Is Didymosphenia geminata an introduced species in New Zealand? Evidence from trends in water chemistry, and chloroplast DNA

Abstract Defining the geographic origins of free‐living aquatic microorganisms can be problematic because many such organisms have ubiquitous distributions, and proving absence from a region is practically impossible. Geographic origins become important if microorganisms have invasive characteristics. The freshwater diatom Didymosphenia geminata is a potentially ubiquitous microorganism for which the recent global expansion of nuisance proliferations has been attributed to environmental change. The changes may include declines in dissolved reactive phosphorus (DRP) to low levels (e.g., <2 mg/m3) and increases in dissolved inorganic nitrogen (DIN) to >10 mg/m3 because both these nutrient conditions are associated with nuisance proliferations of D. geminata. Proliferations of D. geminata have been observed in South Island, New Zealand, since 2004. We aimed to address the ubiquity hypothesis for D. geminata in New Zealand using historical river water nutrient data and new molecular analyses. We used 15 years of data at 77 river sites to assess whether trends in DRP or DIN prior to the spread of D. geminata were consistent with a transition from a rare, undetected, species to a nuisance species. We used new sequences of chloroplast regions to examine the genetic similarity of D. geminata populations from New Zealand and six overseas locations. We found no evidence for declines in DRP concentrations since 1989 that could explain the spread of proliferations since 2004. At some affected sites, lowest DRP occurred before 2004. Trends in DIN also did not indicate enhanced suitability for D. geminata. Lack of diversity in the chloroplast intergenic regions of New Zealand populations and populations from western North America is consistent with recent dispersal to New Zealand. Our analyses did not support the proposal that D. geminata was historically present in New Zealand rivers. These results provide further evidence countering proposals of general ubiquity in freshwater diatoms and indicate that, as assumed in 2004, D. geminata is a recent arrival in New Zealand.


