Multi-Stage Soil-Hydraulic Recovery and Limited Ravel Accumulations Following the 2017 Nuns and Tubbs Wildfires in Northern California

.


10.1029/2022JF006591
2 of 19 Post-fire landscapes therefore experience a window of elevated hazard until they sufficiently recover (Ebel, 2020;Swanson, 1981), and studies of wildfire disturbance show that the recovery time for field-saturated hydraulic conductivity can range from years to decades (Ebel & Martin, 2017).The progression of landscape recovery for the Northern Bay Area of California is not currently known, and the goal of this study is to capture the functional form of post-fire soil-hydraulic recovery in this landscape (Figure 1) through high-frequency repeat surveys of field-saturated hydraulic conductivity (K fs ) following the October 2017 Nuns and Tubbs Fires that occurred in California's Northern Bay Area.With this study, our goals are to (a) determine the rate and functional form of post-fire K fs recovery; (b) understand what processes are driving changes to K fs in this landscape; and (c) quantify the flux of post-fire sediment from burned hillslopes to channels.Below we describe the processes that control post-fire K fs reduction and recovery, hillslope sediment flux and its impact on hazard, the landscape that burned in the 2017 fires, and our study design.
What processes act to reduce field-saturated hydraulic conductivity and increase runoff potential following wildfire?Typically, the burning of a soil ecosystem produces a number of phenomena that collectively serve to alter soil-hydraulic properties.The ignition of plants and organic matter can drive volatilized hydrophobic compounds downward into the soil (Debano, 2000, and references therein).The ash produced from burned organic matter itself can also migrate down into the soil, reducing soil porosity (Woods & Balfour, 2010).Ash deposits atop the mineral soil surface are often highly conductive, increasing infiltration storage and resulting in a layered system where relatively high K fs ash sits above a relatively hydrophobic mineral soil surface (Nyman et al., 2014).Heating of the soil itself can change its bulk density and moisture retention characteristics, often resulting in lower moisture retention for a given matric suction (Stoof et al., 2010).Combustion of plants also exposes bare soil at the ground surface, which leaves it vulnerable to rainsplash impacts that dissagregate soil particles and send the resulting microparticles into available pore space below.This creates a structural seal that has been found to exert a dominant control on K fs reduction compared to burning alone in controlled field experiments (Larsen et al., 2009).Other processes that do not impact the soil can also lead to increases in runoff potential, such as the burning of tree canopies that reduces the canopy interception of rainfall and therefore leads to a greater flux of rain impacting the soil surface (Reid et al., 1999).
Over time, a number of processes that operate over different magnitudes and timescales work to return the soil hydraulic conductivity toward pre-fire levels.For example, as soil moisture increases with time from its initially hyper-dry conditions (Ebel & Martin, 2017;Moody & Ebel, 2012), hydrophobic effects at the mineral surface diminish.As vegetative ground cover rebounds, fine roots may disrupt structural seals (e.g., Larsen et al., 2009) and increase macropore flow (Nyman et al., 2014).The net result of these processes acting on burned soil is an increase in the saturated hydraulic conductivity with time since fire, which tends to follow an asymptotic path toward its unburned conditions (Ebel, 2020;Ebel & Martin, 2017).Typical models for the functional form of K fs recovery predict that a minimum post-fire K fs occurs immediately following the burn, and a path towards recovery that may follow a logistic function (e.g., Ebel & Martin, 2017, Figure 1).If the primary reduction of K fs is due to rainsplash impacts and soil sealing (Larsen et al., 2009), that suggests a model where the minimum post-fire K fs occurs sometime after a sufficient amount of rainfall produces a low-permeability layer in the upper subsurface.In actuality both processes can affect burned soil, and while soil hydrophobic impacts decline as soil moisture increases during the winter following a burn (Huffman et al., 2001), soil sealing impacts should increase as rainsplash processes dominate.
Channel sedimentation from burned hillslopes can also have deleterious impacts on the landscape.While reductions in K fs can increase the likelihood of runoff-generated debris flows in post-fire landscapes, such as the recent 2018 Montecito debris flows (Kean et al., 2019), loading of headwater channels by post-fire dry ravel has also been shown to be a key factor in debris-flow occurrence (DiBiase & Lamb, 2020).The addition of sediment slugs to streams can also negatively impact aquatic habitat (e.g., Florsheim et al., 2017;Lisle, 1989;Lyon & O'Connor, 2008;Wood & Armitage, 1997) and general water quality (Murphy et al., 2012).
The Nuns and Tubbs fires were part of a larger complex of wildfires that impacted the northern Bay Area of California starting on 8 October 2017, and lasting several weeks (Figure 2).Ignited during a high-wind event, the fires caused 46 fatalities and took an immense toll on property and infrastructure, burning over 82,000 acres (Mass & Ovens, 2019).The northern Bay Area experiences a Mediterranean climate, with a mean annual precipitation of 787 mm in the city of Santa Rosa (PRISM Climate Group, Oregon State University, http://prism.oregonstate.edu, created 4 February 2004), and last experienced a comparable wildfire with the 1964 Hanly fire, whose burn 10.1029/2022JF006591 3 of 19 scar exhibited a similar footprint to the Tubbs fire (Keeley & Syphard, 2019).Below we describe our site selection and survey design to capture post-fire changes in field-saturated hydraulic conductivity across the Nuns and Tubbs burn scars.

Infiltration Monitoring Sites
We sought to capture the functional form of post-fire soil-hydraulic recovery through repeat surveys of soil infiltration rates at 41 monitoring sites distributed across four primary arrays.As described below, we sought to isolate the recovery signal from the inherent variability in hydraulic conductivity across the landscape by establishing repeat monitoring sites on specific patches of ground (e.g., Figures 3 and 4).In order to characterize the distribution of K fs at the basin scale, we include a high sample size of repeat monitoring sites throughout our four primary arrays.
These arrays are located within the Pepperwood Preserve (PP), north of Santa Rosa, Annadel State Park (ASP), and Sugarloaf Ridge State Park (SRSP).ASP and SRSP are both southeast of Santa Rosa near the town of Kenwood, CA (Figure 2).Sites at PP are primarily within Franciscan melange and volcaniclastics of the Sonoma Volcanics (McLaughlin et al., 2004).Sites mostly experienced a low to moderate-high soil burn severity (Figure 2), and site vegetation classified from both field observation and the lidar-derived Sonoma Veg Map (https://sonomaopenspace.egnyte.com/dl/1SWyCSirE9)consists of a mix of conifer forest, chapparal, oak woodlands, and grasslands (Perkins et al., 2022, Table S1, Figure 3).Our sites at Sugarloaf Ridge State Park (SRSP) consist primarily of Franciscan serpentinite and melange (Fox et al., 1973), with moderate to high soil burn severity.We set up two arrays within Annadel State Park: one along the lower severity eastern burn perimeter (ASPE), and one in the higher burn severity western interior of the park (ASPW).The ASPE sites are solely within the sedimentary Glen Ellen Formation (Fox et al., 1973), and experienced primarily low burn severity.The sites are within a range of chaparral, grassland, and oak and conifer forest.ASPW sites are solely within the Sonoma Volcanics and, like the SRSP sites, experienced moderate to high burn severity (Table 1).The ASPW array is more densely vegetated than the ASPE sites, with more coniferous and oak forest sites.During the summer of 2018, we also measured three unburned forested sites at ASPW along strike of the same units just north of the burn scar (Figure 2).

