The importance of measuring fire severity—Evidence from microarthropod studies

Forest Ecology and Management 260 (2010) 62–70

Contents lists available at ScienceDirect

Forest Ecology and Management journal homepage: www.elsevier.com/locate/foreco

The importance of measuring fire severity—Evidence from microarthropod studies Anna Malmström ∗ Department of Ecology, Box 7044, SE-750 07 Uppsala, Sweden

a r t i c l e

i n f o

Article history: Received 15 January 2010 Received in revised form 23 March 2010 Accepted 5 April 2010 Keywords: Boreal forest Soil fauna Microarthropods Recovery Depth of burn

a b s t r a c t Fires are considered the most important disturbance regime in many ecosystems, including boreal forest. Fires usually reduce the abundances of soil living animals, but the duration of the fire effect and the recovery rate of soil fauna communities after fire are poorly understood. The species-rich group of microarthropods (collembolans, mites and proturans) constitutes an important part of the soil food-web, contributing to important ecosystem services like decomposition and nutrient mobilization. Recovery rates of microarthropods after fire reported in the literature differ considerable between sites and studies. Here, I examine if variation in fire severity can explain part of the variation in recovery of microarthropods after fire observed among studies. To do so, I have chosen studies done in boreal forests and in which the post-fire situation was described in such a way that fire severity (depth of burn) could be estimated. I also selected studies that used real abundance data and that sampled for animals for at least 2 years after fire. More severe fires were more determinal to soil fauna. Collembola (springtails) recovered within a few years at sites burnt with low severity, but the time frame in most studies (2–5 years) was too short to detect recovery at moderate or severely burnt sites. For mesostigmata and oribatida the recovery patterns were harder to interpret. I argue that fire severity is the most important factor explaining differences in microarthropod responses to fire, and that this is probably true also for other soil dwelling organisms. Because fire severity is often not taken into account when the effects of fire are investigated, generalizations about fire effects are hard to make. © 2010 Elsevier B.V. All rights reserved.

1. Introduction Fire is the most prominent large-scale disturbance regime in many of the world’s ecosystems including forests and grasslands (Liacos, 1977; Hobbs and Atkins, 1990; Hartnett, 1991; Jonson, 1992; Swaine, 1992; Pyne et al., 1996; Granström, 2001; Parr and Chown, 2003). In the boreal forests in northern Sweden fires used to occur with intervals of 50–200 years on average (Zackrisson, 1977; Engelmark, 1984), but due to the stochastic character of fire occurrence parts of the forest can escape major disturbances for long periods of time (Kuuluvainen, 2002; Gromtsev, 2002). Today, fire suppression in Sweden is so effective that the amount of burnt substrate in is only about 1% of the amount that was available in the natural state (Granström, 2001). This has lead to a decrease of species that are dependent on fire for their long-term survival either directly or indirectly by being favoured by the disturbance. In order to maintain these species prescribed burning for conser-

∗ Tel.: +46 18 672443; fax: +46 18 672890. E-mail address: [email protected]. 0378-1127/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.foreco.2010.04.001

vational reasons has become more common (Hörnsten et al., 1995; Granström, 2001). It is therefore important that we understand community responses to natural fires, as well as if these responses are similar to prescribed burns. Microarthropods (collembolans, mites and proturans) constitute an important part of the soil food-web, contributing to important ecosystem services like decomposition (Seastedt, 1984; Barrios, 2007) and nutrient mobilization (Heneghan and Bolger, 1998). Soil organisms make up a substantional part of the world’s biodiversity (Giller, 1996; Adams and Wall, 2000) and by their different feeding habits they can, directly or indirectly, influence the function of the primary decomposers such as fungi and bacteria (Heneghan and Bolger, 1996; Berg et al., 2001). Earlier studies have shown that fire usually reduces the total abundance of soil fauna in most ecosystems (Huhta et al., 1967; Metz and Ferrier, 1973; Metz and Dindal, 1975; Tamm, 1986; Koponen, 1995; Paquin and Coderre, 1997; Broza and Izhaki, 1997; McCullough et al., 1998; Wikars and Schimmel, 2001; Saint-Germain et al., 2005; Barratt et al., 2006; Buddle et al., 2006; Malmström et al., 2008, 2009; Kim and Jung, 2008). Fire destroys the preferred part of the soil habitat for most soil organisms, i.e. the litter and uppermost humus layer,

A. Malmström / Forest Ecology and Management 260 (2010) 62–70

or if the fire is very severe, the entire humus layer. More severe fires will destroy more of the habitat for soil- and litter living animals than less severe fires. This habitat destruction is likely to be an important factor affecting survival in the soil and recovery after a fire. Fire severity has indeed been shown to affect the numbers of various soil-living arthropods (Vlug and Borden, 1973; Radea and Arianoutsou, 2000; Wikars and Schimmel, 2001; Hening-Sever et al., 2001; Kim and Jung, 2008). The recovery rate of soil faunal communities after a fire are, however, poorly understood and recovery rates reported for soil microarthropods in the literature vary from just a couple of years to longer than the time period studied which is up to 5 years (Huhta et al., 1967; Koponen, 1995; Broza and Izhaki, 1997; Malmström et al., 2008, 2009; Kim and Jung, 2008). Often, only total abundances are studied, but total abundances have been shown to be a poor measure of recovery since total abundances often recovers faster than species diversity and species composition (Lindberg and Bengtsson, 2005). Also, different microarthropod groups have been shown to recover at different speed after disturbances (Lindberg and Bengtsson, 2005, 2006; Malmström et al., 2008) with a faster recovery of collembolans and mesostigmatid mites and a slower recovery of oribatid mites. Generally, the mortality of ground-living arthropods seems to be directly related to the combustion of litter and organic soil during the fire (Bellido, 1987) or the heat release during the fire (Malmström, 2008; Malmström et al., 2008). In spite of this, only a few studies have taken fire severity into account (Bellido, 1987; Hening-Sever et al., 2001; Wikars and Schimmel, 2001; Malmström et al., 2008; Kim and Jung, 2008) which makes the results hard to compare and interpret. If the burns are pre-planned the most straightforward way to measure fire severity is to measure the depth of burn, i.e. the destruction of the vegetation and humus layer. When this is not possible, like at wildfire sites, the fires needs to be classified in some other way. Here, I suggest the use of a classification system described by Ryan (2002) that classifies fires into severity classes based on the post-fire appearance of the site. This system also makes it possible to classify old fire studies if only the post-fire conditions are described. To measure recovery is usually not uncomplicated. When studying wildfires in forests this is not very difficult since a complete recovery is when the community is similar to the community in the unburnt forest. Many of the burns in the Swedish forests today are prescribed burns on clear-cuts. This is a more complicated situation because a fire applied on a clear-cut is a disturbance applied on an already disturbed site, which means that we are dealing with interacting disturbances. On those clear-cuts other disturbances like planting of trees and soil scarification can complicate the picture even more. Multiple disturbances are known to not always be merely additive but can act synergistically (Hobbs and Huenneke, 1992; Gagnon and Platt, 2008). In this study I wanted to study recovery of soil microarthropods after fires of different severity to determine if severity could explain the variation in recovery rates found in the literature. I also wanted to see if there was a difference between the recovery of total abundances and the recovery of community composition, and if the recovery differed between different soil faunal groups. My hypotheses were: 1. That recovery would be slower for more severe

