stark and Guzzetti, 2009). Figure 12. (A) Longitudinal slope of individual earthfl ows. (B) Comparison of earthfl ow Once earthfl ows extend from the channel to longitudinal slope and area. (C) Histogram of slope of all 4 m pixels within the study area the ridgeline, they can only increase their size (black line), and all 4 m pixels within active earthfl ows (gray shading). by expanding laterally or incorporating tributary fl ows, which is not mechanically advantageous due to the planar nature of the hillslopes. Given Although the sample size of our earthfl ow our study site is 36,500 m2, which equates to an the 2–3-km-long hillslopes in our study area data set (n = 122) was modest in comparison to earthfl ow with an approximate length and width and a length:width aspect ratio for larger fl ows most landslide inventories (which can number in of 400 m and 100 m, respectively, given a 4:1 of ~7:1, the expected maximum earthfl ow size the thousands), our active earthfl ow map should length-to-width ratio. range would be 0.6–1.2 km2, consistent with be highly robust given the use of displaced fea- Landslide size distributions have received our observations (Fig. 10). At the ~1 km2 size, tures and LiDAR for defi ning slide margins. The considerable attention in the past decade because the cumulative earthfl ow size appears to diverge probability distribution as a function of earthfl ow the magnitude and frequency of slope failure are from the power-law trend (which is infl uenced area is very well described by a lognormal dis- primary controls on the rate of sediment fl ux by the Boulder Creek earthfl ow), suggesting tribution (Fig. 10). Lognormal distributions have from hillslopes (see review in Stark and Guz- that earthfl ows rarely attain such extents in this been proposed for landslides in Redwood Creek, zetti, 2009) and can be used to quantify the risk environment. We also note the Boulder Creek northern California (Kelsey et al., 1995), and for of slope failure hazards. Landslide frequency earthfl ow is an outlier on the lognormal fi t. submarine failures on the U.S. Atlantic continen- generally decays as the inverse power of land- The large size of the Boulder Creek earthfl ow tal slope (Chaytor et al., 2009; ten Brink et al., slide area, both for a total landslide inventory seems attributable to several large tributary 2009). Lognormal distributions have a charac- and an event-specifi c distribution (e.g., Hovius lobes and fortuitous placement in a confi ned teristic size, and the median earthfl ow area for et al., 1997; Stark and Hovius, 2001; Guzzetti basin. We note that larger earthfl ows also have

TABLE 1. PROPORTION OF LITHOLOGY AND ACTIVE EARTHFLOWS (EF) ACROSS THE STUDY AREA Lithology Map unit Area (m2) % study area No. EF EF area (m2) % EF area % lithology active Mélange—penetratively sheared meta-argillite cm1 1.32 × 108 58.2 98 1.33 × 107 80.89 10.08 Broken formation (metasandstone) cb1 5.75 × 107 25.4 16.5 2.10 × 106 12.71 3.65 Mélange (subequal argillite and sandstone) cm2 1.88 × 107 8.3 2.5 5.64 × 105 3.42 3.00 Serpentinite Sp 4.67 × 106 2.1 5 1.63 × 105 0.99 3.49 Greenstone RB 4.14 × 106 1.8 Block (unknown lithology) RB 2.66 × 106 1.2 Yolla Bolly (metasandstone) ss 2.61 × 106 1.2 Radiolarian chert RB 1.02 × 106 0.5 White rock (metasandstone) ss 1.23 × 106 0.5 Basaltic rocks RB 6.86 × 105 0.3 Broken formation (intact metasandstone) cb1 4.16 × 105 0.2 Chert RB 5.57 × 105 0.2 Blueschist blocks RB 1.29 × 105 0.1 2.26 × 108 100.0 122 16.5 × 107 Note: Map unit indicates representation in Figure 3. For detailed lithology descriptions, see McLaughlin et al. (2000).