| INTRODUCTION
Species introductions are a recognized component of humanmediated global change (Ricciardi, 2007;Vitousek, D'Antonio, Loope, Rejmanek, & Westbrooks, 1997). Both deliberately and inadvertently introduced species include those that are considered invasive because of their detrimental impacts on indigenous ecosystems in their new locations (e.g., Leung, Finnoff, Shogren, & Lodge, 2005). The geographic origins of non-native, invasive, macroorganisms are often known, as are those of plant and animal-borne pathogenic microorganisms (but see Caudill & Caudill, 2016). Establishing the geographic origins and distributions of free-living microorganisms (invasive or otherwise) can be more problematic (Van de Vijver, Kelly, Blanco, Jarlman, & Ector, 2008;Vanormelingen, Verleyen, & Vyverman, 2008). Microorganisms frequently have ubiquitous distributions because their small size and often large populations facilitate movement over long distances (Finlay & Clarke, 1999). This "ubiquity hypothesis" is based on the well-known tenet that, for microorganisms, "Everything is everywhere, but the environment selects" (Baas-Becking, 1934). However, identification of distinctive freshwater diatom taxa with restricted distributions alongside truly cosmopolitan taxa (Vanormelingen et al., 2008) reinforces a view that everything is not everywhere (even if the environment is suitable) (O'Malley, 2008) and that the distributions of microorganisms that have different dispersal and survival capabilities cannot be expected to conform to any consistent patterns (Novis, Beer, & Vallance, 2008). A fundamental problem is that proof of non-native status for any microorganism requires demonstration of the absence of viable propagules from a region, which is practically impossible using conventional survey methods (Finlay, Monaghan, & Maberly, 2002).
Ability to determine whether a species is native or non-native becomes important when other components of global change, such as shifts in nutrient pools associated with widespread agriculture and changes in river flows linked to both water diversions and climate change (Foley et al., 2005), create conditions that favor proliferations of particular microorganisms. In this situation, is it possible to identify whether an apparently invasive microorganism new to an area really has been introduced, or is a native taxon that was previously extremely rare? The implications of this question for freshwater management have been highlighted recently by the example of Didymosphenia geminata (Lyngbye) Mart. Schmidt .
Didymosphenia geminata is a large, stalked freshwater diatom that inhabits low-nutrient streams and rivers and is thought to be native to boreal and mountainous regions of the Northern Hemisphere (Blanco & Ector, 2009). The cells attach to surfaces in rivers with adhesive extracellular polymeric substance (EPS, Wetherbee, Lind, Burke, & Quatrano, 1998) then exude an EPS stalk ( Figure 1a). Following cell division, each daughter cell produces a stalk so that colonies eventually comprise bifurcating stalks with a layer of cells at the colony surface ( Figure 1b). D. geminata's ability to form high biomass (Figure 1c) in low-nutrient rivers appears to result from excessive EPS production in nutrient-limiting conditions Bothwell, Taylor, & Kilroy, 2014;. The proposed mechanism is photosynthetic "overflow" production of carbohydrate when insufficient phosphorus is available to sustain cell division, but light levels favor photosynthetic activity (Staats, Stal, de Winder, & Mur, 2000). In New Zealand, the dissolved reactive phosphorus (DRP) threshold below which proliferations can form appears to be approximately 2 mg/ m 3 Kilroy & Bothwell, 2012). D. geminata is therefore of particular concern because its nuisance proliferations occur in oligotrophic waters, in contrast to other algal blooms, which are normally associated with excessive cell growth under the influence of high nutrients (e.g., Dodds & Smith, 2016;Graham, Graham, & Wilcox, 2009, p. 586). High benthic algal biomass dominated by D. geminata may adversely affect the recreational values of rivers (Beville, Kerr, & Hughey, 2012) and alter the structure of higher trophic levels (Jellyman & Harding, 2016). Over the past three decades, proliferations of D. geminata have become more common worldwide including in areas where the species was previously rarely or not reported. Initially, the explanation for the novel occurrence of these blooms was the human-mediated spread of an invasive species (Bothwell, Lynch, Wright, & Deniseger, 2009).
More recently, based on the proposal that very low DRP concentrations are a primary cause of D. geminata proliferations , Taylor and Bothwell (2014) hypothesized that the range expansion of D. geminata blooms since the 1980s was largely a result of environmental change, specifically, declining DRP in rivers, or oligotrophication (Eimers, Watmough, Paterson, Dillon, & Yao, 2009;Stockner, Rydin, & Hyenstrand, 2000). This view was consistent with observations from Vancouver Island, Canada, from where there are records of D. geminata in the 1880s but no reports of proliferations until the 1980s (Bothwell et al., 2009). The first reports of D. geminata blooms on Vancouver Island followed a major nitrogen forest fertilization program . Application of N in the catchment would have led to increased terrestrial assimilation of P and, consequently, reduced export of P to waterways. However, no DRP concentration data were presented to support the hypothesis and there has been subsequent debate over its applicability (Bergey & Spaulding, 2015;Keller, Hilderbrand, Shank, & Potapova, 2017;Taylor & Bothwell, 2015). Robust environmental data either supporting or refuting this ubiquity hypothesis for D. geminata are still lacking.

Didymosphenia geminata proliferations were first reported in New
Zealand in October 2004 in the Waiau River catchment (Southland, South Island) (Kilroy, Snelder, Floerl, Vieglais, & Dey, 2008). Since 2004, several lines of circumstantial evidence supported the view that D. geminata is a recent introduction to New Zealand. These include lack of reliable prior records of this very distinctive species either from living or fossil diatom collections, and the pattern of spread (Kilroy & Unwin, 2011). Nonindigenous status of D. geminata in New Zealand is widely accepted (e.g., Keller et al., 2017). Nevertheless, the Taylor and Bothwell (2014) hypothesis has exposed the possibility that D. geminata has been historically present in New Zealand as an extremely rare species.
Here, we present the results of new analyses that address the ubiquity hypothesis for D. geminata and the probability of its presence in New Zealand well before the first proliferations were discovered in 2004. Two approaches were taken. First, we analyzed data on DRP and dissolved inorganic nitrogen (DIN), which have been collected from a suite of 77 New Zealand rivers since 1989. The data were used to test the hypothesis that the novel appearance of D. geminata proliferations in New Zealand can be linked to declines in DRP to concentrations that favor proliferations. DIN concentrations were also investigated because it has been suggested that low-level increases in DIN in rivers may also stimulate D. geminata proliferations, provided DRP concentrations are sufficiently low (Kilroy & Larned, 2016). Kilroy and Larned (2016) suggested that, in New Zealand, D. geminata requires DIN concentrations >10 mg/m 3 for proliferations to form.
Second, we confirmed genetic findings from the nuclear ITS region of D. geminata (Kelly, 2009) that populations of D. geminata from New Zealand and overseas are genetically very similar, using new sequences of chloroplast intergenic regions. We used the combined environmental and molecular information, along with existing circumstantial evidence (Kilroy & Unwin, 2011), to reassess the probability of presence of D. geminata in New Zealand prior to the first discovery of proliferations in 2004.

