Canadian Journal of Fisheries and Aquatic Sciences
Reach and mat scale differences in Microcoleus autumnalis (cyanobacterium) accrual along velocity and nitrate gradients in three New Zealand rivers
Journal: Canadian Journal of Fisheries and Aquatic Sciences
Manuscript ID cjfas-2019-0133.R2
Manuscript Type: Article
Date Submitted by the 13-Jul-2019 Author:
Complete List of Authors: McAllister, Tara; The University of Auckland, Te Pūnaha Matatini Wood, Susanna; Cawthron Institute Mackenzie,Draft Emma; University of Canterbury Hawes, Ian; University of Waikato
Keyword: Phormidium, growth rates, velocity, NUTRIENTS < General
Is the invited manuscript for consideration in a Special Not applicable (regular submission) Issue? :
https://mc06.manuscriptcentral.com/cjfas-pubs Page 1 of 41 Canadian Journal of Fisheries and Aquatic Sciences
1 Reach and mat scale differences in Microcoleus autumnalis (cyanobacterium) accrual along
2 velocity and nitrate gradients in three New Zealand rivers
3 4 Tara G. McAllister 5 Susanna A. Wood 6 Emma M. MacKenzie 7 Ian Hawes 8 9 Tara McAllister- Te Pūnaha Matatini, University of Auckland, Auckland 10 ([email protected]) 11 Susanna Wood- Cawthron Institute, Private Bag 2, Nelson, New Zealand 12 ([email protected]) 13 Emma MacKenzie- Waterways Centre for Freshwater Management, University of Canterbury, 14 Christchurch, New Zealand ([email protected]) 15 Ian Hawes- Coastal Marine Field Station, University of Waikato, 58 Cross Road, Tauranga, New 16 Zealand ([email protected]) 17 18 19 Contact author: 20 Tara G. McAllister 21 Te Pūnaha Matatini, University of Auckland,Draft Address 22 Email: [email protected] 23 24
https://mc06.manuscriptcentral.com/cjfas-pubs Canadian Journal of Fisheries and Aquatic Sciences Page 2 of 41
25 Abstract 26 Proliferations of the toxic, mat-forming cyanobacterium Microcoleus autumnalis are an
27 increasingly recognized problem in cobble bed rivers worldwide. This study explored how
28 flow and nutrient concentrations influence mat expansion. M. autumnalis was inoculated into
29 cobbles placed in runs, riffles and pools in three rivers with different nutrient conditions and
30 mat size was monitored over 21 days. The following hypotheses were tested: (1) mat
31 expansion will reflect cover increases at the reach scale; (2) biomass and cover will be
32 highest in high velocity habitats; and (3) under similar velocities, nutrient concentrations will
33 be more important than other abiotic and biotic variables in determining expansion rates. Mat
34 expansion accurately reflected the increase in reach-scale cover, and expansion was most
35 rapid at intermediate water velocities (0.25–0.45 m s-1). Mats persisted the longest in riffles. 36 Accrual cycles were terminated earlierDraft in runs than riffles, as high expansion rates resulted in 37 patches reaching maximum mat size rapidly. Although M. autumnalis accrual differed among
38 rivers, this was attributed to differences in shear stress and grazing pressure rather than
39 nutrient concentrations.
https://mc06.manuscriptcentral.com/cjfas-pubs Page 3 of 41 Canadian Journal of Fisheries and Aquatic Sciences
40 Introduction
41 Toxic planktonic cyanobacterial blooms have been a recognized water quality problem for
42 centuries (Francis 1878; Kirkby 1672; Paerl et al. 2001). Numerous studies investigating
43 physicochemical factors influencing the growth and bloom formation of planktonic species
44 have been undertaken, and a relatively robust understanding has developed (see(Oliver et al.
45 2012). In contrast, toxic benthic cyanobacterial proliferations have only recently become
46 recognized as an escalating problem in freshwater environments worldwide (Quiblier et al.
47 2013). Microcoleus autumnalis (formerly Phormidium autumnale) and closely related taxa
48 are benthic, mat-forming cyanobacteria which have become increasingly problematic in
49 cobble-bed rivers worldwide (Aboal et al. 2002; Fetscher et al. 2015; Gugger et al. 2005;
50 McAllister et al. 2016; Quiblier et al. 2013). M. autumnalis and closely related taxa produces
51 a variety of cyanotoxins including anatoxin-aDraft (ATX), homoanatoxin-a (HTX) and their
52 structural variants (Faassen et al. 2012; Fetscher et al. 2015; Gugger et al. 2005; Heath et al.
53 2010; Wood et al. 2007). Despite their ability to produce harmful toxins and the associated
54 health risk, the physicochemical factors causing proliferations in cobble-bed rivers are not yet
55 fully understood.
56 To date, most attempts to understand variables controlling the percentage of M.
57 autumnalis cover have examined reach-scale benthic mats dynamics and related these to
58 physicochemical factors measured in the surrounding environment (e.g.,(McAllister et al.
59 2018a; Wood et al. 2017). Such studies have implicated water chemistry, river flow and fine
60 sediment load as affecting M. autumnalis cover, though to date the predictive power of
61 statistical relationships generated is relatively weak (Heath et al. 2011; McAllister et al.
62 2018a; Wood et al. 2017; Wood et al. 2016). The difficulties in identifying key variables may
63 be complicated by the nature of the M. autumnalis accrual cycle. This involves colonisation,
64 initiation of a benthic mat, growth via patch expansion and eventual detachment. At each of
https://mc06.manuscriptcentral.com/cjfas-pubs Canadian Journal of Fisheries and Aquatic Sciences Page 4 of 41
65 these successional phases, M. autumnalis is likely to respond differently to physicochemical
66 conditions (McAllister et al. 2016).
67 The spatial distribution of M. autumnalis mats in cobble-bed rivers is extremely
68 patchy, which has been partially attributed to small-scale differences in velocity (McAllister
69 et al. 2018a). The importance of velocity was reinforced by Hart et al. (2013) who noted that
70 in the field M. autumnalis was positively associated with velocity and generally dominated at
71 velocities >0.4 m s-1, and by Heath et al. (2015) who highlighted that cover was greatest
72 between 0.6 and 1.1 m s-1. The low-profile, dense, mucilaginous nature of M. autumnalis
73 mats, allows it to withstand higher velocities compared to other higher-profile species (Biggs
74 et al. 1998). High velocity environments are probably also conducive to M. autumnalis
75 accrual as the exchange of solutes at the mat-water interface is enhanced due to the reduction
76 of boundary layer thickness. McAllisterDraft et al. (2018b) showed that M. autumnalis biomass
77 accrual, but not mat expansion, was positively affected by an increase in velocity in
78 experimental mesocosms. However, the difference between velocity treatments was limited
79 (0.1 m s-1) and it is likely that larger differences in velocity would have elicited greater
80 responses in M. autumnalis accrual.
81 Field studies have also suggested that optimal ranges of nutrients may exist for M.
82 autumnalis accrual. Proliferations consistently occur at DRP concentrations below 0.02 mg L-
83 1 (McAllister et al. 2016; Wood et al. 2017; Wood et al. 2016). Optimal ranges for water-
84 column dissolved inorganic nitrogen (DIN) concentrations are less defined. Initially,
85 proliferations were thought to only occur when DIN was greater than 0.1 mg L-1 (Wood and
86 Young 2011; 2012), however, subsequent research has documented proliferations when DIN
87 concentrations are below 0.02 mg L-1 (McAllister et al. 2018a; Wood et al. 2017). Studies
88 investigating the effect of nitrate on M. autumnalis have included culture-based laboratory
89 investigations (Heath et al. 2016; Heath et al. 2014), observational field studies (Heath et al.
https://mc06.manuscriptcentral.com/cjfas-pubs Page 5 of 41 Canadian Journal of Fisheries and Aquatic Sciences
90 2011; McAllister et al. 2018a; Wood et al. 2017) and most recently experimental mesocosms
91 (McAllister et al. 2018b). McAllister et al. (2018b) found that elevating nitrate concentrations
92 from 0.02 mg L-1 to 0.4 mg L-1 did not elicit a significant increase in M. autumnalis biomass
93 accrual (chlorophyll a concentrations and biovolumes) or patch expansion rates.
94 An important commonality among observational-based field studies on M. autumnalis
95 is that even after the inclusion of a wide range of physicochemical factors, “site” remains a
96 key determinant of reach-scale cover (McAllister et al. 2018a; Wood et al. 2017). This
97 suggests that time-independent, site-specific attributes are important in determining M.
98 autumnalis cover, but are missing from analyses to date. One factor, which remains largely
99 unexplored, despite its general importance in controlling periphyton growth, is herbivory
100 (Anderson et al. 1999; Karouna and Fuller 1992). The extent to which herbivorous
101 macroinvertebrate communities influenceDraft algal growth depends on factors including the
102 mobility, feeding rates, life-history timing, body size, and density of each species
103 (Holomuzki et al. 2010; Steinman 1996). Velocity also influences the removal of algae by
104 macroinvertebrates and can influence their density and composition (Hintz and Wellnitz
105 2013).