Geological Society of America Bulletin, July/August 2011 1571 Mackey and Roering

A Median aspect of earthflows B Normalized aspect of study area earthfl ows and the study area supports our ob- 2 (Scaled by log10 area (m )) and earthflow terrain servation that much of the landscape has been modulated by earthfl ow processes, even if they 0 0 are not currently active. A slope of 34%–36% may be an approximate limiting threshold slope 315 45315 45 for earthfl ow activity, and possibly represents the residual shear strength of the mélange at Study area the landscape scale (e.g., Skempton, 1964; Hutchinson, 1967; Carson and Petley, 1970). The relative scarcity of terraces in our study site 270 90270 90 (Fig. 8) supports the contention that many of the 2 3456 0 0.4 0.8 hillslopes adjust to threshold values as the Eel Earthflows River incises (Korup et al., 2010). The fact that the very largest earthfl ows typically have lower 225135 225 135 longitudinal slopes (Fig. 12B) suggests that they are not simply relaxing back to a residual slope Mean and potentially have increased erosional effi - 180 180 resultant ciency, as discussed previously herein. vector The earthfl ows in our study site occur preferentially on south-southwest–facing slopes. Figure 13. Aspect of earthfl ows and the study area. (A) Median aspect of 4 m pixels within Kelsey (1978) suggested that slopes with a each earthfl ow polygon, scaled by log area (m2). Mean resultant vector is 210°. (B) Rose 10 southerly aspect are drier and do not support diagram of aspect of 4 m pixels across the study area, and all 4 m pixels within active earth- conifer growth, which preferentially stabilizes fl ows (1° bins). Mean resultant vector of the earthfl ow terrain is 218°. Study area terrain the north-facing slopes. Across our study area, does not have a signifi cant mean direction. lithology appears to have a stronger infl uence on vegetation than slope aspect, and many north- facing slopes have open grassland susceptible lower longitudinal slopes than average (Fig. processes of erosion in the argillaceous terrain to earthfl ow activity (e.g., Fig. 5). The primary 11C), which may refl ect greater interaction with compared to resistant sandstone outcrops, with difference between northerly and southerly as- channel/gully incision processes due to the in- earthfl ows absent in the latter (Table 1). Earth- pect is variation in incident solar radiation, so creased drainage areas inherent to larger earthfl ows move around and between the harder the preferred southerly aspect is plausibly at- fl ows. Alternatively, the larger fl ows may be mélange blocks, leaving them as steep localized tributable to differences in evaporation and soil thicker, and require less slope to attain the same topographic highs extending out of the mélange. moisture. This seems counterintuitive, given driving stress, or the longer transport zones of The exposed sandstone blocks either weather the dependence of earthfl ow movement on ele- large earthfl ows may have less infl uence from away in place by rockfall processes, or they vated pore pressures (e.g., Iverson and Major, toe buttressing, promoting earthfl ow sliding at can fail catastrophically if they are suffi ciently 1987), which would presumably be greater on lower gradients (Bovis and Jones, 1992). under mined by the surrounding earthfl ow-prone north-facing slopes. McSaveney and Griffi ths Earthfl ows predominantly occur in fi ne- terrain. Larger, resistant sandstone outcrops at (1987) suggested that drought may be a pre- grained, clay-rich lithologies (Keefer and the kilometer-scale have well-incised drainages, cursor to earthfl ow activity. They argued that Johnson, 1983; Hungr et al., 2001), especially and steeper slopes, emphasizing the contrasting deep desiccation cracks penetrate the earthfl ow in argillaceous or altered volcaniclastic sedi- erosional regimes that operate within different mass during drought, providing easy conduits ments (Glastonbury and Fell, 2008). Given the lithologies in the same landscape. for water to travel deep into the earthfl ow mass lithological variability across the study area, it The mean longitudinal slope of the earthfl ows when rainfall resumes. Following rainfall, sur- is unsurprising that active earthfl ows preferen- (31%) compares well with the data of Iverson face moisture levels increase, desiccation cracks tially occur in the sheared argillaceous mélange and Major (1987), who reported a residual fric- close, and the permeability of the earthfl ow (unit cm1 in Fig. 3), and 10% of this lithology tion angle for sheared earthfl ow soil of ~33%. surface decreases markedly. In the Eel River is subject to active slope instability (Table 1). The mean longitudinal slope of the earthfl ows is catchment, our fi eld observations indicate that However, given the lack of bedrock exposure less than the mean slope of all earthfl ow terrain south-facing slopes become highly desiccated in this region, much of the lithological mapping (34%; Fig. 12C), probably due to roughness and over the summer months, and surface cracks was undertaken with aerial photograph interpre- hummocky topography on the earthfl ow sur- up to 50 mm wide extend to depths over 1 m. tation, and slope morphology is a key metric in face. Compared to active earthfl ow terrain, the Northerly facing slopes do not dry to the same distinguishing between different mélange units. study area has a signifi cant fraction of terrain extent. This aspect-governed asymmetry of dry- McLaughlin et al. (2000) noted that the Central steeper than 50%, as refl ected by a heavier tail ing and soil moisture over the summer months belt mélange unit exhibits characteristic earth- in the distribution of slopes (Fig. 12C). This is may explain some of the southwesterly aspect fl ow morphology, and used the rounded, poorly attributable to gorges, cliffs, and steeper sand- preference we observe in the active earthfl ows. incised, lumpy, and irregular topography as a stone units throughout the study area, i.e., loca- Aspect-dependent seasonally increased permea- diagnostic feature during bedrock mapping. This tions not conducive to earthfl ow activity. The bility may also enhance weathering, leading to introduces circularity when comparing earth- mean slope of all earthfl ow terrain in Figure increased availability of readily mobilized rego- fl ow activity across argillaceous mélange units. 12C (34%) is very similar to the study area ter- lith on south-facing slopes. The northerly fac- More robust is the observed dichotomy between rain (36%). The similarity of mean slopes for ing slopes do have the morphology of dormant

