Response of benthic invertebrates to natural versus experimental disturbance in a Swiss prealpine river

frbiol0141

Freshwater Biology (1997) 37, 61–77

Response of benthic invertebrates to natural versus experimental disturbance in a Swiss prealpine river C H R I S T O P H D . M AT T H A E I , * U R S U E H L I N G E R A N D A N D R E A S F R U T I G E R Department of Limnology, Swiss Federal Institute for Environmental Science and Technology (EAWAG), CH-8600 Duebendorf, and Swiss Federal Institute of Technology (ETH), CH-8092 Zurich, Switzerland *Present address: Department of Zoology, University of Otago, PO Box 56, Dunedin, New Zealand

S U M M A RY 1. The crucial point of disturbance experiments in streams is the extent to which they can simulate a natural spate. Ideally, disturbance experiments should proceed side by side with a phenomenological study to allow a direct comparison. In the present study conducted in a prealpine Swiss river, the River Necker, fortuitous events made such a comparison possible. 2. In summer 1994, we took Surber samples one day before and on several sampling dates after a major flood (recurrence interval µ 5 years), which was followed by a long period of uniform discharge in a river characterized by frequent spates. Beginning 19 days after this flood, patches of the stream bed (µ 9 m2) were physically disturbed by kicking and raking. 3. The degrees of reduction in the total number of individuals and the dominant taxa were similar after both types of disturbance, as were the recolonization patterns of Rhithrogena spp., Leuctra spp. and Hydracarina. Chironomidae, Baetis spp., Simuliidae, Pentaneurini and Corynoneura/Thienemanniella spp. showed a distinct lag phase after the flood before recolonization began, whereas there was no such lag phase after the experiment. Therefore, the time needed to recover to pre-flood densities was longer for these taxa. Nevertheless, recolonization rates and patterns after the lag phase were similar to those after the experimental disturbance. 4. Size-class measurements indicated that recruitment from egg hatching may have been more important after the flood than after the experimental disturbance for Rhithrogena spp., but not for Chironomidae, Baetis spp., Simuliidae, Pentaneurini and Leuctra spp. Invertebrate drift was probably the most important pathway of recolonization after both types of disturbance. 5. Our experiment allowed a realistic simulation of several important effects of the large flood on the invertebrate community. Smaller spates that induce substratum movement at a spatial scale similar to our experimental plots are much more common than large floods in the River Necker. For these spates, our experiment should provide an even more realistic simulation of natural disturbance. Introduction Disturbance by spates is a dominant organizing factor in lotic ecosystems (Resh et al., 1988; Townsend, 1989). As a consequence, the recovery of stream communities from spates has been widely researched (Steinman & McIntire, 1990; Wallace, 1990; Yount & Niemi, 1990; © 1997 Blackwell Science Ltd

Mackay, 1992; Detenbeck et al., 1992). The effects of disturbance have been studied by phenomenological accounts of particular events (e.g. Fisher et al., 1982; Boulton et al., 1992; Smock et al., 1994), by comparative studies in streams with contrasting disturbance

61

62 C.D. Matthaei, U. Uehlinger and A. Frutiger regimes (e.g. Robinson, Reed & Minshall, 1992; Scarsbrook & Townsend, 1993; Death & Winterbourn, 1995), and by experimental manipulations. Many stream ecologists choose the experimental approach, because experiments allow replication and are controlled, in contrast to phenomenological and comparative studies (Townsend, 1989). Several disturbance experiments have been conducted in artificial streams (see review by Lamberti & Steinman, 1993). Others have experimented in streams using mesh baskets or bricks (e.g. Reice, 1985; Robinson & Minshall, 1986; Rosser & Pearson, 1995). Few experimental studies have physically disturbed patches of natural streambed larger than the size of individual stones (e.g. Clifford, 1982; Doeg, Lake & Marchant, 1989; Johnson & Vaughn, 1995). However, the crucial point of all such experiments is the extent to which they can simulate a natural spate, because the spatial scales of experimental and natural disturbance usually differ by orders of magnitude. The shortcomings of small-scale experiments that study the recovery of benthic invertebrates from disturbance have been discussed in recent reviews. Lake (1990) argued that such experiments may determine the damage done by disturbance but cannot effectively simulate recovery. Minshall (1988) concluded that manipulative experiments conducted at small spatial scales provide valuable information on the recovery dynamics of individual stones or patches of substratum, but are unsuitable for studying the impact of spates on a whole stream. According to Mackay (1992), small-scale studies are reasonable simulations of the colonizing events that might follow minor spates. Nevertheless, she expected colonization rates in small-scales studies to be far higher than in situations when more of the stream bed is affected. In a previous study that focused on the recolonization of benthic invertebrates after experimental disturbance (Matthaei et al., 1996), we physically disturbed substratum patches of 9 m2 to overcome some of the above shortcomings. Nevertheless, a reliable evaluation of the realism of our experiment requires a direct comparison of the recovery patterns after experimental disturbance and after a natural spate. To our knowledge, such a study has only been done once (Brooks & Boulton, 1991), presumably because of the low predictability of spates. During the summers of 1993 and 1994, we tried to conduct a disturbance experiment shortly before or after a nat-

ural spate. In 1993, our efforts were unsuccessful due to the flashy, unpredictable discharge regime of the River Necker (Uehlinger, 1991; Uehlinger, Bu¨hrer & Reichert, 1996), because twelve substratum-moving spates within 12 weeks kept post-flood series too short to study recovery patterns. However, the fortuitous occurrence of a flood in July 1994 followed by a long period of relatively uniform discharge enabled us to make the comparison. The primary objective of our study was to determine to what extent a natural spate could be simulated by physically disturbing medium-sized experimental plots. Depending on the magnitude of the spate, we expected different results. After a ‘smaller’ spate, with substratum movement at a spatial scale similar to that in our experimental plots, rates and pathways of invertebrate recolonization ought to be similar, and reimmigration from nearby patches and by drift should be more important than recruitment from eggs. In contrast, a large flood ought to affect more strongly the local refugia of colonizers (see Townsend, 1989; Sedell et al., 1990; Lancaster & Hildrew, 1993). Therefore, vertical or horizontal migrations should become less important. Similarly, invertebrate drift ought to be of less significance as a colonization mechanism, because upstream reaches may be also denuded of biota (Townsend, 1989). Consequently, a large flood should have a longer lasting impact on the benthic invertebrate community, and recovery ought to occur mainly through recruitment from eggs.

Materials and methods Study site The study was carried out in the prealpine River Necker in north-eastern Switzerland. This river has a mean discharge of 3.4 m3 s–1 at the hydrograph station Aachsa¨ge, which is situated µ 2 km upstream of the study site. Baseflow discharge during summer is µ 0.4–0.8 m3 s–1. ‘Prealpine’ refers to the geographical region between the Swiss plateau and the Alps. The study site, a riffle µ 70 m long at 590 m a.s.l. (stream order 6; 47°239N, 9°089E), was identical with site 1 in Matthaei et al. (1996) and is described in detail there. For more information on the River Necker and prealpine rivers in general see Frutiger (1983), Uehlinger (1991) and Uehlinger et al. (1996). The median particle size at this site is 3.3 cm (Matthaei et al., 1996). Stream © 1997 Blackwell Science Ltd, Freshwater Biology, 37, 61–77

Invertebrate response to natural v experimental disturbance 63

Fig. 1 Discharge in the River Necker during spring and summer 1994 and sampling periods of the study (further details under Materials and methods). ‘d21’ 5 pre-flood sampling, 1 day prior to the flood; ‘d1’ 5 start of postflood sampling, 1 day after the flood; ‘d19’ 5 start of disturbance experiment; ‘d40’ 5 last sampling occasion.

width varies between 20 and 25 m. The shallow riffle section covered µ 10–15 m of this width, whereas the remaining area was a run too deep for Surber sampling.