106 In the present study, variation in M. autumnalis biomass accrual and expansion was
107 investigated across gradients of velocity (pools, runs and riffles) in three rivers with varying
108 water chemistry. Cobbles were manually inoculated with M. autumnalis (McAllister et al.
109 2018b) which allowed the colonisation step of the accrual cycle to be standardized. The
110 following hypotheses were then tested: (1) M. autumnalis growth assessed at the patch scale
111 (area increase), will reflect cover measured at the reach scale; (2) M. autumnalis biomass
112 accrual and cover will be highest in high velocity habitats (i.e., riffles) and lowest in low
113 velocity environments (i.e., pools); and (3) under similar velocities, site-specific nutrient
https://mc06.manuscriptcentral.com/cjfas-pubs Canadian Journal of Fisheries and Aquatic Sciences Page 6 of 41
114 concentrations will be more important than other environmental factors (i.e.,
115 macroinvertebrates and shear stress) in determining M. autumnalis expansion rates.
116 Materials and methods
117 Study sites and experimental design
118 One site on each of the three predominately gravel and cobble-bed rivers were selected for
119 this study (Fig. 1): Ōpihi (site 1; 44°15′48″ S, 171°16′14″ E), Te muka (site 2; 44°14′42″ S,
120 171°16′07″ E) and Te ana a wai (site 3; 44°18′27″ S, 170°57′06″ E) rivers, in the Canterbury
121 region (South Island, New Zealand). The experiment was conducted from 10 January to 1
122 February 2017, in a pool, run and riffle habitat at each study site. Habitats were selected
123 within sites to be in close proximity to each other to limit variation of physicochemical
124 variables, and habitats among sites wereDraft selected to have similar velocities. Habitats within
125 sites were defined by depth, velocities and surface characteristics (Allen, 1951; Mosley,
126 1982). Riffle habits were classified as relatively shallow and high velocity environments with
127 a broken surface. Whereas runs were identified as being comparatively deeper than riffles,
128 having an intermediate velocity and a smooth surface. Pools were characterized as low-
129 velocity environments and were generally deeper than other habitat types (Fig. S1).
130 One hole (5 mm in diameter and depth) was drilled in the center of each of 135
131 cobbles (surface area: 270–524 cm2), which was then seeded with homogenised M.
132 autumnalis mats (see(McAllister et al. 2018b) collected from the site in which cobbles would
133 be placed. At each site, 45 cobbles were then placed in a slow-moving run (near-bed velocity:
134 ca 0.25 m s-1) for two days to allow for initial attachment of the filaments and expansion out
135 of the hole. Following this initial period of growth, 15 cobbles were moved into each habitat
136 type (pool, run, riffle), at each site.
137 138
https://mc06.manuscriptcentral.com/cjfas-pubs Page 7 of 41 Canadian Journal of Fisheries and Aquatic Sciences
139 Microcoleus autumnalis cover, patch expansion and biomass
140 M. autumnalis cover, in each habitat, was assessed weekly at the reach scale (~ 50 m) as
141 described by Wood et al. (2009). This involved estimating percent cover at five equidistant
142 points along each of four transects, in each of the three habitat types, using a bathyscope
143 (Model 0800, Nuova Rade, Italy), giving a total of 20 values per habitat, from which an
144 average cover was calculated for each.
145 M. autumnalis mat expansion on seeded cobbles was followed by photographing them
146 every two to three days using a COOLPIX S33 camera (Nikon, Japan) and calculating patch
147 area using ImageJ (National Institutes of Health, USA). Images were checked for distortion
148 using a calibrated target imaged from the same distance, and no effect was apparent. Due to
149 destructive biomass sampling of nine of the 15 seeded patches, M. autumnalis expansion on
150 only three cobbles could be followed overDraft the entirety of the experiment. Biomass accrual, as
151 phycoerythrin concentrations, was measured on days 8, 16 and 23 for three pre-seeded
152 cobbles randomly selected from each habitat type. Each cobble was scrubbed with a nylon
153 brush in 200 mL of deionized water for two mins and the cell suspension was homogenised
154 with a handheld blender (Kenwood, UK). Aliquots (5 mL) for phycoerythrin analysis were
155 filtered (GF/C filters; Whatman, USA) and the filters frozen (-20°C) until later analysis.
156 Surface areas of sampled cobbles were estimated using aluminium foil. A single layer of foil
157 was molded and trimmed to the surface area of the cobble and then weighed. The weight of
158 foil was converted to surface area by creating a calibration curve using foil of a known
159 surface area. Biomass measurements were normalised to cobble size.
160
161 Physicochemical measurements
162 Every two to three days we measured; (1) the near-bed velocity (i.e., the velocity at the
163 cobble surface) and depth at each of the 135 experimental cobbles using a Marsh-McBirney
https://mc06.manuscriptcentral.com/cjfas-pubs Canadian Journal of Fisheries and Aquatic Sciences Page 8 of 41
164 Flo-Mate 2000 (Marsh-McBirney Corp., USA), (2) dissolved oxygen (DO), pH and
165 conductivity in each habitat, using a HACH HQ40d portable water quality meter, and (3)
166 turbidity of three triplicate samples collected in each habitat using an AQUAfast AQ4500
167 Turbidimeter (Thermofisher, USA). Continuous river discharge was measured at permanent
168 gauging stations using either Encoder, Accubar or Pressure Transducer sensors coupled with
169 iRIS 150 data loggers (Kisters Pioneering Technologies, USA), which were located in close
170 proximity to the sampling site (see Fig. 1). Temperature was measured underwater in each
171 habitat every 15 mins using HOBO Pendant Data Loggers (Onset Computer Corporation,
172 USA). Surface water samples were collected weekly, from each habitat type, for nutrient
173 analyses. Sub-samples for nitrate + nitrite-N (NO3 + NO2-N) and ammoniacal-N (NH4-N)
174 were filtered through GF/C filters (Whatman, USA), whereas samples for DRP were filtered
175 through 0.45 µm membrane filters (Millipore,Draft UK). Nutrient samples were frozen (-20°C)
176 until further analysis. The slope of the water surface was assessed at each site using a Topcon
177 RL-H4C laser level with an LS-100D mm receiver (Topcon, Japan).
178
179 Macroinvertebrate sampling and analysis
180 Three replicate macroinvertebrates samples were collected with a Surber sampler (0.04 m2,
181 500 µm mesh) from each habitat at each of the three study sites on 1 February 2017. In the
182 pool habitats, where velocity was low, flow was manually created. Samples were placed in
183 containers and preserved using 70% ethanol. Macroinvertebrates were identified to genus
184 level or the lowest practicable taxonomic level according to Winterbourn et al. (2006). Sub-
185 sampling was necessary for selected samples as macroinvertebrate abundances were
186 extremely high (>2,000 individuals per sample). Samples were split into thirds and were
187 counted until a target number of at least 500 organisms was reached. The remainder of the
188 sample was then checked for any unrecorded taxa. Taxa were assigned into the following
https://mc06.manuscriptcentral.com/cjfas-pubs Page 9 of 41 Canadian Journal of Fisheries and Aquatic Sciences
189 functional feeding groups (FFG): collector-browsers, predators, grazers, shredders, parasites,
190 omnivores and filter feeders (see Table S1;(Chadderton 1988; Cowie 1980; Cowley 1978;
191 Jaarsma et al. 1998; Lester et al. 1994; Quinn and Hickey 1990; Winterbourn et al. 1984).
192
193 Laboratory analyses
194 Nitrate + nitrite-N samples were analysed as nitrite after a spongy cadmium reduction
195 (Mackereth et al. 1978). Ammonia concentrations were assessed using the phenate method,
196 but all concentrations were below the detection limit of 0.01 mg L-1. Dissolved reactive
197 phosphorus (DRP) was analysed by the molybdenum blue method (Water Environmental
198 Federation and American Public Health Association 2005). The limits of detection were: 0.02
-1 -1 199 mg L for NO3 + NO2-N and 0.002 mg L for DRP. All absorbances were measured
200 spectrophotometrically using a HACHDraft DR3900 spectrophotometer (Hach, USA).
201 Phycoerythrin was extracted from material collected on filters in 5 mL of potassium
202 phosphate buffer (0.1M; pH=6.8) and mechanically homogenised with a teflon and glass
203 homogeniser. Samples were stored at 4ºC for 24 hours, before being filtered (GF/C,
204 Whatman, USA) and phycoerythrin concentrations assessed using a fluorometer (Aquaflor,
205 Turner Instruments). Calibration of the fluorescence output was accomplished by creating a
206 dilution series from an extract of M. autumnalis, containing sufficient phycoerythrin to allow
207 quantification using spectrometric absorbance. Absorbance was converted to concentration
208 using the specific absorption coefficient of 300,000 (Mol-1 cm-1) taken from Bryant et al.