Infiltration Measurements
To measure a time series of field-saturated hydraulic conductivity across our site arrays, we used Decagon Mini Disk tension disc infiltrometers.Tension disc infiltrometers do not measure K fs directly, but are designed so that the time series of cumulative infiltration can be fit with analytical models to estimate K fs and soil sorptivity S(θ) (where θ is the volumetric moisture content of the soil).A common model applied to these data is that of Philip (1957), further discussed in Philip (1969) and modified by Zhang (1997), where cumulative infiltration I [L] due to three-dimensional spreading in a homogeneous medium can be expressed as where and A 1 and A 2 are dimensionless constants described further below.A more robust fitting procedure proposed by (Vandervaere et al., 2000) linearizes the cumulative infiltration  Ebel and Martin (2017).An initial post-fire reduction in field-saturated hydraulic conductivity gradually recovers through vegetation regrowth and bioturbation among other processes.The corresponding disturbance windows (red dashed lines) denote a period where K fs is insufficiently recovered (in this example, where K post−fire /K pre−fire < 0.9) and runoff-related hazard potential remains elevated.Depending on the recovery rate, the disturbance window may last through just one rainy season (blue sigmoid) in the "fast recovery" case or multiple rainy seasons in the "slow recovery" case.relationship, providing a straightforward way to determine field-saturated hydraulic conductivity and sorptivity from the infiltration time series: In the model of Zhang (1997), the parameters C 1 and C 2 represent the suction-driven and gravity-driven components of groundwater flow, respectively.Both parameters vary with disc radius r o , suction head h o , and soil texture expressed in terms of the van Genuchten parameters α and n that relate matric suction to soil moisture content (van Genuchten, 1980).A 1 (and hence S) also varies as a function of the initial moisture content, θ i , and the moisture content θ 0 determined by the suction head of the infiltrometer: 25  3(−1.9) 0(0) 0.15  (3) Because we did not measure initial soil moisture content for our infiltration surveys, we estimate θ i from interpolated NASA Surface Moisture Active Passive (SMAP) data further described in Section 2.6 (Entekhabi et al., 2016).Because we only took measurements on dry days during the rainy season, here we assume that θ i at the ground surface does not rise above the field capacity (≈0.25) during our measurement periods even though the SMAP data (which is depth-integrated) may show values closer to saturation.
During dry periods, we use the interpolated SMAP measurement directly.
(4)  Nearly all measurements were taken with h o = −1 cm for consistency, and grain size distributions were measured for each site using a Beckman-Coulter laser diffraction particle size analyzer (Gee & Or, 2002) located at the U.S. Geological Survey (USGS) Unsatured Zone Hydrology laboratory.We excavated roughly the upper few cm of soil within 2 feet of our measurement location (Figure 4), so as not to disturb the repeat measurement site, and soil texture was determined through the relative percentages of sand, silt, and clay particles (USDA-NRCS, 2017) (Table S1).Characteristic moisture curve parameters based on soil texture are determined using the model of Carsel and Parrish (1988) (Perkins et al., 2022).
There are both advantages and disadvantages to using tension disc infiltrometers for characterizing K fs .Given our large number of sites, many of which required off-trail hiking to access, the portability and low water usage compared to other single-or double-ring infiltrometers made our experimental design logistically tractable.Mini Disk infiltrometers have a small-diameter footprint of 2.5 cm, however, making them ill-suited for capturing larger-scale or heterogeneously distributed features of the soil.Furthermore, tension disc infiltrometers generally measure the matrix flow component of infiltration, and therefore may not capture significant macropore flow.As we describe later, however, because we ran our measurements with the infiltrometer at a low suction, seasonally varying macropore features with air-entry pressures below h o may impact S and K fs calculations.Occasionally, model fits to tension disc infiltrometer data produce negative values of K fs or S.This can be due to a variety of reasons including either measurement error or a layered system that violates the homogeneous assumption in the model (Vandervaere et al., 2000).Here we omit negative values from our analysis, but the raw data can be found in (Perkins et al., 2022).
Surveys were conducted approximately every 2 months for the first year and a half after the fires starting in March 2018.Due to the large number of sites, USGS scientists partnered with the Sonoma County Water Agency and  Pepperwood Preserve scientists to collect bi-monthly infiltration data.Each survey window was generally less than 1 week, but occasionally longer due to scheduling logistics between survey groups from different agencies.
Values for each survey window are lumped into populations at each survey array (e.g., Figure 5).

Hydrophobicity
While we did not measure hydrophobicity throughout our entire study, we took sporadic hydrophobicity measurements before and during our first round of infiltration measurements at the Pepperwood Preserve.During a site reconnaissance visit in December 2017, we measured hydrophobicity adjacent to our PMD-1 site (a grassland in Franciscan melange).In March 2018, we also measured hydrophobicity for a subset of additional sites.In both cases, we used the method of Doerr (1998), which uses an 8-point hydrophobicity scale based on the molarity of an ethanol droplet that penetrated the soil in less than 3 s.Results are described in Section 3.1 and shown graphically in Figure 5b. .Bold boxes at approximately 0.5 and 1.5 years highlight year-over-year differences at each site array (Figure 8).The blue box in (a) represents the full range of unburned values measured at ASPW during the summer of 2018 (n = 3).Vertical black bars in (b) show times of hydrophobicity measurements taken at PP, with the thickness indicating the relative strength of hydrophobicity.The background curves of each plot show the mean plus or minus one standard deviation of the MODIS-derived Enhanced Vegetation Index (EVI) (Didan, 2015) for each site array, which illustrates the impact of fire on vegetation and its subsequent recovery (EVI values on right axis).

Analysis of Infiltration Model Residuals
To examine whether any systematic trends in the pattern of cumulative infiltration can be used to glean information about the processes that either reduce or recover K fs , we also look at the infiltration residuals from the model fit described in Equation 1.For each survey measurement of infiltration rate, consisting of cumulative infiltration plotted against the square root of elapsed time, we fit the data with Equation 1, then normalize all the data relative to the total elapsed time and total amount of infiltration at each site.This gives us a sense of the "infiltration shape" for each measurement.We then look at the residuals for the normalized data across all our sites for four distinct time periods: February-May 2018, July-August 2018, April-May 2019, and May 2021.

