Post-fire recovery of ecosystem carbon pools in a tropical mixed pine-hardwood forest

Aim of the study : To analyze the recovery pattern of carbon pools in terms of size and the relative contribution of each pool to total ecosystem C along a fire chronosequence of tropical mixed pine-hardwood forest. : Research Station core Biosphere in the western Materials and methods : Carbon stored in aboveground plant biomass, standing dead trees, downed woody debris, forest floor, fine roots and mineral soil, was compared with a nested analysis of variance (ANOVA) in post-fire stands of eight-year-old, 28- and 60-year-old stands of mixed Pinus douglasiana -hardwood forest. Main results : The total ecosystem carbon in eight-year-old stands was 50% lower than that of 60-year-old stands. Carbon content in the biomass and mineral soil increased with stand age. The carbon in the biomass recovered to the undisturbed forest in the 28 years of suc­ cession. The main C storage in the eight-year-old stands were the mineral soil (64%) and downed woody debris (18%), while in the 28- and 60-year-old stands, live tree biomass and mineral soil were the two largest components of the total C pool (43% and 46%, respectively). Research highlights : We found a significant effect of high-severity fire events on ecosystem C storage and a shift in carbon distribution. The relatively fast recovery of C in ecosystem biomass suggests that mixed Pinus douglasiana hardwood forest possess functional traits that confer resilience to severe ﬁre events.


Introduction
Wildfires influence the earth´s climate via the release of carbon and greenhouse gases, atmospheric aerosols, and by the alteration of surface albedo (Bowman et al., 2013). Therefore, has been a common perception that fire suppression decrease the rate of release of CO 2 caused by burning and maximize C storage in ecosys tem pools (Tilman et al., 2000;Hurteau & Brooks, 2011). However, there is enough evidence about how 2 a changing climate and policies that encourage fire suppression has contributed to modify forest structure, fuel loads, and fire behavior, causing a shift from low to high-severity fires (Agee & Skiner, 2005;Schoen nagel et al., 2017). In this sense, managers of fire-prone forests must trade off the promotion of C storage against the reduction of fire hazard (Hurteau et al., 2008;Bowman et al., 2013).
Understanding the role of fire regimes on ecosys tem carbon dynamics is critical for the implementation of climate change mitigation strategies in forested landscapes (Hurteau & Brooks, 2011;Williams et al., 2012;Bowman et al., 2013). The effects of wildfires are strongly influenced by variability in vegetation types, site conditions, fire behavior, and the diversity of fire regimes at the landscape level, which are char acterized by the range of variation in frequency, sea sonality, intensity, severity, size, and spatial pattern of the fire events (Agee, 1993;Scott et al., 2014). For this reason, there is a need to study fire effects in different forest ecosystems and environmental condi tions, particularly in mountain forest ecosystems in tropical regions of the world (Pregitzer & Euskirchen, 2004), where the effects of wildfires on the forest C balance and post-fire response of C pools have been little studied (Pompa-García & Sigala-Rodríguez, 2017). This is crucial in the context of increasing wildfire activity and severity as a consequence of fire suppression policies and human-induced climate change that has been reported for other regions like the western United States (Westerling et al., 2006;Schoennagel et al., 2017).
Worldwide, the majority of pine forests are under a frequent low-severity fire regime (which means a tem poral frequency lower than 35 years). Under this wild fire regime, the majority of fires have low severity, and wildfire events with moderate or high severity occur in a smaller proportion (Agee, 1993). The low-severi ty fires reduce the tree basal area by less than 30% (Boerner, 1981;Keyser et al., 2008), mainly affecting the C storage in shrubs, herbs, and the forest floor layer (Boerner, 1981), while high-severity fires induce high tree mortality (> 70%) and elevated carbon emissions, alter the distribution of carbon pools on forest ecosys tems, and may have long-term effects on post-fire for est structure and function (Kashian et al., 2006;Alex ander et al., 2012;Mitchell, 2015).