1572 Geological Society of America Bulletin, July/August 2011 Earthfl ow activity and erosion, Eel River, California earthfl ows, and have been active in the past, yields from earthfl ows are signifi cantly faster earthfl ows. Figures DR1 and DR2 (see footnote potentially under different climatic conditions. than regional erosion rates, an active earthfl ow 1) show examples from the fi eld site refl ecting Dip-slopes with weak bedding or structural will eventually exhaust its source material and each stage of the process. defects can strongly control the directionality of become dormant. We suggest that reactivation We start in Figure 14A with an active earth- slope stability (e.g., Pettinga, 1987). Structure of an earthfl ow complex is driven from the fl ow, such as the larger fl ows mapped in Figure throughout our study site consists of steeply upper slopes, and earthfl ow activity propagates 8. After prolonged movement, the readily mobi- dipping, highly-sheared, and tightly folded downslope, possibly via undrained loading lized source material becomes exhausted (Fig. rocks striking to the northwest (McLaughlin (Hutchinson and Bhandari, 1971), rather than 14B), and the earthfl ow mass thins and reduces et al., 2000). Structures such as bedding planes through toe perturbation (e.g., channel incision) driving stresses, which leads to stability. Once or foliation do not appear to control earthfl ow propagating upslope. The amount of readily mo- the earthfl ow mass is stable, axial gullies propa- location, and the regional-scale structural grain bilized source sediment that can be incorporated gate up through the easily erodible earthfl ow does not have a signifi cant effect (Fig. 13B). into an earthfl ow, as governed by weathering body, while weathering proceeds in the newly rates and the conversion of bedrock to regolith, exposed rock in the source zone and headscarp Long-Term Earthfl ow Evolution may be the ultimate control on the location and walls. This process continues (Fig. 14C), and a rate at which earthfl ows evolve, or the periodic- network of smaller gullies propagates up into We combined the spatial distribution of earth- ity with which they are active. the earthfl ow source zone, destabilizing the flows, field observations, and analysis of Active earthfl ows frequently exhibit perva- recently weathered rock and regolith and gen- LiDAR-derived topography to develop a con- sive surfi cial gully networks (Figs. 4, 5, and erating several small earthfl ows. These small ceptual model for the long-term evolution of 8; Figs. DR1–DR2 [see footnote 1]) and com- earthfl ows coalesce in the source zone, and move earthfl ow-prone terrain. Due to the pervasive monly have large axial gullies that run down downslope, reactivating the earthfl ow complex. history of slope instability throughout our study the margin or center of earthfl ows. Although In essence, the slopes are delicately balanced, and site, we rarely observe earthfl ows that extend the role of gully networks in removing sedi- some destabilization, redistribution of mass, partially up a slope that are not within a larger, ment from active earthfl ows can be signifi - and loading at the top of the earthfl ow can re- established (if dormant) earthfl ow complex cant (Kelsey, 1978; Roering et al., 2009), the activate the whole feature. (Figs. DR1–DR2 [see footnote 1]). Large active relations and feedbacks between gullying and We envisage a period of earthfl ow activity of earthfl ows mostly extend from the channel to the earthfl ow mobility have not been well quanti- decades to centuries, followed by dormancy ridge or a topographic break in slope (Iverson, fi ed. For example, do gullies cause minor in- of thousands of years. During this time, the 1986b). It is rare to observe a youthful earthfl ow stabilities that coalesce to form major fl ows? channel at the toe of the earthfl ow incises at actively retrogressing into a previously unfailed Alternatively, can gullies, by eroding the earth- rates proportional to regional channel incision, hillslope. The few examples of such earthfl ow fl ow mass, lower the driving stresses on the and this ensures maintenance of the critical behavior in our fi eld site were generally retro- fl ow and contribute to stability? slope required for multiple cycles of earthfl ow gressing up broad ridgelines, establishing new It is clear that gullies and creeks play a major activity. The rate of channel incision likely con- earthfl ow basins (e.g., Mile 201 slide, Fig. 8). role by continuing to perturb and incise into trols the downward propagation of the earthfl ow Once established, the architecture of an individ- earthfl ow material once movement has ceased failure plane or shear zone. Channel incision at ual earthfl ow basin can persist in the landscape (Figs. 4; Figs. DR1–DR2 [see footnote 1]), and the toes of earthfl ow-prone hillslopes is neces- for signifi cant periods of time, as evidenced by we observed surfi cial gully networks propagat- sary to generate the relief required for persistent the many dormant earthfl ow features preserved ing through the transport zones and source zones earthfl ow activity, and yet episodes of rapid (e.g., Fig. 4; Figs. DR1–DR2 [see footnote 1]). of dormant fl ows (Figs. DR1–DR2 [see foot- earthfl ow activity and dormancy appear to be Where active earthfl ows do not span chan- note 1]). We argue that propagation of the gully modulated by the availability of readily mobi- nel to ridge, our data show that they are prefer- network into the amphitheater-like, earthfl ow lized regolith on the upper hillslopes. entially confi ned to the upper sections of material-starved source zones of dormant earth- hillslopes (Fig. 8; Table DR1 [see footnote 1]) fl ows promotes dissection of residual material, CONCLUSION and appear to constitute partial reactivation of enhances bedrock weathering, and eventually older, largely dormant earthfl ow features. There increases the availability of easily mobilized Earthfl ows dominate the erosion, morphol- are numerous examples of earthfl ows initiat- soil and regolith feeding into the source zone. ogy, and evolution of the landscape along the ing high on the hillslope. For example, Kelsey These discontinuous gullies are not connected main stem Eel River. Using airborne LiDAR (1978) described smaller earthfl ows confi ned to the channel network and act to disturb and and aerial photos, we objectively documented to midslopes, and Putnam and Sharp (1940) redistribute material on the upper earthfl ow regional earthfl ow activity and sediment fl ux described earthfl ows in Ventura County, Cali- slopes, rather than act as a conduit for signifi - across 226 km2 of earthfl ow-prone terrain. We fornia, starting on the upper slopes and mov- cant sediment removal. Redistributed material fi nd that 7.3% of the study area was active in ing downhill. In our study site, the New 190.5 can then coalesce in the source zone via small the period 1944–2006, and 6% of the study slide (shown in Fig. 4) reportedly started mov- slumps and fl ows from the headscarp and ex- area is composed of earthfl ows discharging ing in 1952. Slide debris progressively moved posure of bedrock underlying the source zone, into the channel network. Movement, as docu- downslope and eventually reached the railway increase local topographic loading, commence mented from displaced shrubs, had a median near the river in 1966 (Scott, 1973). downslope movement, and ultimately reactivate velocity of 0.4 m/yr over the 62 yr interval, al- We argue that the observations of smaller the whole earthfl ow. In Figure 14, we present though the distribution of displacement vectors fl ows preferentially occurring on upper slopes, a conceptual model of the long-term cy clical was heavily positively skewed and ranged up and the past descriptions of earthfl ow activity evolution of an earthflow, governed by the to 2.8 m/yr. support a top-down driver of long-term earth- availability of source material, and the ability of Earthflows in our study site have char- fl ow evolution. Given that rates of sediment gully networks to propagate through dormant acteristic spatial attributes, which include a

Geological Society of America Bulletin, July/August 2011 1573 Mackey and Roering

A Active earthflow B Source material exhausted

Earthflow movement slows or stops Headscarp slumping exposes fresh rock

River Trees Weathering of exposed rock in upper basin

Earthflow movement suppresses extensive gully Deep axial gully erodes network development earthflow body

C Extensive gully D Rejuvenation development of earthflow

Further erosion of stable earthflow body

Gullies propagate upslope and destabilize small failures in headwall or ampitheater Small flows coalesce and reactivate parent earthflow Channel incision maintains critical hillslope gradient

Figure 14. Conceptual earthfl ow evolution model. (A) Active earthfl ow is being supplied with material from small slumps and failures in the upper amphitheater. (B) Eventually, the source material is exhausted, and gullies begin to erode the earthfl ow body. (C) Gullies propagate into the source area, destabilizing and mobilizing the weathered slopes. (D) Small fl ows coalesce in the upper amphitheater source area and reactivate the earthfl ow transport zone.