Post-Fire Ravel Yield
To evaluate the magnitude of dry ravel yield following the October 2017 fires, we conducted a survey along fire roads just south of what is now Mark West Creek Regional Park and Open Space Preserve (Figure 1).The park has steeply sloping drainages that experienced moderate to high soil burn severity, resulting in an estimate of high post-fire debris flow likelihood (60% to over 90%) from the USGS Landslide Hazards Program Post-fire Debris-Flow assessment (Staley et al., 2017) (map available at https://landslides.usgs.gov/hazards/postfire_debrisflow/detail.php?objectid=163, accessed on 04/18/2022).Conveniently, the network of fire and legacy logging roads that crosscut the property created a series of sediment traps below hillslopes of varying gradients that provided an ideal location to measure the magnitude of slope-dependent dry ravel transport (e.g., Gabet, 2003;Lamb et al., 2011).Here ravel piles did not resemble cones but rather laterally continuous triangular wedges at the base of the road cut (Figure 4).Furthermore, we had no prior information on the geometry of the sediment piles before the fire, but trenching across them revealed a stark contact with the burned, charcoal-rich sediment (Figures 4b and 4c).We measured the thickness of the ravel layer at 10 cm intervals along its length, as well as the pile gradient and slope length.Assuming a right-triangular wedge geometry for the entire pile, we then calculate the cross-sectional area of the ravel layer.We did not conduct these surveys until July 2019, approximately 21 months after the fire, so estimates of ravel accumulations should be considered a maximum.At most sites, subsequent sediment drapes on the piles were not observed, with the exception of one site that showed a strong contrast in sediment color above the ravel layer that was subtracted from the total layer thickness.We estimate contributing slope gradient as the average hillslope angle directly above the roadcut.
We compare our ravel measurements to data from the 2009 Station Fire in the San Gabriel Mountains in Southern California from Lamb et al. (2011).In that study, ravel volumes were calculated as conical sections using surveyed pile heights and ravel cone widths, often on order of 5-10 m wide.To more directly compare these measurements to our 2017 Tubbs fire data, we use the surveyed pile heights and hillslope gradients underlying the piles, together with the authors' assumption of a 30-degree pile slope, to calculate ravel cross-sectional area at the apex of each pile.
For illustrative purposes, we compare both sets of ravel measurements to a generalized model for dry ravel flux by Gabet (2003).In this model, the downslope mass flux of ravel [ML −1 T −1 ] varies as a function of the topographic slope β, kinetic friction μ resulting from ravel transport mechanisms (e.g., grain rolling, bouncing, sliding, and collisions), and the constant κ [ML −1 T −1 ] that represents a range of additional processes such as event frequency, gravity, and mean sediment mass that has been displaced.Because neither study includes specific timing data for the measured ravel accumulations, we cast the model of Gabet (2003) as total ravel accumulation A d , measured as a volume per unit cross sectional area (m 3 m −2 ), and assume a common ravel density ρ r = 1,200 kg 1 m −1 and ravel transport time scale of approximately 1 month (τ).

= ( cos − sin )
. (5) We do not fit Equation 5 to either data set because of either the high measurement variability within the Station fire data or the lack of measurements at high slopes where transport becomes highly nonlinear (our study).Rather, these model scenarios provide a general physical context for the ravel survey data.

Remote Data Analysis
In this study we utilize remote datasets that characterize both the initial degree of post-fire soil burn severity, as well as the relative vegetative recovery.To quantify the initial post-fire soil burn severity for our study sites we use a data set of Burned Area Reflectance Classification (BARC) from the interagency Monitoring Trends in Burn Severity project (www.mtbs.gov)for use by the state of California Watershed Emergency Response Team (WERT) post-fire assessment of the Tubbs and Nuns fires.The U.S. Forest Service first calculates a difference normalized burn ratio map from pre-and post-fire Landsat imagery, then scales these values to a 256-value color map (BARC-256) (Eidenshink et al., 2007).The WERT then uses field observation to classify BARC-256 values to four soil burn severity (SBS) classes: very low/unburned, low, moderate, and high.SBS values for each site are shown in Table S1, and BARC-256 values are compiled and averaged for each study area array (Table 1).
Additionally, we track relative vegetative recovery following wildfire using the NASA MODIS-derived enhanced vegetation index (EVI) (Didan, 2015).EVI values are taken globally at 16-day intervals using the Aqua and Terra satellites, with a grid size of 250 m.We sample EVI values at our survey locations beginning in the Fall of 2016, a year before the Nuns and Tubbs fires occurred, through the end of our surveys in May 2021.Here we show the average (+/− the standard deviation) EVI values across all our sites to illustrate the general trend of vegetative recovery (e.g., Figure 7).
Lastly, we observe monthly changes in regional soil moisture at our study sites using NASA Surface Moisture Active Passive (SMAP) Level 3, 9 km resolution data (Entekhabi et al., 2016).SMAP data provide an estimate of regional soil moisture, and here we compile a time series of average monthly SMAP measurements for each of our study areas to examine broadly how changes in soil moisture relate to changes in soil-hydraulic properties, rainfall, and EVI (Figure 7).We also use the SMAP data to inform our estimates of initial soil moisture content θ i for the calculation of soil sorptivity S (Equations 1-3).We describe this in more detail in Section 2.2.

