Abstract

Hydrologic processes during extreme rainfall events are poorly characterized because of the rarity of measurements. Improved understanding of hydrologic controls on natural hazards is needed because of the potential for substantial risk during extreme precipitation events. We present field measurements of the degree of soil saturation and estimates of available soil-water storage during the September 2013 Colorado extreme rainfall event at burned (wildfire in 2010) and unburned hillslopes with north- and south-facing slope aspects. Soil saturation was more strongly correlated with slope aspect than with recent fire history; south-facing hillslopes became fully saturated while north-facing hillslopes did not. Our results suggest multiple explanations for why aspect-dependent hydrologic controls favor saturation development on south-facing slopes, causing reductions in effective stress and triggering of slope failures during extreme rainfall. Aspect-dependent hydrologic behavior may result from (1) a larger gravel and stone fraction, and hence lower soil-water storage capacity, on south-facing slopes, and (2) lower weathered-bedrock permeability on south-facing slopes, because of lower tree density and associated deep roots penetrating bedrock as well as less intense weathering, inhibiting soil drainage.

INTRODUCTION

Extreme precipitation events disproportionately impact landscape processes including erosion, flooding, and natural hazards such as debris flows (e.g., Ogden et al., 2000; Coe et al., 2014). The importance of extreme precipitation events motivates the need to predict hydrologic response and associated natural hazards. Yet conceptualization and prediction remain elusive for extreme precipitation events (Lyon et al., 2008; Borga et al., 2014). Accurate predictions of hydrologically driven natural hazards are particularly challenging where environments recovering from landscape disturbance are combined with slope-aspect effects.

A historically unprecedented rainfall event in September 2013 along the Front Range of Colorado, United States, provided a unique natural experiment in hillslope hydrologic and geomorphic response to extreme precipitation. The Colorado Front Range is a landscape with mixed levels of disturbance, and patterns of vegetation and soil development that vary systematically with slope aspect, which affect hydrologic processes and associated geomorphologic response. The September 2013 event was a prolonged (∼7 days) storm with rainfall totals near the city of Boulder in excess of 200 mm on 12 September and 100 mm on 13 September (NOAA, 2014). Anomalous amounts of precipitable water in the atmosphere and strongly tropical rainfall characteristics document the extreme nature of this precipitation event (Gochis et al., 2014). The hydrologic consequences of the rainfall included substantial flooding, more than 1100 debris flows that caused three fatalities (Coe et al., 2014), and over US$2 billion in damages (Gochis et al., 2014).

Within the footprint of this extreme precipitation event, we are particularly interested in effects on a 2500 ha sub-watershed in the Front Range foothills that burned in the 2010 Fourmile Canyon fire. Because wildfire tends to reduce infiltration and thus increase runoff (Larsen et al., 2009), concerns over flooding prior to the September 2013 storms motivated a multiyear study aimed at understanding the hydrologic response to rainfall in burned and unburned areas with north- and south-facing slope aspects in and near the 2010 Fourmile Canyon fire (Ebel, 2012, 2013a, 2013b). Hydrologic recovery from wildfire can vary greatly, from 3 yr to decades (Shakesby and Doerr, 2006), so the relative effect of aspect versus wildfire on hydrologic response and related hazards for this rainfall event was not clear a priori.

The large number of slope failures was an unexpected consequence of the September 2013 precipitation event, and Coe et al. (2014) characterized the spatial locations, characteristics, and timing of slope failures. A striking pattern that emerged from this analysis was that over 78% of debris flows initiated on south-facing slopes, whereas recently burned areas had only slightly higher rates of slope failure than unburned areas (Coe et al., 2014). Potential factors suggested by Coe et al. (2014) for the aspect dependence include thinner soils, vegetation differences, rainfall dependence on storm direction, and more abundant rock outcrops on south-facing slopes. There are additional potential geomorphic factors such as soil development differences (Birkeland et al., 2003) that may affect geomechanical parameters including soil cohesion, friction angle, or root cohesion.

