Is there an ideal protocol for sampling macroinvertebrates in springs?

Abstract

Sampling in springs has several technical problems due to their reduced dimensions and habitat heterogeneity. A standardized quantitative method for sampling crenic macroinvertebrates has never been proposed. The aim of this study was to compare different sampling methods and consider their environmental impacts. First, we present a review of sampling methods found in the literature and discuss their advantages and disadvantages with respect to selective collection of the target community and habitat disturbance. Altogether, 10 different methods have been reported, the use of nets being the most common protocol. Second, we report the results of macroinvertebrate samplings performed in three springs, each surveyed twice, using three different methods (multi-habitat proportional hand net, baited traps, and vegetation washing), in order to compare their effectiveness in collecting macroinvertebrates. Overall, 32 macroinvertebrate taxa, mostly identified at family level, were collected in the sampled springs. Significant differences in abundances were found using different methods, while results for community structure were comparable between the hand net sampling and the combined use of the other two methods, notwithstanding slight differences in the composition of Coleoptera and Diptera assemblages. The hand net, with a multi-habitat proportional approach, yielded more thorough results, making it suitable for biodiversity inventories but having some potentially negative effects on spring habitats. Traps and vegetation washing are also reliable methods with negligible impacts on spring ecosystems that can be conveniently used in ecological studies.

Keywords:

• springs
• sampling methods
• macroinvertebrates
• biodiversity
• impact assessment

Introduction

In spite of their small size, the ‘mosaic’ ecotonal structure of springs results in a number of microhabitats that sustain high species richness (Cantonati et al. 2012). Several studies have highlighted the great biodiversity of macroinvertebrates in springs as well as the presence of rare and endemic species (e.g., Takhteev et al. 2010; Maiolini et al. 2011; Kubíková et al. 2012; Martin & Brunke 2012; Spitale 2012; Spitale et al. 2012). The uses and limitations of various methods for sampling benthic macroinvertebrates have been extensively discussed (e.g., Davies 2001). In contrast, despite great interest in spring biodiversity, a standardized, quantitative method for sampling crenic macroinvertebrate taxa has never been developed. Cantonati et al. (2007) suggested effective methods for collecting spring invertebrates but a variety of methods have been adopted in crenic investigations. The technical difficulties of sampling in springs were well summarized by Gerecke et al. (2007): ‘The main dilemma of limnological studies in springs probably derives from the generally reduced dimensions and extreme heterogeneity of the habitat.’ Furthermore, many authors (Gerecke et al. 1998; Zollhöfer 1999; Myers & Resh 2002; Staudacher & Füreder 2007; Tichá et al. 2012) noted that some survey methods, which implied samplings in all microhabitats, could be destructive to the environment and the biota of these fragile ecosystems.

Previous studies on macroinvertebrates in different aquatic ecosystems have shown that sampling methods affect the data precision and the selection of sampling technique is among the most important decisions for freshwater studies (Carter & Resh 2001). Standardizing the sampling procedure is thus necessary in order to obtain precise and comparable biological data for spring surveys and assessment. The aim of this study was to summarize the literature on sampling methods in springs and to compare the effectiveness of some semi-quantitative sampling methods, taking into account their potential impacts on spring habitat and biota.

Methods

Three rheocrenic, permanent springs located between 474 and 589 m above sea level in the Mount Prinzera protected area (44° 37' N; 10° 03' E), an ophiolitic outcrop in northern Italy near Parma, were selected for the study. Samplings were carried out in two seasons (May–June and August–September 2014). Macroinvertebrates were collected using three methods.

1. Multi-habitat proportional net – a hand net (frame dimensions: 10 × 10 cm; mesh size: 255 µm) was used to collect 10 replicates at each site. For each replicate, a substrate area equivalent to the net frame was sampled for 15 seconds. The number of replicates for each microhabitat was proportional to its percentage cover in the spring, up to a total number of 10 replicates. For example, given a substrate composition of 50% gravel, 30% mosses, and 20% silt, five samples were collected in gravel, three in mosses, and two in silt. All 10 replicates were composited into a single sample.

