Burn Severity and Its Impact on Soil Properties: A Study of the 2016 Erskine Fire in the Southern Sierra Nevada, California

: Wildﬁres can cause debris ﬂow events in affected areas due to changes in the physical properties of burned soils, which are linked to burn severity. A study in California’s Sierra Nevada explored the impact of burn severity on soil physical properties using various tests. Results showed that higher burn-severity soils had higher total organic carbon content and liquid limit, and the plastic limit was also higher. The plasticity index was highest among low burn-severity soils


Introduction
Wildfires have long been familiar hazards throughout the western United States. However, in the past few decades, wildfires and their impacts have become more destructive, and the frequency of large wildfires has been increasing [1]. A changing climate and prolonged droughts, coupled with an expanding wildland-urban interface, are exacerbating wildfire effects. Drier conditions are generating larger wildfires and endangering people who have settled in fire-prone regions. Furthermore, wildfire management is often insufficient because the increased destructivity of wildfires is outpacing wildfire management efforts. Soil burn severity is a measure of fire's effects on soil and is categorized into four burn severities [2,3]: very low or unburned (VLBS), low (LBS), moderate (MBS), and high burn severity (HBS). In VLBS soil, the surface organic layers experience negligible to no damage resulting from the fire. In LBS soil, the surface organic layers and surface soil are slightly charred, but the soil structure is preserved. In MBS soil, a majority of the surface organic layers are consumed, and surface soil is charred from brown to black, with soil structure mostly preserved. In HBS soil, the surface organic layers are completely or predominantly consumed, and HBS soil is covered in a layer of ash, indicating the complete destruction of surface vegetation. Soil structure is often reduced or destroyed in HBS soils.
The impacts of wildfires extend beyond the fire itself. The remaining burn scar, for example, is prone to debris flows [4,5]. Vegetation forms a protective layer that limits erosion, and a loss of vegetation can accelerate erosion and runoff, increasing debris flow probability [6,7]. Debris flows in the immediate aftermath of fire are commonly generated by runoff when the landscape is particularly vulnerable to infiltration-when there is excess overland flow and the resistance of the soil to erosion is reduced [8]. Post-wildfire debris flows may not solely form through this mechanism [9]. Rainfall-induced infiltration can result in excessive subsurface pore water pressures that trigger shallow landslides, which Wildfire frequency and the affected burn area have increased significantly over much of the western United States since the 1980s [6]. The average length of the fire season has also extended by seventy-eight days since the 1970s, and these observed trends are attributed to a changing climate [6]. Prolonged droughts and higher average temperatures have generated drier forests, and these conditions have allowed wildfires to intensify. Furthermore, the coupling of a changing climate with poor fire management practices of the past has exacerbated the impacts of wildfires [24]. As the wildland-urban interface continues to expand, humans are increasingly putting themselves at risk in fire-prone regions.
In 2017, wildfires in California's State Responsibility Area burned 2048 km 2 , which is more than double the 989 km 2 burned in 2016 and 2.5 times more acreage than the state's five-year average. As of 2018, seven of California's twenty largest wildfires have occurred in the past eight years, and only four of the twenty most destructive wildfires in California have occurred prior to 1990. Increasing burn area trends are similar for the western United States. For example, the combined total acreage burned in all ecoregions of the western U.S. has increased at a rate of~352 km 2 per year from 1984 to 2011 [6].

2016 Erskine Fire Burn Area
The Erskine fire burned 194 km 2 of the southern Sierra Nevada in eastern Kern County, California [25] (Figure 1). Steep, mountainous terrain dominates the topography, and elevation within the fire perimeter ranges between 762 and 1798 m. The communities of Lake Isabella, Squirrel Mountain Valley, Mountain Mesa, and South Lake lie within the fire perimeter and downstream portions of several watersheds, putting humans and structures at risk of debris flows. The mean annual precipitation in the region ranges between eight to twelve inches. The fire ignited on 23 June 2016 and was contained on 10 July 2016. A faulty power line was found to be the ignition source. In total, 90 km 2 of the National Forest Service, 70 km 2 of Bureau of Land Management, and 34 km 2 of private land were impacted, with 185 km of roads and 95 km of trails affected. Roughly 877 km of stream channels were also affected, with less than 3% of the channels being perennial streams. The 2016 Erskine fire destroyed 285 homes and claimed two fatalities. Suppression costs totaled $23 million [25].