K fs Progression Following Wildfire
The time series of field-saturated hydraulic conductivity across our four measurement arrays show a complex yet robust response as the landscape recovers from the October 2017 Nuns and Tubbs wildfires in Sonoma and Napa counties (Figures 5 and 6) (Perkins et al., 2022).For each site array, the minimum values occur at the first surveys following fire during the Spring of 2018, at the tail end of the initial rainy season.These values are typically quite low, on the order of 1-5 mm/hr.Ethanol tests, which use an 8-point hydrophobicity scale based on the molarity of an ethanol droplet that penetrated the soil in less than 3 s (Doerr, 1998), revealed minimal hydrophobicity at the subset of Pepperwood Preserve sites tested.In December 2017, however, initial ethanol tests taken during a single site reconnaissance visit illustrated Class 6 (very strong) hydrophobicity (Figure 5b).This suggests that chemical hydrophobicity rapidly declined with increased soil moisture around 6 months after the fire, and therefore other processes are likely responsible for the observed low K fs values.
In Summer 2018 heading into Fall, the measured K fs values show a marked, order of magnitude increase to the 20-30 mm/hr range across all sites.As the second rainy season resumes in Fall 2018 (Figure 7), K fs values immediately decrease from their summertime highs but remain elevated from their initial post-fire lows.In late Spring 2019, values once again rise.At this stage the primary deviation from this trend comes from the ASPE array, where K fs continues to dip in Spring 2019 (Figure 5c).From 2019 to 2021, however, most sites increase to their highest median year-over-year values across the surveys.The exception to this is SRSP, which burned a second time in the 2020 Glass Fire (https://www.fire.ca.gov/incidents/2020/9/27/glass-fire/, accessed on 04/18/22) and therefore likely reflects the impact of the subsequent burn on K fs .Unfortunately, given the rapid seasonal shifting of observed K fs from spring to summer, it is difficult to assess the true impact subsequent burning had on our SRSP sites.
Taken together, the K fs recovery time series reveal a characteristic recovery path marked by an initial low value, a rapid increase in K fs values in the first summer following the fire, a marked decrease during the subsequent rainy season, and another increase the following spring 2019.With the exception of SRSP, all sites show an increase in median K fs from the 1.5 to 3.5 years surveys (bold boxes in Figure 5), but unfortunately the lack of data between these two surveys makes it unclear how the seasonal recovery trend observed from 0.5 to 1.5 years continues during the latter half of soil-hydraulic recovery.While the overall trend of K fs recovery may follow a logistic-like function (Figures 1 and 7), the seasonal variations driven by distinct processes superimpose a strong second-order variability.
In Figure 7, we show lumped field-saturated hydraulic conductivity and sorptivity for all sites, alongside the records of EVI, local hourly rainfall, and surface soil moisture derived from SMAP.Like K fs , sorptivity values exhibit large seasonal variability.Sorptivity does not show an overall recovery signal, however, and the values of S over time are anti-correlated with seasonal soil moisture as expected.A wetted soil will have a lower capillary gradient, and this will result in faster transition to gravitationally driven infiltration.
Seasonal maxima in K fs occur in the dry season when surface soil moisture is at a minimum, as do peaks in the measured sorptivity.Field observations of small drying cracks at our measurement sites during July 2018 surveys suggest that the opening of macropores in bare soil may lead to enhanced bulk hydraulic conductivity measured by the Mini Disk.As we conducted our surveys at a low suction (1 cm), it is likely that the water entry pressure for the macropores was exceeded and hence our summertime K fs values reflect the contribution from secondary flow.As soon as the first rains appear in early October 2018, both K fs and sorptivity values decline.This suggests that the drying cracks may have begun to close as the soil surface wets up.With continued rainfall through Winter 2019, SMAP soil moisture reaches its seasonal peak and K fs and S values reach their local minima.As the surface soil begins to dry out in Spring 2019, sorptivity values across all sites begin to rise again, but K fs values maintain their levels.This suggests that observed increases in K fs are not just apparent increases caused by changes in moisture-dependent sorptivity.

K fs Reduction, Soil Burn Severity, and the Hazard Window
We did not measure K fs values at any of our infiltration sites before the Nuns and Tubbs fires, and we therefore cannot directly measure the impact of the fire on prior infiltration rates.Instead, we assume that the infiltration rates have returned to typical conditions after 3.5 years and approximate "pre-fire" conditions using our most recent measurements.This holds true for the ASPW array where we did measure unburned infiltration rates in the summer of 2018 (Figure 5).In Figure 8 we show the year-over-year relative reduction in K fs , equal to the May 2019 values divided by the May 2018 values, as a function of the mean BARC-256 for each array.We find that areas that experienced higher soil burn severity (ASPW, which is primarily forested, and SRSP) show the greatest relative change over the course of the infiltration rate survey.The year-over-year change is comparable to the 3-5x total reduction seen in other studies (Ebel & Martin, 2017, and references therein).ASPE, which is along the burn perimeter of the Nuns fire, does not show much relative change.PP, which has the largest mix of soils and plant communities, shows a modest median change but a large distribution.The large seasonal changes in K fs despite low year-over-year differences suggest that accounting for seasonality may be an important consideration in future studies.
How do our observations of K fs over time relate to the concept of a hazard window?In Figure 7 we show the median K fs values in relation to hourly rainfall data from a nearby rain gauge at Mt.St. Helena (Wuertz et al., 2018).When our surveys began at the end of the first rainy season in 2017/2018, the low recorded K fs values of 2-3 mm/hr are well-exceeded by the hourly rainfall amounts.High-intensity rainfall bursts at the 15-min scale are what typically drive post-fire debris flow initiation (Kean et al., 2011), however, and the northern Bay Area experienced a substantially milder rainy season than the previous winter.At the Mark West Creek Regional Park and Open Space Preserve where we conducted ravel measurements, the USGS post-fire hazard assessment suggests that 15-min rainfall intensities exceeding 24 mm/hr should lead to >60-80% chance of failure in the first year following fire (https://landslides.usgs.gov/hazards/postfire_debrisflow/detail.php?objectid=163).These rainfall intensities were not exceeded at the nearby rain gauge (Figure S1 in Supporting Information S1); however, a notable storm on 22 March 2018 that produced abundant landslides in the Sierran foothills (Collins et al., 2020) did produce 15-min intensities close to this threshold.Even so, it is likely that K fs at the Mini Disk (≈3 cm) scale does not represent K fs at the plot scale where larger macropores and more heterogeneity are present.
If we gauge recovery in relative terms, we can use the example from the ASPW array where unburned measurements of field-saturated hydraulic conductivity were collected in late July and early August 2018 (Figure 5a; Perkins et al., 2022).At the end of the first rainy season, median K fs for ASPW was about 10% of the average unburned K fs .At the end of the second rainy season, K fs was approximately 45% of unburned conditions.By May 2021, at the end of the fourth rainy season, the sites have largely recovered (Figure 5).This suggests that the majority of soil-hydraulic recovery has been accomplished sometime between two to 3 years following the fire.Furthermore, the rate of recovery is fastest within the first year, suggesting that the hazard window likely primarily falls within the rainy season following fire within our study area before significant macropore growth at the Mini Disk scale.This is in line with recent estimates of post-fire debris flow likelihood (Hoch et al., 2021).Implicit in this analysis is the assumption that the unburned field-saturated hydraulic conductivity does not vary seasonally at the same magnitude as the burned sites, but we do not have sufficient data to address whether or not this is the case.Future studies could benefit from a time series of unburned K fs alongside burned sites.

EVI as a Proxy for Soil-Hydraulic Recovery
The EVI data show a significant reduction in vegetative cover following the 2017 fires (Figure 7b), and a gradual recovery to background values that continues through 2021 (e.g., Figure 6).Like the K fs measurements, EVI values show strong seasonal peaks; however, peaks occur 2-3 months prior to K fs maxima and reflect the period of spring regrowth.Even into Spring 2021, the EVI values continue to be reduced substantially from the pre-fire values.
We interpret that the persistently reduced EVI values reflect the magnitude of tree canopy damage particularly to oaks that experience topkills where all above-ground biomass dies and trees recover through basal regrowth (Holmes et al., 2006).
Spectral vegetation indices are often used to gauge the landscape recovery to wildfire (e.g., Ebel, 2020;White et al., 1996), but exactly how changes in vegetation indices like EVI relate to changes in K fs is unclear and remains to be sufficiently quantified.Our analysis shows a positive but complicated relationship between the two data sets, both of which exhibit strong seasonal (yet phase-shifted) signatures (Figure 7).While our K fs analysis suggests that soils have largely recovered, the more heavily burned sites still show a substantial EVI reduction compared to background values.Because many of our sites are forested, this offset likely results from the substantial canopy damage and tree loss sustained from the 2017 fires (Green et al., 2020).As novel work begins to link remotely sensed vegetation recovery with soil-hydraulic recovery (e.g., Thomas et al., 2021), caution should be taken in Mediterranean climates where the lingering canopies of topkilled trees may obscure the basal regrowth from satellite line-of-sight.

