For this study, we collected soils from five sites across an elevation transect along the western slope of the central Sierra Nevada, California (Fig. 1); the sites were previously characterized by Dahlgren et al. (1997). We selected four forested sites that are likely to experience forest fires and a fifth lower-elevation grassland site. The thermal alterations in bulk soil physical and chemical properties from the same study soils were previously reported in Araya et al. (2016).

All the sites have a Mediterranean climate characterized by warm-to-hot, dry summers and cool-to-cold, wet winters. Mean annual air temperature ranges from 16.7 ∘C at the lowest site located at 210 m a.s.l. to 3.9 ∘C at the highest elevation site at an elevation of 2865 m a.s.l. Annual precipitation ranges from 33 cm at the lowest site to 127 cm at the highest site (Dahlgren et al., 1997; Rasmussen et al., 2007) (Table 1).

The lower elevation woodlands of the Sierra Nevada experience less frequent fires than further upslope and the fires are often fast moving and lower severity (Skinner and Chang, 1996). At the middle-elevation zone of the Sierran forest, the mixed conifer zones, frequent fires are low to moderate severity at lower altitudes, but fire frequency generally increases with altitude towards the upper elevation of the mixed conifer forest (Caprio and Swetnam, 1993). Fires are infrequent and low severity within the high altitude, subalpine zone of the Sierra Nevada (Skinner and Chang, 1996).

Soils from the lowest elevation site, Vista series soils (210 m a.s.l.), fall within the oak woodland zone (elevations < 1008 m a.s.l.). This is the only soil in our study that does not have an O-horizon. The soil has dense annual grass cover and its A-horizon SOM originates mainly from root turnover. The Musick series soils (1384 m a.s.l.) lie within oak–mixed conifer forest (1008–1580 m a.s.l.) and mixed conifer forest (1580–2626 m a.s.l.). These soils receive the highest litter fall biomass. The Shaver and Sirretta series soils (1737 and 2317 m a.s.l., respectively) fall within the mixed conifer forest range zone, while the Chiquito series soils (2865 m a.s.l.) lie within the subalpine mixed conifer forest range (2626–3200 m a.s.l.). These soils have lower litter fall compared to the lower elevation soils (van Wagtendonk and Fites-Kaufman, 2006).

The western slope of the central Sierra Nevada presents a remarkable climosequence of soils that developed under similar granitic parent material and are located in landscapes of similar age, relief, slope, and aspect (Trumbore et al., 1996), with significant developmental differences attributed to climate. The soils at mid-elevation range (1000 to 2000 m a.s.l.) tend to be highly weathered, while soils at high and low elevations are relatively less developed (Dahlgren et al., 1997; Harradine and Jenny, 1958; Huntington, 1954; Jenny et al., 1949). Among the most important changes in soil properties along the climosequence are changes in soil organic carbon (SOC) concentration, base saturation, mineral desilication, and hydroxyl-Al interlayering of 2 : 1 layer silicates. Soil pH generally decreases with elevation and the concentrations of clay and secondary iron oxides show a step change at the elevation of the present-day average effective winter snow line, i.e., 1600 m elevation (Tables 1 and 2) (California Department of Water Resources, 1952–1962; Dahlgren et al., 1997).

Triplicate samples (0 to 5 cm depth) were collected at the five sites, approximately 10 m apart from each other. Any overlaying organic layer was removed prior to sampling so that only mineral soil was collected. The soils were air dried at room temperature and passed through a 2 mm sieve. Prior to furnace heating, the soils were oven dried at 60 ∘C overnight. Soil bulk density and field soil moisture were determined from separate undisturbed core samples collected from each site (Table 2).

Subsamples from each soil were heated in a muffle furnace to one of six selected maximum temperatures (150, 250, 350, 450, 550, and 650 ∘C). To ensure uniform soil heating and reduce formation of heating gradient inside, the soils were packed 1 cm high in 7 cm diameter porcelain flat capsule crucibles. Oxygen supply was not limited during the heating – the volume of soil sample to volume air in the furnace was approximately 1 : 50. Furnace temperature was ramped at a rate of 3 ∘C min-1 and soils were exposed to the maximum temperature for 30 min. Once cooled to touch, soils were stored in airtight polyethylene bags prior to analysis.