Wildfire in the Western United States
Wildfire frequency and the affected burn area have increased significantly over much of the western United States since the 1980s [6]. The average length of the fire season has also extended by seventy-eight days since the 1970s, and these observed trends are attributed to a changing climate [6]. Prolonged droughts and higher average temperatures have generated drier forests, and these conditions have allowed wildfires to intensify. Furthermore, the coupling of a changing climate with poor fire management practices of the past has exacerbated the impacts of wildfires [24]. As the wildland-urban interface continues to expand, humans are increasingly putting themselves at risk in fire-prone regions.
In 2017, wildfires in California's State Responsibility Area burned 2048 km 2 , which is more than double the 989 km 2 burned in 2016 and 2.5 times more acreage than the state's five-year average. As of 2018, seven of California's twenty largest wildfires have occurred in the past eight years, and only four of the twenty most destructive wildfires in California have occurred prior to 1990. Increasing burn area trends are similar for the western United States. For example, the combined total acreage burned in all ecoregions of the western U.S. has increased at a rate of ~352 km 2 per year from 1984 to 2011 [6].

2016 Erskine Fire Burn Area
The Erskine fire burned 194 km 2 of the southern Sierra Nevada in eastern Kern County, California [25] (Figure 1). Steep, mountainous terrain dominates the topography, and elevation within the fire perimeter ranges between 762 and 1798 m. The communities of Lake Isabella, Squirrel Mountain Valley, Mountain Mesa, and South Lake lie within the fire perimeter and downstream portions of several watersheds, putting humans and structures at risk of debris flows. The mean annual precipitation in the region ranges between eight to twelve inches. The fire ignited on 23 June 2016 and was contained on 10 July 2016. A faulty power line was found to be the ignition source. In total, 90 km 2 of the National Forest Service, 70 km 2 of Bureau of Land Management, and 34 km 2 of private land were impacted, with 185 km of roads and 95 km of trails affected. Roughly 877 km of stream channels were also affected, with less than 3% of the channels being perennial streams. The 2016 Erskine fire destroyed 285 homes and claimed two fatalities. Suppression costs totaled $23 million [25]. In the Erskine fire area, woodland habitat spanning over 21 km 2 was recorded in the pre-fire state [25]. However, after the fire, there was a reduction of more than 13 km 2 (equivalent to 64%) in upland woodlands overall. It is important to note that this reduction in acreage was not equally distributed across all habitat types. For instance, although Jeffrey pine (Pinus jeffreyi) was not widely present in the region, its acreage was reduced by 98%. On the other hand, California juniper (Juniperus californica), which had a much In the Erskine fire area, woodland habitat spanning over 21 km 2 was recorded in the pre-fire state [25]. However, after the fire, there was a reduction of more than 13 km 2 (equivalent to 64%) in upland woodlands overall. It is important to note that this reduction in acreage was not equally distributed across all habitat types. For instance, although Jeffrey pine (Pinus jeffreyi) was not widely present in the region, its acreage was reduced by 98%.
On the other hand, California juniper (Juniperus californica), which had a much broader distribution, experienced a reduction of 72%. Blue oak (Quercus douglasii) and canyon live oak (Q. chrysolepis) were each reduced by 51%, while the more resilient interior live oak (Q. wislizeni) woodlands recorded a 20% reduction in acreage.
Soils of HBS, MBS, LBS, and VLBS comprise 1%, 43%, 45%, and 11% of the burn area, respectively. Shrubland comprised approximately 55% of the burn area, while coniferous vegetation comprised approximately 24% of the burn area. Mixed, herbaceous, and grassland communities comprised the remaining vegetation types within the fire perimeter. According to USDA Soil Surveys, the soils in the region belong to the general soil unit "Soils on Hillslopes and Mountain Slopes of the Southern Sierra Nevada Range" [25]. These soils are derived from weathered granitoid and metasedimentary rocks. The prevalent soil textural class in the uppermost layer is mostly composed of gravelly loam and coarse sand. The burn area is underlain by the Sierra Nevada batholith, metasedimentary, and metavolcanic units of the Mesozoic Era [26,27]. The middle Cretaceous Kings sequence, comprising quartz-mica schist, quartzite, marble lenses, and calc-silicate rocks, lies along the western portion of the burn area, and remnant bedding of the Kings sequence dips steeply N-NW. Quaternary alluvium also lies within the fire perimeter.