63

fires. 2. That Collembola and Mesostigmata would recover faster than Oribatida, due to differences in life-history traits. 2. Materials and methods 2.1. Definitions used The lack of clear definitions of variables in fire ecology is a problem. For example, the terms fire intensity and fire severity are commonly mixed up in the literature. I have based on more recent literature used the following definitions: Fire intensity refers to the rate at which a fire is producing thermal energy and can be measured in terms of temperature and heat release (DeBano et al., 1998; Keeley, 2009). Fire severity is usually measured as depth of burn (DeBano et al., 1998) and can be measured as organic matter loss (Keeley, 2009). I have chosen to use depth of burn and destruction of the organic soil layer as a measure of fire severity. The most straightforward way to measure fire severity of a prescribed burn is to measure the thickness of the organic soil layer before and after fire to see how deep into the soil the fire burns. When wildfires are concerned, this is impossible to do since the fires are not planned in advanced. Instead, some measure of severity must be used that does not take information from before the fire into account. To classify the fires, I have used the classification categories suggested by Ryan (2002). This classification is easy to use even when actual measures of the humus thickness before and after fire are not possible, like at wildfire sites, and uses the post-fire conditions to sort the fires into categories. For a short and simplified description see Table 1. 2.2. Study areas To investigate microarthropod recovery in response to fire severity I have used data from four previously published studies and seven different areas (Huhta et al., 1967; Malmström, 2006; Malmström et al., 2008, 2009). I wanted to study effects of fire severity on the recovery process of soil microarthropods, and therefore I selected studies with data for a minimum of 2 years after fire. To make the studies comparable only studies that deal with real abundance data, i.e. that was sampled with soil samples and not e.g. pit fall traps were used. In order to reduce variation caused by other environmental factors only studies from boreal coniferous forests were included. I also needed the post-fire situation to be described in such a way that fire severity (depth of burn) could be estimated. Data on Collembola abundances were included in all studies found, at least total abundances. Mesostigmata were not distinguished from Prostigmata and Astigmata in Huhta et al. (1967) and was therefore not included in this study and data on Oribatida was not available from Tyresta LF and Tyresta SF (Malmström, 2006). 2.2.1. Hyytiälä The site is situated in Central Finland and was originally a pine stand of Vaccinium type mixed with spruce and birch. The stand was clear-cut in the beginning of 1961, and burned over on the 5th of June 1962. The size of the stand is not documented. In the fire, most aboveground parts of the ground vegetation were consumed,

Table 1 A short and simplified version of the classification system for fire severity originally proposed by Ryan (2002). The classification system uses changes in above ground vegetation and soil organic matter to determine fire severity.

Light Moderate Hard

Surface litter

Organic soil

Logs

Twigs/branches

Charred to consumed Consumed Consumed

Unaltered Deeply burned, to completely consumed Largely consumed

Blackened, not deeply charred Deeply charred Consumed

Larger branches remain Larger branches mostly consumed Consumed

64

A. Malmström / Forest Ecology and Management 260 (2010) 62–70

but recovery was rapid. By 21st of July 1962 Vaccinium vitis-idaea already occupied 25% of the ground. The forest had a moderately thick humus layer. After burning the deeper parts of the humus layer was hardly affected at all. Only one burnt plot was sampled (n = 1) and the site was sampled several times during the year. Each time 10 soil samples were taken with a soil borer (surface area 10 cm2 ). For a more detailed description of the site and the sampling see Huhta et al. (1967). As judged from the description of the effects of fire on the vegetation and humus layer the fire was classified as light according to Ryan (2002). No species identification was done. Only Collembola, Oribatida and “Other mites” were counted. 2.2.2. Kaltila The site is situated in Central Finland and was originally classified as a spruce stand of Myrtillus type that was clear-cut in winter 1961–1962. The size of the stand is not documented. Most of the felling residues and vegetation was consumed by the fire and only small plots of moss cover were left intact. The site was burnt-over on the 14th of May 1963, and planted with pine seedlings in 1964. The site had a thin humus layer that was almost entirely consumed during fire (see Huhta et al., 1967). The fire was classified as moderate. Only one burnt plot was sampled (n = 1) and the site was sampled several times during the year. Each time 10 soil samples were taken with a soil borer (surface area 10 cm2 ). No species identification was done. Only Collembola, Oribatida and “Other mites” were counted. 2.2.3. Pallasjärvi The site is situated in Northern Finland. It was originally a mixed stand of Pinus, Picea and Betula that was clear-cut in the beginning of 1961 and burnt over on 1st of July 1963. The size of the stand is not documented. The burning destroyed the field and ground layer and killed some of the trees (see Huhta et al., 1967). The fire was classified as moderate. Only one burnt plot was sampled (n = 1) and the site was sampled several times during the year. Each time 10 soil samples were taken with a soil borer (surface area 10 cm2 ). No species identification was done. Only Collembola, Oribatida and “Other mites” were counted. 2.2.4. Bjuråker The site is situated in central Sweden and was a 120-year-old forest consisting of mainly Scots pine Pinus sylvestris (L.) and Norway spruce, Picea abies (L.) Karst., before clear-felling. The central part of the clear-cut that was used for the burning experiment was 7 ha, divided into two blocks and four plots, two burnt (B) and two unburnt (C) (n = 2). The clear-cutting took place in March 1999, and half of the clear-cut was burnt in June the same year. The burning combusted practically all vegetation, harvest residue and litter. For a more detailed description see Malmström et al. (2009). The fire was classified as moderate. Soil samples were taken with a 10 cm × 10 cm steel frame (surface area 100 cm2 ) and at each sampling plot 6 samples were taken. The site was sampled once every year, except in 2002 when the autumn sampling was substituted with a spring sampling in 2003 due to heavy snowfall. All Collembola and Oribatida were identified to species level, for Mesostigmata normally only family or genus was used. 2.2.5. Tierp The site is situated in eastcentral Sweden and was burnt for conservational reasons by the forest company Stora Enso. The original tree stand consisted of 100–120-year-old Scots pine and Norway spruce. It was clear-cut in early 2001 and burnt over in May 2002 (see Malmström et al., 2008). The size of the burnt area was 32 ha. Here, 7 subplots were distributed over the burnt area (n = 7), originally to investigate the effects of fire intensity on soil

