REPORT
Land use influences macroinvertebrate community composition
in boreal headwaters through altered stream conditions
Micael Jonsson, Ryan M. Burrows, Johan Lidman,
Emma Fa
¨
ltstro
¨
m, Hjalmar Laudon, Ryan A. Sponseller
Received: 22 February 2016 / Revised: 2 July 2016 / Accepted: 4 October 2016 / Published online: 1 November 2016
Abstract Land use is known to alter the nature of land–
water interactions, but the potential effects of widespread
forest management on headwaters in boreal regions remain
poorly understood. We evaluated the importance of
catchment land use, land cover, and local stream variables
for macroinvertebrate community and functional trait
diversity in 18 boreal headwater streams. Variation in
macroinvertebrate metrics was often best explained by in-
stream variables, primarily water chemistry (e.g. pH).
However, variation in stream variables was, in turn,
significantly associated with catchment-scale forestry land
use. More specifically, streams running through catchments
that were dominated by young (11–50 years) forests had
higher pH, greater organic matter standing stock, higher
abundance of aquatic moss, and the highest macro-
invertebrate diversity, compared to streams running through
recently clear-cut and old forests. This indicates that
catchment-scale forest management can modify in-stream
habitat conditions with effects on stream macroinvertebrate
communities and that characteristics of younger forests may
promote conditions that benefit headwater biodiversity.
Keywords Aquatic insects Biodiversity Forestry
Functional traits
INTRODUCTION
Headwater streams often account for the majority of net-
work length, making them an important lotic habitat
(Clarke et al. 2008). These small streams represent the
primary interface between terrestrial and aquatic environ-
ments (Lowe and Likens 2005) and support key ecosystem
processes, such as litter decomposition (Bilby and Likens
1980; Wallace et al. 1997) and nutrient retention (Bern-
hardt et al. 2005), that are crucial for the functioning of
downstream lentic and lotic systems (Meyer and Wallace
2001). Further, headwater streams may house diverse
species assemblages that are not only functionally impor-
tant but also contribute to local and regional biodiversity
(Finn et al. 2011). However, changed environmental con-
ditions may lead to the loss of headwater species, altered
community composition (Lowe and Likens 2005), and
homogenization of communities resulting in reduced
regional biodiversity (Meyer et al. 2007), with potential
consequences for the functioning of these habitats (Vaughn
2010).
In boreal Sweden, headwater streams (draining catch-
ments\1500 ha) represent more than 90 % of the total
drainage length, yet remain poorly represented in nation-
wide monitoring and assessment programs (Bishop et al.
2008). Due to strong seasonal climate variability, these
streams tend to be vulnerable to drought, bottom freezing,
and floods (Malmqvist et al. 1999; Hoffsten 2003),
requiring species to be adapted to highly dynamic hydro-
logical conditions. Additionally, northern boreal headwa-
ters are typically humic and naturally acidic (Laudon and
Buffam 2008), nutrient poor (Bergstro
¨
m et al. 2008), and
often shaded by dense, coniferous riparian vegetation
(Naiman et al. 1987). In turn, these conditions regulate
organic and inorganic resource availability and quality to
macroinvertebrate consumers, through the input of rela-
tively low-quality litter (Naiman et al. 1987), light and
nutrient limitation of autotrophic production (Kiffney et al.
2004), and nutrient limitation of microbes (Burrows et al.
2015). Boreal headwater streams therefore represent rather
Electronic supplementary material The online version of this
article (doi:10.1007/s13280-016-0837-y) contains supplementary
material, which is available to authorized users.
Ó The Author(s) 2016. This article is published with open access at Springerlink.com
www.kva.se/en
123
Ambio 2017, 46:311–323
DOI 10.1007/s13280-016-0837-y
unique combinations of harsh and limiting environmental
conditions that likely constrain the productivity and rich-
ness of benthic communities (Annala et al. 2014).
Previous research aimed at understanding the factors
controlling macroinvertebrate community composition in
boreal streams has found combinations of several envi-
ronmental and habitat variables to be important. For
example, latitude, longitude, pH, and stream characteristics
such as water velocity, width, and depth are often impor-
tant determinants of macroinvertebrate community struc-
ture (Heino et al. 2003, 2014; Schmera et al. 2013).
Moreover, variation in substrate composition (Heino et al.
2014) and concentrations of nutrients and dissolved organic
carbon (DOC) (Go
¨
the et al. 2014) may also drive patterns
in benthic community composition.