| Trends in nutrient concentrations and relationships with D. geminata cover
New Zealand's National River Water Quality Network (NRWQN) dataset comprises data on flow, water quality, and estimated periphyton percentage cover collected every month at 77 river sites from throughout the North and South Islands ( Figure 2, Table S1). The dataset includes 32 reference sites, with largely undeveloped catchments. The remaining 45 sites (termed impacted hereafter) have varying amounts of urban or agricultural development in their catchments.
Data collection commenced in January 1989 (Davies-Colley et al., 2011) and is ongoing at most sites.

| Field and laboratory methods
Periphyton cover (including D. geminata) was assessed at each site using visual estimates of percentage cover by periphyton mats and filaments at 10 points along one or two transects in wadeable water depths. Prior to 2012, D. geminata presence was reported in field notes, which allowed estimation of mean percentage cover on each survey. After January 2012, cover by D. geminata was recorded as a separate category. We used the monthly mean cover data from each site and also extracted maximum annual cover (January to December). Sites were assigned to one of four groups according to the D. geminata cover eventually recorded the following: (1) sites with persistent cover (visible cover recorded every year since first observed at the site, with >20% cover in at least half of the years); (2) occasional visible but not persistent cover (visible cover recorded in five or fewer years with maximum cover >20% in no more than two of the years); (3) negligible visible cover (cover <1%, or not recorded, but presence confirmed from microscopic examination of samples); (4) not detected.
As part of the NRWQN monitoring, a 1-L water sample was collected at each site, filtered within 24 h of collection through 0.45μm cellulose membrane filters, and analyzed for NO 3 -N, NH 4 -N, and DRP using a Lachat QuikChem FIA+ 8000 series analyzer (Lachat Instruments, Milwaukee, WI, USA). Analytical detection limits were approximately 0.5, 1.8, and 0.5 mg/m 3 for NO 3 -N, NH 4 -N, and DRP, respectively. DIN was calculated as the sum of NO 3 -N and NH 4 -N.

| Data analysis
We first inspected plots of 12-month moving averages of DRP (geo- where the smoothed value is the value predicted by a LOWESS smoother curve fitted to the nutrient vs. flow data (Smith, McBride, Bryers, Wisse, & Mink, 1996). The data series began in 1995 for DIN, because NH 4 -N data were unreliable for the first five years of the Trends were calculated using the Seasonal Sen Slope estimator, which calculates the median of slopes between all pairs of data points within each month. The importance of the trend was determined by calculating a symmetric 100 (1-2α) % confidence interval on the Sen Slope. Evidence for any trend up or down was inferred if the confidence interval did not contain zero. If the interval contained zero, the interpretation was that there were insufficient data to determine the trend direction. All trend analyses were performed using the freeware TimeTrends v. 5.0, http://www.jowettconsulting.co.nz/home/ software.
The dataset was divided into periods prior to widespread occurrence of D. geminata (up to and including 2005) and following its widespread occurrence in multiple South Island rivers (2006 adjusted value = raw value − smoothed value + median value. F I G U R E 2 Locations of the 77 NRWQN sites in the North and South Islands of New Zealand showing reference and impacted sites. The locations of sample collections for the molecular analysis of New Zealand material are also shown. Rivers shown are stream order 5 or greater. Refer to Table S1 for coordinates and details for each site to 2015). The two periods are referred to subsequently as the pre-D. geminata and post-D. geminata periods. We compared DRP distributions between the two periods and average annual concentrations within groups of sites based on D. geminata abundance. Statistical significance of differences in distributions was determined using two-sample Kolmogorov-Smirnov tests, and between means using two-sample t tests or ANOVA. Data were log-transformed to ensure homogeneous variance. At sites with persistent D. geminata cover in the post-D. geminata period, we used linear regression to assess relationships between maximum annual percentage cover and mean annual DRP over the years since establishment of D. geminata.