lognormal size distribution, preferential south- The sediment yield from earthfl ow areas alone ACKNOWLEDGMENTS south westerly aspect, and strong association was 19,000 t km–2 yr–1 (rock erosion rate 7.6 This research was funded by a National Sci- with sheared argillaceous lithology. The earth- mm/yr), and could potentially double by ac- ence Foundation (NSF) grant (Geomorphology and fl ows have a slope distribution that is simi- counting for fl uvial erosion from the earth- Land Use Dynamics, EAR-0447190) to J. Roering. lar to the slope distribution of the study area, fl ow surface. Given that the sediment yield B. Mackey was partially supported by the “Fulbright although the mean slope of earthfl ow terrain from active earthfl ows can exceed the regional EQC Award in Natural Disaster Research” from New (34%) is slightly less than the study area (36%), sediment yield by an order of magnitude, we Zealand. The LiDAR data were acquired by the Na- tional Center for Airborne Laser Mapping (NCALM) suggesting that many of the slopes approach a envisage earthfl ow activity as being intermit- in 2006. We thank Calvin and Wendy Stewart for critical or threshold value. Larger earthfl ows tent, separated by long periods of dormancy. fi eld access and generous hospitality, and the Lone (>250,000 m2) typically span from channel to Most of the argillaceous terrain in our study Pine and Island Mountain ranches for fi eld access. ridge, whereas the majority of earthfl ows that do site shows topographic evidence of earth- We benefi ted from fruitful discussions with Harvey Kelsey and Jim McKean over the course of this re- not are located higher on the hillslopes and are fl ows, suggesting that locations of activity can search. We thank Harvey Kelsey, Bill Haneberg, Grant not connected to channels. The large earthfl ows migrate through the landscape. In contrast to Meyer, and Asso ciate Editor Jon Major for insightful tend to be less steep and are signifi cantly more many other styles of slope failures, earthfl ows reviews that greatly improved this manuscript. elongate than smaller features. appear to reactivate from the top when there Averaged across the 226 km2 studied section is suffi cient readily mobilized material; a top- REFERENCES CITED of the Eel River, mass movement by earthfl ows down perturbation can reactivate the whole Adam, D.P., and West, G.J., 1983, Temperature and precipita- –2 –1 generated a sediment yield of 1100 t km yr . earthfl ow. We suggest that long-term availabil- tion estimates through the last glacial cycle from Clear This is half of the estimated regional sediment ity of readily mobilized source sediment is the Lake, California, pollen data: Science, v. 219, no. 4581, –2 –1 p. 168–170, doi: 10.1126/science.219.4581.168. yield (2200 t km yr ), despite active earth- primary control on cycles of earthfl ow activity Angeli, M.G., Gasparetto, P., Menotti, R.M., Pasuto, A., fl ows accounting for only 6% of the study area. and evolution. and Silvano, S., 1996, A visco-plastic model for slope