Phenomenological study and disturbance experiment Discharge at Aachsa¨ge during the summer of 1994 is shown in Fig. 1. We conducted the study from 6 July to 16 August. Substratum stability during this period was estimated from personal observations in the field. A large flood with a maximum discharge of 166 m3 s–1 occurred on 7 July, followed by a smaller discharge peak of 16 m3 s–1 on 9 July. The flood had a recurrence interval of µ 5 years and caused extensive substratum movement in the whole stream channel. The smaller spate may have caused some additional substratum movement. Between 9 July and 11 August, there was no substratum movement at the study site, allowing the recovery of the benthic community. A small spate of 10.5 m3 s–1 on 11 August removed algae from stones with a diameter of less than µ 2 cm, but did not move larger particles. The period of stable discharge was terminated 1 day after the last sampling occasion. On 17 and 18 August, two spates with a maximum discharge of 16 m3 s–1 each removed almost all algae from the experimental plots, indicating significant substratum movement or at least surface scouring by fine sediments (Scha¨lchli, 1993). The last spate prior to the large flood had occurred on 10 June, preceded by another large flood on 24 May. Therefore, invertebrates had had almost 4 weeks to recover from the last disturbance before the flood of 7 July. We chose ten plots of µ 9 m2 within the riffle area of µ 70 m 3 10–15 m. To avoid disturbing neighbouring plots while sampling, we arbitrarily defined a © 1997 Blackwell Science Ltd, Freshwater Biology, 37, 61–77

minimum distance of 3 m between plots (laterally and upstream/downstream). Except for this condition, the ten plots were chosen randomly. We divided the ten plots into two groups of five using a randomized block design (Hurlbert, 1984). The first group was used in three different sampling series. By chance, we sampled it on 6 July, 1 day before the large flood (‘pre-flood’ data; see Fig. 1). One Surber sample (area 0.04 m2, mesh size 90 µm) was taken at random within each plot using a marked screwdriver to fix a sampling depth of 10 cm. On days 1, 3, 6 and 12 after the flood, we took another Surber sample in each of the same five plots (‘post-flood’ data from day 1 to day 12). Nineteen days after the flood on 7 July, we disturbed the five plots as described in detail by Matthaei et al. (1996). Starting with the one furthest downstream, four people simultaneously kicked and shuffled vigorously (similar to kick sampling) in four longitudinal rows within the plot. This was repeated four times, with each person changing rows each time. The whole procedure took µ 15 min. From each treated plot we then took a single Surber sample on six occasions. The first of these was taken immediately after the disturbance treatment on day 19, the following five on days 20, 22, 26, 29 and 40 after the flood (‘experimentally disturbed’ data from day 19 to day 40; Fig. 1). On day 19, we also started sampling the untreated second set of five plots. Again, a total of six Surber samples was taken from each plot using the same sampling intervals as for the treated plots. These samples served a dual purpose. First, they were used to continue monitoring the recovery of the invertebrates from the flood on 7 July (‘post-flood’ data from day 19 to day 40). Second, they served as controls for our disturbance experiment. We sampled each control

64 C.D. Matthaei, U. Uehlinger and A. Frutiger plot before kicking and raking the corresponding experimentally disturbed plot to avoid any interference. Within each plot, the positions of the six samples consecutively taken were grouped into three pairs, among which the order of sampling was chosen at random (see Fig. 1 in Matthaei et al., 1996). Within the pairs, the position further downstream was always sampled first to avoid influencing the other position. The positions of the four post-flood samples taken consecutively after the flood on 7 July were similarly chosen among two pairs of positions. After sampling the invertebrates, mean water velocity of each sampling area was measured with an OTT current meter at 60% water depth (Smith, 1975; rotor diameter 3 cm) as the average of measurements at two corners and in the centre of the sampling area. We then noted the water depth and estimated the degree of algal cover on the stones in each sampling area using a simple visual index: 1 5 less than 10% algal cover, 2 5 10– 33%, 3 5 33–66%, 4 5 66–90%, 5 5 .90%. Except for one sample taken on day 22 after the flood, there was no filamentous algal growth during our experiment. We are aware that this visual index can only provide a rough estimation of algal recovery, because the density of algal growth is not taken into account. All samples were frozen in the field with dry ice and later transferred to a –20 °C storage room at the laboratory. After thawing, samples were elutriated with water to separate organic and inorganic components. Using a modified subsampler after Meyer (1990), we halved all samples from control plots up to and including day 19 after the flood and all samples from experimentally disturbed plots up to and including day 22. Because of extremely high numbers of invertebrates (see below), we processed only 25% of subsamples from day 20 to day 40 (control treatment) and day 26 to day 40 (experimentally disturbed treatment). All invertebrates in the subsamples were counted under a stereomicroscope (WILD, 350) and identified to the lowest possible taxonomic level. Each taxon was sorted into 1-mm body length size classes using an ocular micrometer. Because benthos data showed non-normal distributions with unequal variances and outliers even after log transformation or square root transformation, they were analysed with two-sided Mann–Whitney U-tests and Friedman tests (Potvin & Roff, 1993). To minimize type II errors (Toft & Shea, 1983; Rotenberry & Wiens,

1985), exact P-values for non-significant results are given (Rice, 1989; Yoccoz, 1991), and ‘borderline cases’ are interpreted conservatively. The relative invertebrate abundances in post-flood, control and experimentally disturbed treatments (see Table 5) consisted of data from several successive benthic samples that had been taken in the same five plots. To avoid temporal pseudo-replication sensu Hurlbert (1984), we did not use statistical tests when comparing the relative abundances in these treatments with those in drift samples. Instead, we only compared the averages of each group of samples.

Drift sampling Invertebrate drift was sampled using four double-net drift samplers (opening 15 3 30 cm, raised 3–4 cm above the substratum, mesh size of the inner net 400 µm, of the outer net 90 µm). They were positioned µ 50 m upstream of the experimental areas to avoid disturbing the recolonization experiment (for more methodological details see Matthaei et al., 1996). The nets collected drifting animals for a period of 1 h just after dark on day 6 after the flood and on day 22, 3 days after the experimental disturbance. Mean water velocity at the mouth of each drift net was measured with an OTT current meter at 60% water depth (Smith, 1975; rotor diameter 3 cm) for 30 s at the beginning and the end of sampling. Hardly any clogging of the drift nets was observed. On day 6, mean water velocity at the end of sampling was 88% of the value measured immediately after the net was placed in position. The corresponding value for day 22 was 91%. Linear extrapolation was used to calculate the volume of water that passed through the nets during the sampling. All animals in the drift nets were taken to the laboratory alive, washed from the nets and frozen at – 20 °C. We counted all invertebrates in 25% of the subsamples, sorted them into size classes as described above, and determined the relative abundances of the common taxa in the drift for comparison with the benthic samples. For these taxa, we also calculated drift densities per 100 m3 and compared densities on day 6 with those on day 22 and those of a winter drift sampling (previously unpublished data, see Table 4). This winter drift sampling was conducted on 13 February 1993, 6 days after experimental disturbance at the same site using the same methods as in the © 1997 Blackwell Science Ltd, Freshwater Biology, 37, 61–77

Invertebrate response to natural v experimental disturbance 65 present study (Matthaei et al., 1996). The comparisons were done with two-sided Mann–Whitney U-tests.

Results To describe the results of the different sampling series, the following terms are used: ‘pre-flood’ 5 samples taken on day ‘–1’, just before the large flood; ‘postflood’ 5 samples taken on days 1–40 after the flood in plots with no experimental disturbance; ‘controls’ 5 samples taken on days 19–40 after the flood in plots with no experimental disturbance; ‘experimentally disturbed’ 5 samples taken on days 19–40 after the flood in experimentally disturbed plots.

Water velocity, depth and degree of algal cover Water velocities and depths in the experimental plots are summarized in Table 1. The mean degree of algal cover on stones in the sampled areas was less than 10% in post-flood plots on day 1 after the flood (algal cover index ‘1’, see Fig. 2), whereas pre-flood plots averaged more than 90% (algal cover index ‘5’). The disturbance experiment reduced algal cover by a similar degree. Algal cover of the post-flood plots recovered to pre-flood levels within 19 days of the flood. Recovery after the experimental disturbance was faster (10 days). The small spate on 11 August (day 35) reduced the degree of algal cover in both control and experimentally disturbed plots to µ 66– 90% (algal cover index ‘4’), because algae were removed from stones with a diameter of less than µ 2 cm.

