Vadose Zone Journal | Advancing Critical Zone Science

Field and Laboratory Hydraulic Characterization of Landslide-Prone Soils in the Oregon Coast Range and Implications for Hydrologic SimulationBrian A. Ebel,* Jonathan W. Godt, Ning Lu, Jeffrey A. Coe, Joel B. Smith, and Rex L. BaumUnsaturated zone flow processes are an important focus of landslide hazard estimation. Differences in soil hydraulic behavior between wetting and drying conditions (i.e., hydraulic hysteresis) may be important in landslide triggering. Hydraulic hysteresis can complicate soil hydraulic parameter estimates and impact prediction capability. This investigation focused on hydraulic property estimation for soil in a landslide-prone area where the relative importance of hys-teresis is unclear. Laboratory measurements of soil-water retention from field soils in the Oregon Coast Range during wetting and drying show that pronounced hydraulic hysteresis is present. In contrast, a 4-yr field data record of pore-water pressure and soil-water content from multiple soil pits at the same landslide-prone area shows relatively minor hydraulic hysteresis compared with the laboratory estimates. Simulated subsurface hydrologic response parameterized using estimates from field data more closely matched hydrologic observations relative to model parameterization based on laboratory analysis of repacked soil samples. Our results suggest that (i) unsaturated hydraulic parameter estimates based on in situ field data, as opposed to laboratory measurements alone, may lead to more accurate simulation of the hydrologic response to rainfall, (ii) in situ data of soil-water retention may need to include values at both high suctions and near saturation to improve estimates of soil hydraulic parameters for slope fail-ure applications, and (iii) laboratory measurements of soil-water retention made under dynamic conditions may overestimate hydraulic hysteresis.

Abbreviations: FS, Factor of Safety; TRIM, Transient Water Release and Imbibition Method.

Landslides, mass movements of regolith including rock and soil, actively sculpt land-forms of sufficient topographic relief (Larsen et al., 2010) and can negatively impact the human environment, particularly in mountainous regions. In a recent 7-yr period from 2004 to 2010, rainfall-induced landslides were responsible for the loss of 32,000 lives globally (Petley, 2012). Catastrophic landslides, in terms of human and economic costs, such as the 2014 Oso landslide in the state of Washington, highlight the importance of improved slope failure prediction (Iverson et al., 2015; Wartman et al., 2016; LaHusen et al., 2016). Disturbances, such as wildfire (Pierce et al., 2004) and forestry practices (Roering et al., 2003), tend to degrade slope stability by changing hillslope vegetation and the hydrologic and mechanical properties of the mineral and organic soil composite. The intensity and extent of such disturbances is generally increasing because of expand-ing anthropogenic activities in areas that are susceptible to landslides (e.g., Turner, 2010; Millar and Stephenson, 2015; Ebel and Mirus, 2014; Mirus et al., 2017a) with recent evidence suggesting that landslides are persistently remobilized in disturbed locations (Mirus et al., 2017b).

Models for predicting the timing and location of slope failure recognize that the different timescales of vertical water movement and pressure transmission through unsat-urated soil are critical controls (Iverson, 2000; Frattini et al., 2004). The combination of simplified representations of transient unsaturated flow and topographically driven

Core Ideas

•Laboratory hydraulic parameters differ substantially from values esti-mated from field data.

•Hydraulic parameters determined in the laboratory overestimate hyster-esis magnitude.

•Field data-based model parameter-izations may simulate landslide initiation more accurately.

B.A. Ebel, Water Cycle Branch, Water Mission Area, US Geological Survey, Lakewood, CO 80225; J.W. Godt, J.A. Coe, J.B. Smith, and R.L. Baum, Geologic Hazards Science Center, US Geological Survey, Golden, CO 80401; N. Lu, Dep. of Civil and Environmental Engineering, Colorado School of Mines, Golden, CO 80401. *Corresponding author (bebel@usgs.gov).

Received 12 Apr. 2018.Accepted 2 June 2018.Supplemental material online.

Citation: Ebel, B.A., J.W. Godt, N. Lu, J.A. Coe, J.B. Smith, and R.L. Baum. 2018. Field and laboratory hydraulic characterization of lands-lide-prone soils in the Oregon Coast Range and implications for hydrologic simulation. Vadose Zone J. 17:180078. doi:10.236/vzj2018.04.0078

UZIG Research: Land-Use and Climate Change Impacts on Vadose Zone Processes

© Soil Science Society of America. This is an open access article distributed under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Published August 30, 2018

VZJ | Advancing Critical Zone Science p. 2 of 15

saturated flow can reliably predict landslide locations for many settings (Godt et al., 2008). However, overpredictions (i.e., false positives) are common (Baum et al., 2010), and accurate forewarn-ing of both the timing and specific locations that fail for a given storm remains difficult. Unsaturated zone hydrologic processes have been a recent focal area for improving understanding and prediction of landslide initiation (Reid et al., 1997; Ng and Shi, 1998; Collins and Znidarcic, 2004; Baum et al., 2010; Sorbino and Nicotera, 2013; Chen et al., 2017; Thomas et al., 2018).

Unsaturated flow in soil-mantled hillslopes exerts a strong inf luence on the timing and location of landslide initiation. Determining appropriate soil hydraulic properties, in particular soil-water retention, is a major challenge for parameterizing hydrologic response models that simulate f low through this zone. Perhaps the most important complication for determining unsaturated soil hydraulic properties, relative to landslide initiation, is the difference in water storage and transmission in soils under wetting vs. drying conditions, known as hydraulic hysteresis. The phenomenon of hysteresis has received much attention from unsaturated zone scientists and engineers (e.g., Topp, 1969, 1971; Mualem, 1974, 1984) including for landslide initiation applications (e.g., Wheeler et al., 2003; Ma et al., 2011; Tsai, 2011; Likos et al., 2014; Chen et al., 2017).

Hysteresis has been attributed to the dependence of the water–solid–air contact angle on the rate and direction of con-tact movement, pore emptying (during drying) and filling (during wetting) being controlled by pore throats and bodies, and air that becomes entrapped during wetting (Bear, 1972). Maximum storage of water is achieved at complete saturation, when the vol-umetric soil-water content, q, is equal to the soil porosity. This maximum reproducible q is often referred to as qs

D and it defines the upper limit of q on the initial drying soil-water retention curve. Following thorough drying from qs

D, a soil-water content qr

D, or the so-called residual moisture content, is reached. It is often assumed that qr

D is equivalent to qrW, the residual moisture

content for the wetting process, and the notation is shortened to qr. Upon wetting, if the pore-water pressure becomes zero at a soil-water content q s

W less than q sD, this lack of closure in the

soil-water retention curve indicates entrapped air.Drying from qs

W follows a primary drying curve. Wetting and drying paths between these primary endmember curves are called scanning curves (see Kool and Parker, 1987; Likos et al., 2014). Soils infrequently follow the primary wetting or drying curves under natural rainfall and climatic variability. Instead, soil-water reten-tion relations typically lie along the transitional scanning curves. Representation of hysteretic soil-water retention along scanning curves is often achieved by using analytical relations (e.g., Kool and Parker, 1987) with estimates of the soil hydraulic parameters describ-ing the primary wetting and drying curves. The primary wetting and drying curves can be estimated by laboratory or field measurements under conditions of sustained wetting and drying that are assumed to represent the respective endmember behaviors. The importance of accurate representation of hysteretic soil-water retention extends

beyond unsaturated flow simulation because of the link between unsaturated-zone hydrologic conditions and effective stress through the concept of suction stress (e.g., Lu and Godt, 2008, 2013; Godt et al., 2009; Likos et al., 2014; Chen et al., 2017).

The primary objectives of this work were to: (i) compare labo-ratory- and field-derived soil-water retention curves and assess any differences including the magnitude of hydraulic hysteresis; and (ii) determine the relative capabilities of the soil hydraulic param-eters from laboratory estimates vs. in situ soil-water retention estimates derived from field monitoring for numerically simulating the hydrologic response to rainfall in a landslide-prone environ-ment. An additional objective was to evaluate factors that improve both laboratory and field characterization of soil-water retention for landslide applications, such as what laboratory protocols are essential and if specific hydrologic conditions (i.e., very wet or very dry) are critical. Laboratory measurements and hydrologic response data from a landslide-prone headwaters catchment in the Oregon Coast Range (Smith et al., 2014) provided the foundation for this investigation of soil hydraulic properties and simulated hydrologic response.

6Materials and MethodsField Site

The field site is in the Elliot State Forest within the southern Oregon Coast Range (Fig. 1A). The Oregon Coast Range trends north–south in the western portion of the state. The underlying bedrock is the Tyee Formation (Baldwin, 1961), which is primarily marine sandstone with thinner interbedded siltstone layers (Niem and Niem, 1990). Colluvial and residual soils mantle the landscape ranging in USDA textural classification from gravel to clay loam (Johnson et al., 1994; Soil Survey Staff, 2013). Soils tend to be low density (<1 g cm−3 in the near surface; Reneau and Dietrich, 1991), with non-plastic fines, and are well drained as a result of rela-tively large saturated hydraulic conductivities (1.5–15 cm h−1; Soil Survey Staff, 2013). The general soil profile, based on character-ization during the excavation of the eight soil pits for instrument installation, has a surface organic layer of litter and duff up to tens of centimeters thick. Beneath the organic layer are horizons of sandy soil that grade into weathered sedimentary bedrock. Soil profile descriptions were summarized by Smith et al. (2014).

