Copyright is owned by the Author of the thesis. Permission is given for a copy to be downloaded by an individual for the purpose of research and private study only. The thesis may not be reproduced elsewhere without the permission of the Author. THE EFFECTS OF HYDROLOGICAL AND NUTRIENT DISTURBANCE ON STREAM INVERTEBRATE COMMUNITIES USING A TRAIT-BASED APPROACH A thesis presented in partial fulfilment of the requirements for the degree of Doctor of Philosophy in Aquatic Ecology at Massey University, Manawatu, New Zealand Yen Dinh Thi Hai 2018 i Abstract Anthropogenic altered flow regimes and nutrient enrichment can cause significant impacts on stream biota and may lead to species loss if characteristics of the local fauna are not compatible with the new environmental conditions. I used fourth corner models, Bayesian ordination, and regression analysis to assess those potential effects on trait and species composition of invertebrate communities in UK, New Zealand (NZ) and Vietnamese streams. NZ temperate mountain streams with greater substrate disturbance increased the abundance of plastron- respirers, but not those having two aquatic life stages or who were filter feeders. UK temperate rivers with predictable multiple high flows per year supported individuals having highly synchronized life history strategy; rivers with one prolonged rising climb and strong groundwater influence were better for those having a high reproduction strategy, and rivers with a steep peak flow supported both strategies. Nutrients affect functional feeding and life history traits via promoting algal overgrowth in NZ streams. Both periphyton biomass and nutrients increased the abundance of algae piercers, collectors and those having two aquatic life stages, being long-lived and having hermaphroditic ability; but decreased the abundance of shredders, scrapers, and those having univoltine life cycles. The post-flood recovery of invertebrate communities depended on the recovery of the food base of the food web that was, in turn, determined by the presence of a forest canopy cover and nutrient levels in a stream. Within the forest canopy stream, communities in the low nutrient site recovered by week 9 after a 1-in-50-year flood in Wellington, NZ. Without the forest canopy, the recovery of communities in nutrient- ii impacted streams (by 25 weeks) was probably associated with a quicker regrowth of periphyton while communities in the low nutrient site had not recovered even after 40 weeks. Hydrological disturbances, nutrients, and their combination had strong effects on invertebrate communities in temperate streams. Taxa that survive in a site have trait characteristics that are highly compatible with both the hydrological and nutrient conditions at a site. In contrast to temperate invertebrate communities, Vietnamese tropical highland community structure was influenced more by elevation than disturbance. Further studies are required to clarify how flow disturbance may effect invertebrate communities in tropical streams. iii Acknowledgement I have received considerable support from my supervisors, Russell Death and Ian Fuller. To Russell and his lovely family, thanks so much for your time and encouragement. I really enjoyed our “big” and “small” talks both with and without coffee. To all my friends, I thank you for sharing the good and bad memories. To Nat, Chris and Hugo, my Kiwi family, I thank you for your kindness that you provided when we first moved to Palmy. Without you, our settlement would have been extremely difficult. To my parents, my brother Kien Dinh, and my little family, I love you so much and thank you for always being there for me. My journey could not have been completed without the financial support from the New Zealand Asean Scholarship. It gave me the unique opportunity to do my PhD at Massey University. Thank you also to support from ISSO members and Ecology Group staff during my time at Massey University. iv Preface The thesis was written up in the format of a paper-based thesis, therefore, each chapter was presented as a standalone manuscript. Consequently, information in methodology in each chapter may be repeated. Chapter 1-8 are primarily my work with input from my chief supervisor, Professor Russell Death. Professor Russell Death provides critical contribution on manuscript development and editing. Therefore, he is a co-author on all manuscripts prepared. Wendy Monk and Paul Wood are co-authors on Chapter Two because they supplied the hydrological and raw invertebrate data for analysis. Hieu Quang Nguyen is a co- author on Chapter Five because he identified invertebrate samples. Statement of Author contribution can be found in Appendix A. v Table of contents Abstract i Acknowledgement ii Preface iii Chapter 1 General introduction 1 Chapter 2 Do disturbance and periphyton productivity affect stream invertebrate traits? 6 Chapter 3 The effects of flow regime on life history strategies of aquatic invertebrate communities 31 Chapter 4 The effects of nutrient enrichment on invertebrate traits in New Zealand streams 43 Chapter 5 The effects of nutrient enrichment on the recovery of stream invertebrate communities after large floods 61 Chapter 6 Does flow disturbance affect diversity and community composition of tropical stream invertebrate communities 78 Chapter 7 Synthesis 93 References 98 Appendix A Statements of Author contribution 118 Appendix B List of family and their life history strategy in Chapter 3 124 Appendix C Raw invertebrate data 128 1 Chapter 1. General introduction Aquatic ecosystems have been severely impacted by land-use and climate change (Strayer and Dudgeon, 2010; Death et al., 2015; Taniwaki et al., 2017). Intensive agriculture and land use conversion, and rapidly growing world population have caused increased water abstraction and nutrient loads from land into aquatic ecosystems (Selman and Greenhalgh, 2010). Climate-related impacts further deteriorate aquatic ecosystems via increased temperature, and the increasing frequency and magnitude of floods, and droughts (Strayer and Dudgeon, 2010). Increasing water temperature has been extensively studied with about 10,100 studies found in Google scholar up to 2015. However, studies on human and climate-induced changes in flow pattern have had much lower attention with approximately 4,420 studies on floods and 2,490 studies on droughts. The effects of current and projected changes to flow disturbances on aquatic ecosystems, therefore, require more investigation. Anthropogenic impacts have already increased the frequency and magnitude of floods and/or extreme flood events (Brown et al., 2007; Death et al., 2015; O'Connor et al., 2015). Any changes to flow pattern can cause significantly negative impacts on stream biota, and may lead to species loss if those fauna do not have characteristics compatible with the new flow regime (Lytle and Poff, 2004; Dudgeon et al., 2006; Vandewalle et al., 2010). Additionally, these aquatic ecosystems have been degraded by increasing nutrient inputs (Allan, 2004; Morgan and Cushman, 2005; Strayer and Dudgeon, 2010). High nutrient levels are associated with the loss of biodiversity (Morgan and Cushman, 2005; Strayer and Dudgeon, 2010), a shift in community structure toward species that feed on algae, bacteria, and fungi (Gafner and Robinson, 2007; Matthaei et al., 2010), a reduction in recreational and amenity values, and an increase in the costs for drinking water 2 treatment (Foote et al., 2015). The concurrent occurrence of both flood-induced and nutrient-induced stresses, therefore, likely pose an increasing threat to stream biota (Staudt et al., 2013; Death et al., 2015). Biological traits have successfully discriminated the role of potential anthropogenic environmental drivers affecting invertebrate communities (Menezes et al., 2010; Statzner and Beche, 2010; Baird et al., 2011). Results from trait analysis can easily transfer findings from one region or country to another thus the trait-based approach creates a comparative tool for environmental management over a range of scales that taxonomic analysis cannot offer (Dolédec and Statzner, 2010). High flows physically remove individuals and periphyton biofilm, an important food source for many aquatic invertebrates (Scarsbrook and Townsend, 1993; Biggs, 1995; Death and Zimmermann, 2005). Benthic invertebrates possess characteristics or traits (e.g., streamlined bodies, flexible bodies, or small size) that help them maintain themselves in streams with high flows (Lake, 1990; Townsend et al., 1997a), avoid high discharge events (Wallace, 1990), or recolonize quickly after floods (Winterbourn, 1997). However, the trait-based approach has not been used to investigate the relative roles of physical removal and periphyton reduction in the overall effect of flood disturbance (Scarsbrook and Townsend, 1993; Townsend et al., 1997a). Different flow regimes may select for distinctive trait composition of invertebrate communities because invertebrates that survive at a site are likely to have life history traits that reflect the timing, frequency, predictability and severity of flow regimes (Lytle, 2001; Lytle and Poff, 2004). However, the magnitude of change in the proportions of life history traits between flow regimes is typically small, namely less than 15% when trait proportions are compared between Mediterranean and temperate streams (Bêche et al., 2006; Bonada et al., 2007a; Bonada et al., 2007b). The small variation in trait proportions 3 was probably associated with the single trait-based approach that has been commonly used in the literature (Bêche et al., 2006; Bonada et al., 2007a; Statzner and Beche, 2010). Environmental selective forces act on whole organisms with specific life history trait combinations rather than individual traits separately, such that life history traits are evolutionary auto-correlated (Poff et al., 2006; Verberk et al., 2013). Without considering phylogenetic linkages, results from the current trait-based approach might not be adequate to elucidate the overall response of traits to environmental changes (Poff et al., 2006; Verberk et al., 2013). The effect of floods and spates on freshwater animals can be similar, regardless of their geographical location (Lake, 2000; but see Death and Barquín, 2012). Flow disturbances are the principal environmental factor shaping the diversity and community structure in Mediterranean and temperate streams (Chessman et al., 2010; Dolédec et al., 2017; Tonkin et al., 2017). Hydrological indices explain between 41% to 52% of the variation in invertebrate community composition in temperate New Zealand streams (Clausen and Biggs, 1997). However, it is still unclear exactly how flow disturbances structure tropical stream invertebrate communities (Boyero et al., 2009; Md Rawi et al., 2014; Tonkin et al., 2016). Nutrient enrichment results in excessive epilithon growth that can shift community composition of invertebrate communities to those that fed on algae, bacteria and fungi, as well as decrease habitat quality by homogenizing invertebrate habitats, decreasing oxygen concentration and changing pH (Gafner and Robinson, 2007; Yuan, 2010; Lange et al., 2014). However, there is no study that has directly examined the effects of nutrient-induced changes in food sources on stream invertebrate traits (Matthaei et al., 2010; Lange et al., 2014). 