| Molecular analyses
DNA was extracted from the samples shown in Table S2 (see Figure 2 for locations) as described by Novis, Schallenberg, and Smissen (2015).
The published chloroplast genome of D. geminata (Genbank accession KC509523) was used to design primers spanning intergenic regions in both the large and small single-copy regions of the genome. Sequences of amplicons from three primer sets are reported here (Table 1), corresponding to the regions between atpF-atpH, rbcS-rbcL, and secA-rbcR. These primers were designed such that the 3′ end of one of each pair overlapped the spacer region, in order to increase specificity and avoid cloning from environmental samples.
PCR was run according to the following conditions: 94°C for 4 min, followed by 35 cycles of 94°C for 30 s, 57-60°C (depending on the sample) for 30 s, and 72°C for 45 s. Products were visualized on agarose gels using ethidium bromide staining. Sequencing was carried out by Landcare Research, Auckland, New Zealand, using BDT 3.1 (Applied Biosystems, Foster City, CA, USA). Amplifications were successful for all three regions in all samples except for the sample from Norway, in which the rbcS-rbcL failed to amplify. This region was included in the dataset as missing data. All electropherograms were of high quality apart from the rbcS-rbcL sequence of the Iranian sample. This sequence was estimated from the trace with assistance from other Didymosphenia sequences for the same region, but may contain some errors. As Didymosphenia from this site was shown to be the most distant from the New Zealand samples according to the other regions (sequence differences and intron presence), this would have minimal effect on the analysis.
Multiple sequence alignment was carried out using MEGA version 6 (Tamura, Stecher, Peterson, Filipski, & Kumar, 2013) and checked by eye. The resulting dataset contained 12 sequences and 2230 sites, 14 of which were variable (2 parsimony-informative), not counting a 53 base pair insertion unique to the Iranian sample. Polymorphic sites were triple-checked on electropherograms due to the low diversity present in the sequences. We elected to present the results as numbers of differences and p-distances, also as a result of this low diversity. No D. geminata was observed at any of the 44 sites in the North Island (Table S1).

| Trends in nutrient concentrations and
Using raw data, no trend in DRP was detectable between 1989 and 2006 at all 11 South Island sites in which persistent D. geminata cover ultimately developed (Table 2). Declines in DRP concentrations were detected at three of the eight sites with occasional visible D. geminata cover. One site with negligible cover and two sites at which D. geminata was not detected showed an increasing trend.
Using flow-adjusted data revealed an upward trend at four sites with persistent D. geminata (AX1, AX2, AX3, and TK6) and two with occasional cover (CH2 and TK2), and a downward trend at three sites with occasional cover (GY1, TK1, and TK5) ( Table 2).
Across the whole dataset of 77 sites, using flow-adjusted data, there was evidence for a decline in DRP concentrations at nine sites (12%). Approximately half of all sites showed trends whose direction was not detectable from the data, and 38% an upward trend.
The upward trends occurred more frequently at sites classed as impacted rather than at reference sites, particularly in the North Island (Table 3).
The distributions of all DRP data from North and South Island sites in the pre-and post-D. geminata periods appeared similar in the two periods, although K-S tests showed significant differences (p < .001); distributions differed markedly between the North and South Islands (Figure 4). Mean average annual DRP in the pre-and post-D. geminata periods was not significantly different (North Island, 13.0 vs. 11.9 mg/m 3 , respectively; South Island, 3.3 vs. 3.2 mg/m 3 , respectively, two-sample t tests, p > .25). Average annual DRP also did not differ between pre-and post-D. geminata periods within groups of sites based on D. geminata cover ( Figure 5). DRP differed between the groups within each period, except between T A B L E 1 Sequences and binding sites of primers used in this study  Mean annual DRP and maximum annual percentage cover by D. geminata were negatively correlated at two sites: mean annual DRP explained 56% (p < .02, n = 9) and 43% (p < .05, n = 9) of the variance in annual maximum percentage cover, across ranges of DRP of 0.9-1.9 and 0.8-2.3 mg/m 3 , at NN5 and TK6, respectively.  Percentage cover of wadeable area by D. geminata mats high cover; concentrations were generally higher at the sites classed as impacted (Table S1). Of the 12 reference sites at which D. geminata was recorded, there was an upward trend in flow-adjusted DIN at two sites, both of which have subsequently had negligible cover of D. geminata.
The trend was down at seven sites, and no trend was detectable at three sites (Table 3). Flow-adjusted DIN declined at three of the reference sites (NN2, DN2, and DN10) at which D. geminata was not detected, and there was no evidence for a trend at a fourth site (DN6 , Table 4).