The climate is maritime, with the majority of precipitation falling between October and May. The relatively low elevations (i.e., <600 m) and moderate temperatures usually prevent substan-tial snowfall (Andrus et al., 2003). The field instrumentation site is located in a southwest-facing, unchanneled headwater basin with an approximate area of 4350 m2. The small, steep (37–45°) catch-ment has been disturbed by industrial tree harvest (most recently around 2005), and the topography has a variable structure with convergent, planar, and divergent hillslopes (Fig. 1B). Soil samples for laboratory analyses are taken from the same monitoring basin. The study area meets several criteria for factors promoting slope instability (recent forest harvest activity combined with slopes

VZJ | Advancing Critical Zone Science p. 3 of 15

exceeding 30°) and efficacy of monitoring and sample collection (road access at the upslope portion). More detailed descriptions of the study area were presented by Smith et al. (2014).

The Oregon Coast Range has been extensively studied with regard to debris flow processes (Benda and Cundy, 1990; Stock and Dietrich, 2003), hillslope hydrology (e.g., Anderson et al., 1997a, 1997b; Montgomery et al., 1997; Torres et al., 1998; Ebel and Loague, 2008), vegetation contributions to slope stability (e.g., Schmidt et al., 2001; Roering et al., 2003), and landslide initiation (e.g., Montgomery and Dietrich, 1994; Ebel et al., 2010). Steep topography, large annual precipitation amounts (often 1500–3000 mm yr−1), and periodic tree harvest from industrial forestry promote landslides as a major geomorphic process and natural hazard in this area (e.g., Roering et al., 2003). The Oregon Coast Range experienced widespread landslides in 1996 in response to major rainfall events, causing five fatalities (Montgomery et al., 2009; Coe et al., 2011). The study catchment choice reflects prior work in this region given that it was clear-cut

logged and replanted between 2004 and 2006 (Fig. 1A), making it a suitable location to investigate the role of unsaturated zone hydrologic processes in landslide initiation.

Laboratory CharacterizationThe Transient Water Release and Imbibition Method (TRIM)

(Wayllace and Lu, 2012) was used to estimate the soil-water retention curve and hydraulic conductivity function in the laboratory for both wetting and drying conditions. The soil specimens were repacked samples from specific depth ranges at the Elliot State Forest site. Soil samples for the laboratory soil-water retention measurements were taken nearest to the SP2 (0–23-, 23–75-, and 75–155-cm depths) and SP5 soil pits (50–80-cm depth), with soil pit locations shown in Fig. 1B. These sample depths correspond to the organic layer of litter duff (0–23-cm sample at SP2), the darker sandy soil (23–75-cm sample at SP2), and the lighter colored soil with a larger pebble and cobble fraction (50–80-cm sample at SP5 and the 75–155-cm sample at SP2). Particle size distributions were determined for the four samples fol-lowing the techniques of Lu et al. (2013).

The TRIM method is a multistep inflow and outflow test that, when combined with a numerical solution of Richards’ equation describing variably saturated water flow and an inverse solution technique for parameter estimation, provides character-ization of unsaturated soil hydraulic properties. A ceramic disk with a high air-entry value bounds the soil sample within the flow cell. Pressure was controlled using two regulators, one from 0 to 15 kPa to identify the air-entry value for coarse-grained soils and one from 10 to 300 kPa to facilitate characterization of soil-water retention. Water flux in and out of the system was measured using an electronic balance to an accuracy of 0.01 g. The water flux mea-surements were recorded at a temporal discretization of 10 s during rapid changes and up to 10 min during conditions near steady state. Water flows in and out of the reservoir atop the electronic balance, facilitating measurement (by mass) of the volumetric soil-water content of the sample and the water flux (the change in mass with time). The water flux data inform the objective function in the inverse estimation of the soil hydraulic properties, which uses the Levenberg–Marquardt method to minimize the difference between simulated and observed water f luxes for both wetting and drying processes. The finite-element model HYDRUS-1D (Šimůnek et al., 2008a) was used to simulate variably saturated water flow for the TRIM apparatus subject to the applied initial and boundary conditions, with the water flux corrected for the soil sample cross-sectional area. The soil hydraulic parameters esti-mated in the inverse procedure were the van Genuchten (1980) parameters a, n, qr, and qs. The van Genuchten (1980) formulation for soil-water retention is given by

( )( )

s rr 1 1/

1nn -

q -qq y =q +

é ù+ ayê úë û [1]

where q is the volumetric soil-water content (cm3 cm−3), y is matric suction (cm), a (cm−1) is a parameter related to the inverse of the

Fig. 1. (A) Photograph and location map (inset) of the field site in the Elliot State Forest, Oregon, and (B) topographic map (1-m contour intervals) showing the location of the hydrologic response instrumented soil pits SP1 to SP8. Topography is from an on-site theodolite survey.

VZJ | Advancing Critical Zone Science p. 4 of 15

air-entry value, n is a dimensionless parameter related to the pore-size distribution, qr is the residual water content (cm3 cm−3), and q s is the water content at saturation (cm3 cm−3). Note that the van Genuchten (1980) parameter m has been assumed to equal 1 − 1/n, which is consistent with the hydraulic conductivity func-tion from Mualem (1976). The TRIM procedure also estimates the saturated hydraulic conductivity during the wetting (Ks

W) and drying processes (Ks

D). Estimation of the van Genuchten parameters facilitates extrapolation of the soil-water retention and hydraulic conductivity functions beyond the 300 kPa of suction used in the TRIM procedure.

The TRIM procedure has several important steps that ensure correct measurement of the system response to increments and decrements in applied pressure and corresponding water flux. First, the entire system, both the apparatus and the soil sample, must be completely saturated prior to beginning a drying test. Saturation of the sample is accomplished by applying a vacuum to the top of the sample while de-aired water supplied at the sample base displaces water present in the sample that may contain dissolved air. At least one full pore volume of de-aired water is cycled through the soil sample. Then a small decrease in pressure (i.e., an increase in suction) is applied to cause the soil sample to become unsaturated, begin-ning the measurement of the unsaturated drying response. Once the system reaches steady state (12–24 h), a further pressure decrease is applied to further desaturate the sample, with the lower limit of applied pressure dictated by the air-entry value of the ceramic stone that bounds the sample (i.e., desaturating the stone affects measure-ment accuracy). Achieving steady state at the limits of the ceramic stone, typically a pressure of 300 kPa of suction, takes about 48 h. Then the procedure is reversed and a pressure increase is applied (i.e., a smaller suction) and water flows into the sample to characterize soil wetting. This step takes about 7 to 24 h to complete.

Further details on the TRIM procedure and a comparison of TRIM results and soil-water retention curve data using a Tempe cell for a poorly graded sand and a silty clay were shown by Wayllace and Lu (2012). Multistep outflow experiments coupled with inverse parameter estimation methods, similar to the TRIM technique, have been used successfully to estimate soil-water reten-tion curves and hydraulic conductivity functions for a variety of soil types (e.g., van Dam et al., 1994; Eching et al., 1994). Soil hydraulic parameters estimated by multistep outflow tests have been compared with the same parameters derived from field data of colocated soil-water content and matric potential measurements and showed reasonable agreement (Marion et al., 1994). One limi-tation of the TRIM procedure is the impact of dynamic effects in hydrologic conditions (i.e., rapid pressure changes) that alter inversely estimated soil-water retention curves (Hassanizadeh and Gray, 1993; Wildenschild et al., 2001; Hassanizadeh et al., 2002; O’Carroll et al., 2005).

Field MeasurementsThe field data capture the unsaturated zone hydrologic

response to rainfall in relatively undisturbed soils. Locations of the

field instrument soil pits are shown in Fig. 1A and 1B. The term soil pit is defined in this work as an approximately 1-m-wide area that has soil descriptions and depths characterized at the excavated pit wall, tensiometers installed at the upslope pit edge, and soil-water content sensors installed in the upslope pit wall. Soil-water content (cm3 cm−3) was measured using Decagon Devices EC-5 sensors installed in the soil pits and backfilled (Smith et al., 2014). The EC-5 sensor response was not calibrated specifically for soils at this site, therefore the factory calibration was used, resulting in approximately 3% error (Decagon Devices, 2018). Some inves-tigations have suggested that the error in the soil-water content measurements is slightly higher, such as the 4% value noted by Robinson et al. (2008). The support volume of the Decagon EC-5 is approximately 240 mL (Decagon Devices, 2018). Tensiometric response was measured using UMS T-8 sensors (±0.5 kPa error; range from −85 to 100 kPa; UMS, 2014) installed in hand-augered inclined boreholes within 1 m of the soil-water content sensors. The tensiometer support volume is difficult to determine and probably depends on hydrologic conditions; in this study it was assumed that the tensiometer support volume is similar to the soil-water content sensors. The tensiometer boreholes are oriented slope normal, and the depth below the land surface used in all analysis reflects the soil depth directly above the tensiometer cup (see Smith et al., 2014). The tensiometers cavitate and data become unreliable when the pore-water pressure drops below matric suc-tions of 60 to 80 kPa; those data are subject to large measurement errors (UMS, 2014). Automated data retrieval facilitated trips to the site after tensiometer cavitation to add water in the porous cup once soil-water content sensors indicated sufficient soil wet-ting to maintain tensiometer integrity. The tensiometer–soil-water content sensor pairs are slightly separated in the vertical direction from 1 to 39 cm of offset (see Table 1). Only tensiometer–soil water content sensor pairs with vertical separation distances <20 cm were used for analysis of in situ soil-water retention. Additionally, some of the soil-water content or tensiometer data do not pass quality control criteria (e.g., outside physically possible bounds or instru-ment failure values) and were excluded from analysis. Table 1 denotes the eight out of 16 tensiometer–soil water content sensor pairs that passed vertical separation distance and quality control criteria. All data were recorded at 15-min temporal resolution on dataloggers. Rainfall was measured using a tipping bucket rain gauge with 0.254 mm per tip (Hydrological Services Pty. Ltd.) at 1 m above the land surface. The timespan of the field data used here is from October 2009 through October 2013. The installation of field sensors was described in detail by Smith et al. (2014).