microarthropods. The fire was classified as a light fire. Soil samples were taken with a 10 cm × 10 cm steel frame (surface area 100 cm2 ) and at each sampling plot 6 samples were taken. The site was sampled once every year. All Collembola and Oribatida was identified to species level, for Mesostigmata normally only family or genus was used. 2.2.6. Tyresta Two wildfires, one big and one small, took place in, or close to Tyresta National Park and Nature Reserve situated 20 km south of Stockholm in central Sweden. The pine forest affected by fire was dominated by 150–300-year-old trees and the site had not previously been affected by large-scale forestry. 2.2.7. Tyresta big fire The fire started on August 1st 1999 and continued until August 10. The shallow organic soils on the bedrock were to a large extent burnt away, whereas the peat soils in moist depressions were reduced, but seldom completely burnt. The size of the burnt area was 450 ha, and only one burnt plot was sampled (n = 1). Soil samples were taken with a 10 cm × 10 cm steel frame (surface area 100 cm2 ) and at each sampling plot 5 samples were taken. The site was sampled once every year. For a more detailed description of the site and the sampling see Malmström (2006). The fire can be classified as a moderate to deep fire according to Ryan (2002). All Collembola was identified to species level, for Mesostigmata normally only family or genus was used. 2.2.8. Tyresta small fire The small wildfire occurred close to the big one in 2001. The fire was strong enough to burn away thin soil layers covered by mosses and lichens on rocky outcrops, but left the soil in the moist depressions almost intact. Sometimes also vegetation remained, only blackened by fire. 22 ha of forest were burnt. Here, two transects were laid out at the fire field, and thus two burnt sites was sampled within the fire field (n = 2). Soil samples were taken with a 10 cm × 10 cm steel frame (surface area 100 cm2 ) and at each sampling plot 5 samples were taken. The site was sampled once every year. For a description of the site and the sampling see Malmström (2006). The fire can be classified as light. All Collembola was identified to species level, for Mesostigmata normally only family or genus was used. 2.3. Data extraction and statistical analysis The total abundances for Collembola and Oribatida in Huhta et al. (1967) were found in the appendix. Huhta et al. (1967) sampled the sites many times during the year. For Hyytiäla and Kaltila I have chosen to take an average value of the autumn and winter samples (August to December) to make them as comparable as possible to the other investigations in the study. In Pallasjärvi the number of samplings were less numerous, and therefore an average value for the whole year was calculated. To test if the communities at a site had recovered or not I compared the value for each year after the fire with those before the fire, or in the cases where an unburnt control was available with this. I tested the hypothesis that there was a difference between burnt and unburnt plots with a t-test, assuming equal variances. At Bjuråker, Tierp and Tyresta SF there were replicated plots within fires (n = 2 for Bjuråker and Tyresta SF, n = 7 for Tierp) and the average between those were used in the analyses. At sites were no within fire replication was done (Hyytiälä, Kaltila, Pallasjärvi and Tyresta BF) I have used the within plot average in the analyses. The tests were done using Mini-tab 15 Statistical Software for Windows (2006).

A. Malmström / Forest Ecology and Management 260 (2010) 62–70

65

Table 2 Results from the different tests; t-test, return time and Anova used to examine if the different microarthropod groups (Collembola, Oribatida and Mesostigmata) recovered at the different sites. + indicates recovered and − indicates not recovered communities. For the t-tests p-values are presented. Anova are results found in previously published studies from sites. The definition of return time are found in Pimm (1991). Light burnt sites are Hyytiäla (n = 1), Tierp (n = 7), Tyrest SF (n = 2) and moderate burnt sites are Pallasjärvi (n = 1), Kaltila (n = 1), Bjuråker (n = 2) and Tyresta BF (n = 1). Animal group

Fire severity

Site

t-Test

Return time

ANOVA

Collembola

Light

Hyytiäla Tierp Tyresta SF Pallasjärvi Kaltila Bjuråker Tyresta BF

+, p = 0.492 +, p = 0.752 +, p = 0.699 −, p = 008 −, p = 0.000 +, p = 0.054 −, p = 0.001

+ + + − − − −

n.a. + + n.a. n.a. − −

Hyytiäla Tierp Pallasjärvi Kaltila Bjuråker

−, p = 0.000 −, p = 0.000 +, p = 0.107 −, p = 0.000 +, p = 0.770

n.a. n.a. n.a. n.a. n.a.