The six heating temperatures were selected to correspond with fire intensity categories that are based on maximum surface temperature (DeBano et al., 1977; Janzen and Tobin-Janzen, 2008; Neary et al., 1999), that is, low intensity (150 and 250 ∘C), medium intensity (350 and 450 ∘C), and high intensity (550 and 650 ∘C). These fire intensity classes generally correspond with thresholds for important thermal reactions in soils observed by differential thermal analyses (Giovannini et al., 1988; Soto et al., 1991; Varela et al., 2010). A heating rate of 3 ∘C min-1 is preferred in laboratory fire simulation experiments (Giovannini et al., 1988; Terefe et al., 2008; Varela et al., 2010); the slow heating rate prevents sudden combustion when soil ignition temperature is reached at about 220 ∘C (Fernández et al., 1997, 2001; Varela et al., 2010). The samples were exposed to the maximum set temperature for a period of 30 min. This length of time ensures that the entire sample is uniformly heated at the set temperature and is in keeping with the wide majority of similar laboratory soil heating experiments (for example Badía and Martí, 2003a; Fernández et al., 2001; Giovannini, 1994; Varela et al., 2010; Zavala et al., 2010). The duration of soil heating under vegetation fires is highly varied and not uniform across landscape (Parsons et al., 2010). The same heating procedure was used for all the soils so that it would be possible to compare how the soils from different climate regimes are likely to respond to the fires.

Dry-aggregate size distribution was measured by sieving. Samples were dry sieved into three aggregate size classes: 2–0.25 mm (macroaggregates), 0.25–0.053 mm (microaggregates), and < 0.053 mm (silt- and clay-sized particles or composites). These aggregate size classes were selected to enable comparison with other studies that investigated the effect of different natural and anthropogenic properties on soil aggregate dynamics and aggregate-protected organic matter (Six et al., 2000).

C and N concentrations and stable isotope ratios were measured using an elemental combustion system (Costech ECS 4010 CHNSO Analyzer, Costech Analytical Technologies, Valencia, CA, USA) that was interfaced with a mass spectrometer (DELTA V Plus Isotope Ratio Mass Spectrometer, Thermo Fisher Scientific, Inc., Waltham, MA, USA). For the analyses, air-dried soil samples were ground to powder consistency on a ball mill (8000M Mixer/Mill, with a 55 mL tungsten carbide vial, SPEX SamplePrep, LLC, Metuchen, NJ, USA) and oven dried at 60 ∘C for over 36 h. This lower temperature and longer duration of oven drying was used to avoid possible heating-related C or N changes that might occur if drying was done at 105 ∘C (Kaiser et al., 2015). The C and N concentration results were corrected for moisture by oven drying subsamples at 105 ∘C overnight. The C and N concentration results were corrected by adjusting for moisture as Wadj=W×100-Wm, where Wadj is the adjusted percent concentration, W is the concentration before moisture adjustment, and Wm is the percent moisture content. All concentration changes resulting from moisture adjustment were a decrease of less than 1 % of the value. The stable isotope ratios are presented using the δ notation (per mill, ‰) as δ13C and δ15N calculated as δ=Rsample-Rstandard/Rstandard×1000 ‰, where R is ratio of 13C to 12C for δ13C and 15N to 14N for δ15N. The standards used for analyses are atmospheric N2δ15N and Vienna Pee Dee Belemnite (VPDB) δ13C.