4 The dual increases in flood-intensity and nutrient enrichment are likely to become more common (Staudt et al., 2013). Invertebrate communities can take three to ten years to recover after extreme floods (greater than 1-in-50 year events), whereas, it is typically a few weeks after smaller floods (Lake, 2008; Death et al., 2015; Reich and Lake, 2015). In contrast, a more rapid recovery of stream periphyton after floods at high nutrient level, may facilitate a quicker recolonization of invertebrate communities in nutrient-impacted streams (Biggs, 1995; Death and Zimmermann, 2005; Tonkin et al., 2013). Therefore, how these nutrient-enriched ecosystems respond to large floods still remains unclear (Death et al., 2015). Thesis structure and aims Questions to be investigated in this thesis include (i) how trait-based approaches can be used to study the effects of flow disturbance and nutrient enrichment on stream invertebrate communities, (ii) how flow disturbance and nutrient enrichment combine to effect stream invertebrate communities, and (iii) the effects of flow disturbances on stream invertebrate communities in tropical streams. The trait-based analysis was used in Chapter 2, 3 and 4 to study the effects of flow and nutrient disturbance on trait composition of aquatic invertebrate communities in temperate streams. Chapter 5 used the taxonomic-based approach to investigate how different nutrient levels at differing flow disturbance levels influence invertebrate assemblages. Chapter 6 also used a taxonomic approach as there is a lack of Vietnamese trait data. The objective of the main chapters in my thesis are as follows: 5 Chapter 2 Examine the relative importance of substrate disturbance and periphyton biomass removal on invertebrate traits in close-canopied sites and open- canopied sites in ten streams in Egmont National Park, New Zealand. Chapter 3 Investigate whether the life-history strategies of invertebrate trait combinations, in 83 English and Welsh rivers are associated with patterns of river flow. Chapter 4 Examine the effects of periphyton biomass, total nitrogen and dissolved reactive phosphate on functional feeding and life history traits of New Zealand stream invertebrates. Chapter 5 Examine the recovery of invertebrate community composition from a 1- in-50-year event flood in Wellington, New Zealand. The study was conducted at five stream sites with contrasting nutrient concentrations from low to high nutrient concentrations. Chapter 6 Examine the relationship between flow disturbance and diversity and community composition of invertebrate communities in Vietnamese mountain streams. Chapter 2. Disturbance – productivity effects on stream invertebrate traits 6 Chapter 2. Do disturbance and periphyton productivity affect stream invertebrate traits? Authors: Yen T.H. Dinh and Russell G. Death Affiliations: Innovative River Solutions, College of Science, Massey University, Private Bag 11-222, Palmerston North 4442, New Zealand Manuscript status: Published, reference Dinh YTH, Death RG. 2018. Do disturbance and periphyton productivity affect stream invertebrate traits? Freshwater Science. 37(2): 367-379. Chapter 2. Disturbance – productivity effects on stream invertebrate traits 7 Abstract: Disturbance in lotic ecosystems strongly influences which animals can survive and how those ecosystems function. Flow disturbances physically remove animals and periphyton, but it is unclear whether physical removal of individuals or the loss of food is the principal driver of effects of flow disturbances on invertebrate communities. Invertebrates possess traits that help them withstand high flows or recolonize rapidly after floods. At light-limited, closed-canopy sites, periphyton biomass should be unaffected by disturbances, so disturbance should affect invertebrate community trait composition by direct removal of animals. At open-canopy sites, disturbance should affect trait composition by both removing individuals and reducing periphyton food resources. We investigated whether a trait-based approach could elucidate drivers of effects of flow disturbance on benthic invertebrate communities and better identify potential mechanistic linkages. We sampled 10 autotrophic streams that differed in substrate-disturbance regime and varied from 100 to 0% canopy cover. We sampled 2 sites per stream, 1 in forest (closed-canopy) and 1 downstream in low-intensity agricultural grassland (open- canopy). Regardless of canopy, in streams with greater substrate disturbance, the proportion of individuals that respired with a plastron increased and proportions of individuals having 2 aquatic life stages or who were filter-feeders decreased. At open- canopy sites, the proportion of collector-gathers with flattened bodies increased with increased substrate disturbance and decreased periphyton biomass, and the proportion of taxa having high body flexibility and 2 aquatic life stages increased with decreased substrate disturbance and increased periphyton biomass. Trait composition of these mountain stream invertebrate communities is strongly influenced by physical removal via substrate movement and by reduced periphyton resources. Key words: invertebrate traits, flood disturbance, the loss of periphyton food, canopy cover. Chapter 2. Disturbance – productivity effects on stream invertebrate traits 8 INTRODUCTION Floods can limit the distribution, abundance, and composition of animals in lotic habitats (Poff et al., 1997; Lake, 2000). Flood disturbance usually is characterized by the magnitude, frequency, duration, timing, and rate of change of flow (Poff et al. 1997). Any alteration to the magnitude or timing of flow can alter water chemistry, foodweb energy sources, physical habitats, and biotic interactions of aquatic ecosystems (Poff et al. 1997). As a result, flow alteration, particularly from anthropogenic causes, can cause significant negative effects on stream biota (Dudgeon et al., 2006; Poff et al., 2007) that may lead to species loss (Vandewalle et al., 2010). High discharge usually results in decreased benthic invertebrate biomass, abundance, and richness (Lake, 2000; Robinson et al., 2003; Death, 2008). Floods increase shear stress on the stream bed, thereby causing substrate disturbance, which physically removes individuals (Townsend et al., 1997a; Death, 2002). A high proportion (70–95%) of individual invertebrates are removed by floods (Brooks and Boulton, 1991) despite the fact that benthic invertebrates possess characteristics or traits (e.g., streamlined bodies, flexible bodies, or small size) that help them maintain themselves in streams with high flows (Lake, 1990; Townsend et al., 1997a), avoid high discharge events (Wallace, 1990), or recolonize quickly after floods (Winterbourn, 1997). Flow disturbances also remove periphyton, an important food source for many aquatic invertebrates (Scarsbrook and Townsend, 1993; Biggs, 1995; Death and Zimmermann, 2005). Death and Zimmermann (2005) found that a decline in the number of species with increased disturbance frequency occurred only at open-canopy sites where periphyton productivity was not light limited by a dense forest canopy. They postulated that diversity (richness) was reduced principally by removal of periphyton biomass—the disturbance–productivity–diversity model (Death 2002). However, exactly how high Chapter 2. Disturbance – productivity effects on stream invertebrate traits 9 flows affect invertebrate communities is unclear. Biological traits have been used successfully to discriminate the role of potential anthropogenic environmental drivers affecting invertebrate communities (Baird et al., 2008; Menezes et al., 2010; Statzner and Beche, 2010; Baird et al., 2011; Culp et al., 2011). Poff et al. (2010) used the traits, cold stenothermal and obligate rheophily, to assess the vulnerability of benthic communities to projected temperature and runoff change in the western USA. Communities at sites with taxa having a high proportion of these traits were deemed more vulnerable. Dolédec et al. (2011) found 9 traits linked with increasing agricultural intensity in New Zealand. The proportion of taxa laying eggs beneath the water surface increased and the number of taxa laying eggs at the water surface decreased with increasing agricultural intensity. The relationships observed between traits and potential environmental drivers are relatively stable across large spatial scales (Culp et al., 2011; Dolédec et al., 2011). Thus, in contrast to taxonomic analysis, the trait-based approach creates a comparative tool for environmental management over a range of scales (Dolédec and Statzner, 2010). Dolédec et al. (2011) found that the responses of traits to agricultural intensification were consistent across New Zealand, even when taxonomic composition changed. Thus, traits offer the potential to transfer findings easily from one region or country to another. Traits have rarely been used to explain the effects of flood disturbance on invertebrate communities, even in light of the potential effects that increases in large floods from climate changes may have on riverine biological communities (Death et al., 2015). Two studies have been done in which invertebrate traits were examined in relation to substrate disturbance (Scarsbrook and Townsend, 1993; Townsend et al., 1997a). Townsend et al. (1997) examined 6 selected traits in 54 tributary sites of the Taieri River on the South Island of New Zealand and found that taxa were small, habitat generalists, Chapter 2. Disturbance – productivity effects on stream invertebrate traits 10 clingers, streamlined, or flattened, or had high adult mobility were more abundant at sites exposed to a high intensity of streambed substrate movement (Townsend et al., 1997a). However, the traits were not used in either study to investigate the relative roles of physical removal and periphyton reduction in the observed disturbance effects (Scarsbrook and Townsend, 1993; Townsend et al., 1997a). Therefore, whether removal of individual invertebrates and reduction of periphyton by disturbance have the same effect on traits in invertebrate communities affected by floods is unclear. Autotrophic streams on Mount Taranaki in Egmont National Park, New Zealand, are covered by a full canopy of native evergreen forest and emerge from the park boundary into open pasture. The 10 studied streams are in the same geographical and geological region and have similar physicochemical characteristics, but differ in substrate disturbance regimes and vary from 100 to 0% canopy cover within a distance of few hundred meters (Death and Zimmermann, 2005). Five streams experience frequent flood disturbances, whereas the others are spring-fed and more stable. We took advantage of the differing substrate disturbance regimes and the differing limiting factors on periphyton accrual in these autotrophic streams to assess the relative contribution of the 2 potential driving factors on trait compositions. We investigated the ability of a trait- based approach to inform our understanding of the mechanisms of the effects of flow disturbance on benthic invertebrate communities by identifying: 1) the invertebrate traits associated with substrate disturbance, 2) the invertebrate traits associated with periphyton biomass, and 3) the invertebrate traits associated with both substrate disturbance and periphyton food loss. Chapter 2. Disturbance – productivity effects on stream invertebrate traits 11 METHOD Study sites We sampled 10 streams on the plain around Mount Taranaki, New Zealand, a dormant andesitic cone volcano. Mt Taranaki provides sites that are geographically close and of similar geological origin. A large portion of Mt. Taranaki is within the circular Egmont National Park, which is dominated by Rimu-Rata-Kamahi forest. The boundary of the national park is a sharp transition from forest to agricultural grassland, where forest canopy is almost completely absent. We sampled 2 sites on each stream, one ~50 m inside the forest and the other several hundred meters downstream in low-intensity agricultural grassland. Paired sites on each stream were 225 to 3800 m apart and differed significantly only in the presence or absence