Carbon recovery patterns have been widely docu mented from boreal and temperate forests, which are dominated by stand-replacing fire regime (mean return interval 50 to >250 years) (Wang et al., 2001;Litton et al., 2004;Gough et al., 2007;Dore et al., 2008;Alexander et al., 2012;Kashian et al., 2013). Using data from chronosequences, C recovery in aboveground biomass has been described as a relatively slow C ac cumulation during stand regeneration, followed by a fast accumulation during stand development, then reaching a maximum and remaining relatively constant or declining in old-growth forest stands (Bormann & Likens, 1979). Site productivity, stand density and composition play a key role as a source of variation in forest C storage after fires (Litton et al., 2004;Kashi an et al., 2006;Alexander et al., 2012;Spies et al., 1988). C stored in belowground components varies less than that in aboveground pools (Kashian et al., 2006;Gough et al., 2007). However, a significant increase in post-fire mineral soil C has been observed, associated with rapid root turnover and potential incorporation of fire-created debris into the mineral soil C pool over time (Seedre et al., 2014).
Here we present the results of a study on the recov ery of carbon pools after wildfire events that opened up gaps, reinitiating succession in mixed pine-hard wood forests of central-western Mexico. Lower-mon tane moist pine forest in Mexico covers 6.1 million hectares (3.2% of the country's area). These forests thrive in warm temperate humid climates (mean an nual temperature of 11-19 ºC with a potential evapo transpiration: annual precipitation ratio less than 1), with a dry spring season and a summer rain regime (Cuevas-Guzmán & Jardel, 2004). Pinus species are the dominant component (>50% basal area), mixed with broad-leaved tree species of genera like Quercus, Carpinus, Clethra, Cornus, Magnolia, and Styrax, among others (Cuevas-Guzmán & Jardel, 2004). Mixed pine-hardwood stands develop in advanced succes sional stages in mesic sites, and the dominance of pines is maintained by recurrent wildfires (Jardel, 2008). The specific objective of the study was to analyze the re covery of carbon pools, in terms of size and the relative contribution of each pool to total ecosystem C along a fire chronosequence of tropical mixed pine-hardwood forest. Carbon content was quantified in stands of 8˗ and 28-year˗old that originated from fire and a > 60-year-old that typifies mature mixed pine-hardwood forests in the study area. We hypothesize that after wildfire ecosystem C storage will increase with stand age.

Study Area
The study was conducted in Las Joyas Research Station (LJRS), in the zone core of Sierra de Manantlán Biosphere Reserve (SMBR) in the state of Jalisco, central western Mexico (19° 14' 49"-19° 37' 30" N 3 Post-fire recovery of ecosystem carbon pools and 104° 14' 49"-104° 18' 16" W). LJRS has an exten sion of 1245 ha in a mountainous landscape with com plex relief, with an elevation range of 1500 m to 2242 m. The climate is warm temperate with a mean annual temperature of 15 ± 2 °C, ranging from 12.8 °C in January to 20 °C in May. Mean annual rainfall is 1826 ± 94 mm, and the potential evapotranspiration:annual rainfall ratio is 0.5-0.6. According to the Holdridge Life Zone System, the bioclimatic conditions corre spond to a lower montane moist forest (Cuevas-Guzmán & Jardel, 2004).
The soil parent materials are Tertiary extrusive igneous rocks, like basaltic porphyries, basalts, an desitic basalts, and volcanic tuffs. A typical soil cat ena in the area is a gradient from Inceptisols in ridges and upper slopes, Alfisols in middle to lower slopes, and Ultisols in hollows and stream banks. The vegetation is a mosaic of pine-oak forests associated with convex landforms (mountaintops and upper slopes), mixed hardwood forests (cloud forest) in concave landforms (ravines and hollows), mixed pine hardwood forest in intermediate conditions, and sec ondary scrub in abandoned agriculture fields (Cuevas-Guzmán & Jardel, 2004). The dominant species in pine-oak and mixed pine-hardwood forests is Pinus douglasiana Martínez. Most fires in the SMBR occur between April and early June, at the end of the dry season. Fires are rela tively small in area (mean: 189 ha, mode: 50 ha) and the most common ignition factors are human-related activities (Balcázar, 2011). Frequent low-severity sur face fires characterize the historical fire regime in the pine-oak forest, with a fire return interval ranging from three to12 years (Cerano-Paredes et al., 2015).