n.a. − n.a. n.a. +

Tierp Tyresta SF Bjuråker Tyresta BF

−, p = 0.026 +, p = 0.121 −, p = 0.026 −, p = 0.002

− + − −

+ + − −

Moderate

Oribatida

Light Moderate

Mesostigmata

Light Moderate

Pimm (1991) describes return time as the time it takes for the population’s displacement from equilibrium to decay to some specified fraction of its initial displacement. Convention sets this fraction at 1/e, or about 37% (Van Vuuren, 1999). As another measure of recovery I have therefore considered the abundances of the different animal groups as recovered when the displacement were less than 37% of the initial decay. Return times were not considered applicable for Oribatida since the double disturbance with clear-cut and burning on top were affecting them so much (see discussion part). In the cases were the sites were included in an already published study that had reported a test done with an ANOVA the results from this test was also included in Table 2. In those cases the microarthropod community was considered as recovered when there was no longer any significant difference between the burnt plots and the control. In those cases where species data were available (Tierp, Tyresta SF, Bjuråker 256 and Tyresta BF) species composition was studied using ordination technique. I used non-metric multidimensional scaling (NMS) from the PC-ORD software package (McCune and Mefford, 1999). NMS is an iterative ordination method based on ranked distances between sample units in the data matrix. It does not assume normally distributed data and is therefore suited for most ecological data. With NMS, the autopilot in PC-ORD selects the highest dimensionality that reduces the final “stress” by 5 or more. Stress is a measure (on a scale of 0–100) on the difference between the rank order of distances in the data matrix and the rank order of distances in the reduced-dimensional space of the ordination matrix (McCune and Mefford, 1999). I used the “slow and thorough” autopilot mode of NMS, with the Sørensen distance measure. 3. Results 3.1. Fire severity and abundance recovery At sites burnt with light fires total abundances of Collembolans (Fig. 1) recovered shortly after fire (Table 2), but at sites burnt with more severe fires no recovery could be detected after 5 years in Tyresta BF or after 3 years in Kaltila and Pallasjärvi (Table 2). In Bjuråker, there was no significant difference between treatments 5 years after fire when t-test was considered, but according to return time and ANOVA the abundance of Collembola had not recovered (Table 2). Oribatida (Fig. 2) did not recover according to t-test

or ANOVA (Table 2). At the more severely burnt sites Oribatida total abundances recovered in Pallasjärvi and Bjuråker 256 according to t-test and in Bjuråker 256 according to ANOVA (Table 2). Mesostigmatid mites (Fig. 3) recovered in Tyresta SF only according to t-test and return time. ANOVA showed recovery at both Tierp and Tyresta SF which were sites burnt with light severity (Table 2). 3.2. Species composition For both wildfire sites (Tyresta BF and Tyresta SF) no good NMS solution was found for any of the microarthropod groups and only 1-dimensional solutions was recommended. Therefore only results from Tierp and Bjuråker are shown here. At Tierp the Collembola NMS did not separate the burnt plots from the unburnt clear-cut, but in Bjuråker the species composition differed between burnt and unburnt clear-cut (Fig. 4). Species data for Oribatida was only available at two sites, Bjuråker and Tierp. At both sites the species composition was clearly different between burnt and unburnt plots (Fig. 5). Mesostigmata showed no difference between treatments in species composition in Tierp, but the burnt plots were clearly separated from the unburnt plots in Bjuråker (Fig. 6). 3.3. Variation in recovery between groups For Collembola the pattern was rather clear, with a recovery of both total abundances and species composition at sites burnt with less severe fires and no recovery at sites burnt with more severe fires. For Oribatida no clear pattern in recovery was seen. The recovery in Bjuråker was entirely due to one single species (Malmström et al., 2009) and could not be considered as a true recovery, which was also indicated by the NMS that showed no recovery in species composition (Fig. 5) at the site. Total abundances of Mesostigmata recovered in Tyresta SF but not in Tierp (Table 2). No recovery was seen for Mesostigmata when moderate fires were considered (Table 2). The NMS showed a recovery of species composition in the lightly burnt area of Tierp, but not in the moderately burnt Bjuråker (Fig. 6). It seems like Collembola and Mesostigmata are recovering within a few years at sites burnt with low severity, but recovery time at sites burnt with moderate severity is longer than the time period studied (up to 5 years) for all the groups considered in this study.

66

A. Malmström / Forest Ecology and Management 260 (2010) 62–70

Fig. 1. Recovery of total Collembola abundances at seven sites. Lightly burnt sites are Hyytiäla (n = 1), Tierp (n = 7), Tyresta SF (n = 2) and moderately burnt sites are Kaltila (n = 1), Pallasjärvi (n = 1) and Bjuråker (n = 2). n = shows within site replication, the average from those are used in the analyses and for getting the (S.E.). Error bars for n = 1 show within plot replication. Data from Hyytiälä, Kaltila and Pallasjärvi were found in Huhta et al. (1967), Tierp in Malmström et al. (2008), Tyresta SF and BF in Malmström (2006) and Bjuråker in Malmström et al. (2009).

Fig. 2. Recovery of total Oribatida abundances at five sites. Lightly burnt sites are Hyytiäla (n = 1) and Tierp (n = 7). Sites burnt with moderate severity are Pallasjärvi (n = 1), Kaltila (n = 1) and Bjuråker (n = 2). n = shows within site replication, the average from those are used in the analyses and for getting the (S.E.). Error bars for n = 1 show within plot replication. Data from Hyytiälä, Kaltila and Pallasjärvi were found in Huhta et al. (1967), Tierp in Malmström et al. (2008), and Bjuråker in Malmström et al. (2009).

A. Malmström / Forest Ecology and Management 260 (2010) 62–70

67

Fig. 3. Recovery of total Mesostigmata abundances at four sites. Lightly burnt sites are Tierp (n = 7) and Tyresta SF (n = 2), and moderately burnt sites are Bjuråker (n = 2) and Tyresta BF (n = 1). n = shows within site replication, the average from those are used in the analyses and for getting the (S.E.). Error bars for n = 1 show within plot replication. Data from Tierp were found in Malmström et al. (2008), Tyresta SF and BF in Malmström (2006) and Bjuråker in Malmström et al. (2009).