Infiltration Model Residuals: Homogenization by Fire, Soil Sealing, and Subsequent Disruption
The time series of infiltration model residuals across our burned survey sites, shown in Figure 9, highlights a number of interesting patterns that reveal the impact of wildfire on soil hydrology and the processes that govern its recovery.First, we observe that both the magnitude and spread of normalized residuals are their lowest during our initial surveys from February to May 2018 (Figure 9a).This suggests that the Nuns and Tubbs fires substantially homogenized the soil-hydraulic characteristics across our 41 sites.This is consistent with the findings of others (e.g., Ebel, 2012) who observe that the combustion of soil organic matter within soils can reduce differences in soil water retention parameters.Over time, we see that the spread of residuals increases as vegetation recolonizes our sites (Figure 9).This observation is consistent with the interpretation that variable ground cover BARC-256 values are Landsat-derived difference Normalized Burn Ratio (dNBR) data scaled to a 256-value axis for use in defining thresholds for soil burn severity classes (Eidenshink et al., 2007).Y-axis value corresponds to the median and vertical error bars represent the 25th and 75th percentiles to be consistent with Figure 5. BARC values are shown as the mean plus or minus the standard error for each study area (n = 10 for AMDW; n = 8 for SMD; n = 8 for AMDE; and n = 15 for PMD).While ASPE, which experienced a low soil burn severity, did not show much of a K fs reduction, the heavily burned arrays at ASPW and SRSP showed a substantial threefold to sixfold reduction in K fs typical of other heavily burned landscapes.
and organic matter reintroduces heterogeneity to the upper soil surface.Therefore, local vegetation must exert a strong control on the hydrologic character of the uppermost soil whilst recovering from wildfire in our study area (e.g., Cerdà & Doerr, 2005).
In addition to the increased dispersion of residuals over time, we see significant changes in the structure of normalized residuals as the soils recover (Figure 9).In February-May 2018, increased negative residuals occur at the beginning of the infiltration measurements, and level toward a mean of zero during the later phase of cumulative infiltration.This implies that during the infiltration measurement, water is infiltrating into the soil slower than would be predicted for a homogeneous soil in the Philip (1957) model.By July-August 2018, the pattern switches, and positive residuals are seen at the beginning of the infiltration measurements, while negative residuals are focused during the latter half of the measurements (peaking at about   1∕2 norm = 0.75 ).Once this pattern sets in, it becomes amplified with time.Positive residuals get larger at the beginning of infiltration measurements, and slower toward the end of the measurements.
What processes can be gleaned from the changing pattern of infiltration model residuals?The relative slowdown seen during incipient infiltration for the first rounds of post-fire measurements (Figure 9b) suggests that the uppermost mineral soil surface might be acting as a low permeability layer.This may be due to either chemical hydrophobicity or structural soil sealing from rainsplash impacts on bare soil that drive disaggregated soil particles and ash into the uppermost soil surface.While we did not conduct systematic hydrophobicity tests during these measurements, ethanol tests for a subset of sites at the Pepperwood Preserve showed little evidence for chemical hydrophobicity.The relative slowdown in the uppermost soil surface is therefore likely due to porosity reductions from ash clogging and rainsplash impacts (e.g., Larsen et al., 2009).Recent work by Ebel and Moody (2020) shows that reductions in K fs for the 2017 Thomas Fire are indeed accompanied by reductions in the surface soil porosity.The rapid progression to high K fs and S seen in the summer of 2018 is accompanied by For each panel the data are binned into 15 equally spaced subsets along the x-axis, and white circles denote the mean of each bin.Gray circles show the raw data points used in the binning procedure.positive residuals in the infiltration measurements.We interpret this to reflect small macropore opening due to cracking of the dry and relatively bare soil, which was observed at a number of sites during this time.The negative residuals present during the latter half of the cumulative infiltration measurements (Figure 9c) could possibly reflect the downward propagation of fines from the low-permeability upper layer or the lower limits of surface cracking.More in-depth examination of the upper soil porosity structure could help determine the processes that control the changes in cumulative infiltration pattern throughout the post-fire recovery phase.