Slope aspect can have major impacts on hydrologic processes that drive slope failures during extreme rainfall. For example, aspect-dependent root strength and rainfall interception can impact slope failure locations (Jacobson et al., 1989). A study of surface erosion during an extreme rainfall event in New Mexico showed that north-facing slopes were generally wetter and fully saturated for longer than south-facing slopes (Gutiérrez-Jurado et al., 2007). The cause of the strong correlation between slope aspect and debris-flow initiation during the extreme precipitation event of September 2013 in Colorado remains a critical unsolved question with important implications for natural hazards and landscape evolution.

Here we use in-situ field measurements of soil-water content during an extreme rainfall event to examine hydrologic and geomorphologic response for experimental plots with opposing north- and south-facing aspects in burned and unburned conditions. Our analysis focuses on the time frame during which rainfall intensity and extent were greatest, 11–12 September 2013 (Gochis et al., 2014), which coincides with time periods of debris-flow triggering (Coe et al., 2014), to gain insight into aspect controls on hydrologic response and associated slope failures. We find that the extent and duration of soil saturation correlate strongly with slope aspect regardless of wildfire disturbance, which has important implications for aspect-dependent debris-flow initiation.

EXPERIMENTAL METHODS

Research Area

The research area is at the western edge of the 2010 Fourmile Canyon fire on a ridge extending east from Sugarloaf Mountain. The area has an elevation range of 2350–2450 m with relatively steep slopes (15°–28°) covered by predominately gravelly sandy soils with a mean soil depth of ∼40 cm. A relatively permeable weathered bedrock layer (Boulder Creek Granodiorite) underlies the soil. The eastern flank of the Colorado Front Range hosts east-west–trending drainage networks that have predominantly north- and south-facing hillslopes. Vegetation is aspect dependent (Marr, 1961), with aspen (Populus tremuloides), Rocky Mountain Douglas fir (Pseudotsuga menziesii var. glauca), and limber pine (Pinus flexilis) in dense forest stands on north-facing slopes contrasting with sparse ponderosa pine (Pinus ponderosa), Rocky Mountain juniper (Juniperus scopulorum), and grassy understory on south-facing slopes (Ebel, 2013a). Soil formation can also differ by aspect in this area, with north-facing slopes having a prominent litter and/or duff layer and associated thin O horizon, and south-facing slopes having an A horizon (Birkeland et al., 2003).

Rainfall

Rainfall during the September 2013 storms was recorded by tipping-bucket rain gages that telemetered the time of each tip. The closest gage (Sugarloaf, ID number 4730; see Fig. DR1 in the GSA Data Repository1) was 1.5 km from the research area and was operated by the Urban Drainage and Flood Control District (www.udfcd.org). Information on rainfall spatial variability is provided in the Data Repository (Fig. DR1) and by Gochis et al. (2014). Cumulative rainfall from the Sugarloaf gage is converted into 30 min average rainfall rates.

Soil-Water Content

Soil-water content (cm3 cm−3) data indicate the degree of saturation within each of the different experimental plots, and reveal which plots became fully saturated. Increases in the degree of saturation reduce effective normal stress and enhance the propensity for slope failure and debris-flow initiation (e.g., Godt et al., 2009). Volumetric soil-water content was measured at seven plot locations with multiple sensor depths at each plot (Fig. 1; Table 1) by using automated subsurface sensors (Decagon 5TE; Decagon Devices, 2006) at 5 min temporal resolution. These sensors were previously calibrated for soil from each plot and temperature corrected with accuracy to ±0.01–0.02 cm3 cm−3 (Ebel, 2013a). Degree of saturation is calculated as volumetric soil-water content divided by previously measured saturated soil-water content of soil cores for each plot (Table 1) and assuming residual water content is zero (Ebel, 2012).