of forest canopy: sites inside the park are fully covered by forest canopy and paired lower sites are in open pasture with 0% canopy cover. All sites were ~400 to 500 m asl, 1st- to 3rd-order, and with similar physicochemical characteristics and substrate size composition (predominately large cobbles). All sites had annual mean water temperatures of 8 to 10°C and conductivity of 58 to 96 µS/cm (see Death and Zimmermann 2005 for more detail on study sites; Fig. 1A–G). Invertebrate samples We sampled invertebrates in April, July, October 1999 and January 2000. We collected five 0.1-m2 Surber samples (250-µm mesh) from riffles at each site and stored them in 10% formalin. In the laboratory, invertebrates were removed from samples, identified to the lowest possible taxonomic level based on available keys, and counted (Cowley, 1978; Winterbourn et al., 1989; Towns and Peters, 1996). Those taxa that could not be identified to species level were separated into apparent morphospecies. Species abundance was the sum of individuals in the 5 Surber samples. Chapter 2. Disturbance – productivity effects on stream invertebrate traits 12 Figure 1. Box-and-whisker plots for stream width (A), depth (B), water temperature (C), substrate size index (D), particulate organic matter (E), conductivity (F), and chlorophyll a (G) at open- and closed-canopy sites 10 streams in Taranaki, New Zealand, in 1999–2000. Lines in boxes are medians, box ends are quartiles, whiskers are quartiles ± 1.5 * IQR (where IQR is the inter- quartile range), and dots are outliers. Periphyton biomass We assessed periphyton biomass by measuring chlorophyll a (Chl a). We collected 4 stones (maximum planar dimension < 60 cm) at random from riffles at each site monthly from April 1999 to January 2000, and kept them frozen. In the laboratory, Chapter 2. Disturbance – productivity effects on stream invertebrate traits 13 we extracted pigments separately for each stone in known volumes of 90% acetone at 5°C in the dark for 24 h. We read absorbences with a Varian Cary (Melbourne, Australia) 50 UV Visible Spectrophotometer. Values were then converted to pigment concentration following Steinman et al. (2006). We corrected pigment concentrations for the stone surface area determined by wrapping stones in Al foil of known mass/unit area. We divided total surface area by 2 because periphyton generally is found only on the upper exposed surface of stones. Substrate disturbance We assessed the effect of disturbance on aquatic invertebrates by measuring substrate movement caused by high-discharge events in each stream. Subtrate movement was measured with the aid of 15 painted tracer particles placed on the stream bed (Death and Winterbourn, 1994b). We placed stones in each of 3 size classes (91–180, 61–90, and <60 mm) in riffles, in random order in triplets (small, medium, and large) across the main flow path of the stream at marked points on the stream bank at each stream. Every month between April 1999 and March 2000, we recorded the distance travelled by each stone, and we placed any stones that had moved back at their original position. We converted the sum of the distances moved to % stone movement by expressing it as a proportion of the maximum possible mass of stones moved (maximum possible mass of stones moved = 405.5 kg), thereby giving more emphasis to the movement of larger stones. A % stone movement of 100 indicates that all stones were washed away, buried, or moved >50 m, and a value of 0 indicates no stone movement. Substrate disturbance was measured at only the closed canopy site on each stream because of the close proximity of paired sites, the absence of any tributaries between paired sites, and their similar physical characteristics and channel morphology. Chapter 2. Disturbance – productivity effects on stream invertebrate traits 14 Selection of species traits Fifteen feature groups with 50 trait categories have been identified for New Zealand freshwater invertebrates (Dolédec et al., 2011; Schmera et al., 2015). We selected 7 trait groups postulated to be linked with flow disturbance: number of aquatic stages, body flexibility, body form, mobility, dissemination potential, locomotion, potential size, and respiration (Townsend and Hildrew, 1994; Townsend et al., 1997a; Lamouroux et al., 2004; Horrigan and Baird, 2008; Statzner and Beche, 2010; Brooks and Haeusler, 2016); and 2 trait groups postulated to be related to periphyton abundance: dietary preference and feeding habit (Scarsbrook and Townsend, 1993; Brooks and Haeusler, 2016). Literature supporting their inclusion is listed in Table 1 along with our predictions of the trait responses to substrate disturbance or periphyton biomass based on that literature. We avoided, as much as possible, traits constrained by phylogeny (Poff et al., 2006) by excluding all life-history traits except maximum potential size. Taxa having small size are more resilient to flow disturbance than large taxa because of their short generation time (Townsend and Hildrew, 1994; Townsend et al., 1997b; Statzner and Beche, 2010). Chapter 2. Disturbance – productivity effects on stream invertebrate traits 15 Table 1. Predictions of trait linkages between substrate disturbance (F) and periphyton biomass (P) effects from flow disturbance made by other investigators. Trait group/trait Linkage Strategies Reference Aquatic stages 2 aquatic life stages ↓F Longer time in water, higher chance to be exposed to high flows Townsend et al. (1997) 1 aquatic life stage ↑F Body flexibility None (<10°) ↓F Higher flexibility, quicker to find refuge from high flows Townsend and Hildrew (1994), Lamouroux et al. (2004) Low (>10-45°) ↓F High (>45°) ↑F Body form Streamlined ↑F Streamlined or flattened body forms reduce drag force of high flows Townsend et al. (1997), Statzner and Beche (2010) Flattened ↑F Cylindrical ↓F Dissemination potential Low ( <10 m) ↓F Greater dispersal ability, higher capacity for intercatchment flight to avoid high flows Townsend et al. (1997) Moderate (10 m–1 km) ↓F High (>1km) ↑F Locomotion and relation to substrate Swimmers ↓F Firmly attached to the substrate, less risk of being washed away by high flows Townsend and Hildrew (1994), Horrigan and Baird (2008) Crawlers ↓F Burrowers ↓F Attached ↑F Maximum potential size Small size (≤10mm) ↑F Rapid recovery after disturbance, shorter generation time of smaller taxa Townsend et al. (1997) Large size (>10mm) ↓F Mode of respiration Tegument ↓F Plastron or gill respirers can remain underwater indefinitely to avoid high surface flow Brooks and Haeusler (2016) Plastron ↑F Gills ↑F Dietary preferences Strong (specialist) ↓P Generalists do not select their food, possibly feed on periphyton as an alternative food Scarsbrook and Townsend (1993) Moderate ↓P Weak (generalist) ↑P Feeding habit Scrapers ↑P Scrapers feed on periphyton; deposit- feeders, filter-feeders, and collectors likely to be affected by high flows; response of scrapers to increased periphyton is likely to benefit predators Scarsbrook and Townsend (1993), Brooks and Haeusler (2016) Deposit-feeders ↓F Filter-feeders ↓F Collectors ↑F Predators ↑P Chapter 2. Disturbance – productivity effects on stream invertebrate traits 16 Selection of analysis Two common statistical tools for analyzing the relationship between traits and the environment are the fourth-corner problem proposed by Legendre et al. (1997) and the RLQ analysis proposed by Dolédec et al. (1996). These statistical tools possess some significant limitations, such as the analysis of a single trait and a single environment variable at a time and the impossibility of using abundance data (Menezes et al., 2010). These 2 limitations can be addressed by the fourth-corner model framework proposed by Brown et al. (2014), in which multiple traits expressed as a taxon-trait matrix, multiple environment variables expressed by a site–environment matrix, and abundance data expressed by a site–taxon matrix are integrated in the model-based analysis (Warton et al., 2015b). Fourth-corner models use a direct 1-stage, model-based framework to fit abundance or presence/absence data against traits and environmental variables (Pollock et al., 2012; Brown et al., 2014). We kept each trait of each taxon as their original information rather than giving an affinity score based on fuzzy coding (Chevene et al., 1994), which is commonly used in the trait-based approach (Statzner et al., 2005; Bonada et al., 2007a; Dolédec et al., 2011; Pilière et al., 2015) and is the first key difference in the fourth-corner model approach (Brown et al., 2014). Considering abundance or presence/absence data rather than the weighted trait matrix as the response in the models is the 2nd key difference in the fourth-corner model approach (Brown et al., 2014). Thus, traits and environmental variables are linked directly in the models, which provides better quantification of relationships between traits and potential environmental variables. Fourth-corner models also allow users to check assumptions, quantify the nature and the strength of the environment–trait associations, identify the important associations, and forecast (relative) taxon abundance in new environmental scenarios (Brown et al., 2014; Chapter 2. Disturbance – productivity effects on stream invertebrate traits 17 Warton et al., 2015b), which similar analyses do not. Data analysis New Zealand invertebrate communities do not exhibit pronounced seasonal changes in abundance, so we did not consider seasonal patterns (Harding and Winterbourn, 1995; Suren and Jowett, 2006; Stark and Phillips, 2009). Death and Zimmermann (2005) found that flow disturbance at closed-canopy sites acted only by physically removing individuals, but flow disturbance at open-canopy sites removed both individuals and periphyton biomass. We analyzed data from open- and closed-canopy sites to partition the effects of animal removal and periphyton food resource removal. We removed rare taxa (those making up <5% of all individuals collected), early instar, and pupa data from the analysis. This step left 21 taxa at open-canopy sites and 23 taxa at closed-canopy sites for use in the analysis. For each of the selected taxa, we used 9 feature groups (e.g., mode of respiration) with 28 traits (e.g., aerial, gills, plastron, and tegument) to describe the trait characteristics of the communities (Table 1). Trait information was generally coded at the generic level, with the exception of the family Hydraenidae (adult), which was coded at the family level. We obtained trait information from the New Zealand Invertebrate Trait database (NIWA, 2012). We undertook all analyses with the mvabund package (version 3.11.4) in R (version 3.2.5; R Project for Statistical Computing, Vienna, Austria). We fitted models for substrate movement and periphyton biomass independently and separately at open- and closed-canopy sites. We used generalized linear models (GLMs) in the traitglm function to fit abundance of each taxon against substrate disturbance or periphyton biomass, traits, and their interactions, or environment-trait associations. The function uses Chapter 2. Disturbance – productivity effects on stream invertebrate traits 18 multiple GLMs (multiGLMs), which are the extended form of multivariate linear models. The GLMs in the package mvabund are adapted for multivariate abundance data because they can be used with the strong mean–variance relationships and nonnormal data often found in ecology (Wang et al., 2012). A row effect was added in the traitglm model by applying the composition function (composition = TRUE) to adjust for different sampling intensities across different samples. This row effect can be understood as a compositional term in the sense that it models relative abundance at a site (Wang et al., 2012; Warton et