Benthic invertebrates The relative abundances of most common taxa in the benthos samples (mean abundance of at least 250 ind. m–2 in all eighty-five samples) were similar in post-flood, control and experimentally disturbed samples (Table 2). Baetis spp. and Simuliidae were less abundant in pre-flood samples than in post-flood samples, whereas Leuctra spp. and Copepoda were more abundant. All samples were dominated by Chironomidae. Baetis spp. consisted mainly of B. rhodani Pictet, but we also found individuals of B. fuscatus Linne´ and occasionally B. vernus Curtis. We were unable to separate the small larvae of these three species or the two chironomid taxa Corynoneura spp. © 1997 Blackwell Science Ltd, Freshwater Biology, 37, 61–77

Table 1 Physical characteristics of experimental plots (averages 6 SD)

Sampling series

Velocity (cm s–1)

Water depth (cm)

Pre-flood (n 5 5) Post-flood (n 5 50) Controls (n 5 30) Experimentally disturbed (n 5 30)

42 6 7 50 6 17 49 6 17 43 6 17

16 6 3 17 6 8 15 6 5 16 6 5

and Thienemanniella spp., which we consider as a genus pair below. The meiofaunal taxon Copepoda consisted of at least eight species of Cyclopoida and four species of Harpacticoida (V. Kowarc, personal communication). The flood reduced the number of taxa per sample from twenty-two to eighteen (Fig. 2). However, this reduction was not significant (Table 3). After the smaller spate on day 2, taxon richness declined further to fifteen, a number significantly lower than the preflood value. Recovery to the pre-flood value did not occur until day 22. The experimental disturbance significantly reduced taxon richness from nineteen to eleven. Nevertheless, taxon richness attained control levels within 3 days. The total number of individuals was reduced by 90% 1 day after the flood and by 92% immediately following disturbance treatment (Fig. 2). Total invertebrate densities of post-flood plots increased from day 1 to day 22 after the flood (df 5 6, P , 0.001), then declined from day 26 to day 40 (df 5 2, P , 0.05). There was no difference between densities of preflood and post-flood samples 19 days after the flood (Table 3). From day 20 to day 29, post-flood densities exceeded pre-flood values. After the experimental disturbance, densities of experimentally disturbed and control plots were different until the tenth day of the experiment. Total invertebrate densities were generally high. Average pre-flood density was 67 210 ind. m–2, whereas that of all post-flood samples was 88 398 ind. m–2. Samples from all control plots of the disturbance experiment averaged 138 878 ind. m–2, those from all experimentally disturbed samples 76 153 ind. m–2. Individual resistances of the nine most common taxa to both types of disturbance can be seen in Figs 2 and 3. Resistance was generally similar, with the exception of Simuliidae and Copepoda. Simuliidae were hardly

66 C.D. Matthaei, U. Uehlinger and A. Frutiger

Fig. 2 Mean degree of algal cover (1 5 ø10%, 2 5 10–33%, 3 5 33–66%, 4 5 66–90%, 5 5 ù90%), taxon richness of benthic invertebrates, abundances of all invertebrates (Ntotal), Rhithrogena spp., Leuctra spp. and Hydracarina in pre-flood (m), post-flood (d) and experimentally disturbed (s) plots. Error bars represent SEs. Algal cover is shown without error bars, because it was not determined quantitatively. Some of the other errors are too small to be visible. Table 2 Abundances of common taxa in benthos samples as percentage of total invertebrates in the four sampling series (6 SE). Taxa with a mean abundance of less than 250 ind. m–2 (average of all eighty-five samples) are listed together under ‘others’. n 5 number of samples in each series

Chironomidae Baetis spp. Rhithrogena spp. Simuliidae Pentaneurini Leuctra spp. Hydracarina Copepoda Corynoneura/ Thienemanniella spp. Others

Pre-flood (day ‘–1’) (n 5 5)

Post-flood (days 1–40) (n 5 50)

Controls (days 19–40) (n 5 30)

Exp. disturbed (days 19–40) (n 5 30)

68.7 6 7.0 3.5 6 0.9 9.6 6 1.4 0.1 6 0.04 1.6 6 0.1 5.8 6 1.1 0.4 6 0.1 1.7 6 0.4 0.8 6 0.1

64.1 6 25.3 12.2 6 6.2 8.3 6 3.6 6.8 6 7.1 1.9 6 1.2 1.9 6 0.7 0.9 6 0.3 0.4 6 0.2 0.2 6 0.1

64.8 6 11.4 12.5 6 4.1 7.6 6 2.3 7.1 6 5.9 1.9 6 0.9 1.7 6 0.4 0.8 6 0.2 0.4 6 0.2 0.1 6 0.1

62.3 6 18.5 14.9 6 5.5 6.5 6 2.8 4.8 6 4.2 4.4 6 2.0 1.5 6 0.4 0.9 6 0.3 0.9 6 0.3 0.5 6 0.2

7.7 6 2.0

3.2 6 1.4

3.0 6 1.3

3.2 6 1.3

present before the flood and therefore were not reduced at all (see also Table 3), whereas their reduction by the experimental disturbance was close to 100%. In contrast, Copepoda were severely reduced by the flood but did not decrease significantly after the experimental disturbance treatment. The degree of reduction in the other taxa varied from 70 to 92%.

Resilience of taxa to the two types of disturbance differed considerably. Rhithrogena spp. showed similar recolonization rates and patterns after both types of disturbance. Population densities of post-flood plots equalled those of pre-flood plots within 12 days at the latest (Fig. 2, Table 3). Variation between replicates in control plots was very high. There was no difference © 1997 Blackwell Science Ltd, Freshwater Biology, 37, 61–77

Invertebrate response to natural v experimental disturbance 67 Table 3 Comparison of taxon richness, total invertebrates, and the abundances of common taxa before and after the large flood, and in experimentally disturbed v control treatments (Mann–Whitney U-tests). For non-significant results, exact P-values are given. *P , 0.05, **P , 0.01 Day of sampling series Pre-flood (day –1) v post-flood

Taxon richness Total invertebrates Rhithrogena spp. Leuctra spp. Hydracarina Chironomidae Baetis spp. Simuliidae Pentaneurini Corynoneura/ Thienemanniella spp. Copepoda

Exp. disturbed v controls

1

3

6

12

19

20

22

26

29

40

19

20

22

26

29

40

0.11 ** ** ** 0.24 ** ** 0.44 ** **

* ** ** ** 0.16 ** ** 0.82 ** **

* ** 0.08 ** 0.53 ** 0.07 0.41 ** **

* ** 0.25 ** * ** 0.25 0.11 ** 0.11

* 0.12 0.60 0.08 * 0.18 * * ** 0.17

* * 0.47 0.17 ** * ** * 0.92 0.12

0.12 ** 0.60 0.18 ** ** ** ** 0.34 0.09

0.09 ** 0.18 0.75 ** ** ** 0.17 * **

0.24 ** 0.12 0.30 ** * ** ** * *

0.75 0.08 0.25 0.08 ** 0.60 * * 0.12 *

** ** * * * ** ** ** 0.09 **

* ** 0.08 * * ** * 0.60 0.18 0.91

0.92 * 0.18 * * * 0.35 0.47 0.17 0.07

* * 0.12 0.06 0.08 0.18 0.18 0.75 0.68 **

0.06 0.14 0.09 * 0.53 0.08 0.75 0.92 0.12 *

0.14 0.75 0.83 0.46 0.52 0.47 0.60 0.60 * 0.60

**

**

**

0.12

0.09

0.06

0.17

0.92

0.08

*

0.67

0.92

0.21

0.25

0.67

0.13

between experimentally disturbed and control plots at least from the third day until the end of the experiment. Densities in post-flood plots never exceeded pre-flood values. Leuctra spp. also showed similar recolonization rates and patterns after both types of disturbance (Fig. 2). Abundances in postflood plots reached those of pre-flood plots 19 or 20 days after the flood (Table 3). Abundances in experimentally disturbed plots recovered to control levels 21 days after disturbance treatment. Hydracarina was the third taxon that recolonized similarly after both the flood and the experimental disturbance (Fig. 2). However, pre-flood densities were lower than those in post-flood plots from day 12 to day 40 (Table 3). Because of these low densities and variation between replicates, the 74% reduction in mean abundances caused by the flood was not significant. In contrast, experimentally disturbed densities needed at least 7 days to recover to control levels. Several taxa showed a distinct ‘lag phase’ after the flood until densities increased. For the Chironomidae, this lag phase was 6 days (Fig. 3; P 5 0.45 for no change from day 1 to day 6, df 5 2). Densities then increased until day 22 (df 5 4, P , 0.01), reaching preflood levels 19 days after the flood and exceeding them from day 20 to day 29 (Table 3). Population densities of experimentally disturbed plots attained those of controls 7 days after the experimental disturbance. Baetis spp. recolonized after a lag phase from day 1 to day 12 (Fig. 3; P 5 0.12, df 5 3). However, © 1997 Blackwell Science Ltd, Freshwater Biology, 37, 61–77