Fire suppression in LJRS used to encourage the recovery of mixed hardwood forests and mixed pine hardwood forests cover through natural regeneration, has led to decreased fire frequencies and accumulation of fuel. Forest floor and woody debris loads for stands without fires in 20 years have been estimated, respec tively, at 37 to 58 Mg ha -1 , and 31 to 38 Mg ha -1 (Al varado-Celestino et al., 2008). Under these conditions, mixed-severity fire events occurred in 1983 and 2003, causing larger fires; while there is no estimate of the burned area in 1983, the total area burned in the 2003 fire was 390 ha. The result was a heterogeneous spatial pattern formed of low-, moderate-, and high-severity patches, and others did not burn. High-severity fire patches were considered those where intense surface fire with smoldering combustion and torching caused high tree mortality (> 70%), consumption of the entire organic soil layer, and the restarting of succession (Jardel, 2008). The size of patches ranged from 1 to 40 ha and there was not post-fire management, vegetation recovered through natural regeneration. The sites burned at high-severity were selected for the present study (Table 1).

Sampling design
To analyze the recovery of ecosystem carbon pools along a post-fire chronosequence, we compared stands eight and 28 years following a fire, and mature stands (>60 years). The 60-year-old stands have not had low severity fires for at least 30 years and represent late successional condition in the absence of fire (Jardel, 1991;Cuevas-Guzmán & Jardel, 2004). It was assumed that the mature forests are representative forests in this region.
The use of the chronosequence method relies on the assumption that all variation among sites is due to dif ferences in time since the disturbance (Walker et al., 2010). Therefore, a nested sampling design was used to help prove that assumption. Each stand age class was replicated three times. To minimize the impact of spatial autocorrelation, the stands were not samples in close proximity to one another. This was achieved by To minimize the effect of variation in site conditions, stands were located in the same elevational range (1950 to 2150 m), within the same soil type (Alfisols), in the mid-portion of N-facing slopes dominated by Pinus douglasiana (> 70% basal area). Mean slope inclina tions showed no significant difference (from 22 to 38%). Mean stem density did not differ among stands, but tree basal area in the eight-year-old stands was four times lower than in the 28-and 60-year-old stands (Table 1).

Field sampling
The ecosystem carbon pools sampled within each 500-m 2 circular plot were: live and standing dead trees (>2.5 cm of diameter at breast height, DBH), under story (trees < 2.5 cm of DBH, shrubs <1.3 m in height and herbaceous plants), forest floor layer (litter and duff layer), downed woody debris, fine roots and min eral soil. The vegetation and soil samples were taken in the dry season, between January and June 2011. DBH (cm) was measured at 1.30 m height of all live and standing dead trees (>2.5 cm of DBH) using stand ard diameter tape graduated into 0.1-cm. The total height (m) of all live and standing dead trees within the plot was measured using a Suunto clinometer. All live trees were identified to species level. Understory was clipped at the base in four 1-m 2 subplots located 13 m from the center of each plot following the direc tion of the main four cardinal points. Components were stored in paper bags, labeled, and transported 3-4 days after sampling to the laboratory to determine dry weigh.