4. Discussion In this study I have shown that recovery of soil fauna in boreal forests after fire to a large extent depends on fire severity, which

has seldomly been measured in previous studies. It is clear that fire reduces the abundances of soil fauna in general (e.g. Huhta et al., 1967; Muona and Rutanen, 1994; Saint-Germain et al., 2005; Buddle et al., 2006; Malmström et al., 2009). At least for Collem-

Fig. 4. Species composition of Collembola. Non-metric multidimensional scaling (NMS) was used to compare burnt treatment (() with unburnt () at sites burnt with different severity. The sites are Tierp (n = 7) that are lightly burnt and Bjuråker (n = 2) that are burnt with moderate severity. C is the control plots, B is the burnt plots. B1 is 1 year after fire, B2 2 years after fire, etc. Tierp: the final solution was obtained from 63 iterations and a 2-dimensional solution was recommended. The final stress was 10.5, which was significantly lower than 250 randomized runs (p < 0.05 for the Monte Carlo test). Bjuråker: the final solution was obtained from 48 iterations and a 2-dimensional solution was recommended. The final stress was 10.7, which was significantly lower than 250 randomized runs (p < 0.05 for the Monte Carlo test).

Fig. 5. Species composition of Oribatida. Non-metric multidimensional scaling (NMS) was used to compare burnt treatment (() with unburnt () at sites burnt with different severity. The sites are Tierp (n = 7) that are lightly burnt and Bjuråker (n = 2) that are burnt with moderate severity. C is the control plots, B is the burnt plots. B1 is 1 year after fire, B2 2 years after fire, etc. Tierp: the final solution was obtained from 500 iterations and a 3-dimensional solution was recommended. The final stress was 10.7, which was significantly lower than 250 randomized runs (p < 0.05 for the Monte Carlo test). Bjuråker: the final solution was obtained from 158 iterations and a 2-dimensional solution was recommended. The final stress was 7.3, which was significantly lower than 250 randomized runs (p < 0.05 for the Monte Carlo test).

68

A. Malmström / Forest Ecology and Management 260 (2010) 62–70

Fig. 6. Species composition of Mesostigmata. Non-metric multidimensional scaling (NMS) was used to compare burnt treatment () with unburnt () at sites burnt with different severity. The sites are Tierp (n = 7) that are lightly burnt and Bjuråker (n = 2) that are burnt with moderate severity. C is the control plots, B is the burnt plots. B1 is 1 year after fire, B2 2 years after fire, etc. Tierp: the final solution was obtained from 118 iterations and a 2-dimensional solution was recommended. The final stress was 10.2, which was significantly lower than 250 randomized runs (p < 0.05 for the Monte Carlo test). Bjuråker: the final solution was obtained from 500 iterations and a 2-dimensional solution was recommended. The final stress was 7.0, which was significantly lower than 250 randomized runs (p < 0.05 for the Monte Carlo test).

bola and Mesostigmata, fire severity seems to explain the variation in recovery time. Even so, many papers discussing fire effects on litter- and/or humus-dwelling organisms are not taking fire severity into account (e.g. Heyward and Tissot, 1936; Metz and Dindal, 1975; Webb, 1994; Kiss and Magnin, 2003, 2006; Baker et al., 2004; Driessen and Greenslade, 2004; Larrivée et al., 2005; Barratt et al., 2006; Buddle et al., 2006). This is a major drawback that makes generalizations about longer-term fire effects on ground- and soilliving arthropods difficult. Since most of the studies in this paper are based on prescribed burns on clear-cuts the results are influenced by clear-cut effects. In Bjuråker, the burning effect could be separated from the clearcut effect because the experimental design included an unburnt clear-cut. This was not the case in Tierp, where the whole clear-cut was burnt by a forest company. In Hyytiälä, Kaltila and Pallasjärvi no unburnt clear-cut was left, but clear-cut effects of microarthropods were studied simultaneously at different sites (Huhta et al., 1967). No, or small, differences between clear-cut and managed forest are normally reported for total abundances of Collembola and Mesostigmata (Huhta et al., 1967; Bird and Chatarpaul, 1986; Blair and Crossley, 1988; Bengtsson et al., 1997; Siira-Pietikäinen et al., 2001; Malmström et al., 2009), while a shorter or longer decline in the total number of Oribatida seems to be the general rule (Huhta et al., 1967; Vlug and Borden, 1973; Huhta, 1976; Abbott et al., 1980; Bird and Chatarpaul, 1986; Blair and Crossley, 1988; Marshall, 2000; Lindo and Visser, 2004; Malmström et al., 2009). The dramatic reduction in oribatid mites in the burnt clear-cut is

probably a combination of clear-cut and burning effects. Because the recovery values reported here are compared to either unburnt clear-cut or before burning values on clear-cut, the clear-cut effect is incorporated in both the treatment and the control. This is why I decided not to use the return time measure for oribatid mites, since the return to the clear-cut values is still a return to a disturbed situation. The effects of multiple disturbances are not well understood, but it has been shown that the effects of two disturbances often are not merely additive, but can act synergistically (Hobbs and Huenneke, 1992; Gagnon and Platt, 2008). Concerning species composition an effect of double disturbances could probably affect the results also for other microarthropod groups, like Collembola. The species composition between a clear-cut and a forest can differ for Collembola even if the total abundances do not differ (Malmström, unpublished data). Oribatida seems to recover at a slower pace after disturbances than Collembola and mesostigamatid mites (Lindberg and Bengtsson, 2006; Malmström et al., 2008). Here, the possibly larger effect of interacting disturbances on oribatid mites can also influence the recovery patterns. Oribatid mites are thought to have more “K-selected” traits than Collembolans and Mesostigmatid mites (Norton, 1994), but Lindberg and Bengtsson (2005) showed that also traits like habitat specialization and dispersal ability are important for recovery rates. The size of the burnt area could therefore be of importance for recolonisation rates. Also, most fire fields are heterogenous, and recolonisation can occur from light-burnt and unburnt plots within the fire-field. Unfortunately, data was not available on heterogeneity in any of the studies used here, and the size of the fire field was not reported by Huhta et al. (1967), and therefore no further analyses was done on this issue. It is not uncommon that only the total numbers of mites or Acarina are calculated in microarthropod studies (Heyward and Tissot, 1936; Vlug and Borden, 1973; Metz and Ferrier, 1973; Koponen, 1995; Radea and Arianoutsou, 2000), or that Mesostigmata are combined with Prostigmata and Astigmata as “other mites” (e.g. Huhta et al., 1967). Since the different groups of mites can react very differently to a disturbance, and have different recovery rates (Table 2, Fig. 2 vs. Fig. 3) the reactions of the different groups are masked this way. A higher resolution is needed to be able to understand microarthropod responses to fire. Preferably identification to species level is recommended, and at the minimum distinguishing between different mite groups is needed since they respond very differently to disturbances (Lindberg and Bengtsson, 2005; Malmström et al., 2009). Total abundances has been shown to be a poor measure of recovery after disturbances (Lindberg and Bengtsson, 2006), returning to control values much faster than species richness and diversity indices. In Bjuråker Oribatida showed an almost complete recovery in total abundance, but the recovery could be explained by an explosive increase in one single species (Malmström et al., 2009). This emphasizes the importance of studying total abundances and population responses of individual species in combination. Abundances of microarthropods are known to vary considerable both between seasons (Takeda, 1987; Persson and Lohm, 1977) and years (Huhta et al., 1967; Takeda, 1987; Wolters, 1998) although the community structure seem rather stable (Takeda, 1987; Bengtsson, 1994; Kampichler and Geissen, 2005). This is a potential problem when using “before” data to study recovery as I do in this study at Pallasjävi, Hytillä, Kaltila and Tierp. However, Huhta (1967) claims that the years 1963–1965 were quite stable in microarthropod numbers, increasing the usefulness of the data. Another potential problem with the present review is that the studies used all suffer from poor replication. All sites in this study except Tierp (n = 7) suffer from the fact that they are un-replicated or replicated with very low replication (n = 2). This results in a low power to detect differences between “treatments” resulting in type