| Molecular analyses
Genetic variation over the chloroplast regions sequenced was very small. All collections of New Zealand material were identical (Table 5).
These were also identical to material collected from Vancouver Island (Canada) and Colorado (USA). Small differences (1-3 bp) were found between material from these sites and that collected from River Cocquet (United Kingdom), Boulder Creek (Missouri), and the Nidelva Percentage cover of wadeable area by D. geminata mats F I G U R E 3 (Continued) River (Norway). Material from Iran was more divergent from other collections (6-10 bp). In addition, the Iranian material contained a 53 bp insertion in the atpF-atpH intergenic region; this insertion was also largely shared by the species Gomphoneis minuta var. cassieae (not shown) and was mostly alignable between the two, indicating that its inclusion in the genome represents the ancestral state in this group.

| DISCUSSION
Although vigorous debate has taken place in the scientific (Bergey & Spaulding, 2015; and popular literature (Science Media Centre 2014) concerning the role of declining phosphorous in the appearance of problematic blooms of D. geminata, empirical data to test the hypothesis of Taylor and Bothwell (2014; that reduced P favors D. geminata blooms) have been lacking until now. Bergey and Spaulding (2015) noted that "investigations of patterns of nutrient concentrations relative to blooms are needed". Here, we have used a long-term, high-quality, river monitoring dataset from New Zealand to specifically address this knowledge gap.
Before considering water chemistry, several additional lines of evidence are pertinent to the appearance of D. geminata in New Zealand. Taylor and Bothwell (2014) suggested that small populations of D. geminata persisting under conditions unfavorable to blooming could have been missed during historical routine monitoring; this is possible but contentious, due to the species' large, distinctive cells   (Bergey & Spaulding, 2015;Kilroy & Unwin, 2011). Furthermore, the biology of D. geminata, which has now been subject to considerable investigation, would seem to preclude its long-distance dispersal to an isolated archipelago without assistance from humans. It is destroyed by exposure to salt, freezing, drying, and the low pH characteristic of a bird gut (Kilroy, Lagerstedt, & Robinson, 2007), thereby restricting its dispersal by wind, ocean currents, or birds. If this is correct, a strict ubiquity hypothesis for D. geminata requires a vicariant origin for the species (co-occurrence in an ancestral landmass that has since split up). However, as the estimated separation of New Zealand from other land masses is >80 Mya (Laird & Bradshaw, 2004), as a molecular clock (Kumar, 2005). Although molecular clocks are influenced by idiosyncracies of individual lineages, such as generation times, population sizes, and intensity of selection (Ayala, 1999), and divergence estimates are associated with errors (sometimes considerable; Graur & Martin, 2004), isolated populations of D. geminata persisting in New Zealand for substantial time periods would accumulate mutations in the intergenic chloroplast regions sequenced by us. In fact, there is almost no diversity in these regions (Table 5), and little more in the ITS region (Kelly, 2009) further indicating recent dispersal. The chloroplast sequence data (being identical within New Zealand and to sequences from Vancouver Island and Colorado) suggest a dispersal event from North America, where a greater genetic diversity within the species occurs, although the limited geographic sampling of this dataset must also be acknowledged. Similar data have been recently used to assert that D. geminata is a recent invader to parts of eastern North America (Keller et al., 2017); however, the USA situation differs in having prior records further west (e.g., Letham et al., 2016), with no oceanic dispersal barriers to the eastern seaboard. Our limited sampling shows some diversity within North American specimens (Table 5), as also found by others (Keller et al., 2017). Neither earlier records nor genetic diversity were features of the New Zealand data.
Our dataset, with low sequence diversity, satisfies the common assumptions of chloroplast sequence evolution made in phylogenetics (inheritance as a single nonrecombining linkage, i.e., without intergenomic recombination; Christie & Beekman, 2017;Sullivan, Schiffthaler, Thompson, Street, & Wang, 2017). However, it should be noted that the contributions of different processes shaping the evolution of chloroplast DNA sequences in D. geminata are currently uncertain. In particular, it is not clear whether D. geminata undergoes a sexual cycle. Size restoration, usually assumed to occur following sexual reproduction, has been clearly demonstrated in D. geminata (Bishop & Spaulding, 2017), but the possibility of an asexual mechanism to achieve this has not been ruled out, as neither these authors nor others have observed any other sign of sexual reproduction despite examining