Sample Collection
The sample locations are situated within the property of the Bureau of Land Management, and a special-use research permit was issued and obtained prior to sample collection. All samples were taken from soils belonging to the Stineway-Kiscove association soil unit [25], which is derived from weathered metasedimentary rock and characterized as very gravelly sandy loam that grades into a gravelly loam. The Stineway-Kiscove association soil profile is approximately 33 cm thick, and the most limiting soil layer's capacity to transmit water is very low, with a rate of 0.00 to 0.25 cm/h [25]. Very low infiltration rates can facilitate runoff during rainfall events, which can be especially problematic in wildfireaffected areas. Fire-affected soil further reduces infiltration rates, catalyzing erosion and increasing the likelihood of a post-fire debris flow [28].
A total of sixty soil samples were collected approximately fifteen months after the ignition of the Erskine fire, with fifteen samples collected from each burn severity. Samples were taken from soils derived from the Kings sequence, from the surface to a depth visually determined to be unaffected by the fire (approximately 5-8 cm), based on the absence of burned soil, ash, and charred roots. In Southern California, shallow landsliding is a frequent occurrence and may account for about 50% of hillslope erosion [29]. For example, the depth measured at 286 landslide scarps in the field in Southern California ranged from 0.15 to 0.5 m, with an average depth of 0.35 m [30]. Therefore, it is essential to investigate how disturbances such as wildfires can affect their occurrence. Thus, roughly 400 g of soil was collected for each sample from a visually determined depth affected by the wildfire.
Ten samples from each burn severity were analyzed for Atterberg limits, plasticity index, grain size, and total organic carbon Among these samples, three were selected for X-ray diffraction (XRD) analysis, while two were chosen for scanning electron microscopy (SEM) imaging. The XRD analysis was performed using a Panalytical Emprean X-ray diffractometer (Malvern Panalytical, Malvern, UK), and the SEM imaging using a Hitachi S3400N VP SEM (Tokyo, Japan) at California State University Bakersfield (CSUB). Additionally, five samples from each burn severity were reserved exclusively for direct shear tests. The number of samples tested was determined to be the minimum necessary to capture the properties of the soil with respect to each test procedure.

Atterberg Limits Analyses
The Atterberg limits and PI were determined following the ASTM Standard D4318 Wet Preparation Method [31]. Ten samples from each burn severity were air-dried overnight and then passed through a #40 sieve to separate soil particles of size 0.425 mm or smaller from the bulk sample. All particles larger than 0.425 mm were discarded, and Atterberg limits were explored on the portion of the sample that passed through the #40 sieve. The one-point method was used to determine the LL [31]. Incrementally, around 25 mL of water was added to roughly 80-100 g of air-dried soil until the soil exhibited liquid-state properties. Then, the LL was tested on a portion of the wetted soil using a Humboldt hand-operated liquid limit device.
A portion of the soil remaining from LL testing was retained for PL testing. The handrolled method was employed to determine the PL [31]. For each test, a three-to four-gram sample of soil was hand rolled into a linear thread 3.2 mm in diameter. Once a linear thread 3.2 mm in diameter was achieved, the thread was remolded into an ellipsoidal mass and then rerolled. This process continued until the sample's water content had decreased such that a linear thread could no longer be formed. This testing procedure was repeated twice per sample.
After performing LL and PL test procedures, the water content of each sample was determined as specified in ASTM Test Method D2216 [32]. The samples were oven-dried for twenty-four hours at 95 • C. The oven temperature was then lowered to 40 • C, and the samples were cooled at this temperature for one hour prior to being removed. The samples were weighed, yielding the mass of the container and the oven-dried soil. Then the water content, liquid limit, plastic limit, and plastic index were calculated.

Direct Shear Strength Test
Shear strength analysis was performed using the Gilson Direct Shear Testing instrument. Direct shear testing was performed following ASTM Standard D3080 [33]. Five samples from each burn severity were tested. Each sample was sheared under the following normal loads: 48 kPa (0.5 TSF; tons per square foot), 96 kPa (1.0 TSF), and 192 kPa (2.0 TSF). Approximately 100 g of bulk soil sample was used for each shear test. Prior to shearing, to replicate the in situ partial saturation condition due to the water-repellent layers created by the wildfire, the soil in each test was partially saturated with approximately 25 mL of deionized water and consolidated for approximately sixty minutes. A relatively fast shearing rate of 0.13 cm per minute was performed for each test to simulate the in situ undrained conditions. Active shearing was conducted for four minutes, resulting in a total displacement of 0.52 cm. This procedure was repeated on a separate sample aliquot for each test on all samples at different normal loads.
Shear stress versus displacement was manually recorded at 15-second intervals. The mean values for maximum shear stress under the normal loads of 48 kPa, 96 kPa, and 192 kPa were graphed, and a best-fit line was used to connect the data points, whereby the cohesion and maximum angle of internal friction were derived. The y-intercept represents the value of cohesion, and the maximum angle of internal friction was derived from the slope of the line. The shear strength of the soil is expressed with the equation τ = c + σ · tanϕ, where τ is the mean maximum shear strength, c is the cohesion, σ is the normal stress, and ϕ is the maximum angle of internal friction.