The Dynamic, Multi-Stage Post-Fire Recovery of K fs
Our infiltration survey data reveal a seasonally dynamic recovery of field-saturated hydraulic conductivity at all study sites.These seasonal differences are on the order of, or perhaps larger than, the scale of post-fire recovery (Figures 5 and 7).Observations of surface cracks during the late summer surveys of 2018 suggest that secondary flow through macropores whose air entry pressures are less than the 1 cm of head used for our tension disc infiltrometer measurements may be responsible for the dramatic increase of K fs from the initially low Spring values (e.g., Nyman et al., 2014).Assuming a contact angle of 45°, macropores up to 1 mm in diameter could be contributing to the measured infiltration rates.The largest impacts of fire on soil typically occur within the upper 1 cm (Ebel & Moody, 2020), and our data point to a complicated interrelationship between the seasonal growth and reduction of macropore flow and the gradual increase of matrix flow.To what extent does the former influence the latter?It is possible that summertime cracking of bare soil might be the primary mechanism responsible for the observed rapid increase in K fs .Given the clayey nature of the soils at these sites, whose parent lithologies range from Franciscan melange to mafic volcanics (Table S1), the sites would be particularly prone to surface crack opening and closing during the post-fire vegetative recovery when soils are comparatively bare.Does drying and cracking control the post-fire recovery of soils more than vegetation regrowth?More work is necessary to tease out the relative efficacy of soil drying and cracking versus the reestablishment of fine roots on impacting the porosity structure of the upper soil.It is also possible that the clay fraction leads to substantial swelling and shrinking in the upper soil during wet and dry periods (e.g., Fernandes et al., 2015), which also lead to changes in porosity and hence K fs .Regardless of the mechanism, the large difference in dry versus wet season K fs during the first year of recovery suggests that the dynamic infiltration capacity may influence seasonal recharge, particularly during the transition windows in the Fall and late-Spring.
From synthesis of K fs , S, hydrophobicity, soil moisture, and rainfall data, we hypothesize a conceptual multi-stage recovery model for the soils burned by the 2017 Nuns and Tubbs fires (Figure 10).During Stage 1 recovery, which occurs from the cessation of wildfire to the onset of the first rainy season, soils are homogenized through the combustion of organic matter.Here chemical hydrophobicity is likely the dominant K fs -reducing process.While we do not have measurements of K fs during this time, a preliminary ethanol test at the Pepperwood Preserve in December 2017 yielded class 6 hydrophobicity (strongly hydrophobic) (Doerr, 1998).
During Stage 2 recovery, which we suggest occurs during the first post-fire rainy season, we observe some incipient plant growth on the soils, and the lowest K fs measurements across our survey sites (Figures 5 and 7).Our infiltration model-residual analysis reveals relatively slower infiltration at the uppermost soil surface (Figure 9a).The few ethanol tests conducted at the Pepperwood Preserve during this time show limited chemical hydrophobocity, so we therefore hypothesize that porosity reductions are due to ash infiltration and rainsplash-driven soil sealing that act to reduce K fs .
Stage 3 recovery involves the drying and cracking of relatively bare soil during the summer following the first rainy season.Here the regional soil moisture is at a minimum (Figure 7), infiltration model residuals transition from relative slowdown at the soil surface to relative speed-up at the soil surface (Figures 9b and 9c), and both K fs and S increase substantially over this interval (Figures 5 and 7).Stage 4 recovery yields a relative reduction in field-saturated hydraulic conductivity as drying cracks and associated macropores likely are reduced with the onset of seasonal rainfall (Figure 10).The overall K fs trend is still increasing during Stage 4, even as sorptivity values are reduced in the now wetter soil (Figure 7).At this stage we hypothesize that there are contributions from both the partially opened macropores as well as increased matrix flow from root infiltration by increasing groundcover (e.g., Figure 6 bottom-left panel).
While we lack the measurements to show this seasonal recovery cycle over multiple years, we envision that Stages 3-4 repeat in a potentially muted fashion during the wet and dry periods as K fs approaches its initial pre-fire values.This evolution is consistent with the dynamic interplay between matrix and macropore flow seen following wildfire in southeast Australia (Nyman et al., 2014).Advanced imaging and analysis techniques using CT-scans (e.g., Ferguson et al., 2018) or SEM imagery (e.g., Tang et al., 2020) may be able to capture the processes that change porosity structure with time.
Returning again to the concept of the hazard window for flash floods and post-fire debris flows in the Northern Bay Area of California, how do soil hydraulic recovery stages fit in?It is likely that once the soil reaches Stage 3 recovery (marked by the reversal of early time negative infiltration residuals that signify relative slowdown at the soil surface) the hazard impact by K fs reductions from wildfire is greatly diminished.In the sites impacted by the 2017 Nuns and Tubbs wildfires, therefore, the primary hazard window from runoff-generated debris flows most likely encompasses the first post-fire rainy season and then diminishes significantly before the second.Recent work by Thomas et al. (2021), however, highlights the trade-off in relative hazard between debris flows that initiate from runoff, rilling, and bulking, and debris flows that initiate from shallow landslides driven by pore water pressure increases at depth as K fs returns to pre-fire conditions.In our study area, therefore, the relative hazard increase from infiltration-driven debris flows may persist as K fs increases but vegetation, whose roots provide cohesive strength to buffer shallow landslides, is slow to recover (e.g., Figure 7).

Post-Fire Ravel Yield
Our measured ravel yield data show a clear slope-dependent relationship (Figure 11).Here, we show the data alongside predictions of dry ravel flux for a range of parameters by Gabet (2003).Because our steepest contributing hillslopes do not exceed the typical friction angle where flux begins to dramatically increase with slope (e.g., Gabet, 2003;Lamb et al., 2011), we cannot characterize this most important aspect of the sediment flux relationship.
Comparison to ravel yield data for the 2009 station fire (Lamb et al., 2011) shows an order-of-magnitude reduction for a given slope (centimeters vs. meters).There are multiple reasons why the ravel yield might be so different between these two post-fire landscapes.First, the vegetation characteristics of each site are starkly different, with the Mark West hillslopes being dominated by Douglas fir forests (Figure 4) and the San Gabriel Mountains hillslopes largely supporting chaparral.If the dry ravel process is driven by the release of sediment after the burning of vegetation dams (e.g., Florsheim et al., 1991;Lamb et al., 2011), then forested landscapes may preferentially hamper ravel transport because the large-diameter conifers may act as the primary sediment traps and they often do not fully incinerate in a wildfire despite high burn severity (e.g., Figure 4).Large root burnouts from fully incinerated trees may also act as a sizable sediment sink.Forest density may also impact topographic surface roughness, which can strongly control the sediment transport rates of post-fire dry ravel (DiBiase et al., 2017;Roth et al., 2020).Second, observations of post-fire dry ravel transport following the Station fire show that catchments with more granitic composition in the San Gabriel Mountains produced a much higher post-fire erosion rate than the more mafic catchments (DiBiase & Lamb, 2020).The Mark West Creek study area Our study area impacted by the 2017 Tubbs fire (blue circles) has mobile regolith that overlies fine-grained volcanic rock, and experienced moderate to high soil burn severity within a Douglas Fir forest where many of the trees did not fully incinerate (e.g., Figure 4b).Measurements following the 2009 Station fire were in the granitic and dioritic San Gabriel mountains of Southern California (Lamb et al., 2011) (red squares) were in a primarily chaparral vegetation community.Lines represent ravel accumulation scenarios using the model of Gabet (2003) over a period of 1 month with ρ r = 1200 kg 1 m −3 , and κ = 50 kg 1 m −1 yr −1 (solid), κ = 500 kg 1 m −1 yr −1 (dashed), and κ = 5,000 kg 1 m −1 yr −1 (dotted dashed).

Low Ravel Yield and the Lack of Observed Post-Fire Debris Flows in Northern California
Despite recent widespread wildfires in the Northern Bay Area of California over the past decade, there have been very few reports of post-fire debris flows in the region compared to Southern California.This is despite well-vetted model predictions for the substantial likelihood of their occurrence (Staley et al., 2017), and substantially reduced post-fire infiltration rates (e.g., Figures 5-7).This may be due solely to the dearth of high-intensity rainfall necessary to trigger post-fire debris flows in the region (e.g., Figure S1 in Supporting Information S1).
Considering how dry ravel sedimentation may act to substantially reduce critical discharge for in-channel failure, however (Palucis et al., 2021), our observations of limited post-fire dry ravel transport suggest that burned Northern California watersheds may be comparatively less prone to post-fire debris flow hazards even while other sedimentary hazards such as increased reservoir sedimentation may still persist (East et al., 2021).
Although there are multiple processes that can lead to post-fire debris flows (Kean et al., 2013), DiBiase and Lamb (2020) show that for the case of the 2009 Station Fire in Southern California, dry ravel loading of headwater channels played a substantial role determining the location of debris flows.The observed limited ravel flux, possibly due to parent lithology and denser vegetation, may therefore yield limited sediment supply necessary for within-channel failures.Characterizing post-fire topography with lidar (Rengers et al., 2016) or drone-based structure-from-motion imagery acquisition before the first rains (Palucis et al., 2021) could be a primary emergency response goal to help shed light on the relevance of these processes for Northern California.
Detailed characterization of the deadly 2018 Montecito debris flow event resulting from the 2017 Thomas Fire along Southern California's Santa Ynez mountains, however, reveals that these debris flows initiated primarily from surface runoff and rilling in the watershed (Kean et al., 2019).The Montecito event originated from a narrow cold-frontal rainband (NCFR) (Oakley et al., 2018), a mesoscale atmospheric phenomenon that produces extremely high rainfall intensities that can lead to flooding, hillslope erosion, and shallow landslides (Collins et al., 2020), and it is possible that an event with sufficient rainfall intensity to trigger debris flows has just not occurred in our study area (e.g., Figure S1 in Supporting Information S1).
Another potential explanation is that the soil K fs itself has less to do with runoff generation than the soil cover across the watershed.Arid mountain ranges such as those surrounding Southern California often have thin soils with significant bedrock outcrops, and given that bare bedrock should have a substantially reduced water storage relative to mineral soil, a higher percentage of bedrock exposure should shunt more water onto the soil and therefore lead to enhanced runoff potential (Coe et al., 2008).Northern California Bay Area soils are often typified by soil-mantled hillslopes with comparatively limited bedrock exposure, and therefore may be less prone to any enhanced runoff by exposed bedrock.Further analysis could reveal the extent of this process in governing the regionally varying response of post-fire debris flows.