Estimates of soil hydraulic parameters from the field data were conducted to compare soil hydraulic parameter estimates from in situ soil-water retention data with laboratory data. The software RETC (van Genuchten et al., 1991) was used to estimate the drying and wetting van Genuchten (1980) soil hydraulic param-eters. Noise in the soil-water content data made it challenging to use simple automated criteria (e.g., dq/dt > 0 is wetting) to identify wetting vs. drying data. Efforts to reduce this noise by rounding to

VZJ | Advancing Critical Zone Science p. 5 of 15

the instrument error, restricting to thresholds of dq/dt, temporal subsampling, automated smoothing, and time averaging tech-niques were not successful. Therefore, time periods of wetting and drying were manually extracted from the 4 yr of field data by visual examination of time series of rainfall and hydrologic response. The data were not separated into training and evaluation data sets because all data were needed to sample as much of the soil-water content and pressure head range as possible for the RETC fitting. The seasonal rainfall of this area promotes drying to relatively dry conditions only one time per year (the end of the rainy season), which often results in tensiometer cavitation, and the tensiometers were not reestablished until wet conditions returned the following fall. RETC can only incorporate 1000 paired soil-water content and matric suction values; therefore, in situ data for wetting or drying at a given soil pit were down-sampled (i.e., every 20th measurement used) to have the total number <1000. The param-eters estimated in RETC are the standard van Genuchten (1980) parameters a , n, qr, and qs for both drying and wetting data (see Eq. [1]). The assumption that the dimensionless van Genuchten fitting parameter m = 1 − 1/n was also made, which reduces the number of free parameters at the expense of constraining the fit between the data and the van Genuchten (1980) relation. Initial estimates of the van Genuchten (1980) parameters were varied in

RETC to ensure that the final estimates reflect globally optimal parameter estimates, following the suggestion of van Genuchten et al. (1991). The objective function in the RETC fitting is the difference between observed and fitted soil-water content, which is minimized using the Marquardt maximum neighborhood method (Marquardt, 1963) using equal weighting of data points. The goodness of fit of the RETC estimates of the van Genuchten (1980) parameters was assessed using the R2 for the regression of the observed and fitted values, where 1.0 is perfect correlation (van Genuchten et al., 1991). RETC also calculates upper and lower limits of the 95% confidence interval around the van Genuchten (1980) parameter estimates.

Numerical Model of Unsaturated Water FlowThe one-dimensional (i.e., vertical) numerical model

HYDRUS-1D (Šimůnek et al., 2008b) was used to simulate unsaturated flow at the field site. The model uses the finite-element method to solve the Richards equation. A vertical discretization, Dz, of 1 cm was used along with an adaptive time step, Dt. An atmospheric boundary with surface runoff (Šimůnek et al., 2008a) was used as the surface boundary condition, parameterized with the measured precipitation at the site (Fig. 2). Evapotranspiration was neglected in the simulations. Prior work focused on hydrologic

Table 1. Instrumented soil pit characteristics, sensor depths, and ver-tical separation distances between tensiometer and soil-water content measurements. Sensor pairs in bold pass selection criteria for vertical separation distance (<20 cm of separation) and data quality criteria (outside physically possible bounds or instrument failure values). Only these sensor pairs were analyzed in this work; sensor pairs with >20-cm vertical separation distance were not analyzed.

Instrumented soil pit† Local slope

Soil-water content depth

Tensiometer depth

Vertical separation distance‡

° —————————— cm ——————————

SP1, shallow 37 50 63 13

SP1, deep 37 93 113 20

SP2, shallow 45 50 70 20

SP2, deep 45 125 126 1

SP3, shallow 38 50 63 13

SP3, deep 38 94 133 39

SP4, shallow 35 50 61 11

SP4, deep 35 75 97 22

SP5, shallow 32 50 59 9

SP5, deep 32 160 129 31

SP6, shallow 35 50 61 11

SP6, deep 35 90 122 32

SP7, shallow 37 50 63 13

SP7, deep 37 80 119 39

SP8, shallow 39 50 64 14

SP8, deep 39 80 96 16

† See Fig. 1B.‡ The absolute value of the vertical separation distance between the nearest tensiometer and soil-water content sensor at a given depth range.

Fig. 2. Schematic of the HYDRUS-1D model domain and boundary conditions used for both the laboratory- and field-based soil-water retention parameterizations. Precipitation as a function of time is the applied surface boundary condition and if the precipitation rate exceeds the infiltration capacity, surface runoff reduces the flux into the subsurface. The basal boundary condition is free drainage (unity hydraulic gradient). The soil was divided into three layers based on the TRIM characterization and soil profile information at Soil Pit SP2. Soil hydraulic parameters for each model layer are given in Supple-mental Table S1.

VZJ | Advancing Critical Zone Science p. 6 of 15

response and slope stability in the Oregon Coast Range noted minor errors as a result of neglecting evapotranspiration in hydrologic-response numerical simulation, with evapotranspiration amounting to 1 to 3% of the annual water balance during the wetter portion of the year, October through May, when slope failures tend to initi-ate (Ebel et al., 2008). The basal boundary condition was treated as a free drainage boundary condition (Šimůnek et al., 2008b) set at the 3-m depth, determined by sensitivity analysis to minimally affect the simulated response higher up in the soil column (Fig. 2). Initial conditions in the subsurface were set using the measured soil-water content data. Hydraulic hysteresis was represented using the Lenhard approach (Šimůnek et al., 2008a) with a drying initial condition. The goal of the simulations was to compare the simu-lated hydrologic response to rainfall with the observed response

for two scenarios: first with the TRIM parameterization from the laboratory and second with the RETC parametrization from the field data. Because the SP2 instrument soil pit has the most thor-ough laboratory characterization (Table 2), the SP2 soil pit was the focus of the HYDRUS-1D simulations. The soil profile of the model domain was established using the soil layering characteriza-tion and hydraulic parameters in Table 2 from TRIM, with the exception of the 75- to 155-cm soil profile layer that was set to the RETC estimates from Table 3 for the RETC simulation (Fig. 2; Supplemental Table S1). Hydraulic conductivity was assumed to be isotropic and specified using the TRIM estimates from Table 2 (also see Supplemental Table S1). The simulated periods for the parameter estimate comparison is 1 Oct. 2010 to 1 June 2011 and 1 Oct. 2011 to 1 June 2012, which encompasses two wet climatic

Table 2. Unsaturated hydraulic parameters† for the van Genuchten (1980) model from the laboratory measurements using the TRIM method (Wayl-lace and Lu, 2012).

Sample depth Soil pit rb aD aW nD nW qr‡ qsD qs

W KsD Ks

W

cm g cm−3 ———— cm−1 ———— —————— cm3 cm−3 —————— ——— cm h−1 ———

0–23 SP2 0.99 0.012 0.015 1.4 1.35 0.16 0.48 0.38 18.9 14.4

23–75 SP2 1.20 0.01 0.023 1.9 2.0 0.15 0.42 0.39 13.7 9.2

50–80 SP5 1.50 0.014 0.051 2.8 2.6 0.06 0.38 0.27 1.49 0.036

75–155 SP2 1.64 0.008 0.018 2.32 2.3 0.091 0.4 0.26 12.2 10.1

† rb, dry bulk density; aD, drying van Genuchten parameter; aW, wetting van Genuchten parameter; nD, drying van Genuchten parameter; nW, wetting van Genuchten parameter; qD, drying saturated water content; qW, wetting saturated water content; Ks

D, drying saturated hydraulic conductivity; KsW, wetting saturated hydraulic con-

ductivity.‡ qr

D is assumed to be the same as qrW and denoted as qr.

Table 3. The van Genuchten (1980) shape parameters a and n and the residual and saturated water contents (qr and qs, respectively) fitted with RETC (van Genuchten et al., 1991) from in situ data and their 95% confidence intervals.

Instrumented soil pit† a a , 95% CI n n, 95% CI qr‡ qr, 95% CI qs qs, 95% CI R2§

——— cm−1 ——— —————— cm3 cm−3 ——————

SP1, s wetting 0.0547 0.0454–0.0639 1.685 1.355–2.015 0.294 0.268–0.319 0.392 0.3868–0.3918 0.745

SP1, s drying 0.0356 0.0276–0.0435 1.727 1.591–1.863 0.305 0.301–0.308 0.384 0.3743–0.3840 0.898

SP2, d wetting 0.3280 0.1790–0.4770 1.247 1.091–1.402 0.257 0.201–0.313 0.409 0.403–0.414 0.865

SP2, d drying 0.4097 −2.206–3.026 1.473 1.313–1.633 0.313 0.308–0.318 0.410 0.154–0.667 0.878

SP3, s wetting 0.2292 0.0688–0.3897 1.062 1.049–1.075 0 ¶ 0.282 0.277–0.287 0.732

SP3, s drying 0.0066 0.0062–0.0069 1.305 1.294–1.316 0 ¶ 0.253 0.252–0.254 0.981

SP5, s wetting 0.0417 0.0272–0.0562 1.372 1.027–1.717 0.169 0.090–0.249 0.305 0.299–0.311 0.820

SP5, s drying 0.0165 0.0110–0.0220 1.431 1.216–1.647 0.162 0.127–0.198 0.293 0.286–0.301 0.931

SP6, s wetting 0.0296 0.0133–0.0458 1.120 1.102–1.297 0 ¶ 0.332 0.328–0.337 0.561

SP6, s drying 0.0277 0.0128–0.0425 1.145 1.127–1.163 0 ¶ 0.323 0.311–0.335 0.715

SP7, s wetting 0.0335 0.0079–0.0591 1.534 0.676–2.393 0.138 −0.105–0.380 0.307 0.302–0.310 0.351

SP7, s drying 0.0259 0.0199–0.0319 1.153 1.143–1.163 0 ¶ 0.306 0.302–0.310 0.894

SP8, s wetting 0.1186 0.0709–0.1663 1.144 1.120–1.167 0 ¶ 0.267 0.263–0.273 0.853

SP8, s drying 0.0169 0.0152–0.0185 1.328 1.316–1.340 0 ¶ 0.239 0.237–0.242 0.964

SP8, d wetting 0.0353 0.0148–0.0558 2.080 1.138–3.021 0.196 0.071–0.321 0.326 0.323–0.328 0.557

SP8, d drying 0.0158 0.0138–0.0177 1.875 1.648–2.101 0.168 0.153–0.183 0.309 0.304–0.313 0.852

† See Fig. 1B; s denotes shallow instrument pair, d denotes deep instrument pair.‡ For drying q r is qr

D and for wetting is qrW.