Downed woody debris (DWD), defined as all dead wood lying or standing (with a zenith angle 45°) was sampled using the planar intercept method (Van Wag ner, 1982). Starting from the center of each plot, three 20-m transects were established. The direction of the first transect was chosen randomly; the second tran sect was placed 120 degrees from the first and so on. The criteria used for decide which DWD was inter sected by transect were the proposed by Brown (1974). According to its size, the DWD was classified into fine (< 0.6 cm), regular (0.6-2.5 cm), medium (2.5-7.6 cm), and coarse (>7.6 cm). DWD pieces < 2.5 cm in diameter intersected by transect were counted in the last 5-m section. While, DWD pieces 2.6-7.5 cm were counted along the 20-m transect. DWD >7.6 cm were measured in diameter (cm) at the point of intersection and the decay was classified in five categories following the classification of Waddell, (2002). DWD pieces of all size were randomly col lected, stored in paper bags, labeled, and transported 3-4 days after sampling to the laboratory to assess the carbon concentration (%).
The forest floor was divided into two layers: the litter layer (LL) and the duff layer (DL). The LL is composed of fresh plant residues (excluding woody debris) that keep their structure and have an identifi able origin; the DL includes decomposed organic matter that has lost its original structure. The depth (cm) of the LL and DL was measured using a gradu ated needle, every 3-m starting from the center of each plot and along four transects (12.6 m) following the direction of the main four cardinal points (16 points per plot). LL and DL bulk density samples were col lected in three random points per plot following the method proposed by Ottmar & Andreu, (2007) modi fied by Morfín et al., (2012). At each sampling point, 30-cm sharpened steel square was positioned on top of the LL and DL and inserted until the bottom of the square was embedded in mineral soil. Thirteen iron markers were positioned in a systematic pattern within the square and inserted until flush with the top of the litter layer. The litter was carefully removed from the square and placed within a labeled paper bag. The distance between the top of each marker and the top of the DL was measured and recorded. To measure the DL, each marker was inserted further until flush with the top of the duff layer. The duff layer was care fully removed from the square and placed within a labeled paper bag. The distance between the top of each marker and the top of the mineral soil was meas ured and recorded. These thirteen depth measurements were averaged to represent the LL and DL depth for the sample. LL and DL samples were transported every 3-4 days after sampling to the laboratory to determine dry weigh. A total of 81 litter and 81duff bulk density samples were collected.
The fine roots (< 5 mm of diameter) were sampled at eight points systematically selected from the 16 points used for forest floor depth measurements. Soil samples were taken from the top 40 cm of mineral soil with a cylindrical soil corer (5-cm diameter, 15-cm height), placed in plastic bags, labeled, and transported every 3-4 days after sampling to the laboratory to de termine dry weight. Fine roots in the forest floor were not considered in this study.
Mineral soil samples were taken with a cylindrical soil corer (5-cm diameter, 15-cm height) at four depth intervals (0-10, 10-30, 30-50, and 50-70 cm) in eight points systematically selected from the 16 points used for forest floor depth measurements. For each soil depth, four of the eight samples were used to determine 5 Post-fire recovery of ecosystem carbon pools soil bulk density by the core method. Soil was placed in plastic bags, labeled, and transported every 3-4 days after sampling to the laboratory.

Laboratory analyses
The fine roots were separated from soil manually with tweezers and placed in a petri dish with water to eliminate soil particles. The live and dead roots were not separated. To biomass determination, the samples of understory, DWD, LL, DL and fine roots were oven dried to constant weight at 60 °C and weighed. Sub samples (0.5 g) were ashed in a muffle furnace at 500 °C for 4 h to determine inorganic content to report the data on a dry ash-free basis. The samples were ground and passed through a 40-mesh sieve (0.420 mm) and then pooled into one composite sample per plot for total carbon concentration analyses. Mineral soil sam ples were oven-dried at 50 °C for 48 h. Bulk samples were weighed and the volume of coarse fragments (> 2 mm) were determined and used to correct bulk density for each plot; the eight soil subsamples were pooled into one composite sample for each depth per plot, which was then passed through a 100-mesh sieve (0.149 mm) for total carbon concentration analyses.