Several studies from temperate regions show that exter-
nal factors, such as riparian canopy openness, abundance of
deciduous streamside vegetation, and catchment-scale land
use affect stream habitats and communities (Allan 2004). In
this context, research on boreal headwaters is underrepre-
sented (but see Schmera et al. 2013; Heino et al. 2014).
Given the strong reliance of headwater stream macroin-
vertebrates on terrestrial resources (Vannote et al. 1980;
Webster and Benfield 1986; Richardson and Danehy 2007),
any alterations to the terrestrial environment that result in
quantitative or qualitative changes in allochthonous organic
matter (OM) input, or levels of in-stream primary produc-
tion (e.g. via increased canopy openness and/or nutrient
inputs), may affect macroinvertebrate communities.
In addition to affecting the richness of stream assem-
blages, catchment properties also shape the functioning of
these communities through effects on the diversity of
species traits represented locally. Indeed, it is increasingly
clear that the analysis of species traits adds additional
insight to our understanding of how stream communities
respond to environmental pressures and change (e.g. Poff
et al. 2006). Knowing which functional traits are present in
a community (and their relative abundance), and how the
relative abundance of traits may change due to external
influences, leads to a better understanding, and thus pre-
dictive ability, of how ecosystem functioning might be
altered following changed environmental conditions (Poff
1997; Bonada et al. 2007). To enable predictions of how
changed community composition affects ecosystem func-
tioning, it is important to unravel drivers of those traits that
are directly linked to the maintenance of ecosystem pro-
cesses (e.g. filter feeders—filtration rate). Several previous
studies have shown that ecosystem process rates and,
hence, functioning can be related to species diversity
(Vaughn 2010). However, functional traits are often shared
among sets of species, and the occurrence of specific traits
in a community may remain unchanged despite species
losses or gains, due to functional redundancy among
species (Rosenfeld 2002). Therefore, functional trait
diversity is likely a more robust measure, compared to
species richness, for understanding and predicting impacts
of community change on ecosystem functioning (Poff
1997; Bonada et al. 2007).
In the Scandinavian boreal zone, land-use pressures on
streams occur primarily through forest management, and in
particular through clear-cutting (Laudon et al. 2011a),
which increases the short-term concentrations of nutrients
and DOC (Schelker et al. 2012, 2016), potentially elevates
sediment loads (Futter et al. 2016), and reduces canopy
cover and changes community composition of riparian
vegetation (McKie and Malmqvist 2009). All these chan-
ges are known to influence stream macroinvertebrate
structure and function (Zhang et al.
2009; Hoover et al.
2011; Schmera et al. 2013;Go
¨
the et al. 2014; Heino et al.
2014). Effects of clear-cutting may be transient (Hoover
et al. 2011) and/or difficult to detect (McKie and Mal-
mqvist 2009), and likely change as adjacent managed
forests regenerate and stream macroinvertebrate commu-
nities recover towards a pre-disturbance state (Stone and
Wallace 1998; Liljaniemi et al. 2002). However, such long-
term patterns in recovery may not be detected unless later
stages of forest regeneration also are considered. Hence,
studies that encompass all the stages of regeneration of
managed boreal forests are required to detect the cumula-
tive impact of forestry and assess how influential this type
of land use is, compared to other factors, at shaping boreal
headwater environments and macroinvertebrate communi-
ties (Zhang et al. 2009).
Here we ask whether the impacts on benthic invertebrate
communities caused by boreal forest management are
detectable when considered in conjunction with natural
variation in land cover (e.g. percentage of lakes and mires
in catchment), geographical variables (e.g. altitude, catch-
ment size), and in-stream environmental conditions. To do
this, we used 18 boreal headwater catchments in northern
Sweden to investigate the influence of land use, land cover,
and in-stream environmental conditions, in addition to
influences of geographical variables, on stream macroin-
vertebrate community composition, and functional trait
diversity. With this design, our aim was to investigate how
gradients in catchment-scale land use and land-cover
characteristics influence stream environmental conditions
and, subsequently, macroinvertebrate communities.
MATERIALS AND METHODS
Study sites
The 18 study sites and their catchments (Table 1) are all
situated in the boreal forest of northern Sweden (Fig. 1)
312 Ambio 2017, 46:311–323
123
Ó The Author(s) 2016. This article is published with open access at Springerlink.com
www.kva.se/en
Table 1 Geographical, land-cover, and land-use characteristics of the study sites and their catchments
Site Latitude Longitude Elevation
(m a.s.l.)