Bulk soil organic matter composition was analyzed using FTIR spectroscopy on a Bruker IFS 66v/S vacuum FT-IR spectrometer (Bruker Biosciences Corporation, Billerica, MA, USA). We used the diffuse reflectance infrared Fourier transform (DRIFT) technique (Ellerbrock and Gerke, 2013; Parikh et al., 2014). Powder samples were dried overnight at 60 ∘C and scanned in mid-infrared from 4000 to 400 cm-1. We used non-KBr-diluted samples after preliminary analyses showed that dilution was not necessary. KBr dilution is not required for soils with low (< 10 %) organic matter concentrations (Ellerbrock and Gerke, 2013; Reeves III, 2003). The FTIR spectrum was collected using KBr background and was baseline corrected using the rubber band correction method with the default 64 baseline points that is part of the OPUS software (Bruker Corporation, 2009).

All quantitative results are expressed as means of three replicates ± standard error, unless otherwise indicated. Differences in means were tested by analysis of variance (ANOVA) and pairwise comparison of treatments done using Tukey's honest significant difference (HSD) test at p<0.05 significance level. The normality of the data and the homogeneity of variances was checked using Shapiro–Wilk's and Levene's tests, respectively. All statistical analyses were performed using R statistical software (R Core Team, 2014). The Pearson's correlation coefficient was used to examine relationships between C concentration and changes in soil properties.

The initial concentration of C ranged from 1.5 % (Vista soil, 210 m) to 7.7 % (Musick soils, 1384 m). Soil C concentration continuously decreased with increasing temperature. The largest decrease occurred between temperatures of 250 and 450 ∘C. At 450 ∘C, all soils lost more than 95 % of their original C. C concentration changes with heating above 450 ∘C were small and not statistically significant at p<0.05. The C : N ratio ranged from 10 (Vista soils, 210 m) to 29 (Musick soils, 1384 m). Following a similar pattern to C concentration changes, the C : N ratio decreased with an increase in heating temperature (Fig. 2).

The loss of C and N from soils due to heating showed a similar response among all five soils (Fig. 2). After 250 ∘C, all the soils lost more than 25 % of their initial C (except Shaver soils that lost only about 10 %). At 350 ∘C all soils lost 50 to 70 % of C. Heating at 450 ∘C led to the loss of more than 95 % of their initial C for all soils in this study. However, the rate of loss of N was lower than that of C. At temperatures greater than 550 ∘C there was 5 to 15 % of soil N still remaining. Consequently, we observed a decrease in C : N ratio with increased heating temperature. All soils continued to lose about 15 % soil N for every 100 ∘C increase and maintained more than 60 % of their N at heating temperatures up to 350 ∘C. After heating at 450 ∘C, all soils lost more than 60 % of their original soil N and 85 % by 550 ∘C.

The δ13C composition of all soils was indicative of C3 vegetation. Soil δ13C composition was most negative at about -28 ‰ for the lowest elevation Vista site (210 m), and the value became consistently less negative with an increase in elevation, reaching -24 ‰ for the highest two sites (i.e., > 2317 m elevation). For all soils, there was a general trend of δ13C enrichment with temperature increase (Fig. 2). The largest change (2.5 to 3.0 ‰) occurred at heating temperatures between 250 and 450 ∘C for the lower elevation soils and between 150 and 450 ∘C for the two highest elevation soils. For the two highest elevation soils, there was a significant (p<0.05) depletion above that temperature. For all soils, except Musick (1384 m) and Shaver (1737 m), the maximum enrichment occurred at 450 ∘C. All soils showed a similarly patterned δ15N composition change with temperature. The soils were increasingly δ15N enriched with temperature increase up to 350 ∘C. At temperatures above 350 ∘C, the soils got more δ15N depleted, with the most negative δ15N occurring at 650 ∘C (Fig. 2).

C and N concentrations as well as 13C- and 15N-stable isotope ratios were measured for individual soil aggregate size class. The analysis was done on samples heated up to a temperature of 450 ∘C. The concentration of C and N in samples heated above 450 ∘C was too low to measure significant changes in C distribution in the different aggregate size classes.