because pre-flood densities were quite low, they were reached within 12 days at the latest after the flood (Table 3) even with this non-significant increase. Postflood densities rose from day 12 to day 29 (df 5 5, P , 0.001) and exceeded pre-flood densities from day 19 to day 40. The decline from day 29 to day 40 was not significant for either control or experimentally disturbed plots. The latter were as densely populated as controls 3 days after disturbance treatment. Colonization of Simuliidae began after a lag phase of 12 days (P 5 0.24 for no change until day 12, df 5 3). Post-flood densities then increased until day 22 (df 5 3, P , 0.05), followed by a decline from day 22 to day 40 (df 5 3, P , 0.05). Pre-flood densities were lower than post-flood densities from days 19 to 40 (Table 3). Variation between replicates was extremely high during the experimental disturbance, especially in control plots. Densities of experimentally disturbed plots reached those of controls 1 day after disturbance. Pentaneurini (Chironomidae) recolonized after a lag phase of 19 days (Fig. 3; P 5 0.29 for no change until day 19, df 5 4). Post-flood densities then increased from day 19 to day 26 (df 5 3, P , 0.01), reaching preflood densities on day 20 and exceeding them on days 26 and 29 (Table 3). Because of low densities and variation between replicates in controls on day 19, the 70% reduction by the disturbance treatment was not significant. Densities remained similar until day 40 when experimentally disturbed densities were higher than control densities.

68 C.D. Matthaei, U. Uehlinger and A. Frutiger

Fig. 3 Abundances of Chironomidae, Baetis spp., Simuliidae, Pentaneurini, Corynoneura/Thienemanniella spp. and Copepoda in pre-flood (m), post-flood (d) and experimentally disturbed (s) plots. Error bars represent SEs. Some errors are too small to be visible.

Corynoneura/Thienemanniella spp. showed a domeshaped recovery pattern after both types of disturbance (Fig. 3). After a lag phase from day 1 to day 6 (df 5 2, P 5 0.37), densities in post-flood plots increased until day 19 (df 5 2, P , 0.05), reaching preflood densities on day 12 (Table 3). Mean densities declined from day 19 to day 26 (df 5 3, P 5 0.05) and increased from day 26 to day 40 (df 5 2, P , 0.05), but did not attain pre-flood levels. Densities in experimentally disturbed plots increased from day 19 to day 29 (df 5 4, P , 0.05), matching control levels within 1 day and exceeding them on days 26 and 29. Densities of Copepoda increased in post-flood plots from days 1 to 26 (df 5 7, P , 0.001; see Fig. 3), equalled pre-flood densities 12 days after the flood and remained roughly on that level until day 29 (Table 3). The decline from day 26 to day 40 was not significant (df 5 2, P 5 0.21). However, post-flood densities on day 40 were again lower than pre-flood densities. Copepoda abundances never differed between experimentally disturbed and control treatments.

Invertebrate drift Results of the two drift samplings 6 days after the flood and 3 days after the experimental disturbance (5 day 22 after the flood) are summarized in Table 4. Drift samples were dominated by Chironomidae, as were benthic samples (see Table 2). Hydracarina were relatively more common in drift than in benthic samples. Baetis spp., Rhithrogena spp., Simuliidae, Pentaneurini and Leuctra spp. were relatively less abundant in the drift. However, drift densities of these taxa were still quite high in comparison with the sampling on 13 February 1993 (Table 4). Mean total drift density on 13 February was five times lower than total density on day 6 after the flood and thirty-six times lower than total density on day 22. All differences in total drift densities between the three sampling occasions were significant (P , 0.05). Total density on day 6 after the flood was seven times lower than on day 22. Drift densities and percentages of Chironomidae, Simuliidae and Pentaneurini were higher on day 22 (P , 0.05). In spite of their lower © 1997 Blackwell Science Ltd, Freshwater Biology, 37, 61–77

Invertebrate response to natural v experimental disturbance 69 Table 4 Abundances of the common benthos taxa in drift samples as percentage of total invertebrates and drift densities per 100 m3 (averages of four drift nets 6 SE). The three last named taxa were not present in the drift samples taken on 13 February 1993

Total drift densities Chironomidae Baetis spp. Rhithrogena spp. Simuliidae Pentaneurini Leuctra spp. Hydracarina Copepoda Corynoneura/ Thienemanniella spp.

13 February 1993

13 July 1994 (day 6 after flood)

29 July 1994 (day 22)

Percentages

Drift densities

Percentages

Drift densities

Percentages

Drift densities

7.6 6 1.4 1.3 6 0.4 0.3 6 0.1 82.1 6 1.7 0.08 6 0.08 0.3 6 0.1

2878 6 483 224 6 70 36 6 11 11 6 5 2371 6 410 363 965

70.1 6 3.1 3.6 6 0.5 1.0 6 0.2 1.6 6 0.1 0.1 6 0.02 0.2 6 0.05 21.4 6 2.5 0.4 6 0.05 0.2 6 0.05

14846 6 1151 10453 6 1129 530 6 57 149 6 26 240 6 21 15 6 3 34 6 5 3153 6 364 58 6 8 28 6 9

91.4 6 0.8 0.7 6 0.06 0.02 6 0.01 3.0 6 0.4 0.5 6 0.05 0.02 6 0.01 3.8 6 0.5 0.2 6 0.04 0.1 6 0.01

104128 6 5458 95208 6 5354 724 6 90 20 6 8 3110 6 338 542 6 78 23 6 9 3907 6 476 198 6 47 96 6 17

percentages on day 22, drift densities of Baetis spp., Leuctra spp. and Hydracarina were not lower than on day 6 (P 5 0.25 for Baetis spp. and P 5 0.56 for Leuctra spp.). Drift densities of Corynoneura/Thienemanniella spp. and Copepoda were even higher on day 22 (P , 0.05). Only Rhithrogena spp. was less abundant on day 22 (P , 0.05).

Size-class distributions of common taxa in benthos and drift samples Chironomidae, Baetis spp., Rhithrogena spp., Leuctra spp., Pentaneurini and Simuliidae were found in sufficiently different sizes for determining a size-class distribution. For these taxa, the majority of individuals in all benthos and drift samples had a body length of 0–3 mm (Table 5). Consequently, we combined all individuals that were larger than 3 mm into a single size class for further analysis. This size class (. 3 mm, see Fig. 4) may provide information on the emergence of mature nymphs. The size class , 1 mm was also studied in detail, because it may provide information on the importance of hatching from eggs as a source of recolonists (see Fig. 5). We may have missed some Chironomidae hatchlings because they are very thin, but we repeatedly found invertebrate eggs as well as blackfly larvae that were still half within their eggshells. The Chironomidae were the only taxon where individuals . 3 mm differed in their relative abundances between all four sampling series (Fig. 4). Abundances in post-flood plots decreased from day 1 to day 19 © 1997 Blackwell Science Ltd, Freshwater Biology, 37, 61–77