Catchment
size (ha)
Land cover (%) Forest regeneration age class (%)
Forest Mire Lake 0–10 11–50 51–100 101–300
B1 64°12
0
06 19°49
0
43 215 181.9 78.1 21.9 0 12.5 17.3 54.7 9.1
B3 64°00
0
43 18°56
0
32 279 156.0 97.4 2.6 0 2.4 42.8 45.6 7.4
B4 64°00
0
52 18°56
0
50 271 41.0 93.2 6.8 0 57.4 9.3 21.9 9.4
G1 63°52
0
06 18°05
0
23 302 112.0 79.6 20.4 0 5.4 28.6 47.7 10.9
G2 63°51
0
29 18°02
0
25 404 109.0 88.6 7.3 4.0 4.0 54.2 25.0 4.1
G3 63°50
0
43 18°02
0
46 415 50.0 95.6 4.4 0 12.4 22.5 48.9 12.9
KB1 64°05
0
20 18°36
0
15 362 82.8 87.1 8.9 0 2.7 26.6 49.2 8.6
KB8 63°59
0
35 18°48
0
22 241 64.0 79.1 20.9 0 1.2 54.6 18.8 6.3
KR1 64°14
0
55 19°48
0
28 223 45.0 97.9 2.1 0 0.6 3.9 50.3 25.6
KR6 64°15
0
07 19°46
0
16 237 100.0 69.7 27.0 3.3 0 0.4 30.5 52.8
KR7 64°14
0
59 19°46
0
39 232 47.0 82.1 17.9 0 3.0 13.4 28.9 54.6
R1 64°07
0
51 20°00
0
08 172 392.0 88.3 11.2 0.4 10.4 24.2 49.4 7.3
S2 64°04
0
59 19°14
0
24 250 37.0 69.8 30.2 0 0 11.2 60.7 9.5
S6 64°05
0
33 19°10
0
06 254 89.0 96.2 3.8 0 10.2 30.6 49.0 7.0
S16 64°07
0
36 19°11
0
20 222 593.8 59.1 40.5 0.2 10.4 20.2 37.3 13.0
S26 64°06
0
54 19°12
0
28 222 18.0 100.0 0 0 0 36.0 42.4 21.6
V1 64°12
0
00 19°54
0
20 188 167.8 92.2 7.8 0 9.4 26.3 53.7 8.1
V2 64°11
0
18 19°54
0
32 203 253.5 81.2 18.8 0 5.4 30.6 50.4 6.1
x 261 141.1 85.3 14.0 0.4 8.2 25.1 42.5 15.2
Fig. 1 Locations of study sites in northern Sweden, including map coordinates. The inset shows the location of the study region in Sweden
Ambio 2017, 46:311–323 313
Ó The Author(s) 2016. This article is published with open access at Springerlink.com
www.kva.se/en
123
and were selected to represent a land-use gradient while
being similar in slope, width, and depth. For these 1st to
2nd order streams, elevation above sea level (m a.s.l.),
catchment size (ha), land cover (percentage of forest, mire,
and lake), and proportions of different forest regeneration
were determined from 25 9 25 m digital elevation models
using the Watershed tool within the Spatial Analyst tool-
box in ArcMap version 10. For this, two map sources were
used; Swedish Topographic Map (Terra
¨
ngkartan; 1:50 000)
and Forest Map (Skogskarta; 1:50 000). All 18 catchments
were dominated by forest and did not contain agricultural
land use. Forest regeneration classes were organized
according to years following clear-cutting: 0–10, 11–50,
51–100, and 101–300, which represent deciduous-domi-
nated, mixed, coniferous-dominated, and old-growth
stands, respectively.
Data collection
In late September 2012, study sites at each of the 18
streams were selected as a 50-m reach containing riffles.
At both ends and in the middle of each study reach, a
spherical densiometer was used to measure canopy
openness. At the upstream end of each reach, we mea-
sured water temperature and took water samples for
analysis of pH, dissolved organic carbon (DOC), dis-
solved inorganic nitrogen (DIN), and soluble reactive
phosphorus (SRP). Water samples for DOC, DIN, and
SRP were filtered on site (0.45-lm nylon membrane fil-
ters, Sarstedt, Nu
¨
mbrecht, Germany). All samples were
kept cold during the day and later stored in a refrigerator
(pH and DOC) or frozen (–20°C; DIN and SRP) for
analysis within a few days or weeks, respectively. DOC
and total dissolved nitrogen (TDN) were analysed by a
Shimadzu TOC-V
CPH
analyzer (Shimadzu, Duisburg,
Germany). NO
3
-
(Method G-384-08 Rev. 2), NH
4
?