2. Vegetation washing – about 250 mL volume of submerged vegetation was collected and washed in the laboratory through a 255-µm sieve.

3. Traps – traps were constructed following Bottazzi et al. (2011). They were obtained from polyvinyl chloride (PVC) centrifuge tubes (length 100 mm; diameter 28 mm), by cutting the conical end, drilling an opening (0.5 cm of diameter) in its apex, and inserting it, inverted, into one end of the tube. The other end of the tube was closed with a 50 μm net. These traps were filled with washed and sieved gravel (0.3–1.0 cm). Traps were baited with meat, placed at the sediment–water interface, and covered with stones to keep them in place for 7–8 days. Two pairs of traps were deployed in each spring: one pair at the source and the other 2 m downstream. For each pair, one trap was placed with the opening in the flow direction and the other in the opposite direction.

Vegetation washing and macroinvertebrate trapping were performed two weeks after the net sampling.

In the laboratory, collected material was washed through a 255-µm sieve and fixed with 90% ethanol. Macroinvertebrates were identified to the family level for Plecoptera, Trichoptera, Ephemeroptera, Coleoptera, Diptera, and Crustacea but to a coarser taxonomic level for Hirudinea, Gastropoda, Collembola, Hydrachnidiae, Odonata, and Oligochaeta.

Differences in organism abundance between the three methods were tested with an analysis of variance (ANOVA). Logarithmic transformation was used to obtain normal distributions and homogeneity of data, as determined by Shapiro and Bartlett tests (Legendre & Legendre 2012). Non-metric multidimensional scaling (NMDS, Legendre & Legendre 2012) was performed to evaluate possible differences in community structures determined by different methods. Centroids of methods were fitted on NMDS plots in order to identify these differences and then tested with Permanova (Anderson & Walsh 2013). Differences between methods were assessed by considering both the three methods individually (net, vegetation washing, and traps) and the combination of vegetation washing and traps versus net. Differences were also tested for each of the most diverse insect orders (Trichoptera, Coleoptera, and Diptera). Statistical analyses were performed using R software (version 3.0.0, R Development Core Team 2013) and Vegan (version 2.0-7, Oksanen et al. 2013).

Results

Ten different methods were found in the literature. The use of hand or kick net was by far the most commonly used protocol (Table 1). Overall, 32 taxa were collected in our survey (Table 2). Insect orders with the greatest number of families were Diptera (9), Trichoptera (7), and Coleoptera (6). Chironomidae was the most abundant taxon collected with the net (1029 specimens) and vegetation washing (60), whereas traps collected the greatest number of Niphargidae (293). Lepidostomatidae, Chironomidae, Ceratopogonidae, Hirudinea, and Gastropoda were found in all samples collected by the net. The maximum number of taxa collected in one sampling session was 8 using the net, and 11 combining traps and vegetation washing. Thirteen taxa were collected by all methods; net and the trap samplings shared seven taxa, whereas net and vegetation washing shared eight taxa. Finally, Hydropsychidae, Limnephilidae, and Hydrophilidae were only found in net samples, and Empididae was exclusively collected with traps (Table 2).

Table 1. Sampling methods used in springs by other researchers.

Methods	References
Surber sampler	Erman and Erman (1995); Erman (1998); Zollhöfer (1999); Smith et al. (2003); von Fumetti et al. (2006); Barquín and Death (2008); Gerecke et al. (2011)
Hand or kick net	Williams (1991); Gerecke and Cantonati (1998); Hahn (2000); Myers and Resh (2002); Mezzanotte and Sambugar (2004); Mori and Brancelj (2006); Lencioni (2007); Staudacher and Füreder (2007); Ilmonen et al. (2012); Kubíková et al. (2012); Martin and Brunke (2012); Rader et al. (2012); Spitale (2012); Tichá et al. (2012)
Collection at sight	Williams (1991); Bonettini and Cantonati (1998); Gerecke and Cantonati (1998); Myers and Resh (2002); Gerecke and Di Sabatino (2007)
Bou-Rouch pump	Crema et al. (1996)
Sweep net	Crema et al. (1996); Sambugar et al. (2006)
Core sampler/sediment sample	Gooch and Glazier (1991); Myers and Resh (2002); Dumnicka et al. (2007); Staudacher and Füreder (2007); Worthington Wilmer et al. (2008); Takhteev et al. (2010); Koperski et al. (2011); Spitale (2012)
Traps	Bottazzi et al. (2011)
Drift tube or net	Stoch et al. (2008); Bottazzi (2010)
Squeezing mosses or washing vegetation	Bottazzi et al. (2011); Gerecke et al. (2011); Spitale (2012)
Emergence traps	Erman and Erman (1995); Erman (1998); Gathmann et al. (2009)

Table 2. List of taxa and their abundances collected using different methods.