1574 Geological Society of America Bulletin, July/August 2011 Earthﬂ ow activity and erosion, Eel River, California

analy sis applied to a mudslide in Cortina d’Ampezzo, western California: Geology, v. 19, no. 5, p. 526–529, Geophysical Research–Earth Surface, v. 110, no. F2, Italy: Quarterly Journal of Engineering Geology, v. 29, doi: 10.1130/0091-7613(1991)019<0526:MQUATN> F02015, 16 p, doi: 10.1029/2004JF000268. p. 233–240, doi: 10.1144/GSL.QJEGH.1996.029.P3.06. 2.3.CO;2. Iverson, R.M., and Major, J.J., 1987, Rainfall, groundwater- Barron, J.A., Heusser, L., Herbert, T., and Lyle, M., Dwyer, M.J., Scott, R.G., and Lorens, P.J., 1971, Landslide fl ow, and seasonal movement at Minor Creek land- 2003, High-resolution climatic evolution of coastal Conditions and Related Sediment Production on the slide, northwestern California—Physical interpretation northern California during the past 16,000 years: Eel River: Red Bluff, California, California Depart- of empirical relations: Geological Society of America Paleoceanography, v. 18, no. 18, 1020, 19 p., doi: 10.1029/ ment of Water Resources, Northern District, 69 p. Bulletin, v. 99, no. 4, p. 579–594, doi: 10.1130/ 2002PA000768. Fuller, T.K., Perg, L.A., Willenbring, J.K., and Lepper, 0016-7606(1987)99<579:RGFASM>2.0.CO;2. Baum, R.L., Fleming, R.W., and Johnson, A.M., 1993, Kine- K., 2009, Field evidence for climate-driven changes Jayko, A.S., Blake, M.C., McLaughlin, R.J., Ohlin, H.N., Ellen, matics of the Aspen Grove landslide, Ephraim Canyon, in sediment supply leading to strath terrace forma- S.D., and Kelsey, H.M., 1989, Reconnaissance Geologic central Utah, in Landslide Processes in Utah—Obser- tion: Geology, v. 37, no. 5, p. 467–470, doi: 10.1130/ Map of the Covelo 30-by 60-Minute Quadrangle, North- vation and Theory: U.S. Geological Survey Bulletin G25487A.1. ern California: U.S. Geological Survey Miscellaneous 1842, p. F1–F34. Furlong, K.P., and Govers, R., 1999, Ephemeral crustal thick- Field Investigation Map MF-2001, scale 1:100 000. Benda, L., and Dunne, T., 1997, Stochastic forcing of sedi- ening at a triple junction: The Mendocino crustal Keefer, D.K., 1984, Landslides caused by earthquakes: ment supply to channel networks from landsliding and conveyor: Geology, v. 27, no. 2, p. 127–130, doi: Geological Society of America Bulletin, v. 95, debris fl ow: Water Resources Research, v. 33, no. 12, 10.1130/0091-7613(1999)027<0127:ECTAAT> no. 4, p. 406–421, doi: 10.1130/0016-7606(1984)95 p. 2849–2863, doi: 10.1029/97WR02388. 2.3.CO;2. <406:LCBE>2.0.CO;2. Booth, A.M., Roering, J.J., and Perron, J.T., 2009, Auto- Furlong, K.P., and Schwartz, S.Y., 2004, Influence of Keefer, D.K., and Johnson, A.M., 1983, Earth Flows: Mor- mated landslide mapping using spectral analysis and the Mendocino triple junction on the tectonics of phology, Mobilization and Movement: U.S. Geological high-resolution topographic data: Puget Sound low- coastal California: Annual Review of Earth and Survey Professional Paper 1256, 56 p. lands, Washington, and Portland Hills, Oregon: Geo- Planetary Sciences, v. 32, p. 403–433, doi: 10.1146/ Kelsey, H.M., 1977, Landsliding, Channel Changes, Sedi- morphology, v. 109, no. 3–4, p. 132–147, doi: 10.1016/ annurev.earth.32.101802.120252. ment Yield and Land Use in the Van Duzen River j.geomorph.2009.02.027. Glastonbury, J., and Fell, R., 2008, Geotechnical charac- Basin, North Coastal California, 1941–1975 [Ph.D. Bovis, M.J., and Jones, P., 1992, Holocene history of teristics of large slow, very slow, and extremely slow thesis]: Santa Cruz, University of California, 370 p. earthflow mass movements in south-central British landslides: Canadian Geotechnical Journal, v. 45, Kelsey, H.M., 1978, Earthfl ows in Franciscan mélange, Van Columbia—The infl uence of hydroclimatic changes: no. 7, p. 984–1005, doi: 10.1139/T08-021. Duzen River basin, California: Geology, v. 6, no. 6, Canadian Journal of Earth Sciences, v. 29, no. 8, González-Diez, A., Remondo, J., de Teran, J.R.D., and p. 361–364. p. 1746–1755. Cendrero, A., 1999, A methodological approach for Kelsey, H.M., 1980, A sediment budget and an analysis of Brown, W.M., and Ritter, J.R., 1971, Sediment Transport the analysis of the temporal occurrence and triggering geomorphic process in the Van Duzen River Basin, and Turbidity in the Eel River Basin, California: U.S. factors of landslides: Geomorphology, v. 30, no. 1–2, north coastal California, 1941–1975—Summary: Geological Survey Water-Supply Paper 1986, 67 p. p. 95–113, doi: 10.1016/S0169-555X(99)00047-1. Geological Society of America Bulletin, v. 91, no. 4, Brückl, E., and Sehidegger, A.E., 1973, Application of theory Guzzetti, F., Malamud, B.D., Turcotte, D.L., and Reichen- p. 190–195. of plasticity to slow mud fl ows: Geotechnique, v. 23, bach, P., 2002, Power-law correlations of landslide Kelsey, H.M., and Carver, G.A., 1988, Late Neogene and no. 1, p. 101–107, doi: 10.1680/geot.1973.23.1.101. areas in central Italy: Earth and Planetary Science Quaternary tectonics associated with northward growth California Department of Water Resources, 1970, Middle Letters, v. 195, no. 3–4, p. 169–183, doi: 10.1016/ of the San Andreas transform-fault, northern Califor- Fork Eel River Landslides Investigation: California S0012-821X(01)00589-1. nia: Journal of Geophysical Research–Solid Earth and Department of Water Resources, Northern District, Haneberg, W.C., Cole, W.F., and Kasali, G., 2009, High- Planets, v. 93, no. B5, p. 4797–4819, doi: 10.1029/ Memorandum Report, 117 p. resolution lidar-based landslide hazard mapping and JB093iB05p04797. Carranco, L., and Beard, E., 1981, Genocide and Vendetta— modeling, UCSF Parnassus Campus, San Francisco, Kelsey, H.M., Coghlan, M., Pitlick, J., and Best, B., 1995, The Round Valley Wars of Northern California: Nor- USA: Bulletin of Engineering Geology and the En- Geomorphic analysis of streamside landslides in the man, Oklahoma, University of Oklahoma Press, 403 p. vironment, v. 68, no. 2, p. 263–276, doi: 10.1007/ Redwood Creek Basin, northwestern