(df 5 4, P , 0.05) and then increased until day 40 (df 5 5, P , 0.05). Relative abundances in experimentally disturbed plots increased from day 19 to day 29 (df 5 4, P , 0.01) and then declined to day 40 (P , 0.01). Percentages in post-flood plots were always lower than those in pre-flood plots (P , 0.05), whereas experimentally disturbed plots reached pre-flood levels at least on day 29 (P 5 0.83; day 26: P 5 0.08). Percentages in experimentally disturbed plots exceeded those in controls on days 20, 26 and 29 (P , 0.05). The other five taxa differed mainly in their relative abundances of the size class . 3 mm in the preflood samples compared with all other samples (Fig. 4). There were no Chironomidae , 1 mm in pre-flood samples (Fig. 5). After a lag phase of 3 days, relative abundances of this size class in post-flood plots increased until day 12 (df 5 2, P , 0.05), decreased from day 12 to day 19 (P , 0.05), and then remained constant until day 40 (df 5 5, P 5 0.19). Percentages in experimentally disturbed plots never differed from those in controls (P 5 0.46–0.92) and did not change with time (df 5 5, P 5 0.22). Relative abundances of Baetis spp. , 1 mm in post-flood plots increased from day 1 to day 22 (df 5 6, P , 0.01), followed by a decline until day 40 (df 5 3, P , 0.05). Pre-flood values were lower than post-flood and experimentally disturbed values, with the exceptions of day 3 (P 5 0.17) and day 40 (P 5 0.17 and P 5 0.08, respectively). Percentages in experimentally disturbed plots never differed from those in controls (P 5 0.11–0.92) and decreased with time (df 5 5, P , 0.01). Relative abundances of Rhithrogena spp. , 1 mm in post-flood plots

70 C.D. Matthaei, U. Uehlinger and A. Frutiger Table 5 Percentage contributions of the size classes , 1 mm and 0–3 mm to the total abundances of Chironomidae, Baetis spp., Rhithrogena spp., Leuctra spp., Pentaneurini and Simuliidae in drift samples, pre-flood, post-flood, control and experimentally disturbed plots (6 SE). n 5 number of samples in each series

Chironomidae , 1 mm 0–3 mm Baetis spp. , 1 mm 0–3 mm Rhithrogena spp. , 1 mm 0–3 mm Leuctra spp. , 1 mm 0–3 mm Pentaneurini , 1 mm 0–3 mm Simuliidae , 1 mm 0–3 mm

Pre-flood (day ‘–1’) (n 5 5)

Post-flood (days 1–40) (n 5 50)

Drift 1 (day 6) (n 5 4)

Controls (days 19–40) (n 5 30)

Exp. disturbed (days 19–40) (n 5 30)

Drift 2 (day 22) (n 5 4)

0.0 6 0.0 70.3 6 2.9

15.3 6 4.7 90.3 6 2.7

12.3 6 3.2 93.4 6 1.5

17.4 6 2.2 92.8 6 2.0

16.5 6 3.1 85.6 6 4.3

3.6 6 0.1 83.4 6 3.8

23.3 6 4.9 47.7 6 6.9

60.7 6 10.1 91.9 6 4.4

57.2 6 4.3 85.1 6 2.1

67.9 6 9.3 93.4 6 2.8

69.8 6 8.5 93.9 6 3.4

22.8 6 4.7 34.5 6 5.2

8.4 6 1.2 89.4 6 4.4

53.4 6 10.7 97.9 6 3.3

97.0 6 1.7 100.0 6 0.0

67.8 6 8.2 99.9 6 0.2

54.1 6 10.7 98.6 6 2.7

100.0 6 0.0 100.0 6 0.0

6.7 6 0.9 67.8 6 7.2

8.5 6 4.6 80.6 6 9.8

16.7 6 14.4 92.7 6 3.7

7.9 6 3.1 86.2 6 8.5

7.9 6 4.2 85.7 6 8.3

0.0 6 0.0 100.0 6 0.0

0.0 6 0.0 26.5 6 2.0

1.0 6 1.6 68.3 6 12.5

0.0 6 0.0 50.0 6 17.7

0.9 6 1.1 76.9 6 7.5

5.8 6 6.3 82.4 6 7.0

0.0 6 0.0 62.2 6 18.4

25.0 6 12.4 50.0 6 22.4

45.5 6 15.8 86.0 6 11.4

9.0 6 2.3 77.5 6 3.8

42.7 6 14.5 92.9 6 8.4

49.4 6 14.2 96.7 6 2.7

14.3 6 4.1 91.4 6 7.0

Fig. 4 Relative contribution of the size class .3 mm to the total abundances of Chironomidae, Baetis spp., Rhithrogena spp., Leuctra spp., Pentaneurini and Simuliidae in pre-flood (m), post-flood (d) and experimentally disturbed (s) plots. Error bars represent SEs. Some errors are too small to be visible.

© 1997 Blackwell Science Ltd, Freshwater Biology, 37, 61–77

Invertebrate response to natural v experimental disturbance 71

Fig. 5 Relative contribution of the size class ,1 mm to the total abundances of Chironomidae, Baetis spp., Rhithrogena spp., Leuctra spp., Pentaneurini and Simuliidae in pre-flood (m), post-flood (d) and experimentally disturbed (s) plots. Error bars represent SEs. Some errors are too small to be visible.

increased from day 1 to day 26 (df 5 7, P , 0.01) and then declined until day 40 (df 5 2, P , 0.05). With the exception of the experimentally disturbed samples on day 19 (P 5 0.12), pre-flood values were lower than post-flood and experimentally disturbed values (P , 0.05). The latter differed from controls solely on day 20 (P , 0.05), increased from day 19 to day 26 (df 5 3, P , 0.01) and decreased from day 29 to day 40 (P , 0.05). For Leuctra spp., Pentaneurini and Simuliidae, the size class , 1 mm showed few differences with time and between sampling series (Fig. 5).

disturbance to differ considerably from those after our experimental disturbance. Our results provide evidence that the experiment imitated several important aspects of invertebrate recolonization after the flood quite accurately, however, in spite of its smaller spatial scale. Due to the scarcity of similar studies, our results can be directly compared solely with those of Brooks & Boulton (1991) who investigated invertebrate recolonization in an Australian temporary stream after experimentally disturbing areas of 0.25 m2 and after a spate that occurred 7 days after the start of their experiment.

Discussion The main purpose of this work was to test to what extent a spate could be simulated by a disturbance experiment. The flood on 6 July of 166 m3 s–1 exceeded bankfull discharge (76 m3 s–1, calculated with a recurrence interval of 1.5 years). It caused extensive substratum movement not only in the River Necker but also in its smaller tributaries. Therefore, we expected recovery rates and pathways after this large-scale © 1997 Blackwell Science Ltd, Freshwater Biology, 37, 61–77

Similar resistances and recovery patterns—except for a lag phase The invertebrate community and most of the common taxa exhibited low resistance to both types of disturbance (Figs 2 and 3). The total number of individuals was reduced by 90% 1 day after the flood and by 92% immediately after disturbance treatment (Fig. 2). The experiment of Brooks & Boulton (1991) reduced total

72 C.D. Matthaei, U. Uehlinger and A. Frutiger invertebrate densities by 97%, whereas reduction by their spate was 70%. Taxon richness was lowered significantly by our disturbance experiment, but not by the flood on day 0 (Fig. 2). Brooks & Boulton’s experimental disturbance lowered taxon richness by 83% compared with only 45% reduction after their spate. These findings suggest that experimental disturbances have a more severe impact on taxon richness than natural disturbances. The reason for this more severe impact could be that experimental disturbances are very sudden, whereas discharge increases more gradually during a spate. Nevertheless, the second spate on day 2 (see Materials and methods) did reduce taxon richness significantly in comparison with preflood densities in our study, implying that two such events in rapid succession may have a more severe impact than a single one, even if the scale of substratum movement is much smaller. This may have important consequences for the invertebrate fauna in rivers like the River Necker where several spates may occur in rapid succession (Uehlinger et al., 1996). With the exception of the Copepoda, the recolonization patterns of all studied taxa were similar after both types of disturbance (Figs 2 and 3). Chironomidae, Baetis spp., Simuliidae, Pentaneurini and Corynoneura/ Thienemanniella spp. all showed a distinct lag phase after the flood before densities increased, whereas there was no such lag phase after the experiment. Therefore, the time needed to recover to pre-flood densities was longer for these taxa. Nevertheless, recolonization rates and patterns after the lag phase were similar to those after the experimental disturbance. There are three potential explanations for this lag phase of 6–19 days, all fairly tentative. First, the lag phase could be related to the mobility of these taxa, i.e. their colonization abilities (B. Peckarsky, personal communication). The fastest recolonizers after experimental disturbance should then show the shortest lag phase after the larger scale flood disturbance. However, our data do not support this explanation, because most of the fastest recolonizers after the experimental disturbance in summer (Simulidae, Baetis spp. and Corynoneura/Thienemanniella spp.) did show a lag phase, whereas the slowest recolonizers (Hydracarina and Leuctra spp.) did not. Moreover, we had observed the same general recolonization patterns for Simuliidae, Baetis spp. and Leuctra spp. (fast, fast and