209 (1976).
210
211 Data analyses
212 Reach-scale shear stress was estimated using the formula given by Statzner et al. (1988):
213 Shear stress = surface water slope × water depth × 푔 × 푝푤,
https://mc06.manuscriptcentral.com/cjfas-pubs Canadian Journal of Fisheries and Aquatic Sciences Page 10 of 41
-2 -1 214 where gravity g = 9.81 m s and density pw = 1.0 g mL . The average water depth was
215 calculated from over 300 measurements throughout runs and riffles from each respective
216 river. Depths of pools were not incorporated into this calculation as shear stress in these
217 environments is likely to be very different than in runs and riffles.
218 Initially, to explore patch expansion rates the following simple exponential growth model
219 was fitted to the mat expansion data:
(푙푛푁2 ― 푙푛푁1) 220 푏 = , (푡2 ― 푡1)
-1 221 where b is the Exponential Accrual Rate (EAR; day ), lnN2 is the natural log of patch size at
222 time two (t2) and lnN1 is the natural log of patch size at time one (t1).
223 However, this simplified model did not fit data well and logistic and other formulations of 224 exponential growth models were explored.Draft The logistic growth model achieved better fits for 225 site 1 data (Table S2). EARs were calculated for three patches from each habitat type within
226 site 1, by fitting the following growth model:
퐾 227 N = , 1 + 푒푎 ― 푏푡
228 where N is the patch size at day t, K is the carrying capacity (maximum patch size reached), a
229 is a constant which indicates the relative position from the origin and b is the EAR.
230 Not all site × habitat combinations yielded values for EAR. At site 2, the lack of increase in
231 patch size in any habitat types precluded this analysis. At site 3, the short-lived nature of
232 patches in the pool habitat also prevented EARs being calculated. However, patches in the
233 run and riffle habitats of site 3 exhibited exponential growth and an exponential growth
234 model was consequently fitted to describe patch expansion:
235 푦 = 푦0 + 푎푒(푏푡),
2 236 where y is the patch size (cm ) at day t, y0 is the initial patch size at time t = 0, a is a constant
237 and b is the EAR.
https://mc06.manuscriptcentral.com/cjfas-pubs Page 11 of 41 Canadian Journal of Fisheries and Aquatic Sciences
238 Homogeneity of variances of data were tested by inspecting the residual and fitted
239 values as well as a Levene’s test. Normality was checked through the inspection of Quantile-
240 Quantile plots and conducting a Shapiro-Wilk test. Data were transformed when necessary to
241 meet model assumptions. One-way ANOVAs were utilised to assess the effect of habitat type
242 on EAR and K for site 1, whereas EARs calculated in the run and riffle habitats of site 3 were
243 compared using a t-test. One-way ANOVAs were also employed to detect significant
244 differences in the densities of macroinvertebrate communities. To evaluate the effect of site
245 and habitat type on biomass two-way ANOVAs were conducted. For phycoerythrin
246 concentrations ANOVAs were employed for each day individually (i.e., day 8, 16 and 23).
247 Where significant differences among treatments were identified by ANOVAs, post-hoc
248 Tukey’s honest significance test (HSD) tests were used to identify which treatments were
249 significantly different. Generalized additiveDraft mixed models (GAMMs; Hastie and Tibshirani,
250 1990) were used to model the non-linear relationship between M. autumnalis cover and patch
251 size.
252 Variance and normality tests, ANOVA analyses and generalized additive models were
253 performed in the software R Studio (version 3.1.1;(Team 2014). Patch expansion models were
254 conducted using SigmaPlot (Systat Software, United States of America).
255
256 Results
257 Microcoleus autumnalis accrual dynamics at the reach and patch scale
258 Reach-scale observations showed that M. autumnalis mats were present at all sites, but cover
259 varied spatially and temporally (Fig. 2). M. autumnalis was only observed in pool habitats on
260 two occasions, at site 1 (Fig. 2). It was also rare in all habitat types at site 2, where cover was
261 only observed on days 1 and 23 and did not exceed 1% in any habitat type (Fig. 2). In
https://mc06.manuscriptcentral.com/cjfas-pubs Canadian Journal of Fisheries and Aquatic Sciences Page 12 of 41
262 contrast, median cover in the run and riffle habitats of site 1 increased from 17.5% and 36%
263 on day 1 to 93% and 80% respectively by day 23. At site 3, median M. autumnalis cover in
264 the riffle habitat was consistently higher than that of the run habitat. M. autumnalis cover in
265 both these habitat types showed little increase from days 1 to 15 but had increased markedly
266 by day 23, from 5 to 30% in the run habitat and 17 to 65% in the riffle habitat (Fig. 2).
267 In pool habitats, seeded patches generally expanded quickly, forming a very thin
268 round patch on cobbles (Fig. 3; Fig. 4). However, they were short lived in pool habitats,
269 surviving an average of 7, 18 and 5 days at sites 1, 2 and 3 respectively (Fig 4; Table 1). The
270 largest patches across all sites were observed in run habitats, and the increase was fastest and
271 most persistent in site 1 (Fig. 3; Fig. 4). Patch expansion at site 2 was limited, this was
272 especially apparent in the riffle habitat where patch areas did not exceed 2.5 cm2 throughout
273 the entire experimental period (Fig. 3).Draft At site 3, patch expansion was slower than at site 1,
274 and smaller patches eventuated.
275 There was a strong correlation between reach-scale M. autumnalis cover and patch
276 size measured on individual cobbles (R2 = 0.71, p < 0.0001; Fig. 5). Cover was limited and
277 patch sizes were consistently low in pool habitats, whereas in run and riffle habitats both
278 metrics increased simultaneously (Fig. 5). When cover reached greater than 60% in run and
279 riffle habitats, patch sizes began to stabilize (Fig. 5).
280
https://mc06.manuscriptcentral.com/cjfas-pubs Page 13 of 41 Canadian Journal of Fisheries and Aquatic Sciences
281 Patch survival, expressed as the percentage of patches remaining at the end of the
282 experiment, was zero in all pool habitats (Table 1). In runs and riffles survival was higher at
283 sites 1 and 3 than at site 2. Where they differed, riffles had higher survival than runs. Survival
284 was 100% in riffles in sites 1 and 3 and runs at site 1 (Table 1). More of the patches that
285 survived to the end of the experiment had begun to decrease in size in runs compared to
286 riffles, and only in riffles at site 1 was no erosion of patches observed (Table 1).
287 At site 1 the EARs for each habitat were significantly different from each other (One-
288 way ANOVA: F2 = 30.5, p < 0.001; Table 2). The highest EAR at site 1 was measured in the
289 run (5.2 cm2 per day) and the slowest was 2.35 cm2 per day in the pool habitat. There was
290 also a significant difference in maximum patch size among habitat types at site 1 (One-way
2 2 291 ANOVA: F2 = 61.5, p < 0.0001; Table 2), with an average of 22.7 cm in the pool, 357.5 cm
292 in the run and 291.8 cm2 in the riffle habitatDraft (Fig. 3; Table 2). Average EAR, calculated using
293 an exponential growth model, at site 3 in run and riffle habitats were 0.20 and 0.23 cm2 per
294 day respectively and there was no significant difference (T-test: t4 = 1.17, p = 0.307).
295
296 Microcoleus autumnalis biomass among sites and habitat types
297 Two-way ANOVAs showed a significant effect of site on average phycoerythrin
298 concentrations on days 8, 16 and 23 and a significant effect of habitat on days 16 and 23 (Fig.
299 6; Table 3). Tukey’s HSD tests showed that the differences between phycoerthrin
300 concentrations among habitat types, increased through time (Fig. 6). For example, on day 8,
301 phycoerythrin concentrations in habitats within sites were not significantly different from
302 each other, and all site 2 habitats were not different from those of site 3 and the pool and run
303 habitat of site 1 (Fig. 6). On day 16, the mats in the site 1 run had significantly higher
304 biomass than all other habitats, however on day 23 the phycoerythrin concentrations in the
305 site 1 riffle increased markedly and these two habitats were significantly different to all
https://mc06.manuscriptcentral.com/cjfas-pubs Canadian Journal of Fisheries and Aquatic Sciences Page 14 of 41
306 others (Fig. 6). Phycoerythrin concentrations measured in pool habitats were consistently
307 lower than all other habitat types (Fig. 6). When phycoerythrin concentrations were
308 normalised to patch size, rather than cobble size, biomass per unit area was greater in all
309 riffle habitats compared to runs on days 8, 16 and 23. On day 23 average phycoerythrin
310 concentrations, normalised to patch size, at both sites 1 and 3 were higher in riffle habitats
311 (site 1: 517 mg m-2, site 3: 167 mg m-2) compared to run habitats (site 1: 251 mg m-2, site 3:
312 101 mg m-2).