California, in Carson, M.A., and Petley, D.T., 1970, The existence of s10064-009-0204-3. Nolan, K.M., Kelsey, H. M., and Marron, D.C., eds., threshold slopes in the denudation of the landscape: Hilley, G.E., Burgmann, R., Ferretti, A., Novali, F., and Geomorphic Processes and Aquatic Habitat in the Red- Transactions of the Institute of British Geographers, Rocca, F., 2004, Dynamics of slow-moving landslides wood Creek Basin, Northwestern California: U.S. Geo- v. 49, p. 77–95. from permanent scatterer analysis: Science, v. 304, logical Survey Professional Paper 1454–J, p. J1–J12. Carter, W.E., Shrestha, R.L., and Slatton, K.C., 2007, Geo- no. 5679, p. 1952–1955, doi: 10.1126/science.1098821. Korup, O., 2005a, Distribution of landslides in southwest detic laser scanning: Physics Today, v. 60, no. 12, Hovius, N., Stark, C.P., and Allen, P.A., 1997, Sediment New Zealand: Landslides, v. 2, no. 1, p. 43–51, doi: p. 41–47, doi: 10.1063/1.2825070. fl ux from a mountain belt derived by landslide map- 10.1007/s10346-004-0042-0. Chandler, J.H., and Brunsden, D., 1995, Steady-state be- ping: Geology, v. 25, no. 3, p. 231–234, doi: 10.1130/ Korup, O., 2005b, Large landslides and their effect on sedi- havior of the Black-Ven mudslide—The application of 0091-7613(1997)025<0231:SFFAMB>2.3.CO;2. ment fl ux in South Westland, New Zealand: Earth Sur- archival analytical photogrammetry to studies of land- Hovius, N., Stark, C.P., Tutton, M.A., and Abbott, L.D., 1998, face Processes and Landforms, v. 30, no. 3, p. 305–323. form change: Earth Surface Processes and Landforms, Landslide-driven drainage network evolution in a pre– Korup, O., 2006, Rock-slope failure and the river long pro- v. 20, no. 3, p. 255–275, doi: 10.1002/esp.3290200307. steady-state mountain belt: Finisterre Mountains, Papua fi le: Geology, v. 34, no. 1, p. 45–48. Chaytor, J.D., ten Brink, U.S., Solow, A.R., and Andrews, New Guinea: Geology, v. 26, no. 12, p. 1071–1074, Korup, O., Densmore, A.L., and Schlunegger, F., 2010, The B.D., 2009, Size distribution of submarine landslides doi: 10.1130/0091-7613(1998)026<1071:LDDNEI> role of landslides in mountain range evolution: Geo- along the US Atlantic margin: Marine Geology, v. 264, 2.3.CO;2. morphology, v. 120, no. 1–2, p. 77–90, doi: 10.1016/ no. 1–2, p. 16–27, doi: 10.1016/j.margeo.2008.08.007. Hungr, O., Evans, S.G., Bovis, M.J., and Hutchinson, J.N., j.geomorph.2009.09.017. Coe, J.A., Ellis, W.L., Godt, J.W., Savage, W.Z., Savage, 2001, A review of the classifi cation of landslides of the Larsen, I.J., Montgomery, D.R., and Korup, O., 2010, Land- J.E., Michael, J.A., Kibler, J.D., Powers, P.S., Lidke, fl ow type: Environmental & Engineering Geoscience, slide erosion controlled by hillslope material: Nature D.J., and Debray, S., 2003, Seasonal movement of the v. 7, no. 3, p. 221–238. Geoscience, v. 3, no. 4, p. 247–251, doi: 10.1038/ Slumgullion landslide determined from global posi- Hutchinson, J.N., 1967, The free degradation of London ngeo776. tioning system surveys and fi eld instrumentation, July clay cliffs, in Proceedings of the Geotechnical Con- Lave, J., and Burbank, D., 2004, Denudation processes and 1998–March 2002: Engineering Geology, v. 68, no. 1–2, ference: Oslo, Norwegian Geotechnical Institute, rates in the Transverse Ranges, southern California: p. 67–101, doi: 10.1016/S0013-7952(02)00199-0. p. 113–118. Erosional response of a transitional landscape to exter- Craig, D., 1981, Mudslide plug fl ow within channels: Engi- Hutchinson, J.N., and Bhandari, R.K., 1971, Undrained nal and anthropogenic forcing: Journal of Geophysical neering Geology, v. 17, no. 4, p. 273–281, doi: 10.1016/ loading, a fundamental mechanism of mudfl ows and Research–Earth Surface, v. 109, F01006, 31 p. 0013-7952(81)90003-X. other mass movements: Geotechnique, v. 21, no. 4, Lawson, A.C., 1908, The California Earthquake of April Cruden, D.M., and Varnes, D.J., 1996, Landslide types and p. 353–358, doi: 10.1680/geot.1971.21.4.353. 18, 1906: Report of the State Earthquake Investigation processes, in Turner, A.K., and Schuster, R.L., eds., Iverson, R.M., 1986a, Dynamics of slow landslides: A Commission Publication 87: Washington, D.C., Carne- Landslides: Investigation and Mitigation: Washington, theory for time-dependent behavior, in Abrahams, gie Institution of Washington, 1, 451 p. D.C., National Academy Press, p. 36–71. A.D., ed., Hillslope Processes: Boston, Massachusetts, Lehre, A.K., and Carver, G.A., 1985, Thrust faulting and Dietrich, W.E., and Dunne, T., 1978, Sediment budget for a Allen and Unwin, p. 297–317. earthfl ows: Speculations on the sediment budget of small catchment in mountainous terrain: Zeitschrift für Iverson, R.M., 1986b, Unsteady, nonuniform landslide mo- a tectonically active drainage basin, in Savina, M.E., Geomorphologie, v. 29, p. 191–206. tion: 1. Theoretical dynamics and the steady datum ed., Redwood Country: American Geomorphological Dietrich, W.E., Bellugi, D.G., Sklar, L.S., Stock, J.D., Heim- state: The Journal of Geology, v. 94, no. 1, p. 1–15, Conference, Field Trip Guidebook 1985: Arcata, Cali- sath, A.M., and Roering, J.J., 2003, Geomorphic trans- doi: 10.1086/629006. fornia, p. 169–183. port laws for predicting landscape form and dynamics, Iverson, R.M., 1986c, Unsteady, nonuniform landslide mo- Lisle, T.E., 1990, The Eel River, northwestern California; in Wilcock, P., and Iverson, R.M., eds., Prediction in tion: 2. Linearized theory and the kinematics of tran- high sediment yields from a dynamic landscape, in Geomorphology: Washington D.C., American Geo- sient-response: The Journal of Geology, v. 94, no. 3, Wolman, M.G., and Riggs, H.C., eds., Surface Water physical Union, p. 103–132. p. 349–364, doi: 10.1086/629034. Hydrology, v. 0–1, The Geology of North America: Dumitru, T.A., 1991, Major Quaternary uplift along the Iverson, R.M., 2005, Regulation of landslide motion by Boulder, Colorado, Geological Society of America, northernmost San Andreas fault, King Range, north- dilatancy and pore pressure feedback: Journal of p. 311–314.

