Marcos Francos·Xavier U′beda·Paulo Pereira

Abstract Studies of post-fire soil status in Mediterranean ecosystems are common;however,few have examined the effects of long-term forest management after a wildfire on physicochemical soil properties.Here,we analyzed differences in soil properties attributable to long-term postfire management and assessed the sustainability of these management practices in relation to the soil properties.The study area is located in O′dena in the northeast region of the Iberian Peninsula consisted of the control forest(burned more than 30 years ago),low density forest(LD;burned in a wildfire in 1986 and managed in 2005)and high density forest(HD;burned in a wildfire in 1986 and no managed).For soils from each plot,we measured soil water repellency,aggregate stability,total nitrogen(TN),soil organic matter(SOM),inorganic carbon(IC),pH,electrical conductivity, extractable calcium, magnesium, sodium,potassium(K),phosphorus,aluminum(Al),manganese(Mn),iron(Fe),zinc,copper,boron,chrome,silicon and sulfur and calculated the ratios of C/N,Ca+Mg/(Na+K)1/2,Ca/Al and Ca/Mg.Significant differences were found in TN, IC, SOM,pH,K,Al,Mn, Fe and C/N ratio(p ＜0.05).All soil properties were found to have largely recovered their pre-fire values.Soils were affected by the post-fire management practices implemented 20 years after the fire,as reflected in their respective physicochemical properties,so that soil properties at the control and LD sites are more similar today than those at the control and HD sites.Thus,sustainable forest management can overcome soil degradation in areas affected by wildfire in the medium-and long-term by improving soil properties.

Keywords Soil chemical properties·Aggregate stability·Post-fire management·Wildfire risk·Vegetal density

Introduction

Wildfires are natural phenomena in Mediterranean ecosystems(Gill 1975;Bento-Gonc?alves et al.2012).In the Mediterranean region of Catalonia, socioeconomic changes that characterized the end of the twentieth century produced forest structures of great vegetation density,which clearly led to an increase wildfire severity(Ve′lez 2000).Thus,while wildfires usually have short-to medium-term impacts on soil properties, when these burn episodes are especially severe the effects can last much longer.Soil chemical properties,determined by the consumption and depletion of plant nutrients, are closely related to forest characteristics such as vegetation density(Francos et al.2018a)and constitute a critical relationship for determining post-fire management practices in affected areas(Rezaei and Gilkes 2005).

The effects of short-term post-fire management have been widely studied (e.g., Garc?′a-Orenes et al. 2017;Francos et al.2018b;Taboada et al.2018),but few have focused on the long-term effects of fire on soil properties.Studies such as that by Pereira et al.(2018)are essential for summarizing and understanding the effect of forest management in fire affected areas.In addition,Francos et al.(2018a)analyzed soil chemical properties 18 years after a wildfire,Ojima et al.(1994)studied the effect of fire on N cycling over a 50-year period,and Johnson et al.(2012)assessed the effects of two fires(1960 and 1981)on soil chemistry and vegetation.In a limited number of cases,most notably LeDuc and Rothstein(2010),Mun~oz-Rojas et al.(2016)and Yermakov and Rothstein(2006),the authors also carried out a chronosequence of soil status,although in the majority of studies,only long-term analyses of soil samples were done.Here,in common with DeLuca et al.(2006)and McNamara et al.(2015),we compare the properties of a soil long after a fire with those of an unburned area,which we use as a reference for pre-fire values(control),and focus specifically on the role of vegetation density.In this respect,few studies have analyzed the effects of plant densities on soil properties after fire over the long term,the most notable exceptions being Holz et al.(2000),who studied the long-term effects of sewage sludge amendments on barley yield,and Kaye et al.(2010),who studied long-term soil status following differential tree and shrub regeneration in an area affected by wildfire.