313 314 Physicochemical variables
315 Median near-bed velocities at the experimental cobbles over the deployment were similar in
316 each habitat type across sites (Fig. 7A). In pools median velocities varied from 0 (site 2) to 317 0.05 m s-1 (site 3). In runs values rangedDraft from 0.16 (site 3) to 0.25 m s-1 (site 2) and in riffles 318 were from 0.45 (site 3) to 0.55 m s-1 (site 1). Water depths were more variable, although there
319 was a tendency for median depths to decrease within sites with increasing velocity (Fig. 7B).
320 Median DIN concentrations varied markedly among sites. Site 3 had the lowest DIN
321 concentrations (0.07 mg L-1) and site 2 the highest (0.96 mg L-1; Fig. 7C). Median DRP
322 concentrations were also relatively constant among habitats within sites with the exception of
323 the pool habitat in site 1, which had a higher median concentration than the run and riffle
324 (Fig. 7D). With the exception of the pool habitat, site 1 had the lowest median DRP
325 concentration of 0.01 mg L-1, followed by site 3 (0.02 mg L-1) and site 2 had the highest of
326 0.05–0.06 mg L-1 (Fig. 7D). Proliferations and notable patch expansion were observed at both
327 site 1 and 3. Both of these sites had low DRP concentrations (<0.02 mg L-1) and low to
328 intermediate concentrations of DIN (site 1 = 0.4 mg L-1; site 3 = 0.07 mg L-1). Low nutrient
329 concentrations did not preclude the development of M. autumnalis proliferations and cover
330 exceeding 60% was observed in riffle habitats at site 3 (Fig. 2; Fig. 7).
https://mc06.manuscriptcentral.com/cjfas-pubs Page 15 of 41 Canadian Journal of Fisheries and Aquatic Sciences
331 Average water temperature was similar between sites, varying from 15.7 (site 3 pool)
332 to 17.6°C (site 1 riffle; Table 4). Average conductivity varied consistently among sites,
333 ranging from 95.5 µS cm-1 (site 1 run) to 147.9 µS cm-1 (site 2 riffle; Table 4). Average
334 turbidity was generally low across all sites (<1 NTU), varying from 0.3 NTU (site 1 pool) to
335 0.9 NTU (site 2 pool; Table 4).
336 Across all sites, pools had the smallest average wetted widths (range: 3.8–6.2 m) and
337 run habitats were the widest varying from 22.1 to 33.8 m, at site 2 and site 3, respectively
338 (Table 4). Shear stress was markedly higher at site 2 (1.05 kg m s-2) than at sites 1 (0.17 kg m
339 s-2) and 3 (0.11 kg m s-2; Table 4), primarily due to a steeper water surface slope. Average
340 daily discharge followed a similar pattern across sites, being relatively stable up until day 14
341 of the experiment, when discharge increased two to four-fold (Fig. 8) for a period of three
342 days. The spate was proportionally greaterDraft at sites 1 and 3 (three to four times prior
343 discharge) than at site 2 (less than twice prior discharge). During this spate, there was a steep
344 increase in patch size on day 15 in the site 1 run and also a change in growth trajectories in
345 the run and riffle habitats of site 3 (Fig. 3; Fig. 8). 346 347 Macroinvertebrate assemblages and associated functional feeding groups
348 The abundance of invertebrates of the various feeding guilds varied among and across sites in
349 complex ways. The density of collectors/browsers, mostly Deleatidium sp., Orthoclads,
350 Oligochaetes, Pycnocentria and Pycnocentrodes (see Table S1), differed among habitat types
351 and sites (One-way ANOVA: F8 = 15.46, p < 0.0001; Fig. 9A). Density tended to be similarly
352 high at all run habitats across rivers, and less in pools, while densities in riffles were high at
353 site 1 but less at other sites.
354 The density of grazers, which included the gastropods Potamopyrgus, Physa and
355 Gyraulus, and Ephydrids, (see Fig. S2), also differed among habitats and site. It was
356 generally lowest in run and riffle habitats across sites, with the exception of site 1 run (One-
https://mc06.manuscriptcentral.com/cjfas-pubs Canadian Journal of Fisheries and Aquatic Sciences Page 16 of 41
357 way ANOVA: F8 = 16.84, p < 0.0001; Fig. 9B). Site 1 and 3 pools had the highest average
358 density of grazers of 5,181 m2 and 4,681 m2 respectively, whereas the pool at site 2 had a
359 very low grazer density of 86 per m2 (Fig. 9B). No patches in site 1 and 3 pools survived until
360 the end of the experiment and were removed on average within 5 and 7 days respectively
361 (Table 2). The grazing community in the pool habitats of site 1 and 2 were made up of 41 and
362 50% of Potamopyrgus, whereas in the site 3 pool Potamopyrgus contributed 98% (Fig. S2).
363 Hudsonema sp. was the only macroinvertebrate assigned to the omnivorous functional
364 feeding group and its density varied across habitat types and sites (One-way ANOVA: F8 =
365 8.68, p < 0.0001; Fig. 9C). Hudsonema was found at a relatively high density in the site 3
366 pool and riffle habitats (average: 779 per m2 and 1,014 per m2 respectively).
367 Similarly, the only shredder identified was Olinga sp., however densities were low Draft 368 and did not differ significantly among habitat types and sites (One-way ANOVA: F8 = 2.27,
369 p = 0.121; Fig. 9D). The highest average density of Olinga sp. was 60 per m2 in the pool
370 habitat of site 3 (Fig. 9).
371
372
373 Discussion
374 Microcoleus autumnalis expansion dynamics at the patch and reach scale
375 Most studies investigating M. autumnalis (or closely related taxa) proliferations assess the
376 percentage of the stream substrata with benthic cover at the reach scale rather than measuring
377 patch expansion on individual cobbles (Heath et al. 2011; McAllister et al. 2018a; Schneider
378 2015; Wood et al. 2017). The first hypothesis of this study, that M. autumnalis growth
379 assessed at the patch scale, as area increase, would reflect accrual measured at the reach
380 scale, as percent cover, was supported. M. autumnalis has an unusual accrual characteristic,
381 in that well-developed mats spread laterally across the substrata rather increasing in thickness
https://mc06.manuscriptcentral.com/cjfas-pubs Page 17 of 41 Canadian Journal of Fisheries and Aquatic Sciences
382 (McAllister et al. 2016). This growth form may partly explain why patch expansion and
383 reach-scale cover followed similar accrual patterns. For monitoring and management
384 purposes, this finding reinforces that assessment of reach-scale benthic cover provides a
385 useful approximation of accrual dynamics.
386
387 Microcoleus autumnalis accrual among habitat types - the importance of
388 velocity
389 In the present study, there were distinct differences in M. autumnalis accrual that were
390 consistent across rivers. As predicted in our second hypothesis, accrual was smallest in the
391 slow-flowing pool habitats, and highest in the faster flowing run and riffle habitats supporting
392 a subsidy-stress relationship between M. autumnalis accrual and velocity. The subsidy-stress
393 model suggests that accrual may be enhancedDraft by velocity, as it increases the exchange of
394 solutes between the mat and overlying water both from the mat and into the mat, by reducing
395 the thickness of the diffusive boundary layer (Sand-Jensen 1983). Were this due to nutrient
396 limitation, the effect might be expected to be offset at site 2, where nutrient concentrations
397 were much higher than in the other sites. However, in addition to slow growth, the pool
398 habitats also had short patch longevity, which is not consistent with the subsidy-stress model,
399 where abrasion would be expected to be low at low water velocity. We suggest that one
400 plausible explanation for short patch longevity at site 2 in pool habitats is high grazing
401 pressure (discussed further below). Another potential explanation, which could also explain
402 detachment observed in run habitats, is the accumulation of oxygen bubbles. Oxygen bubbles
403 are more likely to accumulate in low and intermediate velocity environments due to enhanced
404 boundary layer thickness (Bosak et al. 2010; Boulêtreau et al. 2006; Hawes et al. 2014). The
405 facilitation of detachment by the accumulation of oxygen is likely to be unimportant in pool
406 habitats because of limited biomass accrual. In run habitats, maximum patch sizes were
https://mc06.manuscriptcentral.com/cjfas-pubs Canadian Journal of Fisheries and Aquatic Sciences Page 18 of 41
407 reached earlier, which in concert with accumulation of oxygen within mats likely led to a
408 state where biomass easily sloughed and accrual cycles were terminated.
409 The subsidy-stress model goes on to posit that as velocity increases, growth rate
410 should increase, through alleviation of resource limitation, but so would abrasion stress and
411 drag, eventually enhancing autogenic detachment, especially in the later stage of the growth
412 cycle. The results of this study contrast with this and show that patch growth was highest in
413 the intermediate velocity run habitat, and there was no evidence that higher velocities in the
414 riffles enhanced autogenic detachment or abrasion over runs, at least for the duration of our
415 experiments.