Grain Size Analysis and SEM Imaging Characterization
Grain size analysis was performed on ten samples from each burn severity using a Mastersizer Hydro G 2000 grain size analyzer (Malvern Panalytical, Malvern, UK) at CSUB. Sample aliquots used for grain size analysis consisted of 1.05 g ± 0.1 g selected from sample material previously used to determine Atterberg limits. The samples were saturated in deionized water for at least twenty-four hours, followed by a five-minute soaking in a solution of Calgon (sodium hexametaphosphate) (4 g/1000 mL distilled H 2 O) to promote de-coagulation of non-cemented and unfused grains. After the deflocculation-Calgon soaking, a five-minute sonicator bath was used to de-coagulate any remaining aggregates. The sample was then diluted to 1/8 of its original concentration with deionized water. Employing the splitter method, the Mastersizer Hydro G 2000 analyzed samples within an obscuration range of 11-20% [34]. Results were processed in a spreadsheet software program.
To further explore potential changes in grain size resulting from wildfire, scanning electron microscopy was used to identify grain fusion, aggregation, and size distribution with respect to burn severity. Bulk soil samples, three from each burn severity, were analyzed using a Hitachi S-3400N scanning electron microscope with an automated stage. Soil samples were mounted on a sample holder, carbon-coated, and then loaded into the SEM vacuum chamber for imaging. The working distance was set at 10 mm, with beam settings at 20 kV and 70 µA.

Total Organic Carbon Measurement
The Heiri loss-on-ignition (LOI) method was employed to explore the TOC content with respect to burning severity [35]. However, the Heiri LOI method does not account for the dehydration of minerals [36]. Nevertheless, based on the soil composition, dehydration is assumed to be negligible, and the change in mass resulting from LOI is assumed to be attributed to TOC. Ten samples from each burn severity were analyzed using the LOI method. Sample aliquots used for TOC analysis were selected from the sample material previously used to determine Atterberg limits. The mass of each crucible was measured, and roughly 2.0 g was obtained for each sample, which was measured to 0.0001 g. All samples were placed in an oven for 48 h at 45 • C to evaporate all moisture. The dry mass of each sample was measured in its respective crucible and recorded. Following dry-mass measurements, all samples were then placed in an oven for one hour at 600 • C. The oven was turned off, and the samples were allowed to cool gradually to 40 • C before measuring and recording the mass of the samples. The TOC content was then calculated based on the mass difference.

X-ray Diffraction Mineralogy Analysis
Sample preparation was conducted using the following processes [37]. Bulk samples were crushed into a fine powder using the SPEX Sample Prep Shatterbox 8530 for two minutes. Bulk samples were then placed into bulk sample holders and analyzed using a Malvern PANalytical X-ray diffractometer. Clay-size fraction samples were soaked in hydrogen peroxide (3% H 2 O 2 ) for at least twenty-four hours to eliminate any organic matter. Then, a 250 mL solution of Calgon (sodium hexametaphosphate) (4 g/1000 mL distilled H 2 O) was added to the beakers. Each beaker was then placed into an ultrasonic bath for several minutes to facilitate deflocculation and disaggregation of sample grains. Samples were then placed into a centrifuge for two minutes, isolating grains of 2 µm and smaller. Oriented aggregates on glass slides were prepared using the filter-peel method and 0.45 µm membranes [38]. Each slide was then placed into the XRD instrument for analysis.
X-ray diffraction analysis was performed on three samples from each burn severity, with a 2θ scanning range from 3 • to 40 • at 45 kV and 40 mA. MacDiff 4.2.6 software was utilized to help identify the minerals in each sample [39]. The relative abundance was indicated by the simple comparison of peak intensities.

One-Way Analysis of Variance
One-way analysis of variance (ANOVA) was performed to investigate the variation in means between soil burn severity (SBS) groups while accounting for variation within groups. The significance level was set at 0.05 (α = 0.05). ANOVA was conducted on Atterberg limits, TOC, grain size, and shear strength results. ANOVA results were generated using the data analysis tool in Microsoft Excel [40].
This statistical technique helps determine whether significant differences exist in soil properties with respect to burn severity, or whether the observed differences may be attributed to random variation or error. ANOVA was performed based on the assumption of the null hypothesis: all SBS groups had the same mean value for a given soil property.
To test for variation in soil properties with respect to SBS, an F-test was conducted, and the resulting p-value was evaluated. A significant p-value indicated that the results were statistically significant and that there was a variation in soil properties with respect to SBS. An F-test computes a p-value using the test statistic: where M B is the mean square value between groups and M W is the mean square value within groups. The p-value is computed using the probability distribution of F-values, which is dependent on the degrees of freedom between groups and within groups. If the generated p-value is lower than the set significance level (α = 0.05), then the null hypothesis is rejected. The lower the p-value, the stronger the evidence that significant differences exist in soil properties with respect to burn severity.  (Table 1). ANOVA was performed at a significance level of 0.05 (α = 0.05). LL had an F-value of 2.479 and a p-value of 0.077 (Table 1).