many thousands of cells. The mode of inheritance of plastids appears to affect the architecture of their genomes (Crosby & Smith, 2012), and inter-and intraspecific recombination has been detected in the plastid genome of Pseudo-nitzschia, another raphid diatom that features biparentally inherited plastids (as might have been expected for D. geminata). The enigmatic life cycle of D. geminata thus also represents an intriguing unsolved mystery regarding its chloroplast genetics. Taylor and Bothwell (2014) posited more than ubiquity of D. geminata; they also proposed a mechanism to explain its apparent absence in many sites prior to its recent proliferation. Under their model, this species has been globally present for some time, but has only been detected recently in many areas due to lowered phosphorous concentrations in freshwaters.
However, we found no correlation between the timing of the first In practice, we expect stochastic effects to operate, even on the distribution and dispersal of microalgae (e.g., Novis et al., 2008), leading to nonuniform dispersal; however, we regard the strong evidence from multiple sites that prolific D. geminata growth did not occur at historically lowest DRP concentrations as confirmation of very recent first dispersal of the species to New Zealand.
While lack of consistent declines in DRP up to the time of the first discovery of visible D. geminata in a New Zealand support the contention that the species is a recent arrival, the combined DRP and percentage cover data do lend support to the Taylor and Bothwell (2014) hypothesis regarding the association between D. geminata prolifera-  , and changes in cover under prolonged elevated DRP and stable flows may take months (Kilroy & Larned, 2016 (Biggs, 1996) and nutrient concentrations (Alexander et al., 2009 River. Dam-regulated rivers are especially favorable for proliferations (Kirkwood, Jackson, & McCauley, 2009).
In a broader context, our findings for D. geminata in New Zealand provide further evidence that counters the proposal of ubiquity for freshwater diatoms in general (Finlay et al., 2002). Finlay et al. (2002) maintained that: "The argument in favor of endemic diatom species is untenable, because it is not possible to disprove their existence elsewhere in the biosphere". We contend that our combined environmental and genetic evidence against the presence of D. geminata in New Zealand prior to about 2004 is close to qualifying as proof of the species' absence in this particular part of the biosphere. The natural range of D. geminata likely extended across broad areas of the Northern Hemisphere (Blanco & Ector, 2009), and the species could be considered endemic to that region. However, extensive bloom formation was evidently rare, although blooms have been reported for many decades in Norwegian rivers (Lindstrøm & Skulberg, 2008). The absence of D. geminata from New Zealand until the early 2000s likely reflected its relative rarity in the Northern Hemisphere, until blooms started to become widespread in the late 1980s (Blanco & Ector, 2009), increasing the likelihood of human-mediated, long-distance transport of D. geminata cells. Once in New Zealand, D. geminata's eventual distribution was determined by environmental suitability so that, currently, the species appears to be restricted to low-phosphorus rivers, which occur mostly in the South Island. In conclusion, we find no evidence either from the time series of monthly DRP and DIN data, or from new molecular analyses, to support the ubiquity hypothesis for D. geminata  in the New Zealand context. None of the data suggest that D. geminata was present in New Zealand freshwaters prior to the early 2000s, indicating that its initial treatment in 2004 as an introduced invasive species (Kilroy & Unwin, 2011) was justified.
However, our data do support the proposal by Taylor and Bothwell (2014) that the dynamics of D. geminata proliferations in rivers over time are likely to be linked to changes in DRP concentrations. We also agree with Taylor and Bothwell (2014) that distinguishing native and non-native invasive species is important, and note that our findings do not address invasion history of the species in areas of the Northern Hemisphere.
Crown Research Institutes from the Ministry of Business, Innovation and Employment's Science, and Innovation Group [PN]. We thank C.
Cary (University of Waikato), for samples of D. geminata collected from outside New Zealand, and M. Schallenberg (University of Otago) for the sample from Wye Creek, Otago. Field sample and data collections for the NRWQN were conducted by NIWA field teams. We thank G.
Bryers (NIWA) for water nutrient analyses and provision of NRWQN data, and Phil Jellyman reviewing an earlier version of the manuscript.

AUTHOR CONTRIBUTIONS
CK performed data analysis and interpretation in the water chemistry component of the study; PN designed, executed and interpreted the molecular component of the study; CK and PN wrote and revised the manuscript.

CONFLICT OF INTEREST
None declared.