416 M. autumnalis is clearly capable of growing and persisting at high velocity. The
417 optimal velocity for M. autumnalis patch expansion in the present study was in run habitats,
418 and patch expansion was lower in the Draftfaster-flowing riffles, though the optimal habitat for
419 persistence was the riffle. Previous studies have suggested broadly similar optimal velocities
420 of 0.4 m s-1 to 1.1 m s-1 (Hart et al. 2013; Heath et al. 2015). Biggs et al. (1998) highlights
421 that the growth form of M. autumnalis, which is dense, low profile and tightly adhering
422 mucilaginous mat, should be advantageous in such high velocity environments.
423 The observation that optimal M. autumnalis growth and persistance is generally found
424 at high velocities (i.e. Heath et al. 2015) was also supported by EARs in this study, which
425 were slightly higher in runs than riffles. However by day 23, reach-scale cover and M.
426 autumnalis biomass did not vary markedly between these habitats. The relationships between
427 patch accrual, cover and velocity are simplified when biomass is normalised to patch area,
428 rather than cobble size. M. autumnalis from riffles had a higher biomass per area of mat
429 compared to mats in runs. M. autumnalis mats from riffles were also darker and thicker than
430 those grown in runs suggesting that patches in riffles increased in thickness rather than
431 spreading laterally. Thus, the absolute biomass accrual may be similar in riffles to that in
https://mc06.manuscriptcentral.com/cjfas-pubs Page 19 of 41 Canadian Journal of Fisheries and Aquatic Sciences
432 runs, and more persistent, despite smaller patch size. Several studies and numerous field
433 observations have indicated that M. autumnalis mats are initially largely confined to riffles
434 (high velocity, turbulent areas) in rivers (Heath 2009; Heath et al. 2015). However, the results
435 of the present study show that these observations could reflect longer patch life spans in riffle
436 habitats, and that M. autumnalis mats reach maximum mat size in runs earlier, with accrual
437 cycles therefore being terminated earlier in run habitats. It is noteworthy that in the pools, at
438 the lowest velocity, mats were at their thinnest and had the lowest concentration of
439 phycoerythrin per unit patch area, supporting an overall tendency for mats to expand rather
440 than thicken as velocity decreases. 441 442 Microcoleus autumnalis accrual across sites 443 The third hypothesis, that accrual rate Draftwould be correlated with nutrient concentrations, was 444 not supported. M. autumnalis accrual differed among sites but was not related to DIN or DRP
445 concentrations. Rate of accrual and accumulated biomass were lowest in the stream with
446 highest DIN concentration, and proliferations were ultimately similar at sites across a range
447 of DIN concentrations. There are however several other potential factors underlying these
448 differences.
449 Nutrient concentrations at site 1 may, however, have contributed to higher EARs and
450 biomass accrual than at the more oligotrophic site 3, where M. autumnalis growth was
451 initially limited and slow. However, towards the end of the experiment patches at site 3
452 expanded at a much higher rate suggesting that if there is a lower DIN limit for formation of
453 proliferations, it is less than 0.07 mg L-1. This finding is consistent with previous research
454 which has highlighted that M. autumnalis is capable of reaching a high biomass under low
455 DIN conditions (McAllister et al. 2018a; Wood et al. 2017).
456 A notable difference between site 2 and the others, which may have contributed to
457 differences in accrual in the faster flowing habitats, was that while velocity was similar shear
https://mc06.manuscriptcentral.com/cjfas-pubs Canadian Journal of Fisheries and Aquatic Sciences Page 20 of 41
458 stress was not. Patch size and persistence at site 2 may have been limited by the physical
459 constraints imposed by the high shear stress. Sporadic incidences of high turbidity were also
460 observed at site 2, potentially due to an upstream discharge or the prevalence of high
461 disturbance recreational activities (i.e., four-wheel driving). This was not captured by spot
462 measurements of turbidity but may have also contributed to the lack of patch expansion.
463
464 Impact of a spate
465 On day 15 there was a distinct change in patch size and growth trajectories in the run habitat
466 of sites 1 and 3 and the riffle habitat of site 3. This increase in expansion and apparent change
467 in rate of growth in cover (Fig. 7) coincided with a temporary increase in river discharge. The
468 average discharge at sites 1 and 3 increased to approximately four times the discharge
469 immediately prior; at site 3 by 2-fold. DraftThe increase of flow at site 1 exceeded the long term
470 median by 1.2 times (McAllister et al. 2018a) and resulted in an increased patch size from 96
471 cm2 to 290 cm2 in a period of only 3 days (site 1 run). Most studies to date have linked high
472 M. autumnalis accrual to prolonged periods of stable or receding flows, and considered
473 abrupt increases to have negative impacts on biomass (Biggs 1990; Cadel-Six et al. 2007;
474 Gugger et al. 2005; Heath et al. 2011; Sabater 2000; Wood et al. 2017). The size of flow
475 increase required to abrade M. autumnalis is, however, site specific (Wood et al. 2017). An
476 increase in accrual rate of M. autumnalis cover following a spate was also observed in one of
477 eight study rivers by McAllister et al. (2018a). Increased flow could indirectly or directly
478 enhance M. autumnalis growth through a variety of mechanisms, including increasing inputs
479 of nitrate and fine sediment from runoff, the latter carrying adhered phosphorus; both have
480 been linked to increases in M. autumnalis proliferations (Wood et al. 2017; Wood et al.
481 2016). Increased velocity may have also reduced grazing pressure and removed other weakly-
482 attached algae thereby reducing competition.
https://mc06.manuscriptcentral.com/cjfas-pubs Page 21 of 41 Canadian Journal of Fisheries and Aquatic Sciences
483
484 Macroinvertebrate communities and M. autumnalis growth
485 Herbivory plays an important role in accrual dynamics for general periphyton (Kohler and
486 Wiley 1997; Lowe and Hunter 1988; Taylor et al. 2002) but, to our knowledge, no studies
487 have yet specifically investigated the impact of macroinvertebrate grazing on M. autumnalis
488 mats. The degree to which macroinvertebrates reduce algal growth is dependent on many
489 interacting factors, including herbivore type, density, mobility, body size, mouthpart
490 morphology, algal type and successional stage (Holomuzki et al. 2010; Steinman 1996). In
491 the present study, densities of herbivorous macroinvertebrates differed among sites and
492 habitats. There were potential relationships between these differences and M. autumnalis
493 initial patch expansion and longevity. The density of grazers (Potamopyrgus, Physa,
494 Ephrididae, Gyraulus) exceeded 4,000Draft individuals per m2 in the pool habitat of site 1 and 3,
495 but was less than 100 individuals per m2 in site 2 and the lack of grazing pressure in the pool
496 habitat of site 2 may have enhanced the longevity of patches. The abundance of grazers (in
497 particular the mud snail, Potamopyrgus anitpodarum), omnivores and shredders was
498 particularly high in the pool habitat of site 3, and may explain the earlier removal of mats at
499 this site and possibly resulted in a higher grazing pressure during the initial growth period.
500 Grazers are likely to have a stronger effect on M. autumnalis growth during early stages of
501 accrual and under high abundances they have the ability to terminate accrual cycles.
502 Potamopyrgus anitpodarum, is known to significantly reduce algal biomass in streams where
503 densities exceed 1,500 per m2 (Biggs and Lowe 1994; Holomuzki and Biggs 2006;
504 Winterbourn and Fegley 1989) and high abundances have been found in M. autumnalis mats
505 previously (Hart et al. 2013). The scraping radula of snails allows them to feed on low-
506 profile, tightly adhered algae, like M. autumnalis mats, which are unavailable to many
https://mc06.manuscriptcentral.com/cjfas-pubs Canadian Journal of Fisheries and Aquatic Sciences Page 22 of 41
507 collector-browser species due to their blade-like or sweeping mouthparts (Holomuzki &
508 Biggs, 2006).
509 Grazing activity is most likely to significantly affect periphyton biomass in physically
510 stable environments (Steinman et al. 1991; Wellnitz and Poff 2006), such as the pool habitats
511 in the present study. Holomuzki and Biggs (2000) compared Potamopyrgus to Deleatidium,
512 Pycnocentrodes and Hudsonema and observed that when exposed to high velocity
513 Potamopyrgus moved to deeper protected habitats, suggesting that they favour more stable,
514 low velocity environments. Despite the high densities of some herbivorous
515 macroinvertebrates in run and riffle habitats in this study, they appeared to not be preventing
516 M. autumnalis accrual. Biggs et al. (1998) suggests that when resource supply and algal
517 growth rates are high, algal growth rates will exceed the consumptive capacity of herbivores.
518 Further investigations into causal relationshipsDraft between various herbivorous
519 macroinvertebrates and M. autumnalis growth are required.