Understory, DWD, LL, DL, fine roots, and mineral soil samples were analyzed for total carbon concentra tion (mg g -1 ) by combustion and coulometric detection using an automated CO 2 analyzer (UIC model CM5012, Joliet, IL, USA).

Biomass and carbon calculation
Plant biomass was considered as the dry weight of living and dead plant material contained above-and below-ground per a unit of surface area at a given point in time. Above-and below-ground plant biomass are expressed in units of Mg ha -1 . Live and dead tree bio mass was calculated using allometric equations pro posed by Vargas-Larreta et al., (2017) for Pinus dougla siana, Pinus herrerae, and Pinus oocarpa and Cruz -Martínez, (2007) for broad-leaved species. DWD biomass was calculated using equations proposed by Van Wagner, (1982). Decay classes registered in field were grouped in three classes: 1, 2, and 3 for sound, 4 for intermediate, and 5 for rotten (Waddell, 2002). The mean diameter quadratic and specific gravity values used are given in Table 2. The biomass of the forest floor was estimated as the product of depth (cm) of the LL and DL layers and bulk density (Mg ha -1 cm -1 ). Bulk density values for the same plots were previously re ported by Quintero-Gradilla et al. (2015). The C content in the different pools was estimated by multiplying the biomass (Mg ha -1 ) with carbon con centration (% of dry weight). C content in the mineral soil was calculated as the product of bulk density (g cm -3 ), depth (cm) and C concentration (%). All C pools were scaled up to Mg C ha -1 , and mean stand-level estimates for each C pool were calculated (n = 3, plots). For live and standing dead trees carbon concentration was assumed to be 50% of biomass (IPCC, 2003). Total ecosystem carbon storage was estimated by summing the individual carbon pools within each plot.

Statistical analyses
All variables were tested for normality and homosce dasticity assumptions using the Kolmogorov-Smirnov and Levene's test, respectively. Data were log-trans formed to meet assumptions when required (Zar, 1999), although they are reported in the original scale of meas ure. A nested analysis of variance (ANOVA) was used to test the effect of stand age after a wildfire on carbon content pools. Stand age (eight, 28, or 60 years) was the main fixed effect, and stands (repetitions) within each age were the nested random effects. After nested ANOVA, the means was compared by Tukey test (p = 0.05). Pearson's correlation was used to analyze the cor relations between tree basal area and aboveground tree biomass and total ecosystem biomass, and the correlation between stand age and total ecosystem C content. All these analyses were done with SPSS software version 16 (SPSS Inc. 1999, IBM, Armonk, NY, USA).

Biomass distribution in ecosystem components
Total ecosystem biomass was lower in eight-year-old stands (152.1 Mg ha -1 ) than in the 28-and the 60-year-old 6 stands (380.4 and 432.8 Mg ha -1 , respectively; Table 3). Aboveground live tree biomass increases significantly with stand age (F 2,6 = 33.41, p = 0.001). Both the above ground pine biomass species and broad-leaved species biomass increase with stand age. However, pine biomass was similar between 28-and 60-year-old stands, while broad-leaved biomass was highest in the 60-year-old stands. The biomass of standing dead tree was not af fected by stand age (Table 3). Understory biomass dif fered significantly with stand age, and peaked early in the chronosequence in the eight-year-old stands, which had 6-to 9-fold greater understory biomass than in the 28-and 60-year-old stands, respectively (Table 3).
Similarly, the DWD biomass in the eight-year-old stands was 5-fold greater than in the 28-and 60-year-old stands (F 2,6 = 11.99, p = 0.008), and made up 56% of the total aboveground biomass (Table 3). A nested stand effect within the stand age was significant (F 6,18 = 5.39, p = 0.002), which means that the site factor was also significant.
The forest floor mass increased with stand age (F 2,6 = 40.58 p ≤ 0.001). Both LL and DL mass were lower in the eight-year-old stands than in the older stands. LL mass peaked in the 28-year-old stands, while the DL mass was similar between the 28-and 60-year old stands. In contrast, fine root biomass was not af fected by stand age.