Long-term post-fire management can lead to changes in plant density that in turn affect soil properties.For example,soil water repellency(SWR)can be affected for periods that range from a few months to years,depending on fire severity(Dyrness 1976).Indeed,the persistence of SWR depends more on fire severity than on the length of time since burning(Huffman et al.2001).However,others have shown that SWR reaches pre-fire values roughly 1 year after the fire,regardless of the severity and no longer differs between burned and unburned areas(MacDonald and Huffman 2004).Likewise,there is no clear pattern for the long-term effects of fire on a soil's aggregate stability(AS)(Mataix-Solera et al.2011).Soil C storage and,consequently,soil organic matter(SOM)and inorganic C(IC)levels varied with plant density(Kaye et al.2010).The time since the last wildfire(Johnson et al.2012)and vegetation density(Zhang et al.2010)also influence soil N, pH, electrical conductivity (EC) and levels of major nutrients.A high C/N ratio can impede plant growth due to the close relationship with SOM and N availability (Jiang et al. 2016). However, we lack information regarding the effects of variations in plant density on the minor elements in the soil and their ratios.

Here we determined the soil status in areas subject to different long-term post-fire management practices after a wildfire event.The objectives were(1)to analyze differences in soil properties associated with post-fire management practices;(2)study how plant density affects soil properties;and(3)identify long-term sustainable post-fire management practices as revealed by soil properties.

Materials and methods

Study site

The study site is located in O′dena in the northeastern region of the Iberian Peninsula(Fig.1)was burned by a forest fire in 1986.The predominant vegetation in the study area is Pinus halepensis Miller,Pinus nigra Arnold and Quercus ilex L.with an understory vegetation of Pistacea lentiscus L.,Genista scopius L.and Hedera helix L.The substrate consists of sediments from Paleocene shale(Panareda-Clope′s and Nuet-Badia 1993).The soil is classified as Fluventic Haploxerept(Soil Survey Staff2014).The site has a typical Mediterranean climate,with an annual temperature of 14.2°C and a mean annual rainfall between 500 and 600 mm.

Experimental design and sampling

Three 1-ha plots were set up and sampled in October 2015,almost 30 years after the fire:a reference plot(control)that was not affected by the 1986 wildfire or subjected to any management;a low density(LD)plot burned in the 1986 wildfire and subjected to management in 2005;and a high density(HD)burned in the 1986 wildfire but not subjected to any management.The management in 2005(almost 20 years after the wildfire)at the LD site involved a clearcutting operation,which left a density of 1000 trees/ha and the cut vegetation on the soil.At each site,we collected nine topsoil samples(0-5 cm depth),for a total of 27 samples for the study area.The three sites occupied areas with a similar soil type,vegetation composition and topographical characteristics (slope ＜3% with a northeast aspect).Wildfire severity in 1986 was classified as high according to Tarrant(1956)and U′beda et al.(2006),given that 100%of tree crowns were burned.

Laboratory methods

Fig.1 Location of study area

The analysis of the physicochemical properties of soil are described in great detail in Francos et al.(2016a)and Francos et al.(2018b).SWR was determined using the water drop penetration time(WDPT)test(Wessel 1988).AS was analyzed using the 10-drop impact(TDI)method(Low 1954).TN of pulverized soil was determined using a Flash EA 112 Series(Thermo-Fisher Scientific,Milan)and Eafer 300 software (Thermo-Fisher Scientific, Milan)(Pereira et al.2012).SOM and IC were determined using the loss-on-ignition(LOI)method described by Heiri et al.(2001).Soil pH and EC were determined by deionized water extraction[1:2.5].Extractable elements(Ca,Mg,Na and K)were determined by ammonium acetate extraction[1:20],in line with the method described by Knudsen et al.(1986)Available P was analyzed following the Olsen Gray method(Olsen et al.1954).Soil Al,Mn,Fe,Zn,Cu,B,Cr,Si, and S were determined using ammonium acetate extraction[1:20].Extractable cations were expressed as mg/kg of soil.The soil C/N ratio was calculated with organic C and TN. The ratio (extractable Na+extractable K)/(extractable Ca+extractable Mg)1/2(SPAR)was obtained using the method of Sarah(2004).We also calculated the ratio Ca/Al and Ca/Mg.