slow, respectively) after our experimental disturbance in winter 1993 (Matthaei et al., 1996). The second explanation attempts to relate the lag phase to the distance from potential sources of recolonists. Because the six taxa showing a lag phase after the flood were mostly faster recolonizers than the three taxa that did not, they may have colonized other areas of stream bed first before reaching our experimental plots (A.G. Hildrew, personal communication). This would imply that their sources of colonists (i.e. their refugia during the flood; see Townsend, 1989; Sedell et al., 1990; Lancaster & Hildrew, 1993) were further from our experimental area than those of the three taxa that recolonized without a lag phase. Examples for potentially important refugia in the River Necker are the areas adjacent to boulders, the bed interstices or the channel periphery. All the same, because we did not sample in any of these habitats, we cannot prove or disprove this hypothesis in the present study. In addition, very few data are available to verify whether benthic organisms really seek the proposed refugia during spates (see Giberson & Hall, 1988; Palmer, Bely & Berg, 1992). We will return to the question of potential refugia in the River Necker during the large flood later. Finally, it is possible that initial recruitment of the six taxa that colonized with a lag phase could have occurred mainly through egg hatching after the flood. This explanation was suggested by the larger percentage of instars of the size class , 1 mm in post-flood samples compared with pre-flood samples (Table 5). Elliott (1972) showed that hatching of Baetis rhodani started after 7 days at 22 °C and was completed 3 days later. Development times from egg laying to emergence can be as short as 6–15 days for Chironomidae (Gray, 1981; Stites & Benke, 1989), 9 days for Simuliidae (Hauer & Benke, 1987) and 10–19 days for Baetis spp. (Gray, 1981; Benke & Jacobi, 1986), at least in the southern U.S.A. Orthocladius calvus developed within 16 days at 52°N in England (Ladle et al., 1985; all references cited in Mackay, 1992). As water temperatures in the lower Necker may exceed 20 °C during summer (Burgherr, 1994), development times could be similarly short. If a ‘seed bank’ of invertebrate eggs existed in the deeper sediments (as proposed, for example, for plecopterans by Zwick, 1996), the flood could have somehow induced hatching of these eggs. Nevertheless, our data on the six taxa for which size © 1997 Blackwell Science Ltd, Freshwater Biology, 37, 61–77

Invertebrate response to natural v experimental disturbance 73 classes could be determined do not support the seed bank theory, as shown in the next paragraph.

Invertebrate drift—the major mode of recolonization For these six taxa, we examined changes in the relative abundances of individuals with a body size of , 1 mm in benthic samples with time after both types of disturbance (Fig. 5) and compared their averages with those in the drift samples (Table 5). We then studied the total drift densities of these taxa and their quotas relative to benthic abundances (Table 4), and finally assessed the most probable mode of recolonization for each taxon. The quotas of individuals , 1 mm of Chironomidae, Baetis spp. and Rhithrogena spp. were higher in post-flood and control samples than in preflood samples, and reached up to 70% of the total numbers of these taxa in post-flood and experimentally disturbed plots (Table 5). Simuliidae , 1 mm also reached high percentage values. The individuals , 1 mm of Chironomidae, Baetis spp. and Simuliidae were similarly common in benthos and drift samples (Table 5). They also reached high total drift densities (Table 4). Therefore, drift was probably the most important means of recolonization for these taxa. The drift densities of Rhithrogena spp. and Leuctra spp. were quite low in comparison with benthic densities in general (Table 4), and on days 6 and 22 (the days when drift was sampled) in particular (Fig. 5). For Rhithrogena spp., this indicates that hatching from a seed bank within the experimental plots could have been the main method of recolonization. All the same, this taxon recolonized without a lag phase after the flood (Fig. 2). For Leuctra spp. and Pentaneurini, the relative abundances of individuals , 1 mm were similarly low as the drift values and did not differ between pre-flood and most remaining samples (Table 4, Fig. 5). Therefore, hatching from a seed bank within the experimental plots was probably not important in spite of low drift densities (Leuctra spp.), and the low relative abundances of hatchlings in the drift (Pentaneurini). Relative abundances of the three taxa for which size classes were not studied (Corynoneura/Thienemanniella spp., Copepoda and Hydracarina) were similar or higher in drift samples than in benthic samples (Table 4). As these taxa are not known to swim or crawl well, they may have also recolonized mainly by drift. The prevailing importance of drift for recolonization © 1997 Blackwell Science Ltd, Freshwater Biology, 37, 61–77

implied by our findings is supported by the results of Werthmu¨ller (1995), who exposed substratum trays shortly after a spate in May 1995, at the same site where our study had been conducted the year before. All studied taxa colonized the trays almost exclusively by drift and not by horizontal or vertical migrations. As in the present study, the relative abundances of several common taxa were lower in the drift than in benthic samples. However, this apparent contradiction could be resolved because the added numbers of common taxa drifting over the trays during his experiment were generally much higher (19–3180 times) than their benthic densities. Another indicator of the importance of drift for recolonization in the River Necker is the high total drift densities (14 846 individuals per 100 m3 on day 6 and 104 128 individuals per 100 m3 on day 22). Slack, Tilley & Kennelly (1991) reported a total invertebrate drift of 14 136 individuals per 100 m3 in a Rocky Mountain stream during summer using a mesh size of 106 µm. Moog & Heinisch (1991) found an average drift density of 1400 individuals per 100 m3 during a 2-year study in an Austrian mountain stream using a mesh size of 112 µm. Williams (1985) studied chironomid drift in the River Chew in England using pump samples that were filtered through a 50-µm net. Chironomid densities in the pumped samples often exceeded 10 000 individuals per 100 m3, with a maximum of 40 000 individuals per 100 m3. Although we used a 90-µm net, chironomid drift densities in the River Necker were similarly high or even higher than in the River Chew (Table 4). Drift was probably similarly important for recolonization after both types of disturbance. All six studied taxa showed few or no differences in relative abundances of individuals , 1 mm between control and experimentally disturbed plots (Fig. 5). For Leuctra spp., Pentaneurini and Simuliidae of this size class, values for control and experimentally disturbed treatments were also similar to those obtained in the first 20 days of the post-flood series. This indicates that colonists of this size class probably came from the same source. Post-flood values of Chironomidae and Baetis spp. , 1 mm were lower on days 1 and 3 after the flood than the corresponding values in experimentally disturbed plots. Individuals , 1 mm of Rhithrogena spp. were relatively more abundant in experimentally disturbed plots 3–10 days after experimental disturbance than 3–12 days after the flood.

74 C.D. Matthaei, U. Uehlinger and A. Frutiger However, all three taxa showed similar mean values in post-flood, control and experimentally disturbed plots (Table 5). Total drift density on day 6 after the flood was seven times lower than on day 22 (Table 4). Total benthic density on day 22 was also much higher than on day 6 (Fig. 2). This implies that drift rates of the studied invertebrates probably did not increase after the flood to allow a faster recolonization of denuded areas (although we do not have drift data for the first 5 days after the flood). It seems that drift densities were simply proportional to benthic densities. The same correlation was found by Sagar & Glova (1992) in a New Zealand river.