520
521 Previous attempts in the field to identify correlations between M. autumnalis cover
522 and physicochemical factors, particularly flow and nutrient concentrations, have been limited
523 and complicated by the covariation of other factors. The aim of this experiment was to
524 advance the current understanding of which factors influence M. autumnalis accrual using a
525 novel method to inoculate cobbles with a standardized amount of M. autumnalis. Both
526 velocity and site were important factors influencing the rate at which M. autumnalis mats
527 expand and accrue biomass. No evidence was found to suggest that nutrients are important in
528 the latter two stages of the M. autumnalis accrual cycle; growth and detachment. However,
529 nutrient concentrations may still be important in the colonisation phase, which was not
530 examined in the present study. The near-bed velocity which supported the highest M.
531 autumnalis accrual was 0.25 to 0.45 m s-1, which represented run habitats. Accrual in pools
https://mc06.manuscriptcentral.com/cjfas-pubs Page 23 of 41 Canadian Journal of Fisheries and Aquatic Sciences
532 was extremely limited with patches being removed quickly, whereas in riffles they grew at a
533 slower rate than runs but had longer accrual cycles. Differences in M. autumnalis accrual
534 among habitat types and sites may be partially explained by differences in grazing pressure
535 and shear stress. While assessing M. autumnalis cover at the reach scale is useful for
536 management and monitoring purposes, measuring patch expansion on individual cobbles
537 provides a more detailed and nuanced understanding of how factors effect M. autumnalis at
538 various stages of the accrual cycle.
539
540 Acknowledgements
541 The authors thank Olivia Rowley, Oliver Gooday and Anna Henderson for their assistance in
542 completing field work. Karen Shearer is thanked for advice on macroinvertebrates. TGM
543 received support for the preparation ofDraft this paper by Te Pūnaha Matatini. SAW received
544 support for this study from the National Institute of Water and Atmospheric Research Ltd.
545 under the causes and effects of water quality degradation: eutrophication risk assessment
546 programme.
https://mc06.manuscriptcentral.com/cjfas-pubs Canadian Journal of Fisheries and Aquatic Sciences Page 24 of 41
547 References
548 Allen, K. R. 1951. The Horokiwi Stream: a study of a trout population. Bulletin of New 549 Zealand Department of Fishery, 10: 1-231. 550 Aboal, M., Puig, M.A., Mateo, P., and Perona, E. 2002. Implications of cyanophyte toxicity on 551 biological monitoring of calcareous streams in north-east Spain. Journal of Applied Phycology 552 14(1): 49-56. 553 Anderson, E., Welch, E., Jacoby, J., Schimek, G., and Horner, R. 1999. Periphyton removal 554 related to phosphorus and grazer biomass level. Freshwater Biology 41(3): 633-651. 555 Biggs, B.J. 1990. Periphyton communities and their environments in New Zealand rivers. New 556 Zealand Journal of Marine and Freshwater Research 24(3): 367-386. 557 Biggs, B.J., Goring, D.G., and Nikora, V.I. 1998. Subsidy and stress responses of stream 558 periphyton to gradients in water velocity as a function of community growth form. Journal of 559 Phycology 34(4): 598-607. 560 Biggs, B.J., and Lowe, R.L. 1994. Responses of two trophic levels to patch enrichment along 561 a New Zealand stream continuum. New Zealand Journal of Marine and Freshwater Research 562 28(2): 119-134. 563 Bosak, T., Bush, J.W.M., Flynn, M.R., Liang, B., Ono, S., Petroff, A.P., and Sim, M.S. 2010. 564 Formation and stability of oxygen‐rich bubbles that shape photosynthetic mats. Geobiology 565 8(1): 45-55. 566 Boulêtreau, S., Garabétian, F., Sauvage, S., and Sánchez‐Pérez, J.M. 2006. Assessing the 567 importance of a self‐generated detachmentDraft process in river biofilm models. Freshwater Biology 568 51(5): 901-912. 569 Bryant, D.A., Glazer, A.N., and Eiserling, F.A. 1976. Characterization and structural properties 570 of the major biliproteins of Anabaena species. Archives of Microbiology 110(1): 61-75. 571 Cadel-Six, S., Peyraud-Thomas, C., Brient, L., De Marsac, N.T., Rippka, R., and Méjean, A. 572 2007. Different genotypes of anatoxin-producing cyanobacteria coexist in the Tarn River, 573 France. Applied and Environmental Microbiology 73(23): 7605-7614. 574 Chadderton, W. 1988. Faunal and chemical characteristics of some Stewart Island streams. 575 New Zealand Natural Sciences 15: 43-50. 576 Cowie, B. 1980. Community dynamics of the bethnic fauna in a West Coast stream ecosystem, 577 Department of Zoology, University of Canterbury, Canterbury, New Zealand. 578 Cowley, D. 1978. Studies on the larvae of New Zealand Trichoptera. New Zealand Journal of 579 Zoology 5(4): 639-750. 580 Faassen, E.J., Harkema, L., Begeman, L., and Lurling, M. 2012. First report of (homo) 581 anatoxin-a and dog neurotoxicosis after ingestion of benthic cyanobacteria in The Netherlands. 582 Toxicon 60(3): 378-384. 583 Fetscher, A.E., Howard, M.D.A., Stancheva, R., Kudela, R.M., Stein, E.D., Sutula, M.A., 584 Busse, L.B., and Sheath, R.G. 2015. Wadeable streams as widespread sources of benthic 585 cyanotoxins in California, USA. Harmful Algae 49: 105-116. 586 Francis, G. 1878. Poisonous australian lake. Nature (London) 18: 11-12. 587 Gugger, M., Lenoir, S., Berger, C., Ledreux, A., Druart, J.-C., Humbert, J.-F., Guette, C., and 588 Bernard, C. 2005. First report in a river in France of the benthic cyanobacterium Phormidium 589 favosum producing anatoxin-a associated with dog neurotoxicosis. Toxicon 45(7): 919-928. 590 Hart, D.D., Biggs, B.J., Nikora, V.I., and Flinders, C.A. 2013. Flow effects on periphyton 591 patches and their ecological consequences in a New Zealand river. Freshwater biology 58(8): 592 1588-1602. 593 Hawes, I., Giles, H., and Doran, P.T. 2014. Estimating photosynthetic activity in microbial 594 mats in an ice‐covered Antarctic lake using automated oxygen microelectrode profiling and 595 variable chlorophyll fluorescence. Limnology and Oceanography 59(3): 674-688.
https://mc06.manuscriptcentral.com/cjfas-pubs Page 25 of 41 Canadian Journal of Fisheries and Aquatic Sciences
596 Heath, M.W. 2009. Mat forming toxic benthic cyanobacteria in New Zealand: Species diversity 597 and abundnce, cyanotoxin production and concentrations, School of Biological Science, 598 Victoria University, Wellington, New Zealand. 599 Heath, M.W., Wood, S., Brasell, K., Young, R., and Ryan, K. 2015. Development of habitat 600 suitability criteria and in‐stream habitat assessment for the benthic cyanobacteria Phormidium. 601 River Research and Applications 31(1): 98-108. 602 Heath, M.W., Wood, S., and Ryan, K. 2010. Polyphasic assessment of fresh-water benthic mat- 603 forming cyanobacteria isolated from New Zealand. FEMS Microbiology Ecology 73(1): 95- 604 109. 605 Heath, M.W., Wood, S., and Ryan, K. 2011. Spatial and temporal variability in Phormidium 606 mats and associated anatoxin-a and homoanatoxin-a in two New Zealand rivers. Aquatic 607 Microbial Ecology 64(1): 69-79. 608 Heath, M.W., Wood, S., Young, R., and Ryan, K. 2016. The role of nitrogen and phosphorus 609 in regulating Phormidium sp.(Cyanobacteria) growth and anatoxin production. FEMS 610 Microbiology Ecology 92(3): fiw021. 611 Heath, M.W., Wood, S.A., Barbieri, R., Young, R., and Ryan, K. 2014. Effects of nitrogen and 612 phosphorus on anatoxin-a, homoanatoxin-a, dihydroanatoxin-a and dihydrohomoanatoxin-a 613 production by Phormidium autumnale. Toxicon 92(17): 179-185. 614 Hintz, W.D., and Wellnitz, T. 2013. Current velocity influences the facilitation and removal of 615 algae by stream grazers. Aquatic Ecology 47(2): 235-244. 616 Holomuzki, J.R., and Biggs, B.J. 2000. Taxon-specific responses to high-flow disturbance in 617 streams: implications for population persistence. Journal of the North American Benthological 618 Society 19(4): 670-679. Draft 619 Holomuzki, J.R., and Biggs, B.J. 2006. Food limitation affects algivory and grazer performance 620 for New Zealand stream macroinvertebrates. In Advances in Algal Biology: A 621 Commemoration of the Work of Rex Lowe. Springer, Dordrecht. pp. 83-94. 622 Holomuzki, J.R., Feminella, J.W., and Power, M.E. 2010. Biotic interactions in freshwater 623 benthic habitats. Journal of the North American Benthological Society 29(1): 220-244. 624 Jaarsma, N., De Boer, S., Townsend, C., Thompson, R., and Edwards, E. 1998. Characterising 625 food‐webs in two New Zealand streams. New Zealand Journal of Marine and Freshwater 626 Research 32(2): 271-286. 627 Karouna, N.K., and Fuller, R.L. 1992. Influence of four grazers on periphyton communities 628 associated with clay tiles and leaves. Hydrobiologia 245(1): 53-64. 629 Kirkby. 1672. A relation of an inland sea, near Danzick, yielding at a certain season of the year 630 a green substance, which causeth certain death; together with an observation about white 631 amber: communicated by Mr. Kirkby, in a letter written to the publisher from Danzick Decemb. 632 19, 1671. Philos. Trans. 7, 4069e4070. 633 Kohler, S.L., and Wiley, M.J. 1997. Pathogen outbreaks reveal large‐scale effects of 634 competition in stream communities. Ecology 78(7): 2164-2176. 635 Lester, P., Mitchell, S., and Scott, D. 1994. Effects of riparian willow trees (Salix fragilis) on 636 macroinvertebrate densities in two small Central Otago, New Zealand, streams. New Zealand 637 Journal of Marine and Freshwater Research 28(3): 267-276. 638 Lowe, R., and Hunter, R. 1988. Effect of grazing by Physa integra on periphyton community 639 structure. Journal of the North American Benthological Society 7(1): 29-36. 640 Mackereth, F.J.H., Heron, J., and Talling, J.F. 1978. Water analysis: some revised methods for 641 limnologists. Freshwater Biological Association Ambleside, Freshwater Biological 642 Association. 643 McAllister, T., Wood, S., Atalah, J., and Hawes, I. 2018a. Spatiotemporal dynamics of 644 Phormidium cover and anatoxin concentrations in eight New Zealand rivers with contrasting 645 nutrient and flow regimes. Science of The Total Environment 612: 71-80.