A. Malmström / Forest Ecology and Management 260 (2010) 62–70

II errors, i.e. not finding a difference when there in fact is one, and interpreting this as recovery. The low within site replication in the studies also made a meta-analysis approach impossible. Metaanalyses use standard deviations in the analyses and therefore at least some within site replication is needed for this approach to work (Cooper and Hedges, 1994). The poor replication is unfortunately a common problem with studies dealing with forest fires. This problem has also been discussed by Parr and Chown (2003) in a paper concerning problems with fire studies undertaken in Southern Africa. Owing to lack of funding, logistic constrains, and/or the heterogeneity of ecosystems, large-scale natural experiments undertaken in natural landscapes are almost impossible to replicate adequately (Diamond, 1986; Carpenter, 1990; Whelan, 1995; Parr and Chown, 2003). Another drawback of natural experiments using e.g., natural wildfires, is the low degree of control over fire severity and/or intensity and replication (see Parr and Chown, 2003). Insufficient funding has the consequence that scientists often need to be opportunistic and use experiments planned for other reasons (e.g. Bjuråker), or sites burned for other reasons. For example, in Tierp, the clear-cut was burnt by a forest company as part of their certification program. This resulted in an experimental set up that lacked unburnt clear-cut as a control site, but as the burning was pre-planned it was possible to take samples before the burning, something that is impossible at wildfire sites. Parr and Chown (2003) suggest the use of non-classical statistics to explore the effects of large-scale or un-replicated fires, like Bayesian methods. Since large wildfires are rare events in the boreal forest today (Granström, 2001) studies on the effects of wildfires are extremely hard to replicate at other sites. How to do good within site replication at a single wildfire site avoiding pseudoreplication is not obvious and should be given more consideration. One of the reasons why data from real wildfire sites are scarce in the literature is probably because un-replicated results are often considered unpublishable. Frequently, replicated fire experiments are undertaken at too small a scale, which limits the applicability of the results (Parr and Chown, 2003). Large-scale experiments that enable the study of fire effects on landscape-level patterning, and provide a different view of ecosystem structuring and functioning to smaller scale studies, are however, often not well replicated (Parr and Chown, 2003). 4.1. Conclusions and implications Using microarthropods as an example, I show that fire severity can have large effects on soil faunal recovery after fires in the boreal forest. The conclusions are also applicable on other soil living organisms. The major factor that influences both survival after fire and recovery rates seem to be depth of burn (i.e. fire severity). Still, this factor is often not considered when fire studies are made. To make better and more robust general conclusions regarding fire effects, fire severity has to be incorporated in future studies. In addition, to get a better general understanding of fire effects future studies on the topic should be better replicated. If not possible, e.g. due to the stochastic nature of fires, one should at least strive towards adequate within site replication. In this way a meta-analysis may be possible in the future, which may facilitate the possibility to make generalizations. The time span of the available studies (2–5 years after fire) is too short to observe any recovery for moderate fires. More long-term studies are needed to understand the soil fauna dynamics following fire disturbance in boreal forest. Acknowledgements I am very grateful to Janne Bengtsson for fruitful discussions and good advice concerning this manuscript. I am also thankful