Where do the drifting invertebrates come from? In our previous winter study (Matthaei et al., 1996), we argued that the harsh, unpredictable disturbance regime of the prealpine River Necker (Uehlinger, 1991; Uehlinger et al., 1996) should select for a benthic fauna that is well adapted to disturbance by bed-moving spates. Indeed, we found that the benthic invertebrate fauna was highly resilient compared with those of other streams. This claim is further supported by the results of the present study, because total invertebrate densities recovered to control levels after the experimental disturbance within 10 days (Fig. 2), whereas they needed 30 days in winter. Even after the severe flood on 7 July that caused a more than 200-fold increase in discharge, recovery to pre-flood levels occurred within only 19 days. This is a much lower figure than after major floods in other streams where invertebrates typically needed two to several months to recover to pre-flood densities (see reviews by Wallace, 1990; Lake, 1990; Mackay, 1992), even if these streams also had a flashy flow regime (Gray, 1981; Fisher et al., 1982). The exceptionally high resilience and the prevailing importance of drift for recolonization after the flood imply that invertebrates in the River Necker must be able to avoid the destructive effects of major spates rather efficiently. This brings us back to the question of which refugia could have been used by the invertebrates during the large flood (Townsend, 1989; Sedell et al., 1990; Lancaster & Hildrew, 1993). As shown above, egg hatching from a seed bank within our experimental plots was probably not important for recovery in the present study. Nevertheless, the size

class , 1 mm made up a substantial percentage of the total number of individuals of common taxa (Table 5). This indicates that egg hatching in general did provide an important pool of colonists. After hatching, the larvulae probably entered the drift for dispersal (Mackay, 1992). Therefore, at the spatial scale of an entire stream reach, seed banks in the deeper substrata (or the presence of ovipositioning adults) appear to be a major refuge for the invertebrate community of the River Necker. All the same, the source of the drifting invertebrates (i.e. where exactly they hatched after the large flood) remains unknown.

Predominance of small body size and its implications for disturbance studies In the pre-flood samples taken on 6 July, only 38% of all invertebrates were . 3 mm. The flood on 7 July reduced this percentage to 10–15% for the three most common taxa, Chironomidae, Baetis spp. and Rhithrogena spp. (Fig. 4). All six taxa whose size classes were studied did not exceed their pre-flood percentages during the whole sampling period and only reached lower values on most sampling occasions. Consequently, the summer fauna of the River Necker probably generally consists largely of very small individuals. This assumption is supported by the much lower invertebrate densities found in other studies conducted in the lower Necker that used a mesh size of 250 µm (Frutiger, 1983; Werthmu¨ller, 1995). In addition to seriously underestimating invertebrate densities, a too coarse mesh size can even prevent the detection of recovery processes in the River Necker. Moos (1994) studied the effects of the two spates on 24 May and 10 June 1994 (see Fig. 1) on the benthic fauna at a site 2 km downstream of our study site. Using a 250-µm net, he counted total invertebrate abundances of µ 1000 ind. m–2 and concluded that there was hardly any recovery 20 days after the second spate. Only 6 days later, we found 67 210 ind. m–2 using a 90-µm net.

Effects of a long period of stable flow on the benthic invertebrate community The recurrence interval of a spate with at least partial substratum movement at our study site is only 22 days (Matthaei et al., 1996). Therefore, the 41 days of stable discharge registered during the present study © 1997 Blackwell Science Ltd, Freshwater Biology, 37, 61–77

Invertebrate response to natural v experimental disturbance 75 are rare in the River Necker. The rise and subsequent fall of the total invertebrate and Chironomidae densities (Figs 2 and 3) may be connected with this long period of stable flow. Our sampling series extended through the whole period without significant substratum movement (Fig. 1). Total invertebrate densities were high, especially in the control samples taken 19– 40 days after the flood (138 878 ind. m–2). We found much lower densities in the control samples of our winter experiment, 25–55 days after a spate with a peak discharge of only 20 m3 s–1 (38 530 ind. m2; Matthaei et al., 1996). This indicates that the carrying capacity of the substratum is higher during summer. Total invertebrate and Chironomidae densities exceeded pre-flood densities from day 20 to day 29 after the flood. However, they returned to pre-flood levels 40 days after the flood. This suggests that densities surpassed the carrying capacity of the substratum during the long period of stable discharge. This ‘overcrowding’ may have caused the rapid subsequent decline that cannot be attributed to emergence, because the percentage of Chironomidae . 3 mm did not decrease at the same time (Fig. 4). No similar decline occurred during the winter series, when total invertebrate densities were much lower.

Conclusion The extent to which our experimental disturbance simulated a natural spate exceeded our expectations. The degree of reduction and the fundamental recolonization patterns of most studied taxa were similar after both types of disturbance, as well as the prevailing importance of drift for recolonization. These results differ from those of Brooks & Boulton (1991), who experimentally disturbed patches of 0.25 m2 and found a different faunal composition as well as different rates and pathways of recolonization after the two types of disturbance. We compared our experimental disturbance with a flood with a recurrence interval of µ 5 years that caused extensive substratum movement in the whole stream channel as well as in the smaller tributaries. Smaller spates with a maximum discharge of 16–30 m3 s–1, that induce substratum movement at a spatial scale similar to our experimental plots (C.D. Matthaei, estimation from field observations), are much more common in the River Necker (Uehlinger et al., 1996). Spates that cause only limited substratum movement have been registered in other prealpine © 1997 Blackwell Science Ltd, Freshwater Biology, 37, 61–77

streams. In the study of Clauss (1992), a spate reduced algal cover on stones and benthic invertebrates solely in the central 9 m2 of a 3 m 3 10 m transect across the whole width of the Lunzer Seebach in Austria. Scha¨lchli (1993) describes the ‘fine sediments transport’ as a typical phenomenon of prealpine and alpine rivers. Fine sediments move over a still stable stream bed of coarser sediments and scour only the very surface of the stream bed. For disturbances on this spatial scale, our experiment should provide an even more realistic simulation of natural spates. Recolonization of denuded patches after smaller spates may be better described as redistribution (Townsend & Hildrew, 1976) than as ‘true recolonization’, e.g. after a catastrophic flood. Nevertheless, such spates are presumably more important for the spatio-temporal dynamics of the benthic community, since they are more frequent than large floods.

Acknowledgments We thank all the enthusiastic ‘disturbers’ who assisted with the field work. A big thanks goes to Dr Verena Kowarc for looking at our Copepoda, even if it turned out that we could not distinguish reliably between the different species. The critical comments of Prof. J.V. Ward and two anonymous reviewers helped us to improve on earlier versions of this manuscript. We are also grateful to Profs B. Peckarsky and A.G. Hildrew for their ideas concerning potential reasons for the lag phase after the flood.

References Benke A.C. & Jacobi D.J. (1986) Growth rates of mayflies in a subtropical river and their implications for secondary production. Journal of the North American Benthological Society, 5, 107–114. Boulton A.J., Peterson C.G., Grimm N.B. & Fisher S.G. (1992) Stability of an aquatic macroinvertebrate community in a multi-year hydrologic disturbance regime. Ecology, 73, 2192–2207. Brooks S.S. & Boulton A.J. (1991) Recolonization dynamics of benthic invertebrates after artificial and natural disturbances in an Australian temporary stream. Australian Journal of Marine and Freshwater Research, 42, 295–308. Burgherr P. (1994) Struktur der MakroinvertebratenGesellschaft und Habitatscharakteristika im La¨ngsverlauf

76 C.D. Matthaei, U. Uehlinger and A. Frutiger eines voralpinen Fliessgewa¨ssers (Necker, SG). Diploma thesis, ETH/EAWAG. Clauss K. (1992) Horizontale ‘patch’-Verteilung in der Schotterfauna eines alpinen Baches 2. Ordnung. Diploma thesis, University of Regensburg. Clifford H.F. (1982) Effects of periodically disturbing a small area of substratum in a brown-water stream of Alberta, Canada. Freshwater Invertebrate Biology, 1, 39–47. Death R.G. & Winterbourn M.J. (1995) Diversity patterns in stream benthic invertebrate communities: the influence of habitat stability. Ecology, 76, 1446–1460. Detenbeck N.E., DeVore P.W., Niemi G.J. & Lima A. (1992) Recovery of temperate-stream fish communities from disturbance: a review of case studies and synthesis of theory. Environmental Management, 16, 33–53. Doeg T.J., Lake P.S. & Marchant R. (1989) Colonization of experimentally disturbed patches by stream macroinvertebrates in the Acheron River, Victoria. Australian Journal of Ecology, 14, 207–220. Elliott J.M. (1972) Effects of temperature on the time of hatching in Baetis rhodani (Ephemeroptera: Baetidae). Oecologia, 9, 47–51. Fisher S.G., Gray L.J., Grimm N.B. & Busch D.E. (1982) Temporal succession in a desert stream following flash flooding. Ecological Monographs, 52, 93–110. ¨ kologie der Frutiger A. (1983) Untersuchungen zur O ra¨uberischen Steinfliege Dinocras cephalotes Curt. (Plecoptera Perlidae) in einem Fliessgewa¨sser der Schweizerischen Voralpen. PhD thesis, ETH. Giberson D.J. & Hall R.J. (1988) Seasonal variation in faunal distribution within the sediments of a Canadian Shield stream, with emphasis on responses to spring floods. Canadian Journal of Fisheries and Aquatic Sciences, 45, 1994–2002. Gray L.J. (1981) Species composition and life histories of aquatic insects in a lowland Sonoran desert stream. American Midlands Naturalist, 106, 229–242. Hauer F.R. & Benke A.C. (1987) Influence of temperature and river hydrograph on black fly growth rates in a subtropical blackwater river. Journal of the North American Benthological Society, 6, 251–261. Hurlbert S.H. (1984) Pseudoreplication and the design of ecological field experiments. Ecological Monographs, 54, 187–211. Johnson S.L. & Vaughn C.C. (1995) A hierarchical study of macroinvertebrate recolonization of disturbed patches along a longitudinal gradient in a prairie river. Freshwater Biology, 34, 531–541. Ladle M., Cooling D.A., Welton J.S. & Bass J.A.B. (1985) Studies on Chironomidae in experimental recirculating stream systems. II. The growth, development and production of a spring generation of Orthocladius