§ R2 is the regression between fitted and observed values calculated by RETC.¶ Fixed to 0 by RETC.

VZJ | Advancing Critical Zone Science p. 7 of 15

seasons when landslide susceptibility is greater than the drier summer season. Note that HYDRUS-1D is the same numerical flow model used in the TRIM inverse estimation procedure.

Effective Stress Framework and Slope Stability Assessment

Landslide hazard evaluation for variably saturated porous media requires an effective stress framework capable of represent-ing both saturated and unsaturated conditions. The generalized effective stress, s¢, framework (Lu and Godt, 2013) used here is

( ) Sau¢s = s- -s [2]

where s is the total stress, ua is the pore-air pressure, and sS is the suction stress. Suction stress is

( )Se a wS u us =- - [3]

where uw is the pore-water pressure and Se is effective saturation:

re

s rS

q-q=

q -q [4]

Typically it is assumed that uatm = ua = 0 (where uatm is the atmospheric pressure, which is the zero reference for gauge pressure).

The infinite slope stability equation can be written (Lu and Godt, 2008) as

( )2 Scos tanFS

sin cos

C z

z

¢+ g b-s f=

g b

¢

b [5]

where FS is the Factor of Safety, C¢ is effective soil cohesion, g is the soil unit weight, z is the soil thickness, b is the slope angle (°), and f¢ is the angle of internal friction (°). An FS £ 1 indicates incipient failure. Soil depths used to estimate slope stability at the SP2 location are the instrument depths (125 cm at SP2), local slope b is 39° at SP2, g is approximately 15.7 kN m−3 (Montgomery et al., 2009), C¢ is 4.6 kPa (root and soil) (Montgomery et al., 2009), and f¢ is 29.1° at SP2 based on direct shear tests (at 10–30 kPa; Smith et al., 2014). Because q s

D does not necessarily equal q sW,

the larger of the two values, which should be qsD, was used as qs

in Eq. [4] to avoid Se exceeding 1. For the estimation of FS from the observed field data, calculation of Se uses qs set equal to the maximum observed q during the period of record (0.432 at the 125-cm depth at SP2) and qr set equal to the minimum observed q during the period of record (0.288 at the 125-cm depth at SP2). Calculation of Se from q in the HYDRUS-1D simulations used qs and qr values from the respective TRIM (laboratory) or RETC (field) values used in the HYDRUS-1D simulations (Supplemental Table S1). Increased air pressure (i.e., ua > 0) from air entrapment was not considered in this study, and ua was taken to be zero in Eq. [3]. The restrictive assumptions associated with the infinite slope approach (Eq. [5]) were further explained by Ebel et al. (2010) and BeVille et al. (2010).

6ResultsLaboratory Soil-Water Retention Data

Substantial hydraulic hysteresis is shown in Fig. 3 for three out of the four laboratory soil-water retention curves fit with the van Genuchten (1980) model. The estimated initial drying curve from complete saturation and the main wetting curve indicate hydraulic hysteresis and lack of closure (i.e., qs

D ¹ q sW) for the 0- to 23-,

50- to 80-, and 75- to 155-cm depth samples. The 23- to 75-cm sample shows a smaller degree of hysteresis, and the wetting and drying curves nearly converge at pore-water pressures near zero. The shallower samples at 0 to 23 and 23 to 75 cm have larger qs

D and higher residual water contents relative to the deeper samples. The trend of increasing bulk density with depth in the soil profile shown in Table 2 is reflected in Fig. 3 by reductions in porosity and therefore also qs

D and qsW. Note that these four samples were

taken from the topographic hollow (see locations of SP2 and SP5 in Fig. 1B), where landslides commonly initiate.

In Situ Soil-Water Retention DataThere are relatively small differences in total rainfall during

the 4 yr of data at the field site. In the 4 yr of field data, total rain-fall ranged from a minimum of 1919 mm in hydrologic year 2010 to a maximum in hydrologic year 2013 of 2299 mm (Fig. 4). The seasonality of yearly rainfall, beginning in mid to late October and largely ending in early to mid June is also shown in Fig. 4. The small amount of interannual variability suggests that the multiple years of combined data can be used for soil-water retention estima-tion without climate influences.

Figure 5 illustrates the magnitude of hydraulic hysteresis for in situ estimates of soil-water retention and soil hydraulic param-eters derived from the in situ soil-water retention data using RETC. Relatively small differences between wetting and drying soil-water retention are shown for all eight of the instrument pairs shown in

Fig. 3. Laboratory soil-water retention estimates from the TRIM method (Wayllace and Lu, 2012) for repacked soil samples from the SP2 and SP5 locations.

VZJ | Advancing Critical Zone Science p. 8 of 15

Fig. 5. Drying data tended to have better RETC fits than wetting data (based on R2 values) between observed and estimated soil-water retention with the van Genuchten (1980) approach for the soil pits (Table 3). The SP2 drying fitted a value is larger than the wetting a value, which is not physically correct (Table 3). The SP6 shallow wetting, SP7 shallow wetting, and SP8 deep wetting sensor data fits have poor R2 values (Table 3). For the SP6 shallow and particularly at the SP7 shallow soil pit, these poor R2 values may reflect scatter in data near zero matric suction (Fig. 5). The SP3, SP6, SP7, and SP8 shallow RETC drying estimates, and the SP8 deep drying sensor data fits have qr values set by RETC during the fitting process (see Table 3). Examination of Fig. 5 suggests that this is the result of a lack of an inflection in the field soil-water retention data at higher suctions (i.e., beyond 50 kPa). A clear issue in the RETC fitting is that qs

W is often larger than qsD, which is

not physically realistic.

Fig. 4. Cumulative rainfall for the hydrologic years (HY) (1 Octo-ber–30 September) for 2010 to 2014 at the Elliot State Forest field site. Cumulative amounts are given next to lines in the legend.

Fig. 5. Best-fit relations of the van Genuchten (VG) (1980) model, using the software RETC, to field data for wetting and drying from paired soil-water content sensors and tensiometers. Instrument depths for each soil pit are given in Table 1, and van Genuchten (1980) parameters from RETC are given in Table 3. Wetting and drying data overlap is partially obscured because drying data are plotted on top of the wetting data. Drying data tended to be at higher suctions because well-defined drying events tended to be in the transition to the dry season when the site was already drying out. Wetting data tended to be at lower suctions because well-defined wetting events tended to be during the winter for prolonged storms. This trend is particularly true for deeper sensors. The earliest storm in each season, when you would expect wetting events at high suctions, tended to take place when the tensiom-eters were at or exceeding the cavitation limit but were not reestablished. Larger versions of each panel, to show more details, are shown in Supplemental Fig. S2A to S2H.

VZJ | Advancing Critical Zone Science p. 9 of 15

Differences between the TRIM soil-water retention estimates from the laboratory and RETC estimates derived from field data are most pronounced in the dry regions of high suction (Fig. 6) for the SP2 and SP5 locations. For the SP2 location portions of the soil-water retention curve corresponding to low suction (wet con-ditions), the RETC estimates for both wetting and drying curves from field data are similar to the drying data from the TRIM labo-ratory estimates. By contrast, at the SP5 location, portions of the soil-water retention curve corresponding to low suction are more similar to the wetting data from TRIM. Overall, there is more pro-nounced hysteresis in soil-water retention in the TRIM estimates relative to the field data based RETC estimates.

Numerical Simulation Comparison of Hydraulic Parameter Estimation

Numerical simulations of the hydrologic response to rainfall show the value of the field instruments for estimating unsaturated hydraulic parameters. Figure 7 shows the graph of the simulated and observed hydrologic response, both soil-water content and matric suction, during the wetter portions of hydrologic years 2011 and 2012 (November–May). The simulations based on the RETC-estimated parameters more closely match the observed data than the TRIM estimates from the laboratory. Measured soil-water content values remain elevated above 0.3 cm3 cm−3 during the wet season; the RETC-based simulation mimics this observed behavior, while the TRIM-based simulation is considerably drier (Fig. 7). The same trend is shown in Fig. 7 for matric suction, where the RETC-based simulation remains between 0 and 10 kPa of suc-tion, similar to the measured field data, while the TRIM-based simulated values are considerably drier and often exceed 10 kPa of suction. Rises in soil-water content following rainfall are spikier in both the observed record and the RETC-based simulation but more subdued in the TRIM-based simulation. These differ-ences in hydrologic simulation fidelity translate into differences in slope failure estimation, as shown by smaller FS values for the RETC-based simulations from field data, which are similar to the FS values estimated from observations, compared with the TRIM-based simulations (Fig. 7). The improvements in simulated soil-water content, suction, and FS are not present solely in the sea-sonal trends shown in Fig. 7 but also at shorter timescales of brief precipitation events associated with slope instability. An example rainstorm from 28–31 Dec. 2012 shows the vast improvement in timing and magnitude of soil-water content, suction, and FS simu-lated values comparing the RETC field-based simulations with the TRIM laboratory-based HYDRUS-1D simulations relative to the FS calculated from observed values (Fig. 8).