69

to Joachim Strengbom and two anonymous reviewers for valuable comments on an earlier version of this manuscript. References Abbott, D.T., Seastedt, T.R., Crossley, D.A., 1980. The abundance, distribution and effects of clear-cutting on Cryptostigmata in the southern Appalachians. Environ. Entomol. 9, 618–623. Adams, G.A., Wall, D.H., 2000. Biodiversity above and below the surface of soils and sediments: linkages and implications for global change. BioScience 50, 1043–1048. Baker, S.C., Richardson, A.M.M., Seeman, O.D., Barmuta, L.A., 2004. Does clearfell, burn and sow silviculture mimic the effect of wildfire? A field study and review using litter beetles. For. Ecol. Manage. 199, 433–448. Barratt, B.I.P., Tozer, P.A., Wiedemer, R.L., Ferguson, C.M., Johnstone, P.D., 2006. Effect of fire on microarthropods in New Zealand indigenous grassland. Rangeland Ecol. Manage. 59, 383–391. Barrios, E., 2007. Soil biota, ecosystem services and land productivity. Ecol. Econ. 64, 269–285. Bellido, A., 1987. Field experiment about direct effect of a heatland prescribed fire on microarthropod community. Rev. Ecol. Biol. Sol. 24, 603–622. Bengtsson, J., 1994. Temporal predictability in forest soil communities. J. Anim. Ecol. 63, 653–665. Bengtsson, J., Persson, T., Lundqvist, L., 1997. Long-term effects of logging residue addition and removal on macroarthropods and enchytraeids. J. Appl. Ecol. 34, 1014–1022. Berg, M., De Ruiter, P., Didden, W., Janssen, M., Schouten, T., Verhoef, H., 2001. Community food web, decomposition and nitrogen mineralisation in a stratified Scots pine forest soil. Oikos 94, 130–142. Bird, G.A., Chatarpaul, L., 1986. Effect of whole-tree and conventional forest harvest on soil microarthropods. Can. J. Zool. 64, 1986–1993. Blair, J.M., Crossley, D.A., 1988. Litter decomposition, nitrogen dynamics and litter microarthropods in a southern Appalachian hardwood forest 8 years following clearcutting. J. Appl. Ecol. 25, 683–698. Broza, M., Izhaki, I., 1997. Post-fire arthropod assemblages in Mediterranean forest soils in Israel. Int. J. Wildland Fire 7, 317–325. Buddle, C.M., Langor, D.W., Pohl, G.R., Spence, J.R., 2006. Arthropod responses to harvesting and wildfire: implications for emulation of natural disturbance in forest management. Biol. Conserv. 128, 346–357. Carpenter, S.R., 1990. Large-scale perturbations: opportunities for innovation. Ecology 71, 2038–2043. Cooper, H., Hedges, L.V., 1994. The Handbook of Research Synthesis. Russel Sage Foundation, New York, NY. DeBano, L.F., Neary, D.G., Ffolliot, P.F., 1998. Fire’s Effects on Ecosystems. John Wiley and Sons, New York, USA. Diamond, J., 1986. Overview: laboratory experiments, field experiments, and natural experiments. In: Diamond, J., Case, T.J. (Eds.), Community Ecology. Harper and Row, New York, New York, USA, pp. 3–22. Driessen, M.M., Greenslade, P., 2004. Effect of season, location and fire on Collembola communities in buttongrass moorlands, Tasmania. Pedobiologia 48, 631– 642. Engelmark, O., 1984. Forest fires in Muddus National Park (northern Sweden) during the past 600 years. Can. J. Bot. 62, 893–898. Gagnon, P.R., Platt, W.J., 2008. Multiple disturbances accelerate clonal growth in a potentially monodominant bamboo. Ecology 89, 612–618. Giller, P.S., 1996. The diversity of soil communities, the ‘poor man’s tropical rainforest’. Biodivers. Conserv. 5, 135–168. Granström, A., 2001. Fire management for biodiversity in the European boreal forest. Scand. J. Forest Res. 16, 62–69. Gromtsev, A., 2002. Natural disturbance dynamics in the boreal forests of European Russia: a review. Silva Fenn. 36, 41–55. Hartnett, D.C., 1991. Effects of fire in tallgrass prairie on growth and reproduction of prairie coneflower (Ratibibda columnifera: Asteraceae). Am. J. Bot. 78, 429– 435. Heneghan, L., Bolger, T., 1996. Effects of component of “acid” rain on soil microarthropods’ contribution to ecosystem function. J. Appl. Ecol. 33, 1329–1344. Heneghan, L., Bolger, T., 1998. Soil microarthropod contribution to forest ecosystem processes: the importance of observational scale. Plant Soil 205, 113–124. Hening-Sever, N., Poliakov, D., Brozoa, M., 2001. A novel method for estimation of wildfire intensity based on ash pH and soil microarthropod community. Pedobiologia 45, 98–106. Heyward, F., Tissot, A.N., 1936. Some changes in the soil fauna associated with forest fire in the longleaf pine region. Ecology 17, 659–666. Hobbs, R.J., Atkins, I., 1990. Fire-related dynamics of a Banksia woodland in Southwestern Australia. Aust. J. Bot. 38, 97–110. Hobbs, R.J., Huenneke, L.F., 1992. Disturbance, diversity, and invasion: implications for conservation. Conserv. Biol. 6, 324–337. Hörnsten, L., Nohlgren, E., Aldentun, Y., 1995. Brand och bränning – en litteraturstudie. SkogForsk, Redogörelse nr 9, 36 pp (in Swedish). Huhta, V., 1976. Effects of clear-cutting on numbers, biomass and community respiration of soil invertebrates. Ann. Zool. Fenn. 13, 63–80. Huhta, V., Karppinen, E., Nurminen, M., Valpas, A., 1967. Effect of silvicultural practices upon arthropod, annelid and nematode populations in coniferous forest soil. Ann. Zool. Fenn. 4, 87–145.

