Key words
Self-organizing map, logistic regression, small stream, land use,
boreal, climate change
Introduction
Studying the relationship between species and their environment is at
the core of ecology. Modelling this relationship has long been
performed, using a wide array of methods (Franklin, 1995; Guisan &
Zimmermann, 2000; Domisch, Jähnig, Simaika, Kuemmerlen, & Stoll, 2015).
The focus in developing these models may be to study species-environment
relationships or to predict the occurrence of the studied species. In
fisheries research, the identification of the environmental variables
that characterize fish distributions has been one of the main objectives
(e.g. Nelson, Plaits, Larsen, & Jensen, 1992; Rieman & McIntyre,
1995). Predictive models may help in fish-based bioassessment (Brosse,
Lek, & Townsend, 2001; Oberdorff, Pont, Hugueny, & Chessel, 2001;
Oberdorff, Pont, Hugueny, & Porcher, 2002), and in focusing inventory
and management activities on areas where species are considered likely
to occur (Porter, Rosenfeld, & Parkinson, 2000).
Several studies have indicated that field-measured site-scale (local)
variables such as stream width, water depth, water chemistry, riverbed
substrate, flowrate, undercut banks, canopy cover, riparian vegetation,
and the slope at the sampling site can predict the occurrence of fish
species (e.g. Gorman & Karr, 1978; Watson & Hillman, 1997; Terra,
Hughes, & Araujo, 2016). However, these field measurements are
laborious and thus demanding for adoption as predictors of species
occurrence in fisheries management, for example. An easier way to
predict species occurrence would be to use large-scale map-based
(regional) variables such as the size of the upper catchment, the
elevation and land use in the upper catchment (Porter et al., 2000).
Indeed, catchment-scale variables can have a greater impact than
site-scale variables on stream fish assemblages (DeRolph, Nelson, Kwak,
& Hain, 2015; Mitsuo, 2017).
The process of taking natural landscapes for human use can cause
detrimental effects on terrestrial and aquatic ecosystems (Huston, 2005;
Pugh, Pandolfi, Franklin, & Gangloff, 2020). For example, increased
land use for agriculture, urban areas, and forestry can impact fish
populations through alterations in stream hydrology, geomorphology,
water quality, sedimentation, riparian vegetation, and habitat
heterogeneity, eventually leading to species loss or replacement (Allan,
Erickson, & Fay, 1997; Lange, Townsend, Gabrielsson, Chanut, &
Matthaei, 2014; Pugh et al., 2020). Recent developments in geographical
information systems (GIS) technology (Lü, Batty, Strobl, Lin, Zhu, &
Chen, 2019) have facilitated easy access to a wide range of catchment
characteristics above any site of a stream network. These catchment
characteristics, typically expressed as the percentage coverage of the
upper catchment, are extensively used in studying the effects of land
use on stream biota.
About 80% of the millions of kilometers of European river networks
consist of small streams, commonly known as brooks, creeks, or
headwaters (Kristensen & Globevnik, 2014). Small headwater streams are
important contributors to aquatic biodiversity and may suppress the
negative impacts of anthropogenic stress on downstream reaches (Burdon
et al., 2016; Baattrup-Pedersen, Larsen, Andersen, Jepsen, Nielsen, &
Rasmussen, 2018). However, in the European Water Framework Directive
(WFD; European Commission 2000), small streams with a catchment size of
less than 10 km2 are mostly omitted from river basin
management plans or merged into larger water bodies (Kristensen &
Globevnik, 2014, Baatturp-Pedersen et al., 2018).
In this study, we chose to examine fish in small streams for some
specific reasons. We inferred that in small streams/catchments, a single
land-use attribute such as an urban area can easily reach high coverage,
and therefore, the effect of land use on fish species occurrence should
be relatively easy to trace. In small streams, the upstream catchment
area is always located relatively near the sampling site, and the impact
of land use should therefore be more direct. Indeed, proximity to the
stream has appeared an important factor in estimating the impact of land
use on stream biota (Wang, Lyons, & Kanehl, 2001). Small streams with a
small volume of water also have only a limited ability to dilute
pollutants such as nutrients from agriculture (Kristensen & Globevnik,
2014). Small tributary streams have appeared to be particularly
sensitive to nutrient enrichment (Bussi et al., 2018). The impact of
human activities is therefore potentially greater on small water bodies
than on larger ones (Kristensen & Globevnik, 2014).
Our main aims in this study were (1) to explore the relationship of
map-based environmental variables and the occurrence of fish species in
small boreal streams; (2) extract fish species clusters and evaluate
their ecological relevance; (3) study species occurrence in relation to
annual mean temperature from the perspective of the climate change in
this region; and (4) identify species-specific responses to man-induced
pressures for the future development of diagnostic indices in
bioassessment of small boreal streams.
Material and methods
Altogether, 11 environmental variables were measured (Table 1). The
studied area covered Southern and Central Finland in the boreal region
from about 60 o to 67 o, which are
mostly covered with coniferous forest. The highest altitude among
sampling sites was about 300 m in the studied territory characterized by
lowlands (Table 1). The variables were map-based, with the exception of
one field-collected variable, water temperature at sampling
(electrofishing). Upstream catchment boundaries were delineated for each
site with Geographical Information System, using the Digital Elevation
Model (DEM) raster database from National Land Survey of Finland (NLS)
and vector data of Drainage Basins in Finland (Finnish Environment
Institute, SYKE). Only sites with a catchment area < 100
km2 were included in the study. The proportions of
different land covers in the catchment areas were extracted from the
CORINE Land Cover 2012 data. The quantity of forest drainage by ditching
was estimated as a percentage of ditched peatlands from the drainage
data of the Finnish Environment Institute. Annual air temperature and
precipitation data were derived from the WorldClim database (Hijmans,
Cameron, Parra, Jones, & Jarvis, 2005).
Electrofishing data from small Finnish streams were gathered mainly from
a national database (Hertta/Koekalastusrekisteri) managed by the Natural
Resources Institute Finland (Luke) and hosted by the SYKE. Additional
data were acquired from Metsähallitus (a state-owned enterprise
responsible for the management of state-owned land and water areas). The
total number of single-run electrofishing samples was 776, conducted at
487 sites, indicating that some of the sites were sampled more than
once. As a rule, repeated sampling at the same site was performed at
different years. Most of the sampling had been performed at the period
2000–2020. The electrofishing sites usually represented wadable riffles
with stony bottoms. Escape nets were not used at any of the sampling
sites, which typically covered 50–150 m2. As the
electrofishing sampling had been performed in July–October, natural
seasonal decline in stream water temperatures was reflected in the
measured temperatures. European standard EN 14011:2003 (Water
quality—sampling of fish with electricity) was followed in sampling.
Fish data were converted to species presence/absence for all analyses in
this study.