Statistical analysis

Data normality and homogeneity were assessed using the Shapiro-Wilk test and Levene's test.We applied a one-way ANOVA test in those cases with a Gaussian distribution and homogeneity of variances.For those that did not satisfy normality and homogeneity requirements,we applied the nonparametric Kruskal-Wallis ANOVA test.If significant differences were identified at p ＜0.05,a Tukey post hoc test was applied to identify differences within treatments.A redundancy analysis(RDA)was carried out to identify the extent to which the variation in one set of variables accounts for the variation in another.The soil properties used in the RDA were SWR,AS,TN,SOM,IC,pH,EC,Ca,Mg,Na,K,P,Al,Mn,Fe,Zn,Cu,B,Cr,Si,S,C/N,SPAR,Ca/Al and Ca/Mg.Statistical analyses were carried out using SPSS 23.0 (IBM, Armonk, NY, USA) and CANOCO for Windows 4.3 (Microcomputer Power,Ithaca,NY,USA).

Results

Soil physicochemical properties

No significant differences in SWR and AS were recorded between the three sites(control,HD and LD).Soil TN was significantly higher at the control site than at the HD site.SOM was significantly higher at the control than at the LD and HD sites.IC levels were significantly higher at the LD site than at the Control and HD sites.Soil pH was significantly higher at the LD than at the HD site and significantly higher at the HD than at the control site.There were no significant differences in EC between the three sites(Table 1).

Table 1 Descriptive statistics for physicochemical characteristics in soils from three plots in northeastern Iberian Peninsula

Major and minor soil elements

The three plots did not differ in Ca,Mg,Na and available P levels.Soil K was significantly higher at the LD than at the HD plot.Soil Al and Fe were significantly higher at the control plot than at the HD.Soil Mn was significantly higher at the HD plot than at the control and LD plots.Soil Zn,Si and S did not differ among the three sites(Table 2).

Soil ratios

C/N ratio presented significantly higher values at the control plot than at the HD.There were no significant differences in SPAR,Ca:Al and Ca:Mg among the three sites(Table 3).

Multivariable analysis

The RDAs for the three sites are shown in Fig.2.The RDA explained a total of 98.7%of the variance:49.7%was explained by factor 1 and 49%by factor 2.The variables with the highest explanatory capacity were pH,Mn and SOM;the variables with the lowest explanatory capacity were Ca/Mg,Mg and SWR.In this principal component analysis(PCA),we found a cluster composed of Ca,Zn,EC,Fe,TN,SOM,C/N and Al.Another cluster was formed by IC,K,Si,Ca/Al and S(Fig.2).

Discussion

Soil physicochemical properties

Although the study sites were affected by a wildfire,almost 30 years later no differences are detected in SWR or AS between the three plots.Other studies have recorded lower long-term values in burned soils after a fire(Mataix-Solera et al.2008),but in our study area we did not observe any changes in these soil properties,with values returning to pre-fire levels just months after the fire(DeBano 2000).Post-fire management has been reported not to affect shortterm AS after a wildfire(Francos et al.2018b).Here,the fact that SWR and AS recovered their pre-fire values and were not affected by differences in forest management or time elapsed since the wildfire can be attributed to their high spatiotemporal variability.

Table 2 Descriptive statistics of major and minor elements characteristics in soils from three plots in northeastern Iberian Peninsula

That soil TN was significantly higher at the control site than at the HD site might be attributed to the fact that losses due to volatilization were more than offset by N fixation by vegetation.Given the absence of any differences with the LD site,the lower TN values can be explained by the higher vegetation density;in other words,the post-fire management almost 20 years after the fire at the LD site did not affect soil TN.