6DiscussionComparison between Previous Laboratory and Field Observations of Soil-Water Retention

The laboratory measurements and the in situ estimates of soil-water retention curves are substantially different in some cases,

with the laboratory data showing more pronounced hydraulic hys-teresis than the field observations (Fig. 6). Comparison between laboratory and field-derived soil-water retention curves can pro-duce a mismatch. For example, Li et al. (2005) compared field and laboratory data derived soil-water retention curves and found that the field data showed negligible hysteresis while the laboratory data showed pronounced hysteresis. The work by Bittelli et al. (2012) showed considerable scatter for in situ soil-water retention curves relative to laboratory measurements; however, the in situ measure-ments were from heat dissipation probes that have a resolution of 1 kPa and are limited to pore-water pressures <10 kPa (Flint et al., 2002). Sorbino and Nicotera (2013) compared laboratory-measured drying soil-water retention curves with in situ data from tensiometers and time-domain reflectometry and showed that the in situ data defined a hysteresis envelope with more scatter than laboratory data. Other studies outside of landslide-prone envi-ronments have shown greater similarity between in situ soil-water retention measurements and laboratory measurements (e.g., Royer and Vachaud, 1975; Wraith and Or, 2001; Vaz et al., 2002). There are a number of reasons why in situ soil-water retention estimates may differ from laboratory estimates for a given soil material.

The use of repacked soil samples is one potential cause of the difference between laboratory-measured and in situ soil-water reten-tion relations. Soil samples used in the TRIM analysis were repacked to bulk densities typical of the soil depths in the Oregon Coast Range from which the samples were extracted (Reneau and Dietrich,

Fig. 6. Comparison between van Genuchten (1980) wetting and dry-ing soil-water retention relations from TRIM laboratory estimates and best-fit relations from the software RETC to field data for wetting and drying from paired soil-water content sensors and tensiometers for the SP2 and SP5 locations.

VZJ | Advancing Critical Zone Science p. 10 of 15

1991; Torres et al., 1998; Montgomery et al., 2000). Soil dry bulk density typically increases with depth in the Oregon Coast Range from 0.8 g cm−3 near the surface to 1.5 g cm−3 (Reneau and Dietrich, 1991). Some studies have measured slightly lower bulk densities, such as Torres et al. (1998), who observed 0.6 g cm−3 in organic-rich mate-rial near the surface and 1.2 g cm−3 near the soil–saprolite interface. The repacked bulk densities for the TRIM samples (see Table 2) are similar in trend and magnitude to the aforementioned values, but there may be differences in soil structure that are detrimentally affected by repacking that impact soil-water retention. There are also differences in stone content between the soils characterized by field instruments installed in situ, where weathered bedrock clasts are present (Smith et al., 2014), vs. the repacked laboratory cores, where such clasts are excluded. The in situ soil-water contents could depend on the volume fraction and timescale of imbibition by stones, which may reduce soil-water storage at storm timescales of hours to days. The influence of stones on reducing the available water storage, thus promoting saturation development and increasing the potential for landslide initiation, was suggested by Ebel et al. (2015) for the Colorado Front Range and could also potentially promote landslides in the Oregon Coast Range. The influence of stone content in soils on landslide initiation, including differences between field vs. labora-tory soil-water retention curves, is worthy of further investigation.

Additional factors that may influence disparities between field-measured and laboratory-measured soil-water retention can include temperature effects on soil-water retention relationships

(Nimmo and Miller, 1986; Hopmans and Dane, 1986), hetero-geneity in soil texture and structure combined with topographic effects between the different field locations that cause soil-water content differences (e.g., Western et al., 1999), the presence of macropores (e.g., rodent burrows and conduits from decayed roots) in field soil, and instrument noise and variability. These difficulties present an ongoing challenge to estimating soil-water retention needed to simulate the hydrologic response to rainfall in landslide-prone landscapes. An additional potential reason for the disparity in the magnitude of hydraulic hysteresis between the laboratory- and field-data-derived soil-water retention curves is nonequilibrium unsaturated zone processes, described below.

Static vs. Dynamic Capillary Pressure–Saturation Relationships

Analysis of unsaturated zone flow processes, including the work presented here, commonly relies on equilibrium concepts of surface tension (Miller and Miller, 1956). Recent work shows that dynamic effects can impact the capillary pressure–saturation relationships embodied in soil-water retention curves (e.g., Weitz et al., 1987; Hassanizadeh and Gray, 1993; Hassanizadeh et al., 2002; Diamantopoulos and Durner, 2012). The importance of including dynamic effects has been directly tied to the rate of change in soil-water content (Sakaki et al., 2010). For example, Wildenschild et al. (2001) observed differences in measured soil-water retention depending on the outflow rate for laboratory outflow experiments

Fig. 7. Time series of observed and simu-lated hydrologic response to rainfall at the SP2 instrument location for periods during the 2011 and 2012 hydrologic years. The soil-water content sensor is at the 125-cm depth and the tensiometer providing matric suction is at the 126-cm depth. The Factor of Safety was calcu-lated using Eq. [5].

VZJ | Advancing Critical Zone Science p. 11 of 15

in sandy soils with high outflow rates, particularly for one-step out-flow experiments. Inverse estimates of soil hydraulic parameters from multistep outflow techniques have also exhibited sensitivity to rates of change in soil-water content (Hollenbeck and Jensen, 1998; O’Carroll et al., 2005).

In the laboratory, dynamic effects can be reduced through slow incremental pressure steps in the multistep TRIM appa-ratus outf low experiments. The field measurements of in situ soil-water retention ref lect uncontrolled wetting and drying by natural hydroclimatic processes and are therefore more sub-ject to uncontrolled dynamic effects. The dynamic effects on capillary pressure–saturation relationships may cause some of the scatter in the RETC fits for the in situ soil-water retention curves and make equilibrium concepts unsuitable for analysis. These same dynamic effects may also contribute to the lack of closure (i.e., q s

W ¹ q sD) at zero suction for the TRIM mea-

surements (Fig. 3 and 6) and explain why the TRIM soil-water retention curves exhibit more pronounced hydraulic hysteresis than the soil-water retention curves from the RETC fits to the field data.

Differences in Numerical Simulations of Hydrologic Response for Field vs. Laboratory Soil Hydraulic Parameterizations

The numerical simulations of the unsaturated zone hydrologic response to rainfall show that the simulations parameterized based on RETC estimates of unsaturated hydraulic parameters from field data outperform the simulations parameterized from TRIM laboratory data. It is not surprising that parameters derived from field measurements tend to reproduce those measurements, but the disparity between the field measurements and the simulated hydrologic response using the TRIM laboratory measurements was not expected (Fig. 7 and 8). Further insight into the reasons for these simulated differences can be gleaned by comparing the unsaturated hydraulic parameters for the 75- to 155-cm depth layer between the TRIM- (Table 2) and RETC-based (Soil Pit SP2, d in Table 3) simulations (also see Layer 3 in Supplemental Table S1). Relative to the simulated response in Fig. 7 and 8, the most perti-nent parameterization differences are the disparities in a , qr, and qs

W. The RETC a values are factors of 18 and 51 larger than the TRIM values for wetting and drying, respectively, and the RETC qr values are factors of 2.8 and 3.4 larger than the TRIM values for wetting and drying, respectively (Tables 2 and 3; Supplemental Table S1). The a parameter differences cause the RETC-estimated soil-water retention curve to be steeper, meaning larger changes with soil-water content for a given decrease in suction, and have near-zero air-entry values. These parameter differences cause the simulated hydrologic response using the RETC curve from the field data to have a more rapid, spiky response that more closely matches the observed response relative to the more muted response of the simulation based on the TRIM parameterization. The dif-ferences in qr cause the RETC-based simulation to remain wetter, better matching the field data, relative to the TRIM-based simu-lation, which dries out considerably between storm events (Fig. 7 and 8). The difference in qs

W is also critical because it determines the water storage of the soil during the wetting process, which is typically when slope failures happen. The lower value of q s

W in the TRIM-based simulation causes under-simulation of the soil-water contents, while the RETC-based simulation has qs

W approximately equal to qs

D to match the wetter soil conditions of the field measurements (Fig. 7 and 8). Taken together, these simulated differences show the value of the field measurements for constraining the unsaturated hydraulic parameters that are criti-cal for simulating the subsurface hydrologic response and driving simulated estimates of slope failure.

These results build on the previous work from Reid et al. (2008) showing the value of hillslope monitoring systems for understanding landslide dynamics and the recent work by Bordoni et al. (2015, 2017) and Thomas et al. (2018) that illustrate the value of field observations of the hydrologic response for slope stability estimates. The implications for hydrologically driven slope stabil-ity simulations are that hydrologic field observations can greatly constrain the hydraulic parameters critical to estimating rates and magnitudes of the subsurface response and that laboratory data

Fig. 8. Time series of observed and simulated hydrologic response to rainfall at the SP2 instrument location for a period from 28 to 31 Dec. 2011. The soil-water content sensor is at the 125-cm depth and the tensiometer providing matric suction is at the 126-cm depth. The Fac-tor of Safety was calculated using Eq. [5].