70

A. Malmström / Forest Ecology and Management 260 (2010) 62–70

Jonson, E.A., 1992. Fire and Vegetation Dynamics: Studies from the North American Boreal Forest. Cambridge University Press, Cambridge. Kampichler, C., Geissen, V., 2005. Temporal predictability of soil microarthropod communities in temperate forests. Pedobiologia 49, 41–50. Keeley, J.E., 2009. Fire intensity, fire severity and burn severity: a brief review and suggested usage. Int. J. Wildland Fire 18, 116–126. Kim, J.W., Jung, C., 2008. Abundance of soil microarthropods associated with forest fire severity in Samcheok, Korea. J. Asia Pacific Entomol. 11, 77–81. Kiss, L., Magnin, F., 2003. The impact of fire on some Mediterranean land snail communities and patterns of post-fire recolonization. J. Mollus. Stud. 69, 43–53. Kiss, L., Magnin, F., 2006. High resilience of Mediterranean land snail communities after fire. Biodivers. Conserv. 15, 2925–2944. Koponen, S., 1995. Postfire succession of soil arthropod groups in a subarctic birch forest. Acta Zool. Fennici 196, 243–245. Kuuluvainen, T., 2002. Natural variability of forests as a reference for restoring and managing biological diversity in boreal Fennoscandia. Silva Fenn. 36, 97–125. Larrivée, M., Fahrig, L., Drapeau, P., 2005. Effects of a recent wildfire and clearcuts on ground-dwelling boreal forest spider assemblages. Can. J. For. Res. 35, 2575–2588. Liacos, L.G., 1977. Fire and fuel management in pine forest and evergreen brushland ecosystems in Greece. In: Mooney, H.A., Conrad, C.E. (Technical coordinators), Proceedings of the Symposium on the Environmental Consequences of Fire and Fuel Management in Mediterranean Ecosystems. USDA For. Serv. Gen. Tech. Rep. WO-3, pp. 289–298. Lindberg, N., Bengtsson, J., 2005. Population responses of oribatid mites and collembolans after drought. Appl. Soil Ecol. 28, 163–174. Lindberg, N., Bengtsson, J., 2006. Recovery of forest soil fauna diversity and composition after repeated summer droughts. Oikos 114, 494–506. Lindo, Z., Visser, S., 2004. Forest floor microarthropod abundance and oribatid mite (Acari: Oribatida) composition following partial and clear-cut harvesting in the mixedwood boreal forest. Can. J. For. Res. 34, 998–1006. Malmström, A., 2006. Effects of wildfire and prescribed burning on soil fauna in boreal coniferous forests. Ph.D. thesis, Faculty of Natural Resources and Agricultural Sciences, Acta Universitatis Agriculturae Sueciae, Uppsala, Sweden. Malmström, A., 2008. Temperature tolerance in soil microarthropods: simulation of forest-fire heating in the laboratory. Pedobiologia 51, 419–426. Malmström, A., Persson, T., Ahlström, K., 2008. Effects of fire intensity on survival and recovery of soil microarthropods after a clearcut burning. Can. J. For. Res. 38, 2465–2475. Malmström, A., Persson, T., Ahlström, K., Gongalsky, K.B., Bengtsson, J., 2009. Dynamics of soil meso-and macrofauna during a 5-year period after clear-cut burning in a boreal forest. Appl. Soil Ecol. 43, 61–74. Marshall, V.G., 2000. Impacts of forest harvesting on biological processes in northern forest soils. For. Ecol. Manage. 133, 43–60. McCullough, D.G., Werner, R.A., Neumann, D., 1998. Fire and insects in northern and boreal forest ecosystems of North America. Annu. Rev. Entomol. 43, 107–127. McCune, B., Mefford, M.J., 1999. PC-ORD. Multivariate Analysis of Ecological Data. Version 5. MjM Software, Gleneden Beach, Oregon, USA. Metz, L.J., Dindal, D.L., 1975. Collembola populations and prescribed burning. Environ. Entomol. 4, 583–587.

Metz, L.J., Ferrier, M.H., 1973. Prescribed burning and populations of soil mesofauna. Environ. Entomol. 2, 433–440. Muona, J., Rutanen, I., 1994. The short-term impact of fire on the beetle fauna in boreal forest. Ann. Zool. Fenn. 31, 109–121. Norton, R.A., 1994. Evolutionary aspects of oribatid mite life histories and consequences for the origin of the Astigmata. In: Houck, M. (Ed.), Mites. Ecological and Evolutionary Analyses of Life-History Patterns. Chapman and Hall, New York, pp. 99–135. Paquin, P., Coderre, D., 1997. Deforestation and fire impact on edaphic insect larvae and other macroarthropods. Environ. Entomol. 26, 21–30. Parr, C.L., Chown, S., 2003. Burning issues for conservation: a critique of faunal fire research in Southern Africa. Austral Ecol. 28, 384–395. Persson, T., Lohm, U., 1977. Energetical significance of the annelids and arthropods in a Swedish grassland soil. Ecol. Bull. 23, 211. Pimm, S.L., 1991. The Balance of Nature? Ecological Issues in the Conservation of Species and Communities. The University of Chicago Press, Chigaco. Pyne, S.J., Andrews, P.L., Laven, R.D., 1996. Introduction to Wildland Fire. John Wiley & Sons, New York. Radea, C., Arianoutsou, M., 2000. Cellulose decomposition rates and soil arthropod community in a Pinus halepensis Mill. Forest of Greece after wildfire. Eur. J. Soil Biol. 36, 57–64. Ryan, K.C., 2002. Dynamic interactions between forest structure and fire behavior in boreal ecosystems. Silva Fenn. 36, 13–39. Saint-Germain, M., Larrivée, M., Drapeau, P., Fahrig, L., Buddle, C.M., 2005. Shortterm response of ground beetles (Coleoptera: Carabidae) to fire and logging in a spruce-dominated boreal landscape. For. Ecol. Manage. 212, 118–126. Seastedt, T.R., 1984. The role of microarthropods in decomposition and mineralization processes. Annu. Rev. Entomol. 29, 25–46. Siira-Pietikäinen, A., Pietikäinen, J., Fritze, H., Haimi, J., 2001. Short-term responses of soil decomposer communities to forest management: clear felling versus alternative harvesting methods. Can. J. For. Res. 31, 88–99. Swaine, M.D., 1992. Characteristics of dry forests in West Africa and the influence of fire. J. Veg. Sci. 3, 365–374. Tamm, J.C., 1986. Fünfjährige Collembolensukzession auf einem verbrannten Kiefernwaldboden in Niedersachsen (BRD). Pedobiologia 29, 113–127. Takeda, H., 1987. Dynamics and maintainance of Collembolan community structure in a forest soil system. Res. Popul. Ecol. 29, 291–346. Van Vuuren, J.H., 1999. Resilience in reaction-diffusion systems. IMA J. Appl. Math. 63, 179–197. Vlug, H., Borden, J.H., 1973. Soil Acari and Collembola populations affected by logging and slash burning in a coastal British Colombia coniferous forest. Environ. Entomol. 2, 1016–1023. Webb, N.R., 1994. Post-fire succession of cryptostigmatid mites (Acari, Cryptostigmata) in a Calluna-heatland soil. Pedobiologia 38, 138–145. Whelan, R.J., 1995. The Ecology of Fire. Cambridge University Press, Cambridge. Wikars, L.-O., Schimmel, J., 2001. Immediate effects of fire-severity on soil invertebrates in cut and uncut pine forests. For. Ecol. Manage. 141, 189–200. Wolters, V., 1998. Long-term dynamics of a collembolan community. Appl. Soil Ecol. 9, 221–227. Zackrisson, O., 1977. Influence of forest fires on the North Swedish boreal forest. Oikos 29, 22–32.