Tree basal area was positively related with above ground tree biomass (R 2 = 0.94, p < 0.001) and total ecosystem biomass (R 2 = 0.92, p < 0.0001). The dead:live biomass ratio decreased with stand age from 1.80 in the eight-year-old stands to 0.09 and 0.23 in the 28-and 60-year-old stands, respectively.

C concentration
The C concentration in the above-and below-ground biomass components, litter, and DWD varied from 39% to 51%, and no one differed significantly among forest stand by age (Table 4). In the three successional ages, the mineral soil C concentration was higher in the first 10 cm of the soil and decreased significantly (to 50%) up to 30 cm depth, and even further to 28% at 70 cm soil depth (Fig. 1). Differences among ages were only significant in the top 0-10 cm (Fig. 1).

C content pools
The C content in biomass components across the age sequence followed a similar pattern compared to the biomass (Table 5, Fig. 2). The C stored in aboveground tree biomass increased 6-fold from the eight-year-old stands to the 60-year-old stands. Similarly, forest floor C content increased by approximately 21.6 Mg C ha -1 from the eight-year-old stands to the 28-and 60-year old stands. In contrast, C stored in understory decreased by 85% between the eight-year-old and 60-year-old Table 3. Total biomass (Mg ha -1 ) in a fire chronosequence of tropical mixed pine-hardwood for est. Values are average with standard error in parentheses. Different letters indicate significant differences between means (p < 0.05) with the post hoc Tukey test.

Components
Stand age (years)  stands; likewise, C in DWD pool decreased with stand age, following an inverse J-shaped trajectory from 38 Mg C ha -1 in the eight-year-old stands to 9 Mg C ha -1 in the 60-year-old stands. The C content in fine root mass averaged 1.4 Mg C ha -1 and did not differ sig nificantly among stand ages. The soil bulk density showed no differences among stand ages at the same depth interval. The mean soil bulk density for each depth was 0.69±0.03, 0.73±0.02, 0.85±0.02, and 0.88±0.02 g cm -3 for 0-10, 10-30, 30-50, and 50-70 cm soil depth, respectively. Total mineral soil C (0-70 cm depth) increased significantly with stand age (Table 5, Fig. 2), from 147 Mg ha -1 in the eight-year-old stands to 161 and 203 Mg ha -1 in the 28-and 60-year-old stands, respectively (Fig. 3).
Total ecosystem C content increased with the stand age from 221 to 423 Mg C ha -1 (F 2,6 = 60.96, p < 0.000, Table 5 and Fig. 4a). The correlation between stand age and total ecosystem C content was significant (R 2 = 0.84 p < 0.001). Eight years after a fire, C was mainly stored in the mineral soil (67%) and the dead Figure 1. Distribution of carbon concentration (mg g -1 ) in the mineral soil (at 10, 30, 50 and 70 cm depth) in a fire chronose quence of tropical mixed pine-hardwood forest. Each data point represents the average of C concentration by soil depth, horizon tal bars are standard error, and lines are used to graphically visu alize the trend in average C concentration with soil depth. Dif ferent letters indicate a significant difference at the same depth between ages (p < 0.05) with the post hoc Tukey test. The lack of letters indicates no significant differences. Table 5. Nested analysis of variance (F and p) for carbon content (Mg ha -1 ) in a fire chronose quence of tropical mixed pine-hardwood forest. Stand age is the fixed effects factor; stand is the nested effect within stand age as a random effect factor. woody debris (18%), both making up 85% of the total ecosystem C content, while the live trees pool con tributed 12% of the total ecosystem C content. In contrast, live trees and mineral soil were the two larg est pools of the total ecosystem C pool in the 28-year old (90%) and 60-year-old (89%) stands, C stored in dead wood accounted only for 3% of the total ecosys tem C pool, in both 28-year-old and 60-year-oldstands and the forest floor represented 8% and 6%, respec tively (Fig. 4b).