VZJ | Advancing Critical Zone Science p. 12 of 15

alone may not lead to accurate simulation of the unsaturated zone response. Furthermore, the field-based soil-water retention esti-mates produce simulated FS values that are nearer to failure and thus conservative relative to slope failure initiation.

Improving Estimation of Soil Hydraulic Properties

The work presented here suggests several key points for improving estimation of soil hydraulic properties (i.e., soil-water retention) from laboratory methods like TRIM and parameter fitting to field data. In porous media that are prone to dynamic effects in capillary pressure–saturation relationships, slow pressure increment steps during drying and especially during wetting may be critical during the multistep TRIM experiments. The RETC fits in Table 3 and Fig. 5 and 6 suggest that field data near qs

D and qr are essential for reliable estimates of soil-water retention. Field data near zero suction during periods of drying are clearly absent in Fig. 5 and are probably the cause of underestimates of qs

D and the reason why qs

D is less than qsW for many of the soil pits in Table

3. This may explain why the laboratory-estimated qsD values are

slightly larger than the field-data-estimated qsD values compared in

Tables 2 and 3. The results presented here reiterate the conclusion of Wraith and Or (2001) that capturing soil-water contents near saturation (also called field-saturated or “satiated” [after Miller and Bresler, 1977]) is essential for accurate characterization of in situ soil-water retention. The tensiometers used in this work cavitate between 60 and 80 kPa, preventing capturing data near qr that are critical when using computer fitting algorithms like those used in RETC. This is clear in Fig. 5 for sites like SP3, SP6, SP7, and SP8, where further data at higher suctions are needed to capture the inflection point in the soil-water retention and pre-vent estimates of qr equal to zero (see Table 3). The importance of field data from drier conditions of high suction and low soil-water content for simulation of the unsaturated hydrologic response and slope failure was recently noted by Thomas et al. (2018) at a site in California. In the future, it may be advantageous to conduct sprin-kling experiments over instrumented soil pits to capture conditions near saturation (like the experiments done by Torres et al., 1998) or to put precipitation exclusion structures over the instrumented soil pits (like experiments in the ecological sciences such as Nepstad et al. [2007] and Cleveland et al. [2010]) to promote drier condi-tions and install equipment capable of monitoring higher suctions beyond the tensiometer range such as the heat dissipation sensors installed by Bittelli et al. (2012). Topographically driven subsur-face flow may reduce the effectiveness of precipitation exclusion unless a large area is treated. Given certain laboratory protocols for slow pressure increments and a more complete range of field data for in situ soil-water retention and fitting with software such as RETC, the techniques used in this study could provide more reliable information for characterizing soil hydraulic parameters for modeling of slope failure initiation. Longer data records would facilitate separating field data into training vs. evaluation data sets to determine the accuracy of training data estimates.

6ConclusionsSoil hydraulic property measurements from soil samples and

a 4-yr field instrumental record during wetting and drying condi-tions in landslide-prone terrain of the Oregon Coast Range were used to examine unsaturated-zone flow hydraulic properties that determine the hydrologic response that can cause landslide ini-tiation. Laboratory measurements of soil-water retention during both drying and wetting processes indicate pronounced hydraulic hysteresis. The large differences in the volumetric water content at saturation between wetting and drying processes are an impor-tant feature of the hysteretic behavior observed in the laboratory measurements. Soil-water retention estimates derived from field data show considerably less hydraulic hysteresis relative to the laboratory estimates, including minimal differences in the volu-metric water content near saturation between wetting and drying processes. Unsaturated-zone flow simulations comparing the labo-ratory- and field-derived unsaturated hydraulic parameterizations show that the field-derived estimates more closely match the mag-nitude and dynamics of the observed hydrologic response. Field monitoring of subsurface hydrology can be valuable for constrain-ing unsaturated hydraulic parameter estimates for hydraulically driven landslide assessment. The results of this work also suggest that laboratory measurements of soil-water retention made under dynamic conditions may overestimate hydraulic hysteresis in areas where representing hysteresis may not be critical for simulating the subsurface response to rainfall. The in situ estimates of soil-water retention from the field data indicate that capturing conditions at both high suctions and near saturation can reduce parameter estimation problems such as qs

W exceeding qsD and qr equal to

zero. Experimental manipulation of instrumented soil pits such as excluding rainfall to promote drier conditions and controlled sprinkling to achieve very wet conditions may be able to achieve the broader range of hydrologic conditions needed for collection of relevant field data.

AcknowledgmentsFunding for this work was provided by NASA Grant NNX12AO19G awarded to Ning Lu; the funding agency did not influence the study design, data collection or processing, manuscript writing, or the decision to submit the manuscript. Data used in this study are available at http://pubs.er.usgs.gov/publication/ofr20131283 and also through the link at https://pubs.usgs.gov/of/2013/1283/downloads/. Any use of trade, firm, or industry names is for descriptive purposes only and does not imply endorsement by the US Government or the Colorado School of Mines.

ReferencesAnderson, S.P., W.E. Dietrich, D.R. Montgomery, R. Torres, M.E. Conrad,

and K. Loague. 1997a. Subsurface flowpaths in a steep, unchanneled catchment. Water Resour. Res. 33:2637–2653. doi:10.1029/97WR02595

Anderson, S.P., W.E. Dietrich, D.R. Montgomery, R. Torres, and K. Loague. 1997b. Concentration discharge relationships in runoff from a steep, unchanneled catchment. Water Resour. Res. 33:211–225. doi:10.1029/96WR02715

Andrus, C., K. Bahus, K. Craig, J. Runyon, R. Schwindt, and J. Souder. 2003. Elliott State Forest watershed analysis. Dep. of Forestry, Salem, OR. https://www.oregon.gov/ODF/Documents/WorkingForests/Elliott-StateForestWatershedAnalysis.pdf (accessed 11 Aug. 2018).

Baldwin, E.M. 1961. Geologic map of the lower Umpqua River area, Or-

VZJ | Advancing Critical Zone Science p. 13 of 15

egon, scale 1:62,500. Oil Gas Investigations Map 204. USGS, Reston, VA. doi:10.3133/om204

Baum, R.L., J.W. Godt, and W.Z. Savage. 2010. Estimating the timing and location of shallow rainfall-induced landslides using a model for tran-sient, unsaturated infiltration. J. Geophys. Res. 115:F03013 [erratum: 118:1999]. doi:0.1029/2009JF001321

Bear, J. 1972. Dynamics of fluids in porous media. American Elsevier, New York.

Benda, L.E., and T.W. Cundy. 1990. Predicting deposition of debris flows in mountain channels. Can. Geotech. J. 27:409–417. doi:10.1139/t90-057

BeVille, S.H., B.B. Mirus, B.A. Ebel, G.G. Mader, and K. Loague. 2010. Using simulated hydrologic response to revisit the 1973 Lerida Court land-slide. Environ. Earth Sci. 61:1249–1257. doi:10.1007/s12665-010-0448-z

Bittelli, M., R. Valentino, F. Salvatorelli, and P.R. Pisa. 2012. Monitoring soil-water and displacement conditions leading to landslide occur-rence in partially saturated clays. Geomorphology 173–174:161–173. doi:10.1016/j.geomorph.2012.06.006

Bordoni, M., C. Meisina, R. Valentino, N. Lu, M. Bittelli, and S. Chersich. 2015. Hydrological factors affecting rainfall-induced shallow landslides: From the field monitoring to a simplified slope stability analysis. Eng. Geol. 193:19–37. doi:10.1016/j.enggeo.2015.04.006

Bordoni, M., C. Meisina, R. Valentino, M. Bittelli, S. Chersich, M. Mu-setti, and M.G. Persichillo. 2017. Analysis of hydro-meteorological monitoring data collected in different contexts prone to shallow landslides of the Oltrepò Pavese (northern Italy). In: M. Mikoš et al., editors, Advancing culture of living with landslides. World Landslide Forum 2017. Vol. 3. Springer Int., Cham, Switzerland. p. 357–364. doi:10.1007/978-3-319-53487-9_41

Chen, P., B. Mirus, N. Lu, and J.W. Godt. 2017. Effect of hydrau-lic hysteresis on stability of infinite slopes under steady infiltration. J. Geotech. Geoenviron. Eng. 143:04017041. doi:10.1061/(ASCE)GT.1943-5606.0001724

Cleveland, C.C., W.R. Wieder, S.C. Reed, and A.R. Townsend. 2010. Experimen-tal drought in a tropical rain forest increases soil carbon dioxide losses to the atmosphere. Ecology 91:2313–2323. doi:10.1890/09-1582.1

Coe, J.A., J.A. Michael, and M.M. Burgos. 2011. Map of debris flows caused by rainfall during 1996 in parts of the Reedsport and Deer Head Point quadrangles, Douglas County, southern Coast Range, Oregon. Open File Rep. 2011-1150. USGS, Reston, VA.

Collins, B.D., and D. Znidarcic. 2004. Stability analyses of rainfall in-duced landslides. J. Geotech. Geoenviron. Eng. 130:362–372. doi:10.1061/(ASCE)1090-0241(2004)130:4(362)

Decagon Devices. 2018. Cover more space: ECH2O EC5. METER Environ-ment, Pullman, WA. https://www.metergroup.com/environment/prod-ucts/ec-5-soil-moisture-sensor/ (accessed 13 Aug. 2018).