Soil organic matter was significantly higher at the control site than at the LD and HD sites,a finding in line with that of Johnson et al.(2012)who determined that,46 years after a fire,soil carbon had not yet reached pre-fire values.Rezaei and Gilkes(2005)observed similarly high values of SOM at sites with a predominant canopy cover.Their result is in line with our findings here,given that the HD site is abundant in herbaceous plants, shrubs and thin pines,which prevent an increase in SOM and TN.TN and SOM was found by Francos et al.(2018a)to need more time than other variables to return to their pre-fire values;thus,the effects of fire were enduring.The same conclusion was reached by Silvana-Longo et al.(2011),who reported that the effects of wildfire on SOM persist more than 10 years.This result may explain why in the present study values at the control site were significantly higher than at sites burned in 1986.A comparison of the low(LD)and high(HD)density sites reveals the same dynamics as described by Zhang et al.(2010)who compared secondary forest sites,with LD having higher TN and SOM than at HD sites,although the differences did not reach statistical significance.Rapid plant regrowth after fire allows C and N to be fixed and ensures that the decrease in these nutrients is not too great(Johnson and Curtis,2001).In the present study,due to rapid plant regrowth,long-term soil properties after the fire had values similar to those recorded for the pre-fire forest.However,this similarity was greater at the LD site than at the HD site.This latter difference can be attributed to the absence of forest management and the consequent high plant density at the HD site,which delays the restoration of soil properties to their pre-fire values.

Table 3 Descriptive statistics of soil ratios

Fig.2 RDA showing the relation between factors 1 and 2.Variables:soil water repellency(SWR),aggregate stability(AS),total nitrogen(TN), inorganic carbon (IC); soil organic matter (SOM); pH;electrical conductivity(EC);extractable Ca,Mg,Na,K,P,Al,Mn,Fe, Zn, Si, S, C/N; (extractable Na+extractable K)/(extractable C+extractable Mg)1/2(SPAR),Ca/Al and Ca/Mg.The three plots are control(unburned in 1986),low density and high density

The fact that IC levels were significantly higher at the LD site than at the other two is in line with studies that show the relation of this soil property with fire severity(Pereira et al.2012).Here,the C(organic and inorganic)balance was higher at the control and LD sites than at the HD.As for pre-fire conditions(Francos et al.2018c),plant regeneration and Ca and Mg dynamics are important for IC levels(Sainju et al.2007).Barbera et al.(2010)observed a similar dynamic to the one we describe here,where high plant density and root biomass led to the precipitation of IC compounds at lower layers.

Our findings in the case of soil pH,which was significantly higher at the LD than at HD plot and at the HD plot compared with the control,are in line with the positive correlation found by Zhang et al.(2010)between the number of years since an area had been abandoned and an increase in pH.In our study,the higher pH values at the LD are probably related to the site's lower SOM and TN values in agreement with the study of Francos et al.(2018c)in which high pH values were correlated with high SOM and TN values and low pH was correlated with low SOM and TN.Although Rezaei and Gilkes(2005)obtained significantly high EC values in areas with high vegetation density,EC value did not differ significantly among plots in our study probably because EC values are closely related to levels of certain soil nutrients,as explained below.

Major and minor soil elements

There were no significant changes in the extractable elements(Ca,Mg,Na and available P)with the exception of K,which was significantly higher at the LD than at the HD site.In our study,the absence of changes in EC is closely related to the similar values presented by Ca,Mg,Na and P among the three sites.In contrast,forest management after a disturbance(such as abandonment or a fire)can produce differences in soil K,with levels being higher in managed areas than in non-managed areas(Zhang et al.2010),as was the case here with the significantly higher K at the LD than at the HD site.Caon et al.(2014)concluded that,following a wildfire,there is likely to be a long-term decrease in nutrients as a result of volatilization and their transformation to recalcitrant forms,all as a result of postfire erosion.