The distribution of C in the three aggregate size fractions followed the same general pattern with increase in the heating temperatures. The macroaggregate size fraction (2–0.25 mm) had the least C concentration and silt–clay-sized particles (< 0.053 mm) had the largest concentration of C (Fig. 3). N concentration for the macroaggregates was below the detection limit at 450 ∘C for Chiquito and Sirretta. The change in C and N concentration across heating temperature was similar for all soils.

The distribution of C and N in different size aggregates did not change noticeably except at 450 ∘C where concentration in all three fractions converged to zero. The distribution of N in the three aggregate size fractions was similar to that of C and followed a similar pattern across all the heating temperatures. Similarly, the macroaggregate size fraction (2–0.25 mm) had the least amount of N concentration, and silt–clay-sized particles (< 0.053 mm) had the largest concentration of N. For Shaver (1737 m), Sirretta (2317 m), and Chiquito (2865 m) soils, the macroaggregate N concentration was too low and could not be detected (Fig. 3). The atomic C : N ratio generally stayed the same for all soils throughout the temperatures. C : N ratio was highest in macroaggregates, which had the lowest C and N concentrations, followed by microaggregates and silt–clay sizes for all soils.

The stable isotope composition of 13C was very similar between aggregate sizes, with silt–clay-sized aggregates being slightly more enriched, except for Shaver (1737 m), which had slightly more enriched macroaggregates. Conversely, the δ15N values showed clear differences among aggregate fractions even though the measured values of δ15N did not change notably with combustion temperatures. δ15N was highest in silt–clay-sized particles and lowest in macroaggregates, with the microaggregates showing intermediate values. The pattern of change in δ15N across combustion temperatures did not affect this order of δ15N values among aggregate fractions. Most of the C and N in the soils was associated with the larger macroaggregate and microaggregate fractions. With the exception of Vista (210 m) soils, the concentrations in macroaggregates continued to decrease with an increase in temperature, and the remaining C and N concentrations were distributed between the smaller aggregate fractions (Fig. 4). At 450 ∘C, most of the C and N of the higher altitude soils (Shaver, Sirretta, and Chiquito) was now associated with the silt–clay-sized fractions.

Changes in chemical composition of SOM due to heating were analyzed by infrared spectroscopy using the DRIFT technique. The spectra and peaks after contrasting levels of thermal treatments exhibited qualitative similarities among the different soils. FTIR spectra for the soils are shown in Fig. 5. One notable change that occurred in the functional group composition of SOM with heating is the lowered absorbance intensity of aliphatic methylene groups (as represented by the aliphatic C–H stretching peak that appears at bands between 2950 and 2850 cm-1) at > 250 ∘C in all soils. When comparing intensity of peaks at 2910–2930 and 2853 cm-1 wave numbers (from aliphatic methyl and methylene groups, band A) with those at 1653 and 1400 cm-1 (oxygen containing carboxyl and carbonyl groups, band B), the decrease in prominence in the aliphatic C-H peak occurs early in the heating sequence, while the C=O band shows little relative change. In addition, after heating at a temperature of 550 ∘C, all soils lost the O-H stretching peaks (between 3700 and 3200 cm-1). In a pattern that is more prominent for the Musick soil that had the highest concentration of OM, the aromatic C=C stretch around 1600 cm-1 gets more resolved with increase in heating temperature. This pattern in the C=C is visible but not as well resolved in the rest of the soils, especially the Vista soil that showed the least-resolved aromatic C=C stretch peak in this region.

Our results show significant effects of combustion temperature on concentration, distribution, and composition of SOM in topsoils that experience the most intense heating during vegetation fires. Topsoils have relatively high OM and low clay content that render them more sensitive to heating since the SOM experiences significant changes during heating. In our study system, the effect of fire heating on SOM ranged from slight distillation (volatilization of minor constituents), typically at temperatures below 150 ∘C, to charring, which typically starts at temperatures above 350 ∘C, and complete combustion, consistent with findings of previous studies (Badía and Martí, 2003b; Certini, 2005). Our findings also confirmed that regardless of the differences our soils had in mineralogy and other soil physical and chemical properties, the heating treatments (as a proxy for wildfires) led to a consistent decrease in concentration of soil C. This was in agreement with previous studies that showed a decrease in soil C concentration in topsoil after fires (for example Badía et al., 2014; Certini, 2005). However, this loss of C is expected to be restricted to topsoil, while it is expected that the C concentration in subsoil is likely to remain unchanged or may even increase (for example Dennis et al., 2013; Kavdır et al., 2005) due to incorporation of necromass from surface biomass (Almendros et al., 1990; Knicker et al., 2005).