Conclusions
Wildfire and its associated impacts have been consistently shown to reduce the infiltration capacity of soils, and although studies have typically focused on soil-hydraulic recovery in more arid landscapes of the American southwest, it is unclear how wetter climate modulates post-fire landscape recovery.This is particularly important as our changing climate is exposing new landscapes to the threat of megafires.This study presents the first documentation of the pattern of post-fire soil-hydraulic recovery in the Northern Bay Area of California.Through conducting a broad field survey of infiltration across dozens of repeat monitoring sites burned by the 2017 Nuns and Tubbs fires, spanning a range of burn severity, lithology, and vegetative community, we found that heavily burned areas experienced a 3-to 6-fold initial reduction in field-saturated hydraulic conductivity.
Our analysis reveals a complex, rapid, multi-stage recovery of K fs following the fire involving interaction between changing matrix and macropore flow over time.Given the fine-grained nature of soils in our study area, the susceptibility of soil type (and thus parent lithology) to cracking and macropore development might exert a control on the functional form of post-fire soil hydraulic recovery.Analysis of our infiltration time series data leads us to suggest that the primary hazard window following wildfires in the Northern Bay Area of California is likely limited to the first rainy season.Infiltration model residuals show that initial post-fire measurements are slowest at the uppermost soil surface, most likely due to rainsplash-induced soil sealing given the lack of observed hydrophobicity during the Spring of 2018.Large increases in K fs a few months later during Summer 2018 coincide with a relative increase in infiltration rate at the ground surface and change in the pattern of infiltration model residuals, suggesting that cracking and macropore growth of relatively bare soil may disrupt any structural seal.While K fs values decline again the following rainy season, they do not approach the very low values observed a year earlier.The pattern of infiltration rate model residuals remains largely unchanged for the remainder of our surveys, which suggests that the initial phase of K fs reduction is accompanied by a unique physical process that uniformly changes before the onset of the second rainy season.
Although our surveys of K fs show a rapid recovery, placing it at the high end of known post-fire recovery rates, the vegetation recovery as measured by satellite indices is muted (Figure 7).This suggests that linking spectral vegetation indices like EVI with soil-hydraulic properties requires a regional understanding of the contribution of EVI reduction from topkilled trees whose burned canopy may hide basal regrowth occurring at the tree base and ground surface.
We also documented post-fire dry ravel yield across a heavily burned watershed using a series of fire road cuts that acted as sediment traps for hillslopes spanning a range of topographic gradients.Although limited in scope, our results show a consistent slope-dependent ravel yield relationship.Comparison across similar slopes to ravel yields following the 2009 Station Fire in the San Gabriel Mountains in Southern California reveals an order of magnitude less ravel yield in our study area.This is likely to do both with the relatively denser and more robust forested vegetation at our study site that traps sediment and increases topographic roughness which can limit nonlocal grain transport, as well as the lithological influence on soil production rates that promote dry ravel in granitic terrain.It is therefore possible that the comparatively low ravel loading to stream networks following the 2017 Northern Bay Area fires helped buffer the landscape from potential post-fire debris flows triggered by in-channel sediment failure, although a more mechanistic regional analysis would help shed light on this issue.Lastly, the relatively mild rainy season immediately following the October 2017 fires failed to produce threshold 15-min rainfall intensities for debris-flow generation predicted by the empirical USGS post-fire debris flow model at our study area.It is therefore likely that both climatic and sedimentary factors played a role in mitigating debris flow likelihood during the post-fire hazard window.
Overall, our study points to regional differences in the rates and processes of post-fire soil-hydraulic recovery and sedimentation, emphasizing the need for a greater understanding of these relationships in landscapes that are now becoming increasingly vulnerable to extreme wildfire and post-fire hydrologic hazards.

Figure 1 .
Figure1.Conceptual post-fire K fs recovery curves.Gray dashed lines show varying soil-hydraulic recovery rates, shown as the ratio of post-fire to pre-fire K fs , based on the model ofEbel and Martin (2017).An initial post-fire reduction in field-saturated hydraulic conductivity gradually recovers through vegetation regrowth and bioturbation among other processes.The corresponding disturbance windows (red dashed lines) denote a period where K fs is insufficiently recovered (in this example, where K post−fire /K pre−fire < 0.9) and runoff-related hazard potential remains elevated.Depending on the recovery rate, the disturbance window may last through just one rainy season (blue sigmoid) in the "fast recovery" case or multiple rainy seasons in the "slow recovery" case.

Figure 3 .
Figure 3. Examples of infiltration monitoring sites throughout our study arrays.(a) Example of a grassland site in Franciscan melange at Pepperwood Preserve.(b) An Oak woodland site at Pepperwood Preserve.(c) Burned chaparral at Sugarloaf Ridge State Park.(d) Burned Bay Laurel and Douglas Fir forest at Annadel State Park.Photos by Jonathan Perkins.

Figure 4 .
Figure 4. Measurement site photos.(a) Photo taken 15 March 2018, showing infiltration site setup.In an effort to control for local soil heterogeneity, roofing nails were placed around the survey site to ensure the same patch of ground was remeasured over time.Infiltration rates were measured using a tension disc infiltrometer (3.1 cm wide at base for scale), and grain size samples were collected nearby at each site.(b) Annotated photo, taken 21 August 2019, showing an example of a surveyed ravel pile following the Tubbs fire at the heavily burned Mark West Creek Regional Park and Open Space Preserve (Figure 1).Sediment piles were about 1 m in length, and dark, charcoal-rich layers were easily identified marking the post-fire ravel contribution.Soil knife is approximately 32 cm long for scale.(c) Zoom in on (b) showing darker post-fire dry ravel layer in excavated pile.Photos by Jonathan Perkins.