Soil Al and Fe were significantly higher at the control than at the HD site.In this case,we cannot attribute the higher values at the control plot to the time since the wildfire because of the absence of any differences between the LD and control sites.As such,these changes can be attributed to the post-fire management and the consequent differences in vegetation density.Soil Fe is especially important for plant fertility,and low Fe levels can hinder recovery(Garc?′a-Marcos and Gonza′lez-Prieto 2008).Here,the high density resulting from the absence of any management led to a combined soil stress for Al and Fe.Unlike Fe,soil Mn was significantly higher at the HD site.This finding is in line with the negative correlation that Jones(2003)reported between Fe and Mn.Garc?′a-Marcos and Gonza′lez-Prieto(2008)found that higher values of Mn may be related to better seed germination and plant growth.In the present study,the higher values at the HD site seem to reflect the fact that higher plant density produces either extremely high or low values for elements,which in turn are indicative of problematic soils.Soil Zn,Si and S levels were similar at all three sites,and according to Johnson et al.(2012),pre-fire values can be restored in less than two decades,while long-term post-fire management does not significantly change soil properties,as appears to have occurred in the present study.One of the causes of the general increase in K,Al and Fe at the LD site may be the presence of wood residues on the topsoil,which may have triggered a long-term increase in these nutrients(Van Lear 1998).This long-term increase also seems to have an impact on C gains,with the result that the IC,SOM and C/N ratio at the LD site are more similar to those of the control than the HD site.

Soil ratios

Of the soil ratios,only the C/N ratio differed significantly,being significantly higher at the control than at the HD site,similar to the results obtained by Francos et al.(2018a)where the control area had a higher long-term(though not significant)ratio than in the burned areas after fire.In the present study,the statistically significant differences can be attributed to the significantly higher SOM values obtained at the control than at the HD site.The same conclusion was drawn by Johnson et al.(2012)who determined that almost 46 years after a wildfire,the burned areas had not reached their pre-fire values.One explanation could be the high resistance of organic material to microbial decomposition,so that the changes persisted over the long term after a wildfire(Santin et al.2016).Francos et al.(2018b)did not find any short-term C/N ratio differences after fire that were attributable to post-fire management.In this study,the differences in the ratio between the LD and HD sites are not significant,even though the values obtained at the LD are more similar to those recorded at the control.No significant differences were recorded in the SPAR,Ca/Al and Ca/Mg between the three sites.Changes in these ratios are confined to just a few months after a wildfire event and,in many cases,appear to be related to erosion,changes in soil structure or strong soil perturbations (Sarah 2005).Accordingly,the absence of differences in these ratios here seems to suggest that the ratios were restored to their prefire values and that long-term post-fire management does not alter soil properties.Soil Ca,Mg and Al were lower in the HD plot than in the control,and the difference for Al were statistically significant.The dynamics of these nutrients are similar to that of Ca/Al and Ca/Mg;the impact of fire on the two ratios is short term(Pereira et al.2017).More studies of the relationship between these ratios and other nutrients are needed to gain a better understanding of their dynamics.To the best of our knowledge,only a few studies(e.g.,Pereira et al.2017)have analyzed the effect of wildfire or post-fire management on soil Ca/Al and Ca/Mg.

Implications for forest management

The RDA explained a total of 98.7%of the variance:49.7%was explained by factor 1 and 49%by factor 2.In the PCA,we found one cluster composed of Ca,Zn,EC,Fe,TN,SOM,C/N and Al,grouped near the control.Another cluster was formed by IC,K,Si,Ca/Al and S(Fig.2)and was grouped close to the LD site.Only one variable,Mn,was close to the HD site.The control and LD sites cluster close to each other because their soil properties are more similar than they are with those at the HD site.The control and the LD sites explained a greater number of properties and a higher variance than the HD site.

Post-fire management may have negative effects on soil properties(Baldini et al.2007;Go′mez-Rey et al.2013).Here,as the post-fire management was carried out 20 years after the wildfire,the soil properties have had 10 years to improve.Indeed,the absence of management in these intervening years is one of the causes of higher vegetation density and consequent soil degradation.Post-fire management actions can have a marked impact on soil properties depending on the particular action,and postponing management for more than a year(Ginzburg and Steinberger 2012)or for longer after wildfire(Francos et al.2018a) has been recommended. Despite this, Mediterranean ecosystems are well adapted to fire,and in the longterm,their soil properties recover pre-fire values(Lo′pez-Poma et al.2014).In the long-term,forests usually regain a high vegetation density(as here,even with a management action),and the best time to carry out post-fire management may be when the direct and indirect effects of wildfire have disappeared(Silvana-Longo et al.2011).