al., 2015b). The relationship between traits and environmental variables are described by interaction terms in the models as environment–trait associations (Brown et al., 2014; Warton et al., 2015b). We tested the significance of the environment–trait associations with the anova.traitglm function. This function used block resampling, in which sites were resampled but all species from a site were kept together in the resample (Warton et al., 2015b). The p-value was calculated from 999 iterative resamplings. p-values < 0.05 indicated significant trait–environment relationships. All assumptions for functions were checked and verified. Traits linked specifically with substrate disturbance or periphyton biomass were identified by the model selection procedure by applying the Least Absolute Shrinkage and Subset Selection Operator (LASSO) penalty in the traitglm function. This penalty set to 0 any interaction coefficients that do not reduce the Bayesian Information Criterion (BIC) and returns the interaction coefficients of the important trait–environment associations (Brown et al., 2014; Warton et al., 2015b). All predictors were standardized, so the sizes of the coefficients measure the relative importance of the predictors (Warton et al., 2015b). Each interaction coefficient can be interpreted as the amount by which a unit (1 SD) change in the trait variable altered the slope of the relationship between Chapter 2. Disturbance – productivity effects on stream invertebrate traits 19 abundance and the environmental variable of interest (Warton et al., 2015b). The magnitude and sign of the interaction coefficients indicated the strength and the direction of the effect of substrate disturbance or periphyton on traits. The higher the value of the interaction coefficients, the stronger the relationship between the traits, and substrate disturbance or periphyton. RESULTS Traits associated with substrate disturbance Substrate disturbance affected traits regardless of whether the sites were in the open or under canopy (p = 0.008 at closed- and p = 0.003 at open-canopy sites) (Table 2). Three traits were linked with substrate movement at both open- and closed-canopy sites. Taxa having 2 aquatic life stages and who were filter-feeders had a negative association with more substrate movement (Table 3, Fig. 2A). Taxa that respired using a plastron were positively linked with more substrate disturbance (Fig. 2B). Nine traits were linked with substrate movement only at the closed-canopy site and 1 trait was linked only at the open-canopy sites. At closed-canopy sites, 6 traits were positively linked with greater substrate disturbance: small size, being a burrower, firm attachment to substrate, high mobility, inflexible bodies, and being a predator. Three traits were negatively linked: moderate mobility, being a swimmer, and being a deposit-feeder. At open-canopy sites, only being a scraper was positively associated with more substrate disturbance (Table 3). Substrate disturbance had a stronger positive effect on traits at closed- than at open-canopy sites. The interaction coefficients of the associations between traits and substrate disturbance ranged from –0.16 to 0.19 at closed-canopy sites, and from –0.23 to 0.1 at open-canopy sites (Fig. 3). Chapter 2. Disturbance – productivity effects on stream invertebrate traits 20 Table 2. Analysis of deviance results testing the significance of interaction terms between substrate movement or periphyton biomass and traits obtained from the fourth-corner models. Samples were collected from open- and closed-canopy sites on 10 streams in Taranaki in 1999– 2000. Res.Df = deviance of residuals, Df.dff = number of parameters taken up by interaction terms, and Dev = deviance. * = p < 0.05, ** = p < 0.01. Model Residual df Interaction terms Deviance p Closed-canopy sites Traits : substrate movement 876 19 106.4 0.008** Traits : periphyton biomass 876 19 43.74 0.32 Open-canopy sites Traits : substrate movement 802 17 67.98 0.003** Traits : periphyton biomass 702 17 55.07 0.01* Figure 2. The proportion of individuals that were filter feeders having 2 aquatic life stages (A) and that used plastron respiration (B) and plotted against substrate disturbance at open- and closed-canopy sites on each sampling occasion at open- and closed-canopy sites in 10 streams in Taranaki, New Zealand, in 1999–2000. Chapter 2. Disturbance – productivity effects on stream invertebrate traits 21 Table 3. Traits associated with substrate disturbance or periphyton biomass. Traits in bold are those affected by substrate disturbances regardless of canopy cover or by both substrate disturbance and periphyton biomass at open-canopy sites. The interaction terms were selected by the Least Absolute Shrinkage and Subset Selection Operator (LASSO) penalty. The nature and the strength of the disturbance-trait associations are identified by the sign and magnitude of the interaction coefficients between substrate movement/periphyton biomass and the traits. Samples were collected from open- and closed-canopy sites on 10 streams in Taranaki in 1999–2000. Traits Closed-canopy sites Open-canopy sites Substrate disturbance Substrate disturbance Periphyton biomass Plastron respirers 0.15 0.07 Filter-feeders –0.13 –0.1 Two aquatic life stages –0.16 –0.13 0.04 High flexible body –0.23 0.14 Collectors 0.1 –0.17 Flattened body form 0.02 –0.08 Scrapers 0.01 Dietary specialist 0.05 Predators 0.01 0.05 High mobility 0.005 Moderate mobility –0.04 Deposit-feeders –0.1 Inflexible body 0.19 Firm substrate attachment 0.08 Burrowers 0.05 Swimmers –0.14 Small size 0.02 Chapter 2. Disturbance – productivity effects on stream invertebrate traits 22 Figure 3. Box-and-whisker plot of the coefficients for interaction between traits and substrate movement at closed- (close.SM) and open-canopy (open.SM) sites, periphyton biomass at open- canopy sites (open.Chla), and substrate movement at open-canopy sites (open.SM) in 10 streams in Taranaki, New Zealand, in 1999–2000. The interaction terms were selected by the Least Absolute Shrinkage and Subset Selection Operator (LASSO) penalty for the fourth-corner models, from fitting abundance data to the environment and traits. Lines in boxes are medians, box ends are quartiles, and whiskers are quartiles ± 1.5 * IQR (where IQR is the inter-quartile range). Traits associated with only periphyton biomass Periphyton biomass was linked with traits only at open-canopy sites (Table 2). Two traits were positively linked with high periphyton biomass: dietary specialization and being a predator (Fig. 4). No trait was negatively associated with high periphyton biomass (Table 3). Chapter 2. Disturbance – productivity effects on stream invertebrate traits 23 Figure 4. The proportion of predators plotted against periphyton biomass at open-canopy sites in 10 streams in Taranaki, New Zealand, in 1999–2000. Traits associated with both substrate disturbance and periphyton biomass The effects of both substrate movement and periphyton biomass on traits occurred only at open-canopy sites (Table 2). Two traits were positively linked with more substrate movement but negatively associated with high periphyton biomass: being a collector and having a flattened body (Table 3, Fig. 5A, B). Two traits were negatively linked with more substrate movement but positively associated with high periphyton biomass: having a flexible body and 2 aquatic life stages (Fig. 5C, D). Traits were more strongly associated with periphyton biomass than substrate movement. The median of the interaction coefficients of the associations between periphyton biomass and traits were higher than those between substrate disturbance and traits (Fig. 3). Chapter 2. Disturbance – productivity effects on stream invertebrate traits 24 Figure 5. The proportion of collectors with flattened body form (A, B) and taxa with highly flexible body and 2 aquatic life stages (C, D) on each sampling occasion plotted against substrate disturbance (A, C) and periphyton biomass (B, D) at open-canopy sites in 10 streams in Taranaki, New Zealand, in 1999–2000. DISCUSSION The effect of flow disturbance on invertebrate traits Flow disturbance can affect invertebrate traits through physical removal of individuals or loss of food. At closed-canopy sites, biomass of periphyton is strongly limited by the overhanging canopy and is unlikely to change with disturbance. Thus, disturbance should affect invertebrate trait composition only through the direct removal of animals. However, at open-canopy sites, both substrate disturbance and periphyton Chapter 2. Disturbance – productivity effects on stream invertebrate traits 25 biomass can affect invertebrate trait composition. In our study streams, periphyton biomass was 3× higher at open- than at closed-canopy streams (Death and Zimmermann, 2005), and median interaction coefficients between traits and periphyton biomass were higher than those between traits and substrate movement. Food loss, therefore, appears to have a stronger influence on inverterbate traits than substrate disturbance, although both of them were important drivers of effects of high-flow disturbances on trait composition in the open-canopy streams. Traits linked with substrate disturbance Substrate disturbance affected invertebrate trait composition regardless of the presence of the overhead canopy. We considered traits indicative of substrate disturbance only if they were affected by substrate disturbance regardless of presence of overhanging cover because effects on traits that occurred only in the open-canopy streams could be a result of either physical removal or loss of periphyton food. These traits were using plastron respiration, having 2 aquatic life stages, and being filter-feeders The observed response of individuals that respired using a plastron was positive at disturbed sites in both open- and closed-canopy streams. A plastron is an array of rigid, closely-spaced hydrophobic hairs (setae), which efficiently extract dissolved O2 from water. A plastron is more resistant to physical damage during floods than unprotected gills (Flynn and Bush, 2008). Tomanova and Usseglio-Polatera (2007) also found that taxa with plastrons were more abundant in streams with high-flow disturbances. The observed response of taxa with 2 aquatic life stages and those that are filter- feeders was negative at disturbed sites. Most aquatic insects have aquatic and terrestrial life stages. These taxa are likely to be more abundant at sites with less disturbance simply because the more time spent in the stream in flood-prone streams, the greater the chance Chapter 2. Disturbance – productivity effects on stream invertebrate traits 26 of being exposed to destructive flows (Southwood, 1977; Townsend and Hildrew, 1994). The negative association between filter-feeders and substrate disturbance can be explained by their need for exposed attachment sites, which position them in places most affected by disturbance (Brooks and Haeusler, 2016). Individuals having both 2 aquatic life stages and being filter-feeders, therefore, were more abundant at less-disturbed sites. Some traits were associated with substrate disturbance at sites with either open or closed canopy. At the closed-canopy sites, the positive response of individuals having small size, high mobility, or firm attachment and the negative response of those having moderate mobility, being a swimmer, and being a deposit-feeder were consistent with our prior predictions (Table 4). We were surprised that individuals with inflexible bodies were more abundant at disturbed sites, but these individuals (in our case: the cased caddisflies Beraeoptera roria) also were more common in disturbed than in undisturbed streams in taxonomy-based studies (Death, 2003). High abundance of individuals with