Figure 5 .
Figure 5.Time series of field-saturated hydraulic conductivity K fs following the October 2017 Nuns and Tubbs fires for each of our four infiltration site arrays: Annadel State Park West (ASPW) (a), Pepperwood Preserve (b), Sugarloaf Ridge State Park (c), and Annadel State Park East (d).Years since fire are shown on the bottom axis, and equivalent calendar dates are shown in the middle of the figure.Boxes represent the median and inter-quartile range of K fs , with blue crosses showing the mean, black x's showing outliers, and numbers indicating the sample size for each box.Smaller sample numbers reflect either measurements with poor model fits that yield negative K fs or S values, or our inability to measure all sites within a single time window (as is the case in (b) for the measurements in Spring 2019).Bold boxes at approximately 0.5 and 1.5 years highlight year-over-year differences at each site array (Figure8).The blue box in (a) represents the full range of unburned values measured at ASPW during the summer of 2018 (n = 3).Vertical black bars in (b) show times of hydrophobicity measurements taken at PP, with the thickness indicating the relative strength of hydrophobicity.The background curves of each plot show the mean plus or minus one standard deviation of the MODIS-derived Enhanced Vegetation Index (EVI)(Didan, 2015) for each site array, which illustrates the impact of fire on vegetation and its subsequent recovery (EVI values on right axis).

Figure 7 .
Figure 7. (a) Time series of K fs combined for all sites are shown as the gray curve with values recorded on the right axis (dark gray box is the median, and gray area shows the inter-quartile range), and hourly rainfall data from the NOAA Mt.St. Helena rain gauge (code MSHC1, lat/lon = 38.67/122.63)(Wuertz et al., 2018) are shown as blue bars whose values are recorded on the left axis.X-axis is portrayed in units of time since the October 2017 wildfires, and approximate calendar dates (MM/YYYY) are shown between panels (a and b).Panel (a) highlights that throughout the second rainy season, K fs values typically exceeded hourly rainfall intensities and increased approximately threefold from values measured a year prior at the beginning of our surveys.Recovery stages correspond to the conceptual model in Figure 10.Panel (b) shows time series of measured sorptivity S(θ) (gray curve, with dark gray boxes representing the median and vertical range representing the inter-quartile range) whose values are shown on the right axis, and time series of Enhanced Vegetation Index (EVI) (Didan, 2015) (green curve, with vertical range equal to the combined range for all survey sites) and Surface Moisture Active-Passive satellite (SMAP)-derived volumetric water content θ (Entekhabi et al., 2016) (blue curve, with vertical range equal to the combined range for all survey sites) whose values are shown on the left axis.

Figure 8 .
Figure8.Plot showing the year-over-year reduction of K fs from our initial measurements in Spring 2017 for ASPW (red), ASPE (light blue), SRSP (green) and PP (purple) as a function of scaled burn severity (BARC-256).BARC-256 values are Landsat-derived difference Normalized Burn Ratio (dNBR) data scaled to a 256-value axis for use in defining thresholds for soil burn severity classes(Eidenshink et al., 2007).Y-axis value corresponds to the median and vertical error bars represent the 25th and 75th percentiles to be consistent with Figure5.BARC values are shown as the mean plus or minus the standard error for each study area (n = 10 for AMDW; n = 8 for SMD; n = 8 for AMDE; and n = 15 for PMD).While ASPE, which experienced a low soil burn severity, did not show much of a K fs reduction, the heavily burned arrays at ASPW and SRSP showed a substantial threefold to sixfold reduction in K fs typical of other heavily burned landscapes.

Figure 9 .
Figure 9. Normalized infiltration model residuals as a function of normalized square root of time (e.g., Equation 1) for each of our study sites at select time intervals: Feb-May 2018 (a), July-August 2018 (b), April-May 2019 (c), and May 2021 (d).For each panel the data are binned into 15 equally spaced subsets along the x-axis, and white circles denote the mean of each bin.Gray circles show the raw data points used in the binning procedure.

Figure 11 .
Figure11.Post-fire ravel accumulation plotted as a function of contributing topographic slope for two very different landscapes.Our study area impacted by the 2017 Tubbs fire (blue circles) has mobile regolith that overlies fine-grained volcanic rock, and experienced moderate to high soil burn severity within a Douglas Fir forest where many of the trees did not fully incinerate (e.g., Figure4b).Measurements following the 2009 Station fire were in the granitic and dioritic San Gabriel mountains of Southern California(Lamb et al., 2011) (red squares) were in a primarily chaparral vegetation community.Lines represent ravel accumulation scenarios using the model ofGabet (2003) over a period of 1 month with ρ r = 1200 kg 1 m −3 , and κ = 50 kg 1 m −1 yr −1 (solid), κ = 500 kg 1 m −1 yr −1 (dashed), and κ = 5,000 kg 1 m −1 yr −1 (dotted dashed).
lithology is primarily andesitic and basaltic flows that comprise the basal units of the 5-3 Ma Sonoma Volcanics(McLaughlin et al., 2004), and it is therefore possible that mobile regolith production rates here are substantially lower.Interestingly, in a recent study examining sedimentation following the 2018 Carr fire surrounding Northern California's Whiskeytown Reservoir,East et al. (2021) found that steep watersheds in similarly granodioritic catchments did not produce similarly large ravel cones to those following the Station Fire.All else being equal, in the nonlocal ravel flux framework described byRoth et al. (2020) it is possible that other factors such as increased topographic relief may lead to enhanced ravel accumulation in channels for the same slope if heavy-tailed particle travel distances are limited only by hillslope length.In other words, if there is runaway transport of post-fire ravel, longer hillslopes may provide a greater effective source area and thus contribute more sediment to channels.Further examination of post-fire sediment loading from hillslopes to streams controlling for vegetation, underlying lithology, and topography should help clarify to what extent these processes determine the variability in post-fire ravel yield more generally.

Kenwood Pepperwood Preserve Pepperwood Preserve Mark West Mark West Sugarloaf Ridge SP Sugarloaf Ridge SP Annadel SP Annadel SP
Figure 2. Map of the study area showing the extent of the Tubbs burn scar (north) and Nuns burn scar (south).Soil burn severity ranges from low (yellow) to moderate (red) to high (dark red).Infiltration survey sites are shown as triangles, and colors correspond to each site array within or near park boundaries (black polygons)-Pepperwood Preserve (PP; purple triangles), Sugarloaf Ridge State Park (SRSP; green triangles), Annadel State Park West array (ASPW; red triangles) with unburned comparison sites (white triangles), and Annadel State Park East burn perimeter array (ASPE; blue triangles).Ravel yield measurements were taken at the southern boundary of Mark West Creek Regional Park and Open Space Preserve (white circles).

Table 1
Summary Characteristics of Site Arrays