Taxa	Net	Traps	Vegetation washing
Leuctridae	6	2	1
Lepidostomatidae	29	4	36
Philopotamidae	10	4	4
Sericostomatidae	104	154	3
Chironomidae	1029	41	60
Stratiomyidae	18	1	3
Psychodidae	26	6	33
Limoniidae	6	10	1
Dixidae	22	14	12
Scirtidae	511	16	23
Hydraenidae	73	1	2
Niphargidae	172	293	2
Gastropoda	581	220	6
Collembola	19	3	3
Veliidae	1	2	0
Ceratopogonidae	50	1	0
Haliplidae	2	1	0
Dytiscidae (adults and larva)	8	2	0
Hirudinea	32	38	0
Hydrachnidiae	2	1	0
Nemouridae	9	0	8
Heptageniidae	4	0	1
Polycentropodidae	4	0	3
Beraeidae	7	0	1
Ptychopteridae	15	0	9
Simuliidae	1	0	1
Odonata	4	0	1
Hydropsychidae	2	0	0
Limnephilidae	1	0	0
Hydrophilidae	1	0	0
Tipulidae	4	0	0
Empididae	0	2	0

Differences in taxa abundance between methods were significant, both considering the three distinct methods (Figure 1; F = 16.18; p < 0.001) and when merging the traps and the vegetation washing (Figure 2; F = 9.46; p = 0.012). Sampled communities formed three distinct groups near their centroids in the NMDS plot (stress = 0.15; Figure 3). This indicates differences in macroinvertebrate assemblages according to the methods, as confirmed by the Permanova test (R² = 0.223; p = 0.010). Stress was 0.13 in the plot of NMDS ordination obtained by merging data collected with traps and vegetation washing (Figure 4). Differences between the two different methods (net versus traps plus washing vegetation) were less detectable. Permanova test (R² = 0.136; p = 0.134) indicated that there was not a significant difference between communities sampled with these two methods. Net sampling and combined traps and vegetation samples showed differences for Coleoptera (R² = 0.219; p = 0.030) and Diptera (R² = 0.250; p = 0.005), but not for Trichoptera (R² = 0.056; p = 0.826) (Figure 5).

Figure 1. Logarithm of taxa abundances for the three methods. Differences in abundances were significant (ANOVA, F = 16.18; p < 0.001).

Figure 2. Comparison of taxa abundances (log transformed) using net sampling or a combination of traps and vegetation washing. Differences in abundance between methods were significant (F = 9.46; p = 0.012).

Figure 3. NMDS ordination of the three methods (stress = 0.15). Bigger points are the centroids of methods (veg = vegetation washing). Smaller points are sampled communities.

Figure 4. NMDS ordination of the traps plus vegetation washing (traps + veg) and nets (stress = 0.13). Bigger points are the centroids of methods and smaller points are sampled communities.

Figure 5. NMDS ordination of the net and the traps plus vegetation washing (traps + veg) for Trichoptera, Coleoptera, and Diptera. Bigger points are the centroids of methods and smaller points are sampled communities.