Diamantopoulos, E., and W. Durner. 2012. Dynamic nonequilibrium of water flow in porous media: A review. Vadose Zone J. 11(3). doi:10.2136/vzj2011.0197

Ebel, B.A., and K. Loague. 2008. Rapid simulated hydrologic response with-in the variably saturated near surface. Hydrol. Processes 22:464–471. doi:10.1002/hyp.6926

Ebel, B.A., K. Loague, and R.I. Borja. 2010. The impacts of hydraulic hys-teresis on variably saturated hydrologic response and slope failure. Environ. Earth Sci. 61:1215–1225. doi:10.1007/s12665-009-0445-2

Ebel, B.A., K. Loague, D.R. Montgomery, and W.E. Dietrich. 2008. Phys-ics‐based continuous simulation of long‐term near‐surface hydrologic response for the Coos Bay experimental catchment. Water Resour. Res. 44:W07417. doi:10.1029/2007WR006442

Ebel, B.A., and B.B. Mirus. 2014. Disturbance hydrology: Challenges and opportunities. Hydrol. Processes 28:5140–5148.

Ebel, B.A., F.K. Rengers, and G.E. Tucker. 2015. Aspect-dependent soil satu-ration and insight into debris-flow initiation during extreme rainfall in the Colorado Front Range. Geology 43:659–662. doi:10.1130/G36741.1

Eching, S.O., J.W. Hopmans, and O. Wendroth. 1994. Unsatu-rated hydraulic conductivity from transient multistep outflow and soil water pressure data. Soil Sci. Soc. Am. J. 58:687–695. doi:10.2136/sssaj1994.03615995005800030008x

Flint, A.L., G.S. Campbell, K.M. Ellett, and C. Calissendorff. 2002. Calibration

and temperature correction of heat dissipation matric potential sen-sors. Soil Sci. Soc. Am. J. 66:1439–1445. doi:10.2136/sssaj2002.1439

Frattini, P., G.B. Crosta, N. Fusi, and P. Dal Negro. 2004. Shallow landslides in pyroclastic soils: A distributed modelling approach for hazard as-sessment. Eng. Geol. 73:277–295. doi:10.1016/j.enggeo.2004.01.009

Godt, J.W., R.L. Baum, and N. Lu. 2009. Landsliding in partially saturated materials. Geophys. Res. Lett. 36:L02403. doi:10.1029/2008GL035996

Godt, J.W., R.L. Baum, W.Z. Savage, D. Salciarini, W.H. Schulz, and E.L. Harp. 2008. Transient deterministic shallow landslide modeling: Require-ments for susceptibility and hazard assessments in a GIS framework. Eng. Geol. 102:214–226. doi:10.1016/j.enggeo.2008.03.019

Hassanizadeh, S.M., M.A. Celia, and H.K. Dahle. 2002. Dynamic effect in the capillary pressure–saturation relationship and its impacts on unsatu-rated flow. Vadose Zone J. 1:38–57. doi:10.2136/vzj2002.3800

Hassanizadeh, S.M., and W.G. Gray. 1993. Toward an improved descrip-tion of the physics of two-phase flow. Adv. Water Resour. 16:53–67. doi:10.1016/0309-1708(93)90029-F

Hollenbeck, K.J., and K.H. Jensen. 1998. Experimental evidence of ran-domness and nonuniqueness in unsaturated outflow experiments designed for hydraulic parameter estimation. Water Resour. Res. 34:595–602. doi:10.1029/97WR03609

Hopmans, J.W., and J.H. Dane. 1986. Temperature dependence of soil water retention curves. Soil Sci. Soc. Am. J. 50:562–567. doi:10.2136/sssaj1986.03615995005000030004x

Iverson, R.M. 2000. Landslide triggering by rain infiltration. Water Resour. Res. 36:1897–1910. doi:10.1029/2000WR900090

Iverson, R.M., D.L. George, K. Allstadt, M.E. Reid, B.D. Collins, J.W. Val-lance, et al. 2015. Landslide mobility and hazards: Implications of the 2014 Oso disaster. Earth Planet. Sci. Lett. 412:197–208. doi:10.1016/j.epsl.2014.12.020

Johnson, D.R., J.T. Haagen, and A.C. Terrell. 1994. Soil survey of Douglas County Area, Oregon. US Gov. Print. Office, Washington, DC.

Kool, J.B., and J.C. Parker. 1987. Development and evaluation of closed form expressions for hysteretic soil hydraulic properties. Water Resour. Res. 23:105–114. doi:10.1029/WR023i001p00105

LaHusen, S.R., A.R. Duvall, A.M. Booth, and D.R. Montgomery. 2016. Sur-face roughness dating of long-runout landslides near Oso, Washing-ton (USA), reveals persistent postglacial hillslope instability. Geology 44:111–114. doi:10.1130/G37267.1

Larsen, I.J., D.R. Montgomery, and O. Korup. 2010. Landslide ero-sion controlled by hillslope material. Nat. Geosci. 3:247–251. doi:10.1038/ngeo776

Li, A.G., L.G. Tham, Z.Q. Yue, C.F. Lee, and K.T. Law. 2005. Com-parison of field and laboratory soil-water characteris-tic curves. J. Geotech. Geoenviron. Eng. 131:1176–1180. doi:10.1061/(ASCE)1090-0241(2005)131:9(1176)

Likos, W.J., N. Lu, and J.W. Godt. 2014. Hysteresis and uncertainty in soil water-retention curve parameters. J. Geotech. Geoenviron. Eng. 140:04013050. doi:10.1061/(ASCE)GT.1943-5606.0001071

Lu, N., and J. Godt. 2008. Infinite slope stability under steady un-saturated seepage conditions. Water Resour. Res. 44:W11404. doi:10.1029/2008WR006976

Lu, N., and J.W. Godt. 2013. Hillslope hydrology and stability. Cambridge Univ. Press, New York. doi:10.1017/CBO9781139108164

Lu, N., M. Kaya, B.D. Collins, and J.W. Godt. 2013. Hysteresis of unsaturated hydromechanical properties of a silty soil. J. Geotech. Geoenviron. Eng. 139:507–510. doi:10.1061/(ASCE)GT.1943-5606.0000786

Ma, K.C., Y.C. Tan, and C.H. Chen. 2011. The influence of water retention curve hysteresis on the stability of unsaturated soil slopes. Hydrol. Processes 25:3563–3574. doi:10.1002/hyp.8081

Marion, J.M., D. Or, D.E. Rolston, M.L. Kavvas, and J.W. Biggar. 1994. Evaluation of methods for determining soil-water retentiv-ity and unsaturated hydraulic conductivity. Soil Sci. 158:1–13. doi:10.1097/00010694-199407000-00001

Marquardt, D.W. 1963. An algorithm for least-squares estimation of nonlinear parameters. J. Soc. Ind. Appl. Math. 11:431–441. doi:10.1137/0111030

VZJ | Advancing Critical Zone Science p. 14 of 15

Millar, C.I., and N.L. Stephenson. 2015. Temperate forest health in an era of emerging megadisturbance. Science 349:823–826. doi:10.1126/science.aaa9933

Miller, E.E., and R.D. Miller. 1956. Physical theory for capillary flow phe-nomena. J. Appl. Phys. 27:324–332. doi:10.1063/1.1722370

Miller, R.D., and E. Bresler. 1977. A quick method for estimating water diffusivity functions. Soil Sci. Soc. Am. J. 41:1020–1022. doi:10.2136/sssaj1977.03615995004100050044x

Mirus, B.B., B.A. Ebel, C.H. Mohr, and N. Zegre. 2017a. Disturbance hydrol-ogy: Preparing for an increasingly disturbed future. Water Resour. Res. 53:10,007–10,016. doi:10.1002/2017WR021084

Mirus, B.B., J.B. Smith, and R.L. Baum. 2017b. Hydrologic impacts of land-slide disturbances: Implications for remobilization and hazard persis-tence. Water Resour. Res. 53:8250–8265. doi:10.1002/2017WR020842

Montgomery, D.R., and W.E. Dietrich. 1994. A physically based model for the topographic control on shallow landsliding. Water Resour. Res. 30:1153–1171. doi:10.1029/93WR02979

Montgomery, D.R., W.E. Dietrich, R. Torres, S.P. Anderson, J.T. Heffner, and K. Loague. 1997. Hydrologic response of a steep, unchanneled valley to natural and applied rainfall. Water Resour. Res. 33:91–109. doi:10.1029/96WR02985

Montgomery, D.R., K.M. Schmidt, W.E. Dietrich, and J. McKean. 2009. Instrumental record of debris flow initiation during natural rainfall: Implications for modeling slope stability. J. Geophys. Res. 114:F01031. doi:10.1029/2008JF001078

Montgomery, D.R., K.M. Schmidt, H.M. Greenberg, and W.E. Dietrich. 2000. Forest clearing and regional landsliding. Geology 28:311–314. doi:10.1130/0091-7613(2000)28<311:FCARL>2.0.CO;2

Mualem, Y. 1974. A conceptual model of hysteresis. Water Resour. Res. 10:514–520. doi:10.1029/WR010i003p00514

Mualem, Y. 1976. A new model for predicting the hydraulic conductiv-ity of unsaturated porous media. Water Resour. Res. 12:513–522. doi:10.1029/WR012i003p00513

Mualem, Y. 1984. A modified dependent-domain theory of hysteresis. Soil Sci. 137:283–291. doi:10.1097/00010694-198405000-00001

Nepstad, D.C., I.M. Tohver, D. Ray, P. Moutinho, and G. Cardinot. 2007. Mortality of large trees and lianas following experimental drought in an Amazon forest. Ecology 88:2259–2269. doi:10.1890/06-1046.1

Ng, C.W.W., and Q. Shi. 1998. Influence of rainfall intensity and duration on slope stability in unsaturated soils. Q. J. Eng. Geol. Hydrogeol. 31:105–113. doi:10.1144/GSL.QJEG.1998.031.P2.04

Niem, A.R., and W.A. Niem. 1990. Explanation of the geologic units for the compilation geologic map of the Southern Tyee Basin, Southern Coast Range, Oregon. Open-File Rep. 0-89-3. Dep. of Geology and Mineral Industries, Portland, OR.