inflexible bodies in streams with high levels of substrate disturbance is difficult to explain. We propose that the positive response of B. roria to substrate disturbance might be related to life-history traits that we did not consider in our study. Further studies, therefore, should be done to investigate the effects of substrate disturbance on invertebrate life-history traits. In addition, more burrowers occurred at disturbed sites than expected based on the literature (Lamouroux et al., 2004; Tomanova and Usseglio-Polatera, 2007). Burrowers typically are found in pool habitats in perennial streams (Walters, 2011). We suggest that pool habitat might act as an important refuge for burrowers during high flows, leading to the positive relationship between burrowers and disturbance in our study. The positive response of scrapers to disturbance at only open-canopy sites probably reflected the response of the most dominant scraper (the mayfly Deleatidium) (Fig. 6). The similarity between the response of a single trait and groups that consist Chapter 2. Disturbance – productivity effects on stream invertebrate traits 27 mostly of organisms that express the trait was expected (Pilière et al., 2015). The greater abundance of Deleatidium in disturbed than undisturbed streams might be associated their behavioural response of drifting during high-flow events, which would enhance their ability to recolonize quickly after substrate disturbance (Townsend et al., 1997a; Death, 2003). Therefore, the positive response of scrapers to substrate disturbance at open- canopy sites may be statistically but not ecologically significant. Our finding highlights the possibility of misinterpreting of environmental response of a single trait, especially if the trait occurs in a highly dominant taxon (Pilière et al., 2015). Figure 6. The proportion of Deleatidium spp. plotted against substrate disturbance on each sampling occasion at open-canopy sites in 10 streams in Taranaki, New Zealand, in 1999–2000. Traits linked with only periphyton biomass More periphyton biomass was positively linked with being a predator and being a dietary specialist. Predators are more abundant at sites with higher periphyton biomass because this biomass provides more food for their prey. The positive association between dietary specialists and periphyton biomass is mainly because the predators in our study were dietary specialists. Chapter 2. Disturbance – productivity effects on stream invertebrate traits 28 Table 4. Comparisons between predicted responses of invertebrate traits and flow disturbance (F) and periphyton biomass (P) and observed responses of invertebrate traits and substrate disturbance (SD) and periphyton biomass. Samples were collected from open- and closed-canopy sites on 10 streams in Taranaki in 1999–2000. Trait group/trait Predictions Observed Canopy Aquatic stages 2 aquatic life stages ↓F ↓SD Both 1 aquatic life stage ↑F ns Body flexibility None (<10°) ↓F ↑SD Closed Low (>10–45°) ↓F ns High (>45°) ↑F ↓SD↑P Open Body form Streamlined ↑F ns Flattened ↑F ↑SD↓P Open Cylindrical ↓F ns Dissemination potential Low ( <10 m) ↓F ns Moderate (10 m–1 km) ↓F ↓SD Closed High (>1 km) ↑F ↑SD Closed Locomotion and relation to substrate Swimmers ↓F ↓SD Closed Crawlers ↓F ns Burrowers ↓F ↑SD Closed Attached ↑F ↑SD Closed Maximum potential size Small size (≤10 mm) ↑F ↑SD Closed Large size (>10 mm) ↓F ns Respiration of aquatic stages (not including eggs) Tegument ↓F ns Plastron ↑F ↑SD Both Gills ↑F ns Dietary preferences Strong (specialist) ↓P ↑P Open Chapter 2. Disturbance – productivity effects on stream invertebrate traits 29 Trait group/trait Predictions Observed Canopy Moderate ↓P ns Weak (generalist) ↑P ns Feeding habit Scrapers ↑P ↑SD Open Deposit-feeders ↓F ↓SD Closed Filter-feeders ↓F ↓SD Both Collectors ↑F ↑SD↓P Open Predators ↑P ↑P Open Traits linked with both substrate disturbance and periphyton biomass New Zealand streams have asynchronous and generally low levels of allochthonous inputs (Winterbourn, 1997). At open-canopy sites, high shear stress from flood events abrade periphyton biofilms that develop on stone surfaces, but algae persist in rock crevices and at low biomass on stone surfaces. Collectors having flattened bodies responded positively at more-disturbed sites even though periphyton food was limited. Horrigan and Baird (2008) found a positive relationship between the proportion of collector taxa and flow velocity. A flattened body reduces drag forces from high flows and enables these invertebrates to maintain themselves in the higher flows during floods. Townsend et al. (1997a) found a positive association between the proportion of individuals with flattened bodies and substrate disturbance. Contrary to our expectations, we found a negative response of individuals with flexible bodies to disturbance (Table 4). Flexible bodies reduce shear stress from high flows, so taxa with this trait can survive in streams with frequent and intense floods. However, these taxa are unlikely to be excluded from less-disturbed streams (Hildrew and Giller, 1987; Townsend and Hildrew, 1994; Townsend et al., 1997a). Furthermore, a positive response of taxa spending longer in the stream at less- than at more-disturbed sites with open canopies was consistent with the pattern observed in the closed-canopy Chapter 2. Disturbance – productivity effects on stream invertebrate traits 30 sites. As a result, individuals having both flexible body and 2 aquatic life stages responded positively at less-disturbed sites. Our trait-based analysis explained flow-disturbance effects on community patterns observed by Death and Zimmermann (2005) in their taxonomically focused study. Filter-feeders having 2 aquatic life stages (the caddisflies Orthopsyche and Aoteapsyche) responded more positively at closed- than open-canopy sites, whereas positive responders at open-canopy sites had both highly flexible bodies and 2 aquatic life stages (chironomids and oligochaetes). These taxa were characteristic of closed- and open-canopy sites in the study by Death and Zimmermann (2005). Linking species traits with the observed species differences, therefore, helps clarify how high flows affect invertebrate communities. To conclude, the principal driver of effects of flow disturbance on trait composition in the light-limiting New Zealand streams was removal of individuals. The loss of periphyton food seemed to be a more important driver of community changes associated with flow disturbances, but both disturbance and periphyton biomass significantly affected trait composition at streams without forest cover. At open-canopy sites, differences in the proportion of plastron respirers and filter-feeders probably reflected the effects of substrate disturbance because these traits were affected by disturbance regardless of canopy cover. On the other hand, differences in the proportion of collectors having flattened body form and those having highly flexible bodies probably indicated the combined effects of disturbance and loss of periphyton biomass. Thus, we were able to separate the influence of substrate disturbance from the combined effects at the open-canopy sites. However, we were unable to partition independent effects of each factor within the combined effects. Chapter 3. Flow regime affects life history traits 31 Chapter 3. The effects of flow regime on life history strategies of aquatic invertebrate communities. Authors: Yen T. H. Dinh1, Wendy A. Monk2, Paul Wood3 and Russell G. Death1 Affiliations: 1 Innovative River Solutions, College of Sciences, Massey University, Private Bag 11- 222, Palmerston North, New Zealand. 2 Department of Biology, Canadian Rivers Institute, University of New Brunswick, Fredericton, New Brunswick, E3B 6E1, Canada. 3 Department of Geography, Loughborough University, Loughborough, Leicestershire, LE11 3TU, UK. Chapter 3. Flow regime affects life history traits 32 Abstract Anthropogenically-modified flow regimes can dramatically change riverine community structure and function if life history traits of the local fauna are not compatible with the new flow regime. However, difference in proportions of individual life history traits between flow regimes is relatively small, less than 15% in the literature. This is probably associated with a single trait-based approach that is commonly used in current studies because that approach cannot account for phylogenetical auto-correlations between life history traits. We therefore investigated the composition of life history strategies, a combination of traits that is defined by evolutionary linkages, in 83 English and Wales rivers with three different flow regimes. Predictable multiple high flow events per year, characteristics of River type A, supported insects with strongly emergent synchronization in strategy S1 (median = 34.29%). The least variable flow due to strong groundwater influence in River type C had the highest abundances of non-flying invertebrates with strongly reproductive investment in strategy R3 (median = 36.34%), which characteristics were long-lived adults, large eggs and parental care. The modest flow variability characterized by a steep peak flow in River type B had an intermediate abundance of both strategy S1 and R3 (median = 25.05% and 21.6%, respectively). Different proportions in strategy S1 and R3 between River type A and C were 16.05 and 30.75%. The life history strategy approach, therefore, better reflected the responses of life history traits to flow regimes. Keywords: body plans, discharge patterns, future flows, geographic constraints Chapter 3. Flow regime affects life history traits 33 INTRODUCTION Ecological responses of biological river communities to flow alterations from anthropogenic impacts, including climate change and water resource management, are far from clear (Poff and Zimmerman, 2010; Death et al., 2015; O'Connor et al., 2015). Stream invertebrates in temperate areas may become extinct or immigrate northward while those in Mediterranean areas will move to the temperate areas as a result of an increased magnitude and frequency of floods and droughts across European in the context of climate change (Bonada et al., 2007a). The supradrought in Australia’s Murray-Darling Basin between 1990 and 2009 reduced the prevalence of 11 invertebrate families; and they are yet to recover (Chessman, 2015). Aquatic invertebrates that survive at a site are likely to have life history traits that reflect the timing, frequency, predictability and severity of flow regimes (Lytle, 2001; Lytle and Poff, 2004). Insects often enter the adult terrestrial phase and/or the diapause phase before flood or drought seasons, respectively. Many non-flying invertebrates (such as flatworms and leeches) enhance their fitness with synchronized reproduction when flows are optimal (Lytle and Poff, 2004). Any changes to flow pattern can, therefore, cause significant impacts on stream biota (Dudgeon et al., 2006; Poff et al., 2007) and may lead to species loss (Vandewalle et al., 2010) if life history traits of the local fauna are not compatible with the new flow regime (Lytle 2001, Lytle and Poff 2004). Different flow regimes may select for distinctive trait compositions (Bêche et al., 2006; Bonada et al., 2007a; Statzner and Beche, 2010). Bonada et al. (2007b) found three trait groups characterised for three flow categories in the Mediterranean climatic area. They found that intermittent streams were dominated by taxa having reproduction by clutches, winged forms, aerial respiration, and swimmers. Ephemeral streams were dominated by taxa having reproduction by eggs or asexually, aquatic forms, tegument Chapter 3. Flow regime affects life history traits 34 respiration, and substrate attachment. Perennial streams had a mixture of both trait groups (Bonada