Atterberg Limits and Plasticity Index
Atterberg limits, TOC, grain size, and shear strength results. ANOVA results were generated using the data analysis tool in Microsoft Excel [40].
This statistical technique helps determine whether significant differences exist in soi properties with respect to burn severity, or whether the observed differences may be attributed to random variation or error. ANOVA was performed based on the assumption of the null hypothesis: all SBS groups had the same mean value for a given soil property To test for variation in soil properties with respect to SBS, an F-test was conducted, and the resulting p-value was evaluated. A significant p-value indicated that the results were statistically significant and that there was a variation in soil properties with respect to SBS An F-test computes a p-value using the test statistic: where MB is the mean square value between groups and MW is the mean square value within groups. The p-value is computed using the probability distribution of F-values which is dependent on the degrees of freedom between groups and within groups. If the generated p-value is lower than the set significance level (α = 0.05), then the null hypothesis is rejected. The lower the p-value, the stronger the evidence that significant differences exist in soil properties with respect to burn severity.  (Table 1) ANOVA was performed at a significance level of 0.05 (α = 0.05). LL had an F-value of 2.479 and a p-value of 0.077 (Table 1).   The average PL of HBS soils was 34.  (Table 2). ANOVA was performed at a significance level of 0.05 (α = 0.05). PL had an F-value of 4.291 and a p-value of 0.011 ( Table 2).

Atterberg Limits and Plasticity Index
The average PI for HBS soils was −0.6, with a minimum value of −7.0 and a maximum value of 5.0 ( Figure 2C). The average PI for MBS soils was 3.8, with a minimum value of 1.0 and a maximum value of 8.0. The average PI for LBS soils was 5.7, with a minimum value of 3.0 and a maximum value of 8.0. The average PI for VLBS soils was 3.9, with a minimum value of 2.0 and a maximum value of 6.0. The PI in seven out of ten HBS samples was found to be less than or equal to zero. No samples in less severely burned soils were found to have a PI less than or equal to zero. Table 2. The ANOVA results for plastic limit with the significance level set at 0.05 (α = 0.05). PL variation between burn severities was statistically significant. SS = sum of squares; df = degrees of freedom; MS = mean squares; F = test statistic; p-value = probability of data value occurring from random error; and F crit (F critical) = value obtained by computing the probability distribution of F-values.

Total Organic Carbon
The average TOC content in percent by mass for HBS, MBS, LBS, and VLBS soils were 12.0%, 10.7%, 10.1%, and 8.7%, respectively (Figure 3). A maximum TOC content of 27.7% and a minimum TOC content of 6.1% were found in HBS soils. A maximum TOC content of 14.1% and a minimum content of 8.0% were found in MBS soils. A maximum TOC content of 14.5% and a minimum content of 8.3% were found in LBS soils. A maximum TOC content of 10.6% and a minimum content of 6.9% were found in VLBS soils.
Geotechnics 2023, 3, FOR PEER REVIEW and a minimum TOC content of 6.1% were found in HBS soils. A maximum TOC con of 14.1% and a minimum content of 8.0% were found in MBS soils. A maximum T content of 14.5% and a minimum content of 8.3% were found in LBS soils. A maxim TOC content of 10.6% and a minimum content of 6.9% were found in VLBS soils. ANOVA was performed on TOC at a significance level of 0.05 (α = 0.05). Varianc highest among HBS soils and lowest among LBS soils, with values of 38.4% and 1 respectively. ANOVA revealed an F-value of 1.608 and a p-value of 0.204 (Table 3). Table 3. ANOVA results for total organic carbon with the significance level set at 0.05 (α = 0 TOC variation between burn severities was statistically insignificant. SS = sum of squares; df = grees of freedom; MS = mean squares; F= test statistic; p-value = probability of data value occur from random error; and F crit (F critical) = value obtained by computing the probability distribu of F-values.