Post-fire management treatments should not be carried out immediately after the fire to protect the seed bank(Madrigal et al.2011),but are best implemented in the medium to long-term(Francos et al.2016b).This strategy also lessen fire risk by reducing the accumulation of fuel and disrupting the fuel continuity,while at the same time reducing the overall costs of forest management(Francos et al.2016b).Almost 30 years after the fire,despite differences between the treated and untreated sites,there were no signs of post-fire erosion or soil degradation.This absence might be attributable to the rapid regrowth of vegetation after wildfire that characterizes Mediterranean ecosystems and protects soils(Cerda`and Doerr,2005).Forest management actions are necessary(and flagged as a priority at the HD site)to reduce the density of small trees and consequent fire risk(van Mantgem et al.2011).The fact that burned areas over the long-term after a fire have been shown to recover their pre-fire nutrient values is one of the best examples for promoting post-fire management,as for the LD site in our study area(Bento-Gonc?alves et al.2012).Post-fire forest management of this kind is not detrimental to the soil properties;on the contrary,the managed areas had values more similar to pre-fire values than in the nonmanaged control site. In addition, the managed areas present a lower fire risk due to the nonaccumulation of forest fuel and their lower vegetation density.The reduced plant density at the LD site in 2005 benefitted some soil properties and reduced root competition for water and nutrients(Castro et al.2011).In the same vein,Francos et al.(2018c)proposed forest management to prevent the outbreak of new wildfires, being able to maintain the effect of these actions for 10 years,the period of time that salvage logging remains effective(Donato et al.2006,2015).According to Francos et al.(2016b),18 years after a fire,the vertical and horizontal accumulation of fuel represents a high fire risk.Sustainable forest management strategies are necessary if we consider future climate change scenarios and associated changes in fire activity(Girardin et al.2013)using forest management tools to reduce fuel loads and restore historical fire conditions(Hunter et al.2011).

Conclusions

Our study on soil properties at two sites subjected to different post-fire management actions revealed that two qualities,SWR and AS,were notably unaffected.The absence of differences in these and other soil properties between the control and the two burned sites(LD and HD)allows us to conclude that the soils have completely recovered.More specifically,the LD and control sites have very similar soil properties,and are more similar than are the soils of the control and HD sites,on the one hand,and those of the HD and LD sites,on the other.The differences that emerged,however,between the low(LD)and high density(HD)sites can be directly attributed to the forest treatment in 2005 and the consequent differences in plant density.Significant differences were recorded between the LD and HD sites in the following soil properties:IC,pH,K and Mn,while non significant differences were recorded for TN,P,Al,Fe and C/N.All in all,our findings point to the positive effects of a management action conducted almost 20 years after a wildfire and the more harmful(though not dramatically so)effects of high plant density at the HD site,which reveal problems in the recovery of certain soil properties.Indeed,various authors advocate forest management to keep fire risk low and promote better soil properties.Further studies,however,are needed to establish when the differences between the soil properties of the LD and HD sites become significant.

AcknowledgementsThis study was supported by the POSTFIRE Project(CGL2013-47862-C2-1 and 2-R)and the POSTFIRE_CARE Project(CGL2016-75178-C2-2-R[AEI/FEDER,UE]),financed by the Spanish Research Agency(AIE)and the European Union through European Funding for Regional Development(FEDER)and the FPU Program(FPU 014/00037)of the Ministry of Education,Culture and Sports and Program 2014SGR825 and 2017SGR1344 of the Generalitat de Catalunya.We also thank our scientific and technical services for analyses of certain soil variables and for the English revision of this manuscript.We also thank the Diputacio′n de Barcelona for facilitating access to the study site to complete the fieldwork.