et al., 2007b). Highly predictable seasonal intermittent flows in Mediterranean streams favour taxa of a small size, aquatic adults, winged life stages and a swimming locomotion; whereas, less variable flows and intermittent flows in temperate streams support taxa having an intermediate size, aquatic larva, and a crawling locomotion (Bonada et al., 2007a). However, difference in proportions of life history traits between these flow regimes is typically small, less than 15% in the literature (Statzner et al., 2004; Bêche et al., 2006; Bonada et al., 2007a). The single trait-based approach that is commonly used in the literature typically analyses individual trait performance assuming identical adaptive value of particular traits across species (Bêche et al., 2006; Bonada et al., 2007a; Bonada et al., 2007b). Analysing changes in life history traits associated with changing flow regimes needs to consider evolutionary inter-relationships because these traits are phylogenetically auto-correlated (Poff et al., 2006). The individual trait-based approach may, therefore, fail to identify an adaptive value of an animal having a particular life history trait combination (Poff et al., 2006). This might explain why differences of individual traits associated with changing flow regimes reported in the literature have been small (Statzner et al., 2004; Bêche et al., 2006; Bonada et al., 2007a). A life history strategy approach proposed by Verberk et al. (2008b) grouped invertebrates into 13 strategies based on evolutionary linkages between traits. Therefore, these strategies may better describe associations between life history traits and flow regimes (Poff et al., 2006; Verberk et al., 2013). We thus studied the trait composition of invertebrate communities in 83 English and Welsh rivers using the life history approach. Monk et al. (2006) characterised the long-term flow regimes of these rivers into three distinctive types and found that Lotic-invertebrate Index for Flow Evaluation (LIFE; Chapter 3. Flow regime affects life history traits 35 Extence et al. (1999)) scores significantly differed among river types (Monk et al., 2007). River type A had multiple predictable high flows, river type B had a steep flow peak, and river type C had a prolonged rising climb with major groundwater aquifers (Monk et al., 2006). METHOD The distribution of river types Monk et al. (2006) classified the flows of 83 streams in England and Welsh into three river types based on the form and timing of the annual hydrograph discharge pattern. Daily hydrological data recorded by the Environment Agency of England and Wales between 1980 and 1999 formed the base for the regime classification. Sites within River type A exhibited three high flow events per year (in October, December and March) and were located on impermeable geology in the wetter northwest of England and one site in south Wales (higher elevation streams). River type B sites exhibited one steep, high flow event in January and was located throughout north-eastern, central and southern England over a range of geologies. Sites within River type C exhibited a prolonged rising climb in March and were associated with major groundwater aquifers and mainly located in eastern and southern England (for more details refer to Monk et al., 2006). Sample collection Benthic invertebrate samples were paired with adjacent hydrological gauging stations. Biomonitoring samples were collected using a standard Freshwater Biological Association net using a three-minute kick sample (<1 mm mesh net) with an additional one-minute hand search, requiring collectors to sample instream habitats in proportion to Chapter 3. Flow regime affects life history traits 36 their occurrence. Samples were collected annually between September and November from 1989 to 2000 before being identified to family-level where abundance data were recorded in log10 abundance categories (for more details refer to Monk et al., 2006). Data analysis Each taxon was assigned to a life history strategy based on the defined species traits in each of those strategies (Table 1 and Appendix S1). Trait information and strategies were obtained from Tachet et al. (2010), Verberk et al. (2008a), Poff et al. (2006), and Vieira et al. (2006). Invertebrates were identified to family level, except for non-insects that were recorded to order. When a family exhibited multiple traits, we defined the trait of the family as the trait of the most common genera. We then calculated the abundance of each strategy at each site by summing abundance of all families within that strategy. Samples collected at each site in different years were treated as replicates. We undertook all analyses with the mvabund package (version 3.11.4) in R (version 3.2.5, R Project for Statistical Computing, Vienne, Austria). We used generalized linear models (GLMs) in the mvabund function to fit abundance of each strategy against the three river types (Wang et al., 2012). The function uses multiple GLMs (multiGLMs), which are the extended form of multivariate linear models. The GLMs in the package mvabund are adapted for multivariate abundance data because they can be used with strong mean–variance relationships and non-normal data that is often found in ecology (Wang et al., 2012). We added a row effect in the model to account for the resampling of each site. We tested the significance of the river type–strategy associations with the anova.manyglm function (Wang et al., 2012). We then obtained the p-value of each strategy by adding p.uni=adjusted in the anova.manyglm function (Wang Chapter 3. Flow regime affects life history traits 37 et al., 2012) and then selected strategies whose p-value ≤ 0.05. The p-value was calculated from 999 iterative re-samplings. p-values < 0.05 indicated significant river type–strategy relationships. All assumptions for functions were checked and verified. Table 1. Characteristics of the eight life history strategies used in our study. Trait information and strategies were obtained from Tachet et al. (2010), Verberk et al. (2008a), Poff et al. (2006), and Vieira et al. (2006). Life history Characteristics D1 Active flight, strong dispersal, short development time, long-lived adults, mainly multi-voltine and large clutch size. D2 Active flight, strong dispersal, short-lived adults, univoltine, and large clutch size. D3 Active flight, weak-moderate dispersal, short-lived adults, mainly multi-univoltine, and early age at first reproduction. R1 Active flight, weak-moderate dispersal, long-lived adults, sequential reproduction, rapid juvenile development R2 No flight, short-lived adults, sequential reproduction with many small eggs, slow juvenile development, and high clutch size. R3 No flight, long-lived adults, sequential reproduction with large egg size, parent care R4 No flight, short generation time, asexual propagation and resistant stages S1 Active flight, weak-moderate dispersal, short-lived adults, strong synchronised adult emergence and mainly univoltine. S2 Active flight, weak-moderate dispersal, long-lived adults, mainly multivoltine, effectively synchronised adult emergence and resistant diapausing stage. S4 Active flight, weak-moderate dispersal, long-lived adults, short synchronized juvenile development T1 Active flight, weak-moderate dispersal, small size, and highly tolerant to harsh environment. T2 No flight, long-lived adult, asexual reproduction, highly tolerant to long periods of harsh environment. Chapter 3. Flow regime affects life history traits 38 RESULTS Strategy composition differed between the three river types, especially abundance of six life history strategies D1, D3, R3, S1, S2 and T2 (Table 2 and Fig. 1). River type A had a low abundance (median = 5.59%) of strategy R3 and a high abundance (median = 34.29%) of strategy S1. River type C had a high abundance (median = 36.34%) of strategy R3 and a low abundance (median = 18.24%) of strategy S1. An intermediate abundance of strategy S1 and R3 (median value 25.05% and 21.6%, respectively) characterized river type B. River type A had higher abundance of strategy D3 and T2, whereas, river type B and C had higher abundance of strategy D1 and S2 (Fig. 2). Table 2. Results of analysis of deviance from the anova.manyglm function testing for a significant difference of life history strategies between river types. Data were collected in 83 English and Welsh rivers between 1989 and 2000. The global p value was given for the life history strategy composition. Each life history strategy was given a p value indicating whether its abundance differed significantly between the three river types. *** = p ≤ 0.001; ** = p<0.01. df Deviance p The strategy compositions 2 215.5 0.001*** Each life history strategy D1 2 48.73 0.001*** D2 2 4.7 0.39 ns D3 2 30.87 0.001*** R1 2 7.62 0.23 ns R2 2 10.35 0.17 ns R3 2 53.57 0.001*** R4 2 1.56 0.5 ns S1 2 34.94 0.001*** S2 2 36 0.001*** S4 2 7.04 0.23 ns T1 2 8.06 0.22 ns T2 2 20.59 0.01* Chapter 3. Flow regime affects life history traits 39 Figure 1. Histogram of percentage of individuals in each life history strategy in three river types collected in 83 English and Wales rivers between 1989 and 2000. * = life history strategy that relative abundance significantly differed between three river types (p < 0.05). Their p values were obtained by adding p.uni=adjusted in the anova.manyglm function (See detail in Table 2). Figure 2. Box-and-whisker plots of the percent of individuals in each life history strategies that differed between river types in 83 UK rivers between 1989-2000. Lines in boxes are medians, box ends are quartiles, whiskers are quartiles ± 1.5*IQR (where IQR is the inter-quartile range), and dots are outliers. Chapter 3. Flow regime affects life history traits 40 DISCUSSION The three river types supported different life history strategies. Similar to our findings, Zuellig and Schmidt (2012) found life history traits varied across nine eco- regions identified by climatic, physiographic and hydrological data in the United States. Poff et al. (2010) also found three trait community types for invertebrates in 279 western United States rivers in response to hydrological and climatic variables. Multiple predictable high flow events, characteristics of River type A, supported highly synchronized strategy S1 because the strong synchronized adult emergence allows these invertebrates to escape from streams before high flow events, as long as those high flows are predictable (Lytle, 2001). It may explain the dominance of strategy S1 found in River type A sites. Traits characterized in strategy S1 were similar to trait characteristics in rivers with predictable floods of Western Mountain, US (Zuellig and Schmidt, 2012). Furthermore, invertebrates in highly tolerant strategy T2 can be tolerant to long periods in a harsh environment, such as multiple high flows in River type A, due to asexual reproduction and the lack of synchronized development in juveniles (Verberk et al., 2008b). This could contribute to a higher abundance of individuals in strategy T2 in River type A. Additionally, weak flyers with short-lived adults in strategy D3 would better suit River type A sites because these sites were restricted in high elevation areas (Monk et al., 2006; Sarremejane et al., 2017). As a result, characteristics of River type A resulted in the dominance of insects with strongly emergent synchronization in strategy S1 and higher abundances of insects with high tolerance in strategy T2 and D3. In contrast, the least variable flows associated with strong groundwater influence in River type C were better for greater investment in reproduction of strategy R3 because these non-flying invertebrates can spread reproductive effort over a longer period and Chapter 3. Flow regime affects life history traits 41 invest