Soil-Water Storage Analysis

A simple one-dimensional soil-water storage analysis provides insight into the relative available storage at the plot scale (Table 1). Total available soil-water storage relative to cumulative rainfall is a first-order control on the development of fully saturated soils. The local surface slope angles are similar (Table 1), suggesting minimal differences between plots in subsurface downslope flow from the soil column driven solely by differences in slope angle. The plots are on quasi-planar hillslopes, suggesting that there are not substantial inflows from surface or subsurface flow driven by convergent topography. Measured soil depth (Ebel, 2013a) was multiplied by measured saturated soil-water content for each plot (Ebel, 2012) to estimate total soil-water storage (mm). Initial filled storage is estimated by multiplying initial volumetric soil-water content (9 September 2013 at 0:00 h Mountain Time [MT]) by the soil depth range closest to a given soil-water content sensor. The total available storage is estimated as total storage minus the initial filled storage at 9 September 2013 at 0:00 h MT (Table 1).

RESULTS AND DISCUSSION

Rainfall and soil moisture data (Fig. 2) reveal an unexpected correlation between slope aspect and soil-moisture dynamics at the hillslope scale, regardless of wildfire disturbance, during the 2013 storm. On north-facing slopes, burned soils (north-facing ridge [NFR], difference infiltrometer 1 [DI1], and north-facing midslope [NFM]) reached higher degrees of saturation than unburned soils (unburned north-facing [UBNF]) at 5 and 10 cm depths (Figs. 2B and 2C). North-facing slope soils did not fully saturate during the extreme rainfall event (Figs. 2B–2D), even during peak rainfall intensity at ∼01:00 MT on 12 September (30 min intensity of 30.7 mm hr−1). In contrast, burned (south-facing ridge [SFR]) and unburned (UBSFR) south-facing slope soils did fully saturate during the rainfall event (Figs. 2F–2H). Both the burned SFR and unburned UBSFR plots fully saturated, to within sensor accuracy, on 12 September 2013 at 1:00 MT during the period of peak rainfall intensity (Figs. 2F–2H). A second time period of elevated degree of saturation, just below fully saturated, in south-facing soils occurred at ∼21:00 h on 12 September when peak 30 min rainfall intensity was 19.3 mm hr−1 (Figs. 2G and 2H). These time periods of elevated degrees of saturation on south-facing slopes match time periods of observed debris-flow initiations by Coe et al. (2014) noted in Figures 2B–2D and 2F–2H (green bars). The observed aspect control on the development of fully saturated soils is consistent with observations of debris-flow initiations driven by reductions in effective normal stress that were concentrated on south-facing slopes (Coe et al., 2014). The timing of south-facing sensors reaching fully saturated conditions (Fig. 2) lies within the window of 23:30 h 11 September to 01:00 h 12 September given by Coe et al. (2014, his landslide number 3) for a nearby debris flow (∼8 km away).

The simple water storage analysis (Table 1) shows that there are no systematic differences in soil depth, saturated soil-water contents (i.e., porosity), pre-event soil-water content, or total available storage between north- and south-facing slope plots that explain the aspect differences in the degree of saturation. The total available storage at the start of the rainfall event was similar for south-facing plots that fully saturated (SFR and UBSFR) and north-facing plots that did not fully saturate (UBNF, DI1, and NFM) (Table 1; Fig. 2). Total rainfall exceeds total available storage at peak degree of saturation for the UBSFR, UBNF, DI1, NFM, and SFR plots (Table 1; Fig. 2), indicating possible differences in the soil drainage rates. In these relatively planar hillslopes, there are not likely to be appreciable differences between plots for convergent or divergent subsurface or surface flow (similar local slope angles; see Table 1), and subsurface downslope flux is approximately balanced by subsurface upslope flux as a first-order approximation. Prior analysis of subsurface soil saturated hydraulic conductivities did not show systematic differences between aspects (Ebel, 2013b). A critical unknown is the vertical leakage rate from soils into the underlying weathered bedrock.