https://mc06.manuscriptcentral.com/cjfas-pubs Canadian Journal of Fisheries and Aquatic Sciences Page 26 of 41
646 McAllister, T.G., Wood, S.A., Greenwood, M.J., Broghammer, F., and Hawes, I. 2018b. The 647 effects of velocity and nitrate on Phormidium accrual cycles: a stream mesocosm experiment. 648 Freshwater Science 37(3): 496-509. 649 McAllister, T.G., Wood, S.A., and Hawes, I. 2016. The rise of toxic benthic Phormidium 650 proliferations: A review of their taxonomy, distribution, toxin content and factors regulating 651 prevalence and increased severity. Harmful Algae 55: 282-294. 652 Oliver, R.L., Hamilton, D.P., Brookes, J.D., and Ganf, G.G. 2012. Physiology, blooms and 653 prediction of planktonic cyanobacteria. In Ecology of cyanobacteria II. Springer. pp. 155-194. 654 Paerl, H.W., Fulton, R.S., Moisander, P.H., and Dyble, J. 2001. Harmful freshwater algal 655 blooms, with an emphasis on cyanobacteria. The Scientific World Journal 1: 76-113. 656 Quiblier, C., Wood, S., Echenique-Subiabre, I., Heath, M., Villeneuve, A., and Humbert, J.-F. 657 2013. A review of current knowledge on toxic benthic freshwater cyanobacteria–ecology, toxin 658 production and risk management. Water Research 47(15): 5464-5479. 659 Quinn, J.M., and Hickey, C.W. 1990. Characterisation and classification of benthic 660 invertebrate communities in 88 New Zealand rivers in relation to environmental factors. New 661 Zealand Journal of Marine and Freshwater Research 24(3): 387-409. 662 Sabater, S. 2000. Diatom communities as indicators of environmental stress in the Guadiamar 663 River, SW. Spain, following a major mine tailings spill. Journal of Applied Phycology 12(2): 664 113-124. 665 Sand-Jensen, K. 1983. Physical and chemical parameters regulating growth of periphytic 666 communities. In Periphyton of Freshwater Ecosytems Springer, Dordrecht. pp. 63-71. 667 Schneider, S.C. 2015. Greener rivers in a changing climate?—Effects of climate and 668 hydrological regime on benthic algal Draftassemblages in pristine streams. Limnologica-Ecology 669 and Management of Inland Waters 55: 21-32. 670 Statzner, B., Gore, J.A., and Resh, V.H. 1988. Hydraulic stream ecology: observed patterns 671 and potential applications. Journal of the North American Benthological Society 7(4): 307- 672 360. 673 Steinman, A. 1996. Effects of grazers on freshwater benthic algae. In Algal Ecology: 674 Freshwater Benthic Ecosystens. Edited by R. Stevenson, M. Bothwell and R. Lowe. Academic 675 Press, San Diego, CA. pp. 341-373. 676 Steinman, A., Mulholland, P., and Kirschtel, D. 1991. Interactive effects of nutrient reduction 677 and herbivory on biomass, taxonomic structure, and P uptake in lotic periphyton communities. 678 Canadian Journal of Fisheries and Aquatic Sciences 48(10): 1951-1959. 679 Taylor, B.W., McIntosh, A.R., and Peckarsky, B.L. 2002. Reach‐scale manipulations show 680 invertebrate grazers depress algal resources in streams. Limnology and Oceanography 47(3): 681 893-899. 682 Team, R.C. 2014. R: A language and environment for statistical computing. R Foundation for 683 Statistical Computing, Vienna, Austria, 2012. ISBN 3-900051-07-0. 684 Water Environmental Federation, and American Public Health Association. 2005. Standard 685 methods for the examination of water and wastewater. American Public Health Association 686 Washington, DC, USA. 687 Wellnitz, T., and Leroy Poff, N. 2006. Herbivory, current velocity and algal regrowth: how 688 does periphyton grow when the grazers have gone? Freshwater Biology 51(11): 2114-2123. 689 Winterbourn, M., Cowie, B., and Rounick, J. 1984. Food resources and ingestion patterns of 690 insects along a west coast, South Island, river system. New Zealand Journal of Marine and 691 Freshwater Research 18(3): 379-388. 692 Winterbourn, M., and Fegley, A. 1989. Effects of nutrient enrichment and grazing on 693 periphyton assemblages in some spring-fed, South Island streams. New Zealand Natural 694 Sciences 16: 57-65.
https://mc06.manuscriptcentral.com/cjfas-pubs Page 27 of 41 Canadian Journal of Fisheries and Aquatic Sciences
695 Winterbourn, M., Gregson, K.L., and Dolphin, C.H. 2006. Guide to the aquatic insects of New 696 Zealand. Bulletin of the Entomological Society of New Zealand 14. 697 Wood, S.A., Atalah, J., Wagenhoff, A., Brown, L., Doehring, K., Young, R.G., and Hawes, I. 698 2017. Effect of river flow, temperature, and water chemistry on proliferations of the benthic 699 anatoxin-producing cyanobacterium Phormidium. Freshwater Science 36(1). 700 Wood, S.A., Depree, C., Brown, L., McAllister, T., and Hawes, I. 2016. Entrapped sediments 701 as a source of phosphorus in epilithic cyanobacterial proliferations in low nutrient rivers. PLoS 702 one 10(10). 703 Wood, S.A., Selwood, A.I., Rueckert, A., Holland, P.T., Milne, J.R., Smith, K.F., Smits, B., 704 Watts, L.F., and Cary, C.S. 2007. First report of homoanatoxin-a and associated dog 705 neurotoxicosis in New Zealand. Toxicon 50(2): 292-301. 706 Wood, S.A., and Young, R.G. 2011. Benthic cyanobacteria and toxin production in the 707 Manawatu-Wanganui region Cawthron Report No. 1959. Cawthron Institute, Nelson, New 708 Zealand. 709 Wood, S.A., and Young, R.G. 2012. Review of Benthic Cyanobacteria Monitoring Programme 710 2012. Horizons Regional Council. 711
712 Draft
https://mc06.manuscriptcentral.com/cjfas-pubs Canadian Journal of Fisheries and Aquatic Sciences Page 28 of 41
713 Figure legends 714 715 Fig. 1. Location of sampling sites in the three study rivers (Canterbury, New Zealand). Map 716 created using ArcGIS (www.arcgis.com). 717 718 Fig. 2. Boxplots of reach-scale Microcoleus autumnalis cover (n=20) measured 719 approximately weekly in each habitat type among sites. Horizontal lines in boxplots represent 720 medians, box ends arequartiles and whiskers extend to the lowest or highest data point, which 721 are within 1.5 times the interquartile range. Black dots represent outliers, which are outside 722 1.5 times the interquartile range. 723 724 Fig. 3. Patch area (n=3) in each habitat type throughout the experimental period. Patch areas 725 were calculated from three mats within each habitat type. The blue line is a smoothed local 726 polynomial regression and grey shading represents standard error. 727 728 Fig. 4. Examples of Microcoleus autumnalis patch expansion on three seeded cobbles, from 729 each habitat type at site 1. 730 731 Fig. 5. Correlation between average M. autumnalis cover and average M. autumnalis patch 732 size across sites. The black line is a smoothed local polynomial regression and grey shading 733 represents standard error. 734 735 Fig. 6. Average phycoerythrin concentrationsDraft normalised to cobble size on days 8, 16 and 23. 736 Error bars represent standard error. Letters indicate significance at the p < 0.05 level 737 according to a post-hoc Tukey’s Honest Significant Difference (HSD) test. 738 739 Fig. 7. Boxplots of; (A) near-bed velocity (n=150), (B) water depth (n=150), (C) dissolved 740 inorganic nitrogen (DIN) concentrations (n=4), and (D) dissolved reactive phosphorus (DRP) 741 concentrations (n=4) measured in each habitat type at each site. Horizontal lines in boxplots 742 represent medians, box ends are quartiles and whiskers extend to the lowest or highest data 743 point, which are within 1.5 times the interquartile range. Black dots represent outliers, which 744 are outside 1.5 times the interquartile range. 745 746 Fig. 8. Average daily river discharge at each of the three study sites throughout the study 747 period. See Figure 1 for site locations. 748 749 Fig. 9. Average density of macroinvertebrates (n=3) belonging to various functional feeding 750 groups (± one standard error) in each habitat at the three study sites (see Fig. 1). Note y-scale 751 varies between plots. 752