Discussion
The changes observed in aboveground plant biomass after stand-replacement fires reveal a fast recovery of C pools in the mixed pine-hardwood forests in LJRS. A significant reduction of total C ecosystem pools by plant mortality and forest floor layer consumption is a widely documented effect of high-severity fires in boreal and temperate forests (Litton et al., 2004;Mac-Kenzie et al., 2004;Kashian et al., 2006; Seedre et al., 2014). In the study area, eight years after a fire the total ecosystem C stored was 48% of the C content in 60-year-old stands. Al though the chronosequence is relatively short com pared with other studies reported for temperate and boreal forests (Wang et al., 2003;Litton et al., 2004;Kashian et al., 2006;Dore et al., 2008;Alexander et al., 2012;Carlson et al., 2012;Kashian et al., 2013), it suggests that the warm temperate and moist climate of montane forests at tropical latitudes facilitates a fast post-fire recovery of forest stands. C storage in eco system biomass increased with stand age, reaching in the first 28 years following fires a level similar to the 60-year-old stands. This rapid C recovery was mainly driven by accumulation of C in live trees and the for est floor recovery and followed the typical pattern of rapid C biomass accumulation in early stages of stand development (Seedre et al., 2011;Alexander et al., 2012;Kashian et al., 2013).
The pattern of post-fire C pools recovery in the mixed Pinus douglasiana-hardwood forest can be de scribed following the stand development stages of Oliver & Larson, (1990) and the results of previous studies about forest succession in these forests (Cuevas-Guzmán & Jardel, 2004): (1) The eight-year-old stands correspond to the stand initiation stage, where P. douglasiana saplings are the most abundant species in the lower vegetation layer and pines represent 95% of live tree biomass (33.5% of total biomass). A high proportion of total biomass (50.5%) is concentrated in downed woody debris and forest floor biomass represents only 6.6%.
(2) Twenty-eight years after a fire, the developing stands are in the transition of the stem exclusion and
The C stored in aboveground live tree biomass is higher in 60-year-old stands than in 28-year-old stands due to the biomass increase of the broad-leaved tree species growing under the canopy. The establishment of shade-tolerant species under the canopy of the first cohort of trees that regenerated after a stand-replace ment disturbance is a common feature of the stand development process in dense forests (Oliver & Larson, 1990) and is favored by fire exclusion (Kane et al., 2013). Carbon storage in the understory layer was higher in the eight-year-old stands. The high C storage in herbaceous and shrub biomass reached in the early stages of forest succession have been associ ated with an open tree canopy that maintains favora ble conditions for its establishment, such as light, moisture, and nutrients (Gilliam, 2007;Jules et al., 2008;Seedre et al., 2011). Favorable conditions that are decreased with forest succession after canopy closure (Gilliam, 2007).
Carbon content in fine roots did not show any cor relation with stand age, consistent with previous stud ies (Howard et al., 2004;Gough et al., 2007). Carbon storage in standing dead trees did not change with stand age, suggesting that, eight years after the fire, C had already transferred from fire-killed trees to the downed woody debris pool. C content in DWD pool followed a pattern of high initial amounts of fire-killed fallen trees in the first years after a fire and then a reduction in C storage due to decomposition of dead wood in the intermediate stage, followed by an increase in the late stage of succession as a result of the natural mortality of trees (Spies et al., 1988;Brais et al., 2005;Kashian et al., 2013). The significant nested site effect on C storage in downed woody debris suggests that there was high variability that could not be explained only by stand age and that could be associated with microsite variation, decay heterogeneity, and management his tory (Harmon et al., 1986).
C stored in the forest floor layer was 80% lower in eight-year-old than in 28-or 60-year-old stands, an effect related to fire consumption of surface fine fuels. Forest floor starts to accumulate C with regeneration and increases throughout the life of a stand until it reaches the maximum accumulation, where there is a balance between production and decomposition (Gough et al., 2007;Seedre et al., 2011), as described by Quintero-Gradilla et al., (2015).