(Euorthocladius) calvus Pinder. Freshwater Biology, 15, 243–255. Lake P.S. (1990) Disturbing hard and soft bottom communities: a comparison of marine and freshwater environments. Australian Journal of Ecology, 15, 477–488. Lamberti G.A. & Steinman A.D. (1993) Research in artificial streams: applications, uses, and abuses. Journal of the North American Benthological Society, 12, 313–384. Lancaster J. & Hildrew A.G. (1993) Flow refugia and microdistribution of lotic macroinvertebrates. Journal of the North American Benthological Society, 12, 385–393. Mackay R.J. (1992) Colonization by lotic macroinvertebrates: a review of processes and patterns. Canadian Journal of Fisheries and Aquatic Sciences, 49, 617–628. Matthaei C.D., Uehlinger U., Meyer E.I. & Frutiger A. (1996) Recolonization of benthic invertebrates after experimental disturbance in a Swiss prealpine river. Freshwater Biology, 35, 233–248. Meyer E. (1990) A simple subsampling device for macroinvertebrates with general remarks on the processing of stream benthos samples. Archiv fu¨r Hydrobiologie, 117, 309–318. Minshall G.W. (1988) Stream ecosystem theory: a global perspective. Journal of the North American Benthological Society, 7, 263–288. Moog O. & Heinisch W. (1991) Macroinvertebrate drift in a fourth-order crystalline mountain stream (Wagreiner Aache, Austria). Verhandlungen der Internationalen Vereinigung fu¨r Theoretische und Angewandte Limnologie, 24, 1897–1907. Moos T. (1994) Die Wirkung unterschiedlich starker, geschiebefu¨hrender Hochwasser in einem voralpinen Fliessgewa¨sser auf die benthische Biozo¨nose und den zeitlichen Verlauf der Wiederbesiedlung. Diploma thesis, ETH/EAWAG. Palmer M.A., Bely A.E. & Berg K.E. (1992) Response of invertebrates to lotic disturbance: a test of the hyporheic refuge hypothesis. Oecologia, 89, 182–194. Potvin C & Roff D.A. (1993) Distribution-free and robust statistical methods: viable alternatives to parametric statistics? Ecology, 74, 1617–1628. Reice S.R. (1985) Experimental disturbance and the maintenance of species diversity in a stream community. Oecologia, 67, 90–97. Resh V.H., Brown A.V., Covich A.P., Gurtz M.E. Li H.W., Minshall G.W., Reice S.R., Sheldon A.L., Wallace J.B. & Wissmar R.C. (1988) The role of disturbance in stream ecology. Journal of the North American Benthological Society, 7, 433–455. Rice W.R. (1989) Analyzing tables of statistical tests. Evolution, 43, 223–225. Robinson C.T. & Minshall G.W. (1986) Effects of © 1997 Blackwell Science Ltd, Freshwater Biology, 37, 61–77

Invertebrate response to natural v experimental disturbance 77 disturbance frequency on stream benthic community structure in relation to canopy cover and season. Journal of the North American Benthological Society, 5, 237–248. Robinson C.T., Reed L.M. & Minshall G.W. (1992) Influence of flow regime on life history, production, and genetic structure of Baetis tricaudatus (Ephemeroptera) and Hesperoperla pacifica (Plecoptera). Journal of the North American Benthological Society, 11, 278–289. Rosser Z.C. & Pearson R.G. (1995) Responses of rock fauna to physical disturbance in two Australian tropical rainforest streams. Journal of the North American Benthological Society, 14, 183–196. Rotenberry J.T. & Wiens J.A. (1985) Statistical power analysis and community-wide patterns. American Naturalist, 125, 164–168. Sagar P.M. & Glova G.J. (1992) Invertebrate drift in a large, braided New Zealand river. Freshwater Biology, 27, 405–416. Scarsbrook M.R. & Townsend C.R. (1993) Stream community structure in relation to spatial and temporal variation: a habitat template study of two contrasting New Zealand streams. Freshwater Biology, 29, 395–410. Scha¨lchli (1993) Die Kolmation von Fliessgewa¨ssersohlen: Prozesse und Berechnungsgrundlagen. PhD thesis, ETH. Sedell J.R., Reeves G.H., Hauer F.R., Stanford J.A. & Hawkins C.P. (1990) Role of refugia in recovery from disturbances: modern fragmented and disconnected river systems. Environmental Management, 14, 711–724. Slack K.V., Tilley L.J. & Kennelly S.S. (1991) Mesh-size effects on drift sample composition as determined with a triple net sampler. Hydrobiologia, 209, 215–226. Smith I.R. (1975) Turbulence in lakes and rivers. FBA Scientific Publication 29. FBA, Ambleside. Smock L.A., Smith L.C., Jones J.B. & Hooper S.M. (1994) Effects of drought and a hurricane on a coastal headwater stream. Archiv fu¨r Hydrobiologie, 131, 25–38. Steinman A.D. & McIntire C.D. (1990) Recovery of lotic periphyton communities after disturbance. Environmental Management, 14, 589–604. Stites D.L. & Benke A.C. (1989) Rapid growth rates of

© 1997 Blackwell Science Ltd, Freshwater Biology, 37, 61–77

chironomids in three habitats of a subtropical blackwater river and their implications for P-B ratios. Limnology and Oceanography, 34, 1278–1289. Toft C.A. & Shea P.J. (1983) Detecting community-wide patterns: estimating power strengthens statistical inference. American Naturalist, 122, 618–625. Townsend C.R. (1989) The patch dynamics concept of stream community ecology. Journal of the North American Benthological Society, 8, 36–50. Townsend C.R. & Hildrew A.G. (1976) Field experiments on the drifting, colonization and continuous redistribution of stream benthos. Journal of Animal Ecology, 45, 759–772. Uehlinger U. (1991) Spatial and temporal variability of the periphyton biomass in a prealpine river (Necker, Switzerland). Archiv fu¨r Hydrobiologie, 123, 219–237. Uehlinger U., Bu¨hrer H. & Reichert P. (1996) Periphyton dynamics in a floodprone prealpine river: evaluation of significant processes by modelling. Freshwater Biology, 36, 249–263. Wallace J.B. (1990) Recovery of lotic macroinvertebrate communities from disturbance. Environmental Management, 14, 605–620. Werthmu¨ller D. (1995) Die Bedeutung der Drift fu¨r die kurzfristige Wiederbesiedlungs-Dynamik im Necker. Diploma thesis, ETH/EAWAG. Williams C.J. (1985) A comparison of net and pump sampling methods in the study of Chironomidal larval drift. Hydrobiologia, 124, 243–250. Yoccoz N.G. (1991) Use, overuse, and misuse of significance tests in evolutionary biology and ecology. Bulletin of the Ecological Society of America, 72, 107–111. Yount J.D. & Niemi G.J. (1990) Recovery of lotic communities and ecosystems from disturbance—a narrative review of case studies. Environmental Management, 14, 547- 569. Zwick P. (1996) Variable egg development of Dinocras spp. (Plecoptera, Perlidae) and the stonefly seed bank theory. Freshwater Biology, 35, 81–100. (Manuscript accepted 7 September 1996)