753
754
755
756
https://mc06.manuscriptcentral.com/cjfas-pubs Page 29 of 41 Canadian Journal of Fisheries and Aquatic Sciences
757
758
Draft
https://mc06.manuscriptcentral.com/cjfas-pubs Canadian Journal of Fisheries and Aquatic Sciences Page 30 of 41
759 Table 1. Summary attributes of patch dynamics in each habitat type within the three study sites. 760 The first data row indicates the percent of patches still present at the end of the experiment, and 761 the second row the percent of those surviving patches that had decreased in size. SD = standard 762 deviation. 763
Attribute
Site/Habitat % of patches % of survivors Patch life span (days:
surviving eroding mean ± one SD)
Site 1
Pool 0 0 7 (5)
Run 100 44 -
Riffle 100 0 -
Site 2 Pool 0 0 Draft18 (4) Run 8 100 17 (3)
Riffle 56 100 21 (2)
Site 3
Pool 0 0 5 (3)
Run 78 100 19 (6)
Riffle 100 14 21 (1)
764 765
https://mc06.manuscriptcentral.com/cjfas-pubs Page 31 of 41 Canadian Journal of Fisheries and Aquatic Sciences
766 Table 2. Average Exponential Accrual Rates (EAR) and the maximum patch size (± one standard 767 deviation) for each habitat type at site 1 calculated using a logistic growth model. See Table S2 768 for model fits and errors associated with calculations. 769 Habitat type EAR (cm2 per Maximum patch size
day) (cm)
Pool 2.35 (0.15) 22.66 (5.14)
Run 5.23 (0.79) 357.51 (19.35)
Riffle 3.32 (0.49) 291.76 (64.86) 770 771 Table 3. Summary of two-way ANOVAs comparing phycoerythrin concentrations among sites 772 and habitat types. Significant p values (< 0.05) are given in bold. df = degrees of freedom. 773 Day Site Habitat Site × Habitat
df F p value df F p value df F p value Day 8 2 38.16 <0.001 2 Draft14.11 <0.001 4 4.97 0.007 Day 16 2 7.71 0.009 2 99.29 <0.001 - - -
Day 23 2 207.40 <0.001 2 523.18 <0.001 - - -
774
https://mc06.manuscriptcentral.com/cjfas-pubs Canadian Journal of Fisheries and Aquatic Sciences Page 32 of 41
775 Table 4. Average water temperate, conductivity, turbidity and wetted width (± one standard 776 deviation) measured at each habitat type at the three sampling sites. A single shear stress value 777 was calculated for each site. Temperature was measured every 15 mins. For conductivity, mean 778 turbidity and wetted width n=11. Turbidity measurements were taken in triplicate. See Figure 1 779 for site locations. 780 Site/Habitat Temperature Conductivity Turbidity Wetted Shear
781 °C µS cm-1 NTU width m stress
kg m s-2
Site 1
Pool 17.0 (0.5) 97.2 (2.9) 0.3 (0.2) 6.2 (0.3)
Run 17.7 (0.7) 95.5 (3.4) 0.4 (0.2) 24.5 (2.1) 0.17
Riffle 17.6 (0.7) 95.8 (3.7) 0.4 (0.1) 19.4 (2.3)
Site 2
Pool 16.6 (0.5) 143.2 (1.6) 0.9 (0.6) 3.8 (0.3)
Run 16.3 (0.7) 147.7 (3.2)Draft0.5 (0.2) 22.1 (1.8) 1.05
Riffle 16.3 (0.7) 147.9 (2.6) 0.5 (0.1) 12.3 (1.4)
Site 3
Pool 15.7 (0.7) 139.4 (9.4) 0.5 (0.2) 5.8 (0.5)
Run 16.5 (0.9) 132.9 (6.9) 0.4 (0.1) 33.8 (1.6) 0.11
Riffle 16.7 (0.9) 132.5 (6.3) 0.4 (0.1) 7.2 (1.5)
https://mc06.manuscriptcentral.com/cjfas-pubs Page 33 of 41 Canadian Journal of Fisheries and Aquatic Sciences
Draft
Fig. 1. Location of sampling sites in the three study rivers (Canterbury, New Zealand).
322x235mm (300 x 300 DPI)
https://mc06.manuscriptcentral.com/cjfas-pubs Canadian Journal of Fisheries and Aquatic Sciences Page 34 of 41
Fig. 2. Boxplots of reach-scale Microcoleus autumnalis cover (n=20) measured approximately weekly in each habitat type among sites. Horizontal lines in boxplots represent medians, box ends arequartiles and whiskers extend to the lowest or highest data point, which are within 1.5 times the interquartile range. Black dots represent outliers, whichDraft are outside 1.5 times the interquartile range. 2116x1058mm (72 x 72 DPI)
https://mc06.manuscriptcentral.com/cjfas-pubs Page 35 of 41 Canadian Journal of Fisheries and Aquatic Sciences
Draft
Fig. 3. Patch area (n=3) in each habitat type throughout the experimental period. Patch areas were calculated from three mats within each habitat type. The blue line is a smoothed local polynomial regression and grey shading represents standard error.
2116x2116mm (72 x 72 DPI)
https://mc06.manuscriptcentral.com/cjfas-pubs Canadian Journal of Fisheries and Aquatic Sciences Page 36 of 41
Fig. 4. Examples of Microcoleus autumnalis patch expansion on three seeded cobbles, from each habitat type at site 1. Draft
https://mc06.manuscriptcentral.com/cjfas-pubs Page 37 of 41 Canadian Journal of Fisheries and Aquatic Sciences
Draft
Fig. 5. Correlation between average M. autumnalis cover and average M. autumnalis patch size across sites. The black line is a smoothed local polynomial regression and grey shading represents standard error.
https://mc06.manuscriptcentral.com/cjfas-pubs Canadian Journal of Fisheries and Aquatic Sciences Page 38 of 41
Draft
Fig. 6. Average phycoerythrin concentrations normalised to cobble size on days 8, 16 and 23. Error bars represent standard error. Letters indicate significance at the p < 0.05 level according to a post-hoc Tukey’s Honest Significant Difference (HSD) test.
752x564mm (72 x 72 DPI)
https://mc06.manuscriptcentral.com/cjfas-pubs Page 39 of 41 Canadian Journal of Fisheries and Aquatic Sciences
Draft
Fig. 7. Boxplots of; (A) near-bed velocity (n=150), (B) water depth (n=150), (C) dissolved inorganic nitrogen (DIN) concentrations (n=4), and (D) dissolved reactive phosphorus (DRP) concentrations (n=4) measured in each habitat type at each site. Horizontal lines in boxplots represent medians, box ends are quartiles and whiskers extend to the lowest or highest data point, which are within 1.5 times the interquartile range. Black dots represent outliers, which are outside 1.5 times the interquartile range.
2116x1552mm (72 x 72 DPI)
https://mc06.manuscriptcentral.com/cjfas-pubs Canadian Journal of Fisheries and Aquatic Sciences Page 40 of 41
Fig. 8. Average daily river discharge at each of the three study sites throughout the study period. See Figure 1 for site locations.
2328x1058mm (72 x 72 DPI) Draft
https://mc06.manuscriptcentral.com/cjfas-pubs Page 41 of 41 Canadian Journal of Fisheries and Aquatic Sciences
Draft
Fig. 9. Average density of macroinvertebrates (n=3) belonging to various functional feeding groups (± one standard error) in each habitat at the three study sites (see Fig. 1). Note y-scale varies between plots.
406x339mm (72 x 72 DPI)
https://mc06.manuscriptcentral.com/cjfas-pubs