We observed significant changes in concentration, distribution, and composition of SOM with increasing heating temperature. The steep decline in concentration of C in soil that we observed in this study is consistent with a decrease of about 25 % C at 250 ∘C and an almost 99 % loss at 450 ∘C (Fig. 6). The magnitude of C loss with heating we observed is similar to the findings of Terefe et al. (2008); and Ulery and Graham (1993), who investigated changes in soil C using artificial heating experiments. Similarly, Giovannini et al. (1988) also found that OM decrease started at 220 ∘C with about 15 % loss of OM and about 90 % OM loss at 460 ∘C. Fernández et al. (1997) reported 37 % of SOM loss at 220 ∘C and 90 % at 350 ∘C. Furthermore, along with the change in C concentration, between 150 ∘C and before almost total loss of C above 450 ∘C, the SOM went through significant qualitative changes that included decrease in C : N ratio, enrichment in δ13C isotope, changes in δ15N isotope, and changes in FTIR spectra. Loss of N after fire heating is the result of combustion and volatilization (Fisher and Binkley, 2000). In this study, we observed that N is not as significantly reduced until 350 ∘C, with about 75 % N remaining as opposed to a loss of C concentration greater than 50 % at the same temperature (Fig. 6). Previous studies had shown that moderate- to high-intensity fires convert most organic-N into inorganic forms of N, such as ammonium (Certini, 2005; Huber et al., 2013). Ammonium is the immediate combustion product that contributes to formation of nitrate (NO3-) by nitrification reactions in the weeks or months after a fire. Other studies have shown that a considerable amount of N is transferred into pyrogenic OM products, to black N (de la Rosa and Knicker, 2011; Knicker, 2010), which would also explain the decrease in the C : N ratio. Decrease in the C : N ratio with fire heating has previously been observed in both laboratory and field fire studies (Badía and Martí, 2003a; Certini, 2005; Fernández et al., 1997; González-Pérez et al., 2004).

SOM has a C isotopic composition that reflects the δ13C signature of native vegetation. Plants are depleted in δ13C relative to atmosphere. The δ13C composition for our soils indicated that the dominant source of OM in all soils was C3 plant biomass that had an average δ13C of -27 %, with the higher-elevation soils having more positive δ13C than the low-elevation soils. Enrichment of 13C with heating is consistent with the loss of plant-derived C. In addition, the fact that lipids (that have relatively more δ13C depleted than the woody materials) are combusted at lower temperatures than woody materials (such as cellulose and lignin) might contribute to the enrichment of δ13C with heating (Czimczik et al., 2002). The stable C and N isotope composition of our soils showed significant fractionation with temperature. δ13C values became more positive (enriched in δ13C) up to 450 ∘C, where up to 99 % of C was lost (Fig. 6). At higher temperatures there was a less uniform pattern among the soils. For the last < 1 % C, Sirretta and Chiquito soils continued to be more negative (depleted in δ13C) at higher temperatures, while for the rest of the soils there was a slight depletion at 550 ∘C followed by a slight enrichment at 650 ∘C (Fig. 2). The depletion of δ13C at 550 and 650 ∘C we found in this study is likely a result of SOM charring since there was little or no decrease in C concentration between these temperatures. In a wood charring experiment (non-oxygen atmosphere) at 150, 340, and 480 ∘C, Czimczik et al. (2002) observed an enrichment of δ13C at 150 ∘C where there was no C concentration change but a depletion of δ13C at 340 and 480 ∘C, with charring where the C concentration increased over 50 % due to charring.