Vertical leakage from the soil into the underlying weathered bedrock is a first-order control on the development of fully saturated soils. More intense physical weathering by freezing processes (Anderson et al., 2013) and chemical weathering (Langston et al., 2015) on north-facing slopes and deeper weathered profiles on north-facing slopes (Befus et al., 2011) in the nearby Boulder Creek Critical Zone Observatory suggest that aspect-dependent weathering processes affect the hydrologic interface between soil and weathered bedrock. It is unclear if weathered bedrock is more permeable on north-facing slopes, thus promoting vertical soil drainage and inhibiting the development of fully saturated soils on north-facing slopes in contrast to south-facing slopes. Weathered bedrock permeability may also be impacted by root penetration, which implies that higher north-facing slope tree density (Ebel, 2013a) could cause greater bedrock permeability. Berndt and Gibbons (1958) noted similar rooting depths and lateral spreads for the dominant tree species for the north- and south-facing aspects in the montane zone in the Colorado Front Range, suggesting that tree density rather than root structure is the primary difference between aspects.

Another potential explanation for the development of fully saturated soils on south-facing slopes may be that large stone fragments in the soil column reduce effective porosity and therefore water storage capacity at storm time scales. Analysis of gravel content in soils for this area by Moody and Nyman (2013) suggests that south-facing soils have larger gravel fractions (particles >2 mm diameter). They examined the top 10 cm of soil and found mean gravel fractions by mass of 33.4% on north-facing slopes (N = 12) and 39.4% on south-facing slopes (N = 12) (Table DR1 in the Data Repository); a two-tailed t-test without assuming equal variances showed a significant difference between aspects at the 7.5% significance level (i.e., threshold p = 0.075). Saturated soil-water contents shown in Table 1 are based on 4-cm-diameter cores from 1 to 8 cm depth and exclude the influence of large stones. A 5% aspect difference in stone content by mass could result in an ∼5% aspect difference in saturated soil-water contents (see calculation in the Data Repository). The reduction of porosity and available soil-water storage by large stones on south-facing slopes because of gravel and large stones may only be critical for extreme rainfall events but could promote fully saturated soils and associated slope failures in concert with other factors.

Other factors may also have contributed to slope-aspect dependence for debris-flow initiation during this event. It is possible that the south-facing slopes receive greater incident rainfall because of the north and westward movement of the storms (Coe et al., 2014). The rain gage network is not sufficient to test or rule this out. Vegetation density and rooting habit have been shown to affect slope failure from root strength contributions (e.g., Rice et al., 1969). The lack of wildfire signature in debris-flow initiation may result from substantial understory vegetation recovery (Coe et al., 2014) and the relatively low rainfall rates relative to infiltration rates. The low-intensity, long-duration rainfall favored subsurface hydrologic response and associated debris-flow generation where aspect factors predisposed hillslopes to hydrologically driven slope failure.

CONCLUSIONS AND IMPLICATIONS

Here we show that the degree of saturation and the development of fully saturated soils during an extreme rainfall event were dominated by slope-aspect effects. South-facing slopes developed fully saturated soils regardless of wildfire disturbance, while north-facing slopes did not fully saturate. At our experimental plots, the aspect dependence could not be attributed to aspect-related differences in soil depth, saturated soil-water contents, or slope angle. We suggest that (1) soil-water storage on south-facing slopes may be reduced by large stones in the soil profile, relative to north-facing slopes, which leads to fully saturated soils and enhanced slope failure probability, and (2) the rate of vertical leakage from the soil into the underlying weathered bedrock is greater on north-facing slopes because of greater weathering intensity and tree density, thus inhibiting the development of fully saturated soils on north-facing aspects and reducing slope failure probability. Further investigation of these two hypotheses is warranted for hazard predictions for hillslopes and headwater catchments in areas where unexpected aspect effects may emerge during extreme rainfall events.

This work benefitted from discussions with J. Moody, B. Mirus, D. Stonestrom, and comments by anonymous reviewers. Any use of trade, firm, or industry names is for descriptive purposes only and does not imply endorsement by the U.S. Government. Ebel was supported by the CIRES Visiting Fellow Program.

1GSA Data Repository item 2015233, analysis of the spatial variability in rainfall and calculations regarding stone impacts on soil porosity, including the saturation time series at each plot, is available online at www.geosociety.org/pubs/ft2015.htm, or on request from editing@geosociety.org or Documents Secretary, GSA, P.O. Box 9140, Boulder, CO 80301, USA.