Grain Size Analysis and SEM Imaging Characterization
The mean percentage of clay-size particles (≤3.99 µm) in HBS soils was 17.5%, w the mean percentage of clay-size particles in MBS, LBS, and VLBS soils was 17.1%, 18 and 19.1%, respectively ( Figure 4A). ANOVA was performed at a significance level of (α = 0.05), which revealed that the variance was highest among HBS soil and lowest am LBS soil, with values of 23.4% and 5.6%, respectively. ANOVA further revealed an F-v of 0.566 and a p-value of 0.641 (Table 4). ANOVA was performed on TOC at a significance level of 0.05 (α = 0.05). Variance is highest among HBS soils and lowest among LBS soils, with values of 38.4% and 1.5%, respectively. ANOVA revealed an F-value of 1.608 and a p-value of 0.204 (Table 3). Table 3. ANOVA results for total organic carbon with the significance level set at 0.05 (α = 0.05). TOC variation between burn severities was statistically insignificant. SS = sum of squares; df = degrees of freedom; MS = mean squares; F = test statistic; p-value = probability of data value occurring from random error; and F crit (F critical) = value obtained by computing the probability distribution of F-values.

Grain Size Analysis and SEM Imaging Characterization
The mean percentage of clay-size particles (≤3.99 µm) in HBS soils was 17.5%, while the mean percentage of clay-size particles in MBS, LBS, and VLBS soils was 17.1%, 18.6%, and 19.1%, respectively ( Figure 4A). ANOVA was performed at a significance level of 0.05 (α = 0.05), which revealed that the variance was highest among HBS soil and lowest among LBS soil, with values of 23.4% and 5.6%, respectively. ANOVA further revealed an F-value of 0.566 and a p-value of 0.641 (Table 4).  Table 4. ANOVA results for clay-size particles with the significance level set at 0.05 (α = 0.05). Claysize particle variation between burn severities was statistically insignificant. SS = sum of squares; df = degrees of freedom; MS = mean squares; F= test statistic; p-value = probability of data value occurring from random error; and F crit (F critical) = value obtained by computing the probability distribution of F-values.
(a) Clay-size particle Summary  Figure 4B). ANOVA was performed at a significance level of 0.05 (α = 0.05), which revealed that the variance was highest among MBS soil and lowest among LBS soil, with values of 35.2% and 10.3%, respectively. ANOVA also revealed an F-value of 2.524 and a p-value of 0.073 (Table 5).  Table 4. ANOVA results for clay-size particles with the significance level set at 0.05 (α = 0.05). Claysize particle variation between burn severities was statistically insignificant. SS = sum of squares; df = degrees of freedom; MS = mean squares; F = test statistic; p-value = probability of data value occurring from random error; and F crit (F critical) = value obtained by computing the probability distribution of F-values.
(a) Clay-size particle Summary The mean percentages of silt-size particles (4.00-62.50 µm) for HBS, MBS, LBS, and VLBS soils were 37.2%, 37.6%, 42.4%, and 38.0%, respectively ( Figure 4B). ANOVA was performed at a significance level of 0.05 (α = 0.05), which revealed that the variance was highest among MBS soil and lowest among LBS soil, with values of 35.2% and 10.3%, respectively. ANOVA also revealed an F-value of 2.524 and a p-value of 0.073 (Table 5). SEM images show that particle size distributions among all samples are consistent with unburned (VLBS) soil. Clay-, silt-, and sand-size particles were observed in all samples ( Figure 5). Thus, no significant variance was observed in grain size distribution with respect to burn severity. Evidence of grain fusion and aggregation was not observed in any of the samples. Table 5. ANOVA results for silt-size particles with the significance level set at 0.05 (α = 0.05). Siltsize particle variation between burn severities was statistically insignificant; SS = sum of squares; df = degrees of freedom; MS = mean squares; F = test statistic; p-value = probability of data value occurring from random error; and F crit (F critical) = value obtained by computing the probability distribution of F-values.
(a) Silt-size particle Summary  Table 5. ANOVA results for silt-size particles with the significance level set at 0.05 (α = 0.05). Siltsize particle variation between burn severities was statistically insignificant; SS = sum of squares; df = degrees of freedom; MS = mean squares; F= test statistic; p-value = probability of data value occurring from random error; and F crit (F critical) = value obtained by computing the probability distribution of F-values.
(a) ----SEM images show that particle size distributions among all samples are consistent with unburned (VLBS) soil. Clay-, silt-, and sand-size particles were observed in all samples ( Figure 5). Thus, no significant variance was observed in grain size distribution with respect to burn severity. Evidence of grain fusion and aggregation was not observed in any of the samples.