(Method G-171-96 Rev. 12), and SRP (Method G–297-03
Rev. 1) were analysed using a SEAL Analytical AutoA-
nalyzer 3 (SEAL Analytical, Wisconsin, USA).
In late September, we used a Surber sampler with a
basal area of 20 9 25 cm (0.05 m
2
) to collect stream
macroinvertebrates. At each site, five samples were taken
at randomly selected locations. Stream depth and water
velocity (Electromagnetic Open Channel Flow Meter,
Model 801; Valeport, Totnes, UK) were also measured at
each sampling location. The samples were obtained by
disturbing the substrate within the Surber sampler by
hand for 60 s. Gravel and fine inorganic and organic
streambed materials were collected in the Surber net.
Cobbles were transferred to a water-filled bucket and
scrubbed separately to collect animals attached to those
surfaces. All the collected material from each sample
was placed in a separate Whirl–Pak
Ò
, along with 10 ml
of 96 % ethanol. Samples were stored at 6 °C before
being sorted.
In the laboratory, samples were separated into
macroinvertebrates and coarse-particulate organic matter
(CPOM). The CPOM was further divided into deciduous
leaf litter, coniferous needle litter, cones and twigs (here-
after, ‘small woody debris’ [SWD], i.e.\2 cm in diame-
ter), and aquatic moss for estimates of litter standing stock
of different qualities and aquatic moss abundance at each
site. Each class of CPOM was dried (60 °C) to a constant
biomass, weighed, ashed (550 °C for 40 min), and then re-
weighed to obtain the ash-free dry mass (AFDM). The
macroinvertebrates were preserved in 70 % ethanol, before
being sent to a certified taxonomist for determination.
In total, 73 taxa were identified and these were used to
calculate total taxonomic richness and diversity (Shannon
Wiener index, H
0
), community composition using princi-
pal component analysis (PCA) on absolute (pooled)
abundances, and the proportional abundance of Simuliidae
and Chironomidae, as these two taxonomic groups were
among the most abundant. For all community measures,
the five subsamples at each site were pooled, to obtain
measures at the site level. Further, we assigned functional
traits to the macroinvertebrate taxa (Poff et al. 2006)
using an extensive European freshwater database (Sch-
midt-Kloiber and Hering 2012). Functional traits were not
assigned to taxa not identified to a high enough resolution
(e.g. Nematoda). This process rendered 21 functional
traits, each with two to five modalities, for 41 macroin-
vertebrate taxa (Supplementary Tables S1, S2). These data
were used to calculate functional trait diversity (Shannon
Wiener index, H
0
) and the proportion of individuals with
low pH sensitivity.
In July 2013, we characterized the benthic substrate
composition at each stream. For this, the intermediate axis
of 200 gravel/cobbles was measured using random walk
sampling. The mineral substrate was classified into dif-
ferent size categories with particles\2 mm (i.e. sand) as
the smallest category. In cases where only fine organic
particles were found at the random location, particles were
classified as zero (and later as ‘organic fines’). Data on the
mineral substrate size classes were used to calculate med-
ian substrate size and substrate heterogeneity (i.e. Shannon
Wiener index, H
0
).
Statistical methods
We used partial least squares (PLS) regression to explore
relationships between different invertebrate metrics, in-
stream habitat variables, and catchment attributes. More
specifically, we analysed how catchment-scale land use
(i.e. forest regeneration age classes) and land cover
explained variation in both in-stream physico-chemical
314 Ambio 2017, 46:311–323
123
Ó The Author(s) 2016. This article is published with open access at Springerlink.com
www.kva.se/en
conditions and headwater macroinvertebrate metrics. Fur-
ther, to assess the relative importance of catchment-scale
descriptors versus in-stream variables for headwater
invertebrate metrics, we performed separate analyses with
in-stream variables as the only predictor variables. PLS
relates two data matrices (including predictor and depen-
dent variables) to each other by a linear multivariate model
and produces latent variables (PLS components) extracted
from predictor variables that maximise the explained
variance in the dependent variables. PLS is especially
useful when predictor variables are correlated and when the
number of predictor variables is high (Carrascal et al.
2009). The evaluation of the PLS models was based on the
level of variance explained (R
2
), loadings of the indepen-
dent variables, and the variable influence on projection
(VIP). The independent variable loading describes the
relative strength and direction of the relationship between
independent and response variables. The VIP value sum-
marises the importance of each variable, and, as a limit for
when a predictor variable is important in a model, we chose
VIP[ 1.0.