in parental brood care to increase the survival rate of their offspring (Verberk et al., 2008b). Spreading reproductive effort can give their offspring a competitive advantage over other species if they occupy more benign environments in River type C (Townsend and Hildrew, 1994; Verberk et al., 2008b), leading to the dominance of strategy R3 in River type C sites. Therefore, strong groundwater influence, characteristic of River type C, supported the dominance of non-flying invertebrates with highly reproductive investment in strategy R3. The moderately variable flows in river type B supported life history investments in both synchronization and reproduction. The strongly synchronized adult emergence would benefit insects in strategy S1 from a predictable and steep high flow in River type B. An intermediate period of stable flow in River type B might be associated with an intermediate abundance of strategy R3 because these animals need stable flow to reproduce efficiently. Additionally, summer droughts are more common in River type B than River type A because of the wetter weather in the northwest of England, and rare in River type C because of the ground water influence (Monk et al., 2006; Monk et al., 2007). Summer droughts supported higher abundances of invertebrates in strategy S2 and D1 in River type B sites. Summer droughts would benefit strategy S2 because these insects have resistant diapause stages, either as eggs or aquatic adults, allow species to bridge long periods of unfavorable conditions in summer droughts (Verberk et al., 2008b). Similarly, summer droughts would benefit strategy D1 because these invertebrates spend most of their lifetime outside of rivers, reducing opportunities to be exposed to droughts (Townsend and Hildrew, 1994; Townsend et al., 1997a; Verberk et al., 2008b). Consequently, the moderate flow variability in River type B supported the dominance of invertebrates in both synchronized strategy S1 and reproductive strategy Chapter 3. Flow regime affects life history traits 42 R3 while more common summer droughts supported higher abundances of invertebrates in strategy S2 and D1. We found relatively large differences in the proportions of strategies S1 and R3 between the three different hydrological river types. Median values for the proportions of strategies S1 and R3 between river types A and C were 16.05 and 30.75%. Previous studies found less than 15% proportional differences in life history traits in 384 running water sites across 14 European biogeographical regions and 527 reference-condition sites of Mediterranean and temperate regions in South Europe, Middle East and North Africa (Statzner et al., 2004; Bonada et al., 2007a; Bonada et al., 2007b). Flow variability differences between river types in our study were much less pronounced than those in previous studies (Monk et al., 2007). However, proportional differences in strategy S1 and R3 between river types A and C were much larger than differences in traits found between rivers of different flow typology in previous studies. We, therefore, conclude that the life history strategy approach improved the power to link traits with flows (Verberk et al., 2008b; Verberk et al., 2013). The life history approach appeared to better describe differences in trait composition of invertebrate communities between flow regimes. This supported the view that environmental selective forces act on whole organisms via trait combinations that represent an adaptive value to environmental gradients rather than individual traits acting separately (Poff et al., 2006; Verberk et al., 2013). Combination of traits, therefore, likely have greater fitness than the individual trait separately (Verberk et al., 2013). The life history approach that integrate evolutionary inter-relationships between traits seemed to better describe differences in trait composition of invertebrate communities along environmental gradients than the single trait-based approach. Chapter 4. Nutrients affect invertebrate traits 43 Chapter 4. The effects of nutrient enrichment on invertebrate traits in New Zealand streams Authors: Yen T.H. Dinh and Russell G. Death Affiliations: Innovative River Solutions, College of Science, Massey University, Private Bag 11-222, Palmerston North 4442, New Zealand Manuscript status: Under review at Hydrobiologica Chapter 4. Nutrients affect invertebrate traits 44 Abstract Eutrophication from land-use intensification can decrease biodiversity and ecological health in rivers. Nutrients affect invertebrates by altering the quality and quantity of food and modifying habitats because epilithon overgrows the substrate. The response of invertebrates, judged by changes in functional feeding traits to nutrient enrichment, appears equivocal in the current literature. However, no study has directly examined the effects of nutrient-induced changes in food sources on stream invertebrate traits. We used fourth corner models to assess the effects of periphyton biomass, total nitrogen, and dissolved reactive phosphate on life history and functional feeding traits in invertebrates collected in 17 New Zealand streams. Periphyton biomass and nutrients had similar effects on functional feeding and life history traits. The responses of shredders more likely reflected nutrient effects on food source, whereas, the responses of algae piercers reflect both food source and habitat changes. Collectors and scrapers having different life history traits responded differently to periphyton biomass and nutrients. As a result, both functional feeding and life history traits are important in monitoring nutrient effects on stream invertebrate communities. Evaluating invertebrate responses to nutrients may be sufficient to reflect the effects of nutrient-induced changes in periphyton biofilm on stream communities. Keywords: Nutrient enrichment, periphyton overgrowth, trait combinations, stream invertebrates. Chapter 4. Nutrients affect invertebrate traits 45 INTRODUCTION Many of the world’s waterways have been degraded by increased nutrient inputs from agriculture and urbanisation (Quinn, 2000; Allan, 2004; Strayer and Dudgeon, 2010). Nitrogen and phosphorus are the two most deleterious and ubiquitous of those nutrients (Carpenter et al., 1998; Elser et al., 2007). High nutrient levels are associated with a loss of biodiversity (Morgan and Cushman, 2005; Strayer and Dudgeon, 2010), and a shift in community structure of consumers (Gafner and Robinson, 2007; Matthaei et al., 2010). There are also reductions in the recreational and amenity value of waterways as well as increased costs for drinking water treatment (Foote et al., 2015). Eutrophication, therefore, not only reduces the ecological health of rivers and streams, but can also generate serious financial and social cost for many nations (Pretty et al., 2003; Dodds et al., 2009; Jarvie et al., 2013). Nutrient enrichment can change the type of invertebrates that occupy a stream, particularly the composition of functional feeding and life history traits, by promoting epilithon overgrowth on the substrate. An increased biomass and productivity of epilithon biofilm can shift the invertebrate community composition to those that fed on algae, bacteria and fungi (Gafner and Robinson, 2007; Matthaei et al., 2010; Liess et al., 2012). Lange et al. (2014) found that abundances of scrapers, deposit feeders and predators were mostly strongly associated with the concentration of total nitrogen in waterways when evaluating changes in 52 biological traits. Additionally, epilithon overgrowth can homogenize habitats, decrease oxygen concentration, and change pH (Yuan, 2010). This can lead to changes in life history traits with increased nutrients, such as the maximum number of reproductive cycles per year, life duration of adults, reproductive techniques and aquatic stages (Lange et al., 2014). Chapter 4. Nutrients affect invertebrate traits 46 However, the links between nutrient concentration and functional feeding trait composition is inconsistent in the literature (Dolédec et al., 2006; Townsend et al., 2008; Matthaei et al., 2010; Statzner and Beche, 2010; Wagenhoff et al., 2012; Lange et al., 2014). Shredder abundance may decline relative to other functional feeding groups in streams of the Taieri River, South Island, New Zealand, but in mesocosm experiments with the same taxa, there was no effect of increased nutrients (Townsend et al., 2008). In contrast to the study of Lange et al. (2014), Wagenhoff et al. (2012) found no relationship between scarpers, deposit feeders and predators and nutrients in mesocosm experiments at the same location in South Island, New Zealand. It, therefore, remains unclear exactly how nutrients can affect functional feeding trait composition in stream invertebrate communities. Environmental selective forces act on whole organisms with specific life history trait combinations rather than individual life history traits separately because these traits can be phylogenetically autocorrelated (Poff et al., 2006; Verberk et al., 2013). However, the current studies that investigated nutrient effects on trait composition only examined the response of individual life history traits at a time (Townsend et al., 2008; Matthaei et al., 2010; Lange et al., 2014). Without considering responses of multiple traits at a time, findings from these studies might be constrained by phylogenetic effects and could not be adequate to elucidate trait responses to nutrient changes (Poff et al., 2006; Verberk et al., 2013). Nutrients affect invertebrates by promoting the growth of epilithon, except at very high concentrations when nutrients can be toxic (Yuan, 2010). However, we are not aware of any study that has directly examined the effects of changes in food resources on functional feeding and life history traits. The main food resource of invertebrates in open New Zealand streams is periphyton (Winterbourn et al., 1981; Winterbourn, 1997). We, Chapter 4. Nutrients affect invertebrate traits 47 therefore, examined how functional feeding and life history traits responded to changes in periphyton biomass and nutrients. METHOD Study sites and nutrient measurements We sampled 24 sites in 17 streams in the Manawatu region, New Zealand. Total nitrogen (TN) and dissolved reactive phosphorus (DRP) were collected yearly using standard methods (APHA method) between 1999 and 2007 (Death and Death, 2008). Means of nutrient concentrations and numbers of samples per year at each site are presented in Table 1. Table 1. Mean for TN and DRP collected at 24 sites on 17 streams in Manawatu, New Zealand between 1999 and 2007. Site DRP (mg/L) TN (mg/L) No. samples per year Tamaki River @ Reserve 0.007 0.095 27 Tamaki River @ State Highway 2 0.008 0.500 28 Mangatainoka River @ Putara 0.003 0.020 21 Pohangina River @ Piripiri 0.005 0.049 27 Tokomaru River @ Horseshoe Bend 0.005 0.080 24 Oroua River @ Fielding Street 0.010 0.240 193 Pohangina River @ Raumai 0.012 0.080 24 Manawatu River @ State Highway 2 0.009 0.396 202 Oruakeretaki Stream @ State Highway 2 0.023 0.000 2 Raparapawai Stream @ Jacksons Road 0.030 0.150 2 Mangapapa Stream @ State Highway 2 0.024 0.900 24 Manga-atua Stream 0.031 0.270 30 Chapter 4. Nutrients affect invertebrate traits 48 Site DRP (mg/L) TN (mg/L) No. samples per year Manawatu River @ Hopelands 0.024 0.900 139 Porewa Stream @ Onepuhi Road 0.030 0.275 40 Waikawa Stream 0.027 1.350 24 Mangarangiora Stream 0.039 0.950 19 Mangatera Stream @ State Highway 2 0.066 0.435 34 Makino Stream @ South Street 0.043 1.700 12 Makaretu Creek @ Ballance Valley Road 0.069 1.137 80 Mangatera Stream @ Timber Bay 0.141 0.810 90 Mangaone Stream @ Milson Line 0.130 0.640 24 Oroua River @ Awahuri 0.101 0.530 147 Tutaenui Stream @ Curls Bridge 0.480 1.900 109 Mangapapa Stream @ Troup Rd. 