Discussion

The lack of a standardized sampling protocol for springs has led to the use of a wide variety of methodologies. Standard Surber samplers (sampling area: 0.09–0.1 m²) have been rarely used (Smith et al 2003; Barquín & Death 2008). More frequently, smaller samplers were preferred (Erman & Erman 1995; Erman 1998; Zollhöfer 1999; von Fumetti et al. 2006; Gerecke et al. 2011). The mesh size of Surber, kick, or hand nets varies from 100 µm to 1 mm. Although Gerecke et al. (2007) recommended sampling different microhabitats in springs, at their relative microhabitat proportion, few studies have used a proportional multi-habitat approach (Crema et al. 1996; Zollhöfer 1999; Martin & Brunke 2012) or sampled all available substrates (Bonettini & Cantonati 1998; Mezzanotte & Sambugar 2004; Ilmonen et al. 2012). In addition, combined methods have been frequently used in the same study (Williams 1991; Erman & Erman 1995; Crema et al. 1996; Bonettini & Cantonati 1998; Erman 1998; Myers & Resh 2002; Sambugar et al. 2006; Staudacher & Füreder 2007; Bottazzi et al. 2011; Gerecke et al. 2011; Spitale 2012).

Each method has advantages and disadvantages that may be dependent on the specific aims of the study. For example, the use of sweep nets or emergence traps allows to sample only organisms with aerial stages, whereas drift tubes/nets underestimate taxa with low tendency to drift. Also, methods that require collection by sight could be biased against small, less-mobile, and less-visible organisms. The Surber net, Bou–Rouch pump, and core sampler may allow the collection of quantitative data but the Surber net is usually too large to be used in springs (see Gerecke et al. 2007) and the Bou–Rouch pump and core sampler only collect sediment and interstitial samples.

Our results showed that macroinvertebrate community structure estimated by the combination of traps and washing vegetation can be considered comparable to those obtained with net. The four taxa exclusively collected by the net, Hydropsychidae, Limnephilidae, Tipulidae, and Hydrophilidae, have body sizes larger than the opening of the traps (Tachet et al. 2000). Furthermore Hydropsychidae, Limnephilidae, and Tipulidae rarely can be found aquatic vegetation, and Hydrophilidae organisms are very mobile and could escape during vegetation collection (Tachet et al. 2000). Although similar communities were collected by both net sampling and combined vegetation washing and trap sampling, there were some differences. The two methods produced different results for Diptera and Coleoptera. This result was probably related to issues with single-habitat protocols. For example, traps and washing vegetation probably underestimated the presence of taxa not associated with vegetation or not attracted by meat. The abundances of organisms collected by traps and vegetation washing were significantly lower than those collected by net. Therefore, the impact of these protocols on spring fauna would be expected to be lower, at least on some taxa. In addition, net sampling requires brushing, scraping, digging, and squeezing of different microhabitats and substrata, causing disturbance of the spring habitat with unknown recovery time. The use of traps is more time consuming than other methods, because they require an additional visit to the springs to be removed. Finally, some sampling methods would not be suitable in particular habitat types. For example, some springs lack any kind of vegetation, and traps cannot be placed in hygropetric springs, where the sediment layer is too thin, or in helocrene springs, that often are too deep.

Since spring fauna show evident habitat-preferences (von Fumetti et al. 2006), single microhabitat protocols should be used only to survey specific target taxa or habitats. In contrast, a multi-habitat methodology allows a better estimation of the overall biodiversity. In order to obtain more comparable results, Gerecke et al. (2007) recommended sampling all available habitats, using proportional sampling time for each substratum and including transitional zones among different substrata since they may host specialized taxa. In addition, the multi-habitat proportional sampling is considered by the Water Framework Directive (Directive 2000/60/EC) as the best approach for assessing macroinvertebrate diversity.

Proposed methods could be improved in order to be more effective. Traps with different opening dimensions may be able to collect the whole range of organism body sizes. The volume (or number of replicates) of vegetation-washing samples can be increased whenever possible.

In conclusion, the net and the vegetation washing with traps have different characteristics and effectiveness, even though both protocols give very similar qualitative results. The use of the net in a multi-habitat proportional approach provides more accurate and complete information, but also significantly impacts the biotic and abiotic components of springs. For these reasons, this method is only recommended for biodiversity inventories. On the other hand, traps and vegetation washing are reliable methods with fewer negative effects on spring ecosystems; thus, they are more suitable for ecological studies focused on the analysis of the community structure.

Acknowledgements

The authors thank the anonymous reviewers for their helpful comments.

Disclosure statement

No potential conflict of interest was reported by the authors.