To visualize relationships between macroinvertebrate
taxa and environmental conditions, canonical correspon-
dence analysis (CCA) was performed (and plotted), using
the predictor variables that were the most important (i.e.
VIP[ 1.0) in the PLS models for macroinvertebrate com-
munity PC1 and PC2. Dependent variables were ln trans-
formed, if necessary, to meet the assumptions of normality
and equal variance, and assumptions were checked using
standard diagnostics. PLS regression analyses were per-
formed using XLSAT (XLSTAT 2015.2.01, Addinsoft
SRAL, Germany), and CCA were performed using the
vegan library (Oksanen et al. 2014) in R (R Core Team
2012).
RESULTS
Among sites, elevation varied by a factor of 2.4, while
catchment size varied by a factor of 33 (Table 1). The
proportion of mire in catchments ranged from 0 to 40.5 %
and was not significantly related to any of the forest-age
categories. Lakes were absent in most catchments and were
therefore not included in the statistical analyses. Hence, all
catchments were dominated by forest, and in these forests,
stands of 11–50 and 51–100 years in age were the most
common (Table 1). At two sites (KR6 and KR7), mature
forests (101–300 years) dominated, while recently clear-cut
forest (0–5 years) was the most common regeneration class
at one site (B4; Table 1).
Mean depth, water velocity, and water temperature were
similar among sites (Table 2). Most canopies were rela-
tively closed (\20 % openness), apart from B4, which was
a recently clear-cut site, whose canopy was largely open
(83.3 %). There was a positive relationship between pro-
portion of young forest (0–10 years) in the catchment and
reach-scale canopy openness (data not shown), and
although this relationship was driven by one site (B4), it
indicates that catchment-scale forest-age composition can
be broadly reflected in reach-scale canopy openness. Sites
varied from acidic to almost circumnetural (i.e. pH of
4.4–6.3) and concentrations of DOC and SRP varied from
9.8 to 42.4 mg C L
-1
and 2.7 to 11.0 lgPL
-1
, respec-
tively (Table 2). Importantly, pH, DOC, and SRP tended to
co-vary among sites, such that sites with low pH tended to
have both high DOC and SRP. Concentrations of DIN were
generally less than 50 lgNL
-1
with the exception of B4
(151.0 lgNL
-1
). The standing stock of organic matter
(OM) was comprised mostly of SWD (54.4 ± 0.3 %
[mean ± 1 SD]), while coniferous needle litter was the
least abundant, and aquatic moss biomass varied substan-
tially among sites (Table 2). There was some variation in
median substrate size and substrate diversity among sites,
but only two sites (G3 and KB8) showed a substantial
cover ([50 %) by organic fines (Table 2).
The proportion of younger forest (i.e. 11–50 years) was
the most important catchment-scale predictor for explain-
ing variation in in-stream conditions (Table 3). Specifi-
cally, proportion of younger forest was negatively related
to concentrations of DOC and SRP and positively related to
pH. Further, streams in catchments dominated by younger
forest had higher aquatic moss abundance and greater
standing stock of SWD. Land-cover characteristics were
also important for several in-stream variables, but catch-
ment size was significantly associated with only physical
characteristics (i.e. substrate, depth, and water velocity),
while elevation and percent mire in the catchment were
also related to water–chemical properties (Table 3).
Macroinvertebrate taxonomic richness based on pooled
samples at each site varied from 12 to 38 taxa and among-
site variation in total abundance was considerable (84 to
2475 individuals per 0.25 m
2
, i.e. the sum of all subsamples
per site; Table 2). PC1 and PC2 explained 32 and 17 % of
the variation in macroinvertebrate community composition,
respectively. PC1 was positively related to abundances of a
diverse assemblage of taxa (e.g. Brachyptera risi, Baetis
rhodani, Bardeniella freyi, Hydraena gracilis) and pri-
marily negatively related to the abundance of Nemurella
picteti. PC2 was positively related to the abundance of
Plectrocnemia conspersa and Limnephilidae and nega-
tively related to primarily the abundance of Silo pallipes
and Jungiella longicornis. As for the in-stream variables,
proportion of younger forests was a strong predictor vari-
able and positively related to all measures of macroinver-
tebrate diversity and PC1 and negatively related to the
abundance of taxa with low pH sensitivity (Table 3). In
Ambio 2017, 46:311–323 315
Ó The Author(s) 2016. This article is published with open access at Springerlink.com
www.kva.se/en
123