Geological Society of America Bulletin, July/August 2011 1575 Mackey and Roering

Lock, J., Kelsey, H., Furlong, K., and Woolace, A., 2006, Perron, J.T., Kirchner, J.W., and Dietrich, W.E., 2009, Forma- Stark, C.P., and Guzzetti, F., 2009, Landslide rupture and Late Neogene and Quaternary landscape evolution of tion of evenly spaced ridges and valleys: Nature, v. 460, the probability distribution of mobilized debris vol- the northern California Coast Ranges: Evidence for no. 7254, p. 502–505, doi: 10.1038/nature08174. umes: Journal of Geophysical Research–Earth Surface, Mendocino triple junction tectonics: Geological So- Pettinga, J.R., 1987, Waipoapoa landslide—A deep-seated v. 114, F00A02, 16 p, doi: 10.1029/2008JF001008. ciety of America Bulletin, v. 118, no. 9–10, p. 1232– complex block slide in Tertiary weak-rock fl ysch, south- Stark, C.P., and Hovius, N., 2001, The characterization of land- 1246, doi: 10.1130/B25885.1. ern Hawkes Bay, New Zealand: New Zealand Journal slide size distributions: Geophysical Research Letters, Mackey, B.H., Roering, J.J., and McKean, J.A., 2009, of Geology and Geophysics, v. 30, no. 4, p. 401–414. v. 28, no. 6, p. 1091–1094, doi: 10.1029/2000GL008527. Long-term kinematics and sediment fl ux of an active Putnam, W.C., and Sharp, R.P., 1940, Landslides and earth- Stock, J.D., and Dietrich, W.E., 2006, Erosion of steepland earthfl ow, Eel River, California: Geology, v. 37, no. 9, fl ows near Ventura, southern California: Geographical valleys by debris fl ows: Geological Society of America p. 803–806, doi: 10.1130/G30136A.1. Review, v. 30, no. 4, p. 591–600, doi: 10.2307/210535. Bulletin, v. 118, no. 9–10, p. 1125–1148, doi: 10.1130/ Malamud, B.D., Turcotte, D.L., Guzzetti, F., and Reichen- Roering, J.J., Stimely, L.L., Mackey, B.H., and Schmidt, B25902.1. bach, P., 2004a, Landslide inventories and their statisti- D.A., 2009, Using DInSAR, airborne LiDAR, and Swanson, F.J., and Swanston, D.N., 1977, Complex mass- cal properties: Earth Surface Processes and Landforms, archival air photos to quantify landsliding and sedi- movement terrains in the western Cascade Range, v. 29, no. 6, p. 687–711, doi: 10.1002/esp.1064. ment transport: Geophysical Research Letters, v. 36, Oregon: Geological Society of America Reviews in Malamud, B. D., Turcotte, D. L., Guzzetti, F., and Reichen- L19402, 5 p., doi: 10.1029/2009GL040374. Engineering Geology, v. 3, p. 113–124. bach, P., 2004b, Landslides, earthquakes, and erosion: Savage, W., and Wasowski, J., 2006, A plastic fl ow model for Swanston, D.N., Ziemer, R.J., and Janda, R.J., 1995, Rate Earth and Planetary Science Letters, v. 229, no. 1–2, the Acquara-Vadoncello landslide in Senerchia, south- and mechanics of progressive hillslope failure in the p. 45–59. ern Italy: Engineering Geology, v. 83, no. 1–3, p. 4–21, Redwood Creek basin, northwestern California, in Malet, J.P., Maquaire, O., and Calais, E., 2002, The use of doi: 10.1016/j.enggeo.2005.06.024. Nolan, K.M., Kelsey, H.M., and Marron, D.C., eds., global positioning system techniques for the continu- Savage, W.Z., and Chleborad, A.F., 1982, A model for creep- Geomorphic Processes and Aquatic Habitat in the ous monitoring of landslides: Application to the Super- ing fl ow in landslides: Bulletin of the Association of Redwood Creek Basin, Northwestern California: U.S. Sauze earthfl ow (Alpes-de-Haute-Provence, France): Engineering Geologists, v. 19, no. 4, p. 333–338. Geological Survey Professional Paper 1454, p. E1–E16. Geomorphology, v. 43, no. 1–2, p. 33–54, doi: 10.1016/ Schulz, W.H., 2007, Landslide susceptibility revealed by ten Brink, U.S., Barkan, R., Andrews, B.D., and Chaytor, S0169-555X(01)00098-8. LIDAR imagery and historical records, Seattle, Wash- J.D., 2009, Size distributions and failure initiation of Maquaire, O., Malet, J.P., Remaitre, A., Locat, J., Klotz, S., ington: Engineering Geology, v. 89, no. 1–2, p. 67–87, submarine and subaerial landslides: Earth and Plan- and Guillon, J., 2003, Instability conditions of marly doi: 10.1016/j.enggeo.2006.09.019. etary Science Letters, v. 287, no. 1–2, p. 31–42, doi: hillslopes: Towards landsliding or gullying? The case Schulz, W.H., Kean, J.W., and Wang, G.H., 2009a, Land- 10.1016/j.epsl.2009.07.031. of the Barcelonnette Basin, south east France: En- slide movement in southwest Colorado triggered by van Asch, T.W.J., Van Beek, L.P.H., and Bogaard, T.A., gineering Geology, v. 70, no. 1–2, p. 109–130, doi: atmospheric tides: Nature Geoscience, v. 2, no. 12, 2007, Problems in predicting the mobility of slow- 10.1016/S0013-7952(03)00086-3. p. 863–866, doi: 10.1038/ngeo659. moving landslides: Engineering Geology, v. 91, no. 1, McKean, J., and Roering, J., 2004, Objective landslide Schulz, W.H., McKenna, J.P., Kibler, J.D., and Biavati, p. 46–55, doi: 10.1016/j.enggeo.2006.12.012. detection and surface morphology mapping using G., 2009b, Relations between hydrology and veloc- Van Den Eeckhaut, M., Poesen, J., Verstraeten, G., Vanacker, high-resolution airborne laser altimetry: Geomor- ity of a continuously moving landslide—Evidence of V., Moeyersons, J., Nyssen, J., and van Beek, L.P.H., phology, v. 57, no. 3–4, p. 331–351, doi: 10.1016/ pore-pressure feedback regulating landslide motion?: 2005, The effectiveness of hillshade maps and expert S0169-555X(03)00164-8. Landslides, v. 6, no. 3, p. 181–190, doi: 10.1007/ knowledge in mapping old deep-seated landslides: McLaughlin, R.J., Kling, S.A., Poore, R.Z., McDougall, s10346-009-0157-4. Geomorphology, v. 67, no. 3–4, p. 351–363, doi: K., and Beutner, E.C., 1982, Post–Middle Miocene