Fires tend to lead to enrichment of 15N. This is particularly observed in soils immediately in the aftermath of fires (Boeckx et al., 2005; Grogan et al., 2000; Herman and Rundel, 1989; Huber et al., 2013), but there is limited information available on the exact temperature ranges that cause specific levels of 15N enrichment. In this study, we observed enrichment of 15N up to 350 ∘C and depletion after 350 ∘C for all soils (Fig. 2). It is likely that the continued 15N enrichment with heating is the result of fractionation due to combustion and volatilization of organic matter, which discriminate against 15N. However, the exact mechanism behind continued depletion of 15N when heated above 350 ∘C remains unclear. One potential explanation for the 15N depletion at higher temperatures could be indiscriminate removal of N since higher temperatures cause the combustion and volatilization process to happen instantly, compared to charring of OM at lower temperatures. In a post-fire analysis of δ15N on a subalpine ecosystem in Australia, Huber et al. (2013) found that the 15N enrichment of bulk surface soil (from unburnt leaves) was higher than that of the charred OM, which was again higher than that of the ash. They attributed this difference in enrichment level to be the result of the lower heating intensity experienced by the bulk soil, which provided slower processes for greater fractionation. Conversely, higher heat intensity experienced by the ash results in full combustion of plant material, providing little opportunity for isotopic discrimination. The temperature range where we observed the depletion of 15N in our experiment corresponds with the range where a steep decline in N concentration happened (Fig. 6), which would be consistent with the explanation.

Investigation of the response of climosequence soils to different heating temperatures in this study enables us to infer how changes in climate are likely to alter the effect of fires on topsoil physical and chemical properties in the long term. Along our study climosequence, we observed critical differences in response of topsoils based mostly on concentration of OM in soil and soil development stages of each soil. Soil OM concentration and composition in particular have been shown to respond to changes in precipitation amount and distribution, as is expected in the Sierra Nevada (Berhe et al., 2012b). Consequently, changes in soil C storage associated with climate change are expected to lead to different amounts of C loss due to fires. This is evidenced by the observed highest total mass of C loss from the mid-elevation Musick soil that had the highest carbon stock, compared to soils on either side of that elevation range. Anticipated changes in climate in the Sierra Nevada mountain range are expected to include upward movement of the rain–snow transition line, exposing areas that now receive most of their precipitation as snow to rainfall and associated runoff (Arnold et al., 2015, 2014; Stacy et al., 2015). Upward movement of the rain–snow transition zone under anticipated climate change scenarios and associated more intense weathering at higher elevation zones can render more C to be lost during fires. More than 80 % of the variability in mass loss, aggregate strength, SSA, pH, CEC, and N concentrations is associated with changes in C concentration (Table 3). Hence, as the vulnerability of these ecosystems to increased fire frequency increases due to climate change (Westerling et al., 2006), we can expect more soil C loss with fires, along with associated changes in soil chemical and physical properties. In particular, our findings of important changes in soil physical and chemical properties occurring between 250 and 450 ∘C are important for recognizing that critical transformations of topsoil SOM are likely to occur when, as a result of climate change, systems that are adapted to low-severity fires experience medium- to high-severity fires.

The different responses of soil aggregation in our climosequence to the treatment temperatures also suggest potential loss and transformation of the physically protected C pool in topsoil. Degradation of aggregates during fire (Albalasmeh et al., 2013) is likely to render aggregate-protected C susceptible to potential losses through oxidative decomposition, leaching, and erosion. Moreover, in systems such as the Sierra Nevada, which are dominated by steep slopes, movement of the rain–snow transition zone upward is likely to increase proportion of precipitation that occurs as rain. The kinetic energy of raindrops and the observed increase in hydrophobicity of soils after fires (Johnson et al., 2007, 2004) can lead to higher rates of erosional redistribution, especially for the free light fraction or particulate C that is not associated with soil minerals (Berhe et al., 2012a; Berhe and Kleber, 2013; McCorkle et al., 2016; Stacy et al., 2015).