Direct Shear Strength
Under a normal load of 48 kPa, the average peak shear stress values for HBS, MBS, LBS, and VLBS soils were 74.5 kPa, 63.4 kPa, 60 kPa, and 60 kPa, respectively ( Figure 6). HBS soil had the highest peak shear stress of 116.5 kPa under a normal load of 96 kPa.

Direct Shear Strength
Under a normal load of 48 kPa, the average peak shear stress values for HBS, MBS, LBS, and VLBS soils were 74.5 kPa, 63.4 kPa, 60 kPa, and 60 kPa, respectively ( Figure 6). HBS soil had the highest peak shear stress of 116.5 kPa under a normal load of 96 kPa.    ANOVA was performed at a significance level of 0.05 (α = 0.05). ANOVA results for a normal stress of 48 kPa revealed an F-value of 7.671 and a p-value of 0.002 ( Table 6). The variance was highest among MBS soil and lowest among VLBS soil, with values of 1.213 and 0.047, respectively. ANOVA results for a normal stress of 96 kPa revealed an F-value of 6.487 and a p-value of 0.004 (Table 7). The variance was highest among HBS soil and lowest among MBS soil, with values of 2.097 and 0.505, respectively. ANOVA results for normal stress of 192 kPa revealed an F-value of 3.103 and a p-value of 0.056 (Table 8). The variance was highest among VLBS soil and lowest among HBS soil, with values of 16.208 and 2.505, respectively. ANOVA was performed at a significance level of 0.05 (α = 0.05). ANOVA results for a normal stress of 48 kPa revealed an F-value of 7.671 and a p-value of 0.002 ( Table 6). The variance was highest among MBS soil and lowest among VLBS soil, with values of 1.213 and 0.047, respectively. ANOVA results for a normal stress of 96 kPa revealed an F-value of 6.487 and a p-value of 0.004 ( Table 7). The variance was highest among HBS soil and lowest among MBS soil, with values of 2.097 and 0.505, respectively. ANOVA results for normal stress of 192 kPa revealed an F-value of 3.103 and a p-value of 0.056 (Table 8). The variance was highest among VLBS soil and lowest among HBS soil, with values of 16.208 and 2.505, respectively.

X-ray Diffraction Analysis
The mineralogical analysis revealed that the abundant minerals included illite, smectite, and kaolinite/chlorite ( Figure 9). Based on the simple comparison of peak intensity, it was observed that the abundance of smectite and kaolinite/chlorite was lower in soils with higher burn severities. No other significant trends were observed.

Total Organic Carbon
The results of this study show that TOC content is highest in HBS soils, and the trends revealed that the lower the burn severity, the lower the TOC content, as illustrated in Fig-Figure 9. (A) Bulk XRD profiles. Smectite abundance was lower in soils with higher burn severities. (B) Clay-size XRD profiles. Lower contents of smectite and kaolinite/chlorite were found in soils with higher burn severities.

Total Organic Carbon
The results of this study show that TOC content is highest in HBS soils, and the trends revealed that the lower the burn severity, the lower the TOC content, as illustrated in Figure 4A. These findings are inconsistent with other studies that found that TOC decreases with increasing burn severity [13,15,41]. Organic material from peripheral lands can be introduced to burned soils, which can increase soil TOC, leaving the possibility of external organic inputs in this study's sample locations [15]. However, pre-fire satellite imagery ( Figure 1B) reveals a variation in vegetative density concerning the sampling locations of each burn severity, with the highest vegetation density within the HBS sampling area. This imagery indicates that TOC contents are likely inherent to the sample location, so the introduction of organic matter from external inputs may not be required to explain the observed TOC values. Although there is variation in TOC among different soil burn severities, ANOVA revealed these results to be statistically insignificant, indicating the need for more analyses to better understand the correlation between changes in TOC content and soil burn severity.

Soil Particle Size
No trend was observed in the abundance of clay-and silt-sized grains with respect to burn severity ( Figure 4A,B). HBS and MBS soils have a lower abundance of clay-sized grains than LBS and VLBS soils. This may be due to greater vegetative denudation within HBS and MBS sample areas, which could have facilitated the erosion of clay-sized grains from the burn scar, leading to lower abundances among these soils [42]. Additionally, scanning electron microscopy was employed to visually detect evidence of grain fusion, but no diagnostic evidence of grain fusion was observed in the samples ( Figure 5). This finding is inconsistent with previous studies, which found a coarsening trend with increasing burn severity [7,18]. The absence of a grain-coarsening trend suggests that the temperatures in the Erskine fire were not high enough to cause the fusion of clays [15,18]. ANOVA analysis revealed these results to be statistically insignificant, and they are also supported by the presence of TOC in all samples. Furthermore, since the majority of the Erskine Fire was shrubland, this further supports the assertion that temperatures were not high enough to cause grain fusion.