Schwab, M., Rieke-Zapp, D., Schneider, H., Liniger, M., 10.1016/j.geomorph.2004.11.001. accretion of Franciscan rocks, northwestern Califor- and Schlunegger, F., 2008, Landsliding and sediment Van Den Eeckhaut, M., Poesen, J., Verstraeten, G., Vanacker, nia: Geological Society of America Bulletin, v. 93, fl ux in the central Swiss Alps: A photogrammetric V., Nyssen, J., Moeyersons, J., van Beek, L.P.H., and no. 7, p. 595–605, doi: 10.1130/0016-7606(1982)93 study of the Schimbrig landslide, Entlebuch: Geo- Vandekerckhove, L., 2007, Use of LIDAR-derived <595:PMAOFR>2.0.CO;2. morphology, v. 97, no. 3–4, p. 392–406, doi: 10.1016/ images for mapping old landslides under forest: Earth McLaughlin, R.J., Ellen, S.D., Blake, M.C.J., Jayko, A.S., j.geomorph.2007.08.019. Surface Processes and Landforms, v. 32, no. 5, p. 754– Irwin, W.P., Aalto, K.R., Carver, G.A., and Clark, Scott, R.G., 1973, Geology and Sediment Production for 769, doi: 10.1002/esp.1417. S.H.J., 2000, Geology of the Cape Mendocino, Ten Eel River Landslides: Red Bluff, CA, California Vulliet, L., and Hutter, K., 1988, Viscous-type sliding laws Eureka , Garberville, and Southwestern Part of the Department of Water Resources, Northern District, for landslides: Canadian Geotechnical Journal, v. 25, Hayfork 30 × 60 Minute Quadrangles and Adjacent Memorandum Report, 29 p. no. 3, p. 467–477, doi: 10.1139/t88-052. Offshore Area, Northern California: U.S. Geological Skempton, A.W., 1964, Long-term stability of clay slopes: Wheatcroft, R.A., and Sommerfi eld, C.K., 2005, River sedi- Survey Miscellaneous Field Studies Map MF-2336, Geotechnique, v. 14, p. 77–102, doi: 10.1680/ ment fl ux and shelf sediment accumulation rates on the 1:100 000 scale. geot.1964.14.2.77. Pacifi c Northwest margin: Continental Shelf Research, McSaveney, M.J., and Griffi ths, G.A., 1987, Drought, rain, Slama, C.C., 1980, Manual of Photogrammetry: Falls v. 25, no. 3, p. 311–332, doi: 10.1016/j.csr.2004.10.001. and movement of a recurrent earthfl ow complex in Church, Virginia, American Society of Photogram- Wieczorek, G.F., 1984, Preparing a detailed landslide-inven- New Zealand: Geology, v. 15, no. 7, p. 643–646, doi: metry, 1056 p. tory map for hazard evaluation and reduction: Bulletin 10.1130/0091-7613(1987)15<643:DRAMOA> Slatton, K.C., Carter, W.E., Shrestha, R.L., and Dietrich, of the Association of Engineering Geologists, v. 21, 2.0.CO;2. W., 2007, Airborne laser swath mapping: Achieving no. 3, p. 337–342. Merritts, D., and Bull, W.B., 1989, Interpreting Quater- the resolution and accuracy required for geosurfi cial Wills, C.J., and McCrink, T.P., 2002, Comparing landslide nary uplift rates at the Mendocino triple junction, research: Geophysical Research Letters, v. 34, no. 23, inventories: The map depends on the method: Environ- northern California, from uplifted marine terraces: L23S10, 5 p, doi: 10.1029/2007GL031939. mental & Engineering Geoscience, v. 8, no. 4, p. 279– Geology, v. 17, no. 11, p. 1020–1024, doi: 10.1130/ Sloan, J., Miller, J.R., and Lancaster, N., 2001, Response 293, doi: 10.2113/8.4.279. 0091-7613(1989)017<1020:IQURAT>2.3.CO;2. and recovery of the Eel River, California, and its Wolf, P.R., and Dewitt, B.A., 2000, Elements of Photogram- Muhs, D.R., Thorson, R.M., Clague, J.J., Mathews, W.H., tributaries to fl oods in 1955, 1964, and 1997: Geo- metry, with Applications in GIS: Boston, Massachu- McDowell, P.F., and Kelsey, H.M., 1987, Pacifi c Coast morphology, v. 36, no. 3–4, p. 129–154, doi: 10.1016/ setts, McGraw-Hill, 624 p. and mountain system, in Graf, W.L., ed., Geomorphic S0169-555X(00)00037-4. Zhang, X.B., Phillips, C., and Marden, M., 1991a, Internal Systems of North America: Boulder, Colorado, Geo- Smith, B.J., Tilton, W.C., Elko, J.M., and Jacobus, C.D., deformation of a fast-moving earthfl ow, Raukumara logical Society of America, p. 517–581. 1974, Eel River Basin Environmental Studies: Red Peninsula, New Zealand: Geomorphology, v. 4, no. 2, Nolan, K.M., and Janda, R.J., 1995, Movement and Sedi- Bluff, California, California Department of Water Re- p. 145–154, doi: 10.1016/0169-555X(91)90025-6. ment Yield of Two Earthfl ows, Northwestern Califor- sources, Northern District, 93 p. Zhang, X.B., Phillips, C., and Pearce, A., 1991b, Surface nia: U.S. Geological Survey Professional Paper 1454, Sommerfi eld, C.K., and Wheatcroft, R.A., 2007, Late Holo- movement in an earthfl ow complex, Raukumara Penin- no. F, p. F1–F12. cene sediment accumulation on the northern California sula, New Zealand: Geomorphology, v. 4, no. 3–4, Ohlmacher, G.C., 2007, Plan curvature and landslide prob- shelf: Oceanic, fl uvial, and anthropogenic infl uences: p. 261–272, doi: 10.1016/0169-555X(91)90009-Y. ability in regions dominated by earth fl ows and earth Geological Society of America Bulletin, v. 119, slides: Engineering Geology, v. 91, no. 2–4, p. 117– no. 9–10, p. 1120–1134, doi: 10.1130/B26019.1. 134, doi: 10.1016/j.enggeo.2007.01.005. Sommerfi eld, C.K., Drake, D.E., and Wheatcroft, R.A., MANUSCRIPT RECEIVED 26 MARCH 2010 Palmquist, R.C., and Bible, G., 1980, Conceptual modeling 2002, Shelf record of climatic changes in fl ood mag- REVISED MANUSCRIPT RECEIVED 10 OCTOBER 2010 MANUSCRIPT ACCEPTED 12 OCTOBER 2010 of landslide distribution in time and space: Bulletin of nitude and frequency, north-coastal California: Geol- the International Association of Engineering Geology, ogy, v. 30, no. 5, p. 395–398, doi: 10.1130/0091-7613 v. 21, p. 178–186, doi: 10.1007/BF02591559. (2002)030<0395:SROCCI>2.0.CO;2. Printed in the USA

1576 Geological Society of America Bulletin, July/August 2011