Nimmo, J.R., and E.E. Miller. 1986. The temperature dependence of iso-thermal moisture vs. potential characteristics of soils. Soil Sci. Soc. Am. J. 50:1105–1113. doi:10.2136/sssaj1986.03615995005000050004x

O’Carroll, D.M., T.J. Phelan, and L.M. Abriola. 2005. Exploring dynamic effects in capillary pressure in multistep outflow experiments. Water Resour. Res. 41:W11419. doi:10.1029/2005WR004010

Petley, D. 2012. Global patterns of loss of life from landslides. Geology 40:927–930. doi:10.1130/G33217.1

Pierce, J.L., G.A. Meyer, and A.J.T. Jull. 2004. Fire-induced erosion and millennial-scale climate change in northern ponderosa pine forests. Nature 432:87–90. doi:10.1038/nature03058

Reid, M.E., R.L. Baum, R.G. LaHusen, and W.L. Ellis. 2008. Capturing land-slide dynamics and hydrologic triggers using near-real-time moni-toring. In: Z. Chen et al., editors, Landslides and engineered slopes: From the past to the future. Vol. 2. CRC Press, London. p. 179–191. doi:10.1201/9780203885284-c10

Reid, M.E., R.G. LaHusen, and R.M. Iverson. 1997. Debris-flow initiation experiments using diverse hydrologic triggers. In: C. Chen, editor, Pro-ceedings of the First International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment, San Francisco. 7–9 Aug. 1997. Am. Soc. Civ. Eng., Reston, VA.

Reneau, S.L., and W.E. Dietrich. 1991. Erosion rates in the southern Oregon

Coast Range: Evidence for an equilibrium between hillslope erosion and sediment yield. Earth Surf. Processes Landforms 16:307–322. doi:10.1002/esp.3290160405

Robinson, D.A., C.S. Campbell, J.W. Hopmans, B.K. Hornbuckle, S.B. Jones, R. Knight, et al. 2008. Soil moisture measurement for ecological and hydrological watershed-scale observatories: A review. Vadose Zone J. 7:358–389. doi:10.2136/vzj2007.0143

Roering, J.J., K.M. Schmidt, J.D. Stock, W.E. Dietrich, and D.R. Montgomery. 2003. Shallow landsliding, root reinforcement, and the spatial distribu-tion of trees in the Oregon Coast Range. Can. Geotech. J. 40:237–253. doi:10.1139/t02-113

Royer, J.M., and G. Vachaud. 1975. Field determination of hyster-esis in soil-water characteristics. Soil Sci. Soc. Am. J. 39:221–223. doi:10.2136/sssaj1975.03615995003900020006x

Sakaki, T., D.M. O’Carroll, and T.H. Illangasekare. 2010. Direct quantification of dynamic effects in capillary pressure for drainage–wetting cycles. Vadose Zone J. 9:424–437. doi:10.2136/vzj2009.0105

Schmidt, K.M., J.J. Roering, J.D. Stock, W.E. Dietrich, D.R. Montgomery, and T. Schaub. 2001. The variability of root cohesion as an influence on shallow landslide susceptibility in the Oregon Coast Range. Can. Geotech. J. 38:995–1024. doi:10.1139/t01-031

Šimůnek, J., M. Šejna, H. Saito, M. Sakai, and M.Th. van Genuchten. 2008a. The HYDRUS-1D software package for simulating the move-ment of water, heat, and multiple solutes in variably saturated media, version 4.08. HYDRUS Software Ser. 3. Dep. of Environ. Sci., Univ. of California, Riverside.

Šimůnek, J., M.Th. van Genuchten, and M. Šejna. 2008b. Development and applications of the HYDRUS and STANMOD software packages and related codes. Vadose Zone J. 7:587–600. doi:10.2136/vzj2007.0077

Smith, J.B., J.W. Godt, R.L. Baum, J.A. Coe, W.J. Burns, N. Lu, et al. 2014. Hy-drologic monitoring of a landslide-prone hillslope in the Elliott State Forest, Southern Coast Range, Oregon, 2009–2012. Open-File Rep. 2013-1283. USGS, Reston, VA. doi:10.3133/ofr20131283

Soil Survey Staff. 2013. Web soil survey. Natl. Soil Surv. Ctr., Lincoln, NE. http://websoilsurvey.nrcs.usda.gov/ (accessed 12 Mar. 2013).

Sorbino, G., and M.V. Nicotera. 2013. Unsaturated soil mechan-ics in rainfall-induced flow landslides. Eng. Geol. 165:105–132. doi:10.1016/j.enggeo.2012.10.008

Stock, J., and W.E. Dietrich. 2003. Valley incision by debris flows: Evidence of a topographic signature. Water Resour. Res. 39:1089. doi:10.1029/2001WR001057

Thomas, M.A., B.B. Mirus, B.D. Collins, N. Lu, and J.W. Godt. 2018. Variability in soil-water retention properties and implications for physics-based simulation of landslide early warning criteria. Landslides 15:1265–1277. doi:10.1007/s10346-018-0950-z

Topp, G.C. 1969. Soil-water hysteresis measured in a sandy loam and compared with the hysteretic domain model. Soil Sci. Soc. Am. Proc. 33:645–651. doi:10.2136/sssaj1969.03615995003300050011x

Topp, G.C. 1971. Soil water hysteresis in silt loam and clay loam soils. Water Resour. Res. 7:914–920. doi:10.1029/WR007i004p00914

Torres, R., W.E. Dietrich, D.R. Montgomery, S.P. Anderson, and K. Loague. 1998. Unsaturated zone processes and the hydrologic response of a steep unchanneled catchment. Water Resour. Res. 34:1865–1879. doi:10.1029/98WR01140

Tsai, T.-L. 2011. Influences of soil water characteristic curve on rain-fall-induced shallow landslides. Environ. Earth Sci. 64:449–459. doi:10.1007/s12665-010-0868-9

Turner, M.G. 2010. Disturbance and landscape dynamics in a changing world. Ecology 91:2833–2849. doi:10.1890/10-0097.1

UMS. 2014. UMS T8 tensiometer manual. UMS GmbH, Munich. http://manuals.decagon.com/Manuals/UMS/T8_Manual.pdf (accessed 27 Dec. 2014).

van Dam, J.C., J.N.M. Stricker, and P. Droogers. 1994. Inverse method to determine soil hydraulic functions from multi-step outflow experiments. Soil Sci. Soc. Am. J. 58:647–652. doi:10.2136/sssaj1994.03615995005800030002x

van Genuchten, M.Th. 1980. A closed-form equation for predicting the hydraulic conductivity of unsaturated soil. Soil Sci. Soc. Am. J.

VZJ | Advancing Critical Zone Science p. 15 of 15

44:892–898. doi:10.2136/sssaj1980.03615995004400050002x

van Genuchten, M.Th., F.J. Leij, and S.R. Yates. 1991. The RETC code for quantifying the hydraulic functions of unsaturated soils. EPA/600/2091/065. USEPA, Robert S. Kerr Environ. Res. Lab., Ada, OK.

Vaz, C.M., J.W. Hopmans, A. Macedo, L.H. Bassoi, and D. Wildenschild. 2002. Soil water retention measurements using a combined tensiometer–coiled time domain reflectometry probe. Soil Sci. Soc. Am. J. 66:1752–1759. doi:10.2136/sssaj2002.1752

Wartman, J., D.R. Montgomery, S.A. Anderson, J.R. Keaton, J. Ben-oît, J. de la Chapelle, and R. Gilbert. 2016. The 22 March 2014 Oso landslide, Washington, USA. Geomorphology 253:275–288. doi:10.1016/j.geomorph.2015.10.022

Wayllace, A., and N. Lu. 2012. A transient water release and imbibitions method for rapidly measuring wetting and drying soil water retention and hydraulic conductivity functions. Geotech. Test. J. 35:103–117. doi:10.1520/GTJ103596

Weitz, D.A., J.P. Stokes, R.C. Ball, and A.P. Kushnick. 1987. Dynamic capillary pressure in porous media: Origin of the viscous-fingering length scale. Phys. Rev. Lett. 59:2967. doi:10.1103/PhysRevLett.59.2967

Western, A.W., R.B. Grayson, G. Bloschl, G.R. Willgoose, and T.A. Mc-Mahon. 1999. Observed spatial organization of soil moisture and its relation to terrain indices. Water Resour. Res. 35:797–810. doi:10.1029/1998WR900065

Wheeler, S.J., R.S. Sharma, and M.S.R. Buisson. 2003. Coupling of hydraulic hysteresis and stress–strain behavior in unsaturated soils. Geotech-nique 53:41–54. doi:10.1680/geot.2003.53.1.41

Wildenschild, D., J.W. Hopmans, and J. Šimůnek. 2001. Flow rate depen-dence of soil hydraulic characteristics. Soil Sci. Soc. Am. J. 65:35–48. doi:10.2136/sssaj2001.65135x

Wraith, J.M., and D. Or. 2001. Soil water characteristic determination from concurrent water content measurements in reference porous media. Soil Sci. Soc. Am. J. 65:1659–1666. doi:10.2136/sssaj2001.1659