Soil Mineralogy
The impact of soil heating on soil stability can vary, as shown in previous research [15]. Soil heating can affect the primary binding agents in soils and their mineral composition. For instance, soils mainly bound by organic material have exhibited decreases in soil aggregate stability after a wildfire, while soils mainly bound by various mineral cements can exhibit increases in soil aggregate stability [13,15]. Products of kaolinite decomposition can act as cementing agents that increase aggregate stability in fire-affected soils, as suggested by previous studies [13,15].
The results of this study indicate that mineral transformations may have occurred only in HBS soil. As shown in Figure 9, kaolinite/chlorite abundance is lower in HBS soil, suggesting that temperatures reached 550 • C, the temperature at which kaolinite is destroyed. However, the presence of TOC in all samples suggests that temperatures did not exceed 450 • C unless external inputs contributed to the TOC content. Although minerals can decompose at temperatures lower than 550 • C, no other trends in mineral composition and abundance were observed in this study with respect to burn severity. Thus, it is possible that temperatures required for kaolinite decomposition were limited, which could have allowed TOC to remain in these areas. These findings are consistent with other studies, indicating that a decrease in kaolinite is usually restricted to moderate and severe fire zones [15] and this process of kaolinite decomposition is not a factor to cause the highest strength of HBS soils. Instead, the destruction of clay minerals can make soils more susceptible to erosion [7].

Soil Atterberg Limits and Plasticity Index
Atterberg limits results revealed a trend of higher LL values as burn severity increased (Figure 2A). PL shared a similar trend with LL ( Figure 2B). PL is highest among HBS soils and lowest among unburned soils. ANOVA analysis revealed the results for LL to be statistically insignificant, but the results for PL were found to be statistically significant with a p-value of 0.01. HBS soils have the highest TOC, and there is a trend of decreasing TOC with decreasing burn severity. The correlation between burn severity and Atterberg limits [6], as seen in previous research where the destruction of organic matter resulted in a decrease in Atterberg limits [13], could potentially be accounted for by this particular trend. It appears that TOC content is the factor that most influences the Atterberg limits. The soil's vulnerability to slaking and dispersion on wetting increased as the organic matter content decreased with decreasing burn severity, indicating that it is very easy for the soils to reach a plastic limit in the case of short-term rainfall, and coupled with looser soil and lower soil shear strength (Figure 6), it is easy for them to be eroded by runoffs. Soil shear strength is determined by the cohesive forces between soil particles [43,44]. The strength is directly linked to the soil's aggregate stability, which, in turn, is affected by intrinsic properties such as clay and organic matter [43]. The texture of the soil also has an impact on its strength. Coarse-textured soils rely on frictional forces, while fine-textured soils depend on cohesive forces. The concentration of organic matter in the soil affects its shear strength by modifying the cohesion between soil particles and structural stability [45]. The impact of soil organic matter on shear strength is somewhat inconsistent according to various studies. For instance, some researchers [45,46] have reported that an increase in soil organic matter concentration leads to a decrease in shear strength. They attributed this to the organic matter's ability to reduce bulk density [47,48], which results in a weaker soil structure. However, other studies [49][50][51][52][53] have found that increasing soil organic matter concentration enhances shear strength.

Soil Direct Shear Strength
Our findings suggest that TOC is the primary factor influencing shear strength and cohesion, as evidenced by the correlation between high TOC content and high shear strength and cohesion in HBS soils (as shown in Figures 3 and 6). This is because soil organic matter enhances the cohesive forces among soil particles (Figure 7), which has a direct effect on soil shear strength [15,54]. Instead, the slight difference in cohesion among the MBS, LBS, and VLBS samples could be due to their comparable particle size and relatively lower TOC content.

Conclusions
The findings from the 2016 Erskine Fire sample analysis show that: (1) LL and PL decrease with lower burn severity.
(2) TOC is less abundant with decreasing burn severity, and XRD analysis reveals lower smectite and kaolinite/chlorite abundances in HBS soils compared to soils with lower burn severities. Mineralogical trends may have been influenced by localized fire temperatures. (3) No correlation was found between burn severity and the size of clay-or silt-size grains.
Direct shear test results indicate higher shear strengths with greater burn severity. Funding: This work was supported by the U.S. National Natural Science Foundation (NSF HRD-1547784).

Data Availability Statement:
The authors can provide access to test data upon request.