The carbon pool in the mineral soil is the long-term product of history, decomposition, and vegetation changes (Harden et al., 2000). The higher total soil C content at 70 cm of depth observed in 60-year-old stands may be associated with long-term processes strongly associated with clay particles and non-crys talline minerals that play an essential role in the protec tion and stabilization of organic matter in the soil (Six et al., 2002). Since soils in the study area are Alfisols with a characteristic subsurface argillic horizon (Mar tínez et al., 1993), they play a key role in C accumula tion. On the other hand, the effects of wildfires have been observed in the first years after fire and mainly in the superficial horizons, as reported by Quintero-Gradilla et al., (2015) and others (Wang et al., 2001;Litton et al., 2004;Kashian et al., 2013).
Total ecosystem carbon increased with stand age due to additional C content in the broadleaf trees' strata and the higher content in mineral soil in 60-year-old stands. The pattern of change of the total ecosystem carbon storage with stand age coincided with previous post-fire chronosequences that showed slow C accumulation during stand regeneration, fol lowed by a fast accumulation during stand develop ment, then reaching a maximum and remaining rela tively constant or declining in old-growth forest stands, reported for tropical forests (Pregitzer & Euskirchen, 2004) and for temperate and boreal for ests (Wang et al., 2003;Pregitzer & Euskirchen, 2004;Kashian et al., 2013;Seedre et al., 2014).
Regeneration of vegetation exerts strong control on post-fire C recovery, capturing carbon lost during or after burning (Litton et al., 2004;Kashian et al., 2006). Although tree mortality was high (> 70% reduction in basal area) in the sites examined in this study, mixed severity fires created small gaps (< 40 ha) where the trees that survived the fire and the trees of surrounding stands functioned as seed sources, which, together with favorable climatic conditions, promoted the subsequent fast regeneration and then the recovery of C stocks. The relatively fast recovery of ecosystem C in biomass suggests that mixed Pinus douglasiana-hardwood for est possesses functional traits like small winged seeds dispersed at long distances, fast growth rate, and shade intolerance, which confer resilience to severe fire (Kee ley & Zedler, 1998).
The management of fire-prone forest must consid ering a trade-offs between maximizing carbon storage by increasing carbon density through fire suppression and maintaining long-term carbon stability which 11 Post-fire recovery of ecosystem carbon pools means minimizing loss of carbon from the system by reducing the risk of high-severity fires (Hurteau & Brooks, 2011), as well as other forest management goals such as biodiversity conservation, watershed protection, timber production, and other ecosystem services (Hurteau et al., 2008;Bowman et al., 2013;Syphard et al., 2016). Fire-induced change in carbon distribution among ecosystem pools could have sev eral implications for C storage at the landscape level (Kashian et al., 2006). In the first years after a fire, when the majority of the carbon is stored in DWD and the mineral soil, a fire recurrence could expose the soil to erosion and loss of C (Neary & Overby, 2006;Gough et al., 2007). Based on this assumption, it is recommended the protection of those sites to promote regeneration. Not protecting the stands from repeated fires could lead to a failing recovery, with consequent changes in the forest structure, soil carbon, and spe cies composition (Kashian et al., 2006). Some exam ples are the transition from forest to grassland or shrubland, with a diminished capacity for carbon storage, observed by Savage & Mast, (2005) in south western United States ponderosa pine forest and Alanís-Rodríguez et al., (2012) in a forest of Pinus pseudostrobus forest in northern Mexico.
However, long periods of fire suppression in fire prone ecosystems, has resulted in an increased fuel loads and a change in vegetation structure, that can contribute to intense wildfires that produce greater levels of CO 2 and other greenhouse gas emissions (Hurteau & North, 2009;Hurteau & Brooks, 2011). It is critical to understand the implications of fire manage ment actions to carbon management and other goals as conservation and climate change mitigation.