0.024 0.900 24 Periphyton biomass The periphyton biomass was assessed by measuring Chlorophyll a on five stones (maximum planar dimension <60 cm) collected yearly at random in riffles at each site between 1999 and 2007, and kept frozen. In the laboratory, pigments were extracted separately for each stone in known volumes of 90% acetone at 5°C in the dark for 24 hours. Absorbencies of the solution were read using a Varian Cary 50 UV Visible Spectrophotometer. Values were then converted to pigment concentration following Steinman et al. (2006). The pigment concentration was adjusted by dividing half of the stone surface area (calculated following Graham et al. (1988)) because periphyton is generally only found on the upper exposed surface of stones. The means of chlorophyll at each site were calculated and used for analyses. Chapter 4. Nutrients affect invertebrate traits 49 Invertebrate samples Invertebrate samples were collected in 2007. Five 0.1m2Surber samples (250µm mesh) were collected from riffles at each site and stored in 10% formalin. In the laboratory, invertebrates were removed from samples, identified to the lowest possible taxonomic level using available keys (Cowley, 1978; Winterbourn et al., 1989; Towns and Peters, 1996) and counted. Those taxa which could not be identified to species level were separated into apparent morphospecies. Abundance per taxon was the sum of individuals of the five Surber samples. Any taxon recorded less than twice was considered a rare species and excluded from the data set. Data analysis For each taxon, we used ten traits in four life history feature groups (maximum number of reproductive cycles per year, life duration of adults, reproductive techniques and aquatic stages) and seven functional feeding traits (scrapers, shredders, algal piercers, filter-feeders, deposit-feeders, collectors, and predators) to describe the trait characteristics of the communities. Trait information was obtained from the New Zealand Invertebrate Trait database (NIWA, 2012). Trait information was generally at the generic level. Fourth corner models use a direct 1-stage, model-based framework that can fit abundance data against multiple traits and multiple environmental variables (Pollock et al., 2012; Brown et al., 2014; Warton et al., 2015b). Analysing the original information of multiple traits of each taxon, rather than an affinity score based on fuzzy coding, is the first key difference of the fourth-corner models compared to the current trait-based approach (Brown et al., 2014). The second key difference in fourth corner models was that abundance data, rather than a weighted trait matrix, is the response in the models Chapter 4. Nutrients affect invertebrate traits 50 (Brown et al., 2014; Warton et al., 2015b). Fourth corner models, therefore, can address two significant limitations of common trait-based analyses, that is, the analysis of a single trait and a single environment variable at a time and the impossibility of using abundance data (Dolédec et al., 1996; Legendre et al., 1997; Menezes et al., 2010). Additionally, multiple traits and environmental variables are linked directly in fourth corner models, which provide better quantification of relationships between traits and potential environmental variables. Being a model-based framework, fourth corner models also allow users to check assumptions, quantify the nature and the strength of the environment–trait associations, identify the important associations, and forecast (relative) taxon abundance in new environmental scenarios (Brown et al., 2014; Warton et al., 2015b), which similar analyses do not. We undertook fourth corner models using traitglm function in the mvabund package (version 3.11.4) in R (version 3.2.5; R Project for Statistical Computing, Vienna, Austria). We used generalized linear models (GLMs) in the traitglm function to fit abundance of each taxon against chlorophyll a, TN or DRP, traits, and their interactions, or environment-trait associations. Functional feeding and life history traits were assessed independently. The function uses multiple GLMs (multiGLMs), which are the extended form of multivariate linear models. The GLMs in the package mvabund are adapted for multivariate abundance data because they can be used with strong mean–variance relationships and non-normal data often found in ecology (Wang et al., 2012). A row effect was added in the traitglm model by applying the composition function (composition = TRUE) to adjust for different sampling intensities across different samples. This row effect can be understood as a compositional term in the sense that it models relative abundance at a site (Wang et al., 2012; Warton et al., 2015b). The relationship between traits and environmental variables are described by Chapter 4. Nutrients affect invertebrate traits 51 interaction terms in the models as environment–trait associations (Brown et al., 2014; Warton et al., 2015b). We tested the significance of the environment–trait associations with the anova.traitglm function. This function used block resampling, in which sites were re-sampled, but all species from a site were kept together in the re-sample (Warton et al., 2015b). The p-value was calculated from 999 iterative re-samplings; p-values ≤ 0.05 indicated significant trait–environment relationships. All assumptions for functions were checked and verified. Traits linked specifically with chlorophyll a, TN or DRP were identified by the model selection procedure by applying the Least Absolute Shrinkage and Subset Selection Operator (LASSO) penalty in the traitglm function. Both functional feeding and life history traits were fitted in the traitglm function at a time. The LASSO penalty set to 0 any interaction coefficients that do not reduce the Bayesian Information Criterion (BIC) and returns the interaction coefficients of the important trait–environment associations (Brown et al., 2014; Warton et al., 2015b). The focus of the study is on analyzing the effects of nutrient-induced changes in food sources on invertebrate traits, therefore, we excluded four traits of having short-lived, sexual reproduction, being deposit feeders and/or filter feeders because they were linked with TN and/or DRP, but not chlorophyll a. All predictors were standardized, so the sizes of the coefficients measure the relative importance of the predictors (Warton et al., 2015b). Each interaction coefficient can be interpreted as the amount by which a unit (1 SD) change in the trait variable alter the slope of the relationship between abundance and the environmental variable of interest (Warton et al., 2015b). The magnitude and sign of the interaction coefficients indicate the strength and the direction of the effect of chlorophyll a, TN or DRP on traits. The higher the value of the interaction coefficients, the stronger these relationships are. Chapter 4. Nutrients affect invertebrate traits 52 RESULTS Functional feeding and life history traits were linked with both chlorophyll a and nutrients (Table 2). Associations between shredders and scrapers with chlorophyll a and two nutrients were negative, whereas, algal piercers and collectors were positively associated (Fig 1 and Table 3). Individuals having univoltine life cycles had a negative link with chlorophyll a, TN and DRP, while those having two aquatic life stages, being long-lived, and/or having hermaphroditic ability were positively associated. Individuals having asexual reproduction were negatively associated with chlorophyll a, but positively linked with TN (Table 3). Table 2. Analysis of deviance testing the significance of interaction terms between Chlorophyll a (Chla), dissolved reactive phosphate (DRP), total nitrogen (TN) and functional feeding and life history traits obtained from fourth-corner models. Samples were collected at 24 sites in 17 streams in Manawatu, New Zealand in 2007. * = p< 0.05, ** = p< 0.01. Models Interaction terms Deviance p Functional feeding traits : Chla 6 32.99 0.04* Functional feeding traits : TN 6 47.75 0.005** Functional feeding traits : DRP 6 29.12 0.05* Life history traits : Chla 6 75.03 0.007* Life history traits : TN 6 119 0.003** Life history traits : DRP 6 85.86 0.006** Chapter 4. Nutrients affect invertebrate traits 53 Figure 1. The proportion of algal piercers, collectors, scrapers and shredders plotted against periphyton biomass, total nitrogen, and dissolved reactive phosphate collected at 24 sites on 17 streams in Manawatu, New Zealand in 2007. Chapter 4. Nutrients affect invertebrate traits 54 Table 3. Traits associated with Chlorophyll a (Chla), total nitrogen (TN) or dissolved reactive phosphate (DRP). The interaction terms were selected by the Least Absolute Shrinkage and Subset Selection Operator (LASSO) penalty. The nature and the strength of the disturbance-trait associations are identified by the sign and magnitude of the interaction coefficients between Chlorophyll a, total nitrogen (TN) or dissolved reactive phosphate (DRP) and invertebrate traits. Samples were collected at 24 sites on 17 streams in Manawatu, New Zealand in 2007. Trait group/trait Chla TN DRP Life history traits Aquatic stages 2 aquatic life stages 0.07 0.13 0.15 Life duration Long 0.09 0.01 0.06 Reproduction frequency Univoltine -0.17 -0.30 -0.19 Reproduction technique Asexual -0.2 0.07 0 Hermaphrodism 0.58 0 0.21 Functional feeding traits Algal piercers 0.04 0.05 0.03 Collectors 0.09 0.26 0 Scrapers -0.1 -0.02 0.02 Shredders -0.25 -0.18 -0.25 DISCUSSION Associations between the responses of functional feeding and life history traits to periphyton biomass and/or nutrients were most likely a result of the link between periphyton biomass and nutrients (Fig. 2). This supports the idea that nutrients promote algal growth that, in turn, change the composition of functional feeding groups toward those that feed on algae (Matthaei et al., 2010; Yuan, 2010; Lange et al., 2014). Chapter 4. Nutrients affect invertebrate traits 55 Figure 2. Periphyton biomass plotted against total nitrogen and dissolved reactive phosphate. Samples were collected at 24 sites on 17 streams in Manawatu, New Zealand in 2007. Nutrient enrichment may shift algal communities from diatom dominance at low- nutrients to filamentous algae and cyanobacteria-dominated communities in high nutrient levels (Suren and Riis, 2010). Algal piercers directly feed on algae; therefore, a positive response of algal piercers with both periphyton biomass and nutrients was mostly because of an increased periphyton biomass at high nutrient levels. Two life history traits of having two aquatic life stages and being plurivoltine that showed a positive response to periphyton biomass and nutrients also contributed to higher abundance of algal piercers at high periphyton biomass. Therefore, the positive response of algal piercers likely reflected nutrient effects acting on both food source and habitat quality. The increased periphyton biomass and nutrients resulted in a decrease in the abundance of shredders because they feed on coarse organic matter, which was consistent with the observed pattern in the study of Townsend et al. (2008) and Lange et al. (2014). Shredders have a wide range of life history traits that showed both negative and positive Chapter 4. Nutrients affect invertebrate traits 56 responses to periphyton biomass and nutrients. For example, the trait of having two aquatic life stages in the shredder caddisfly Olinga feredayi showed a positive response to periphyton biomass and nutrients, whereas, the trait of having one aquatic life stage in shredder stoneflies did not. Among these stoneflies, some are univoltine that showed a negative response to periphyton biomass and nutrients, whereas, others are plurivotine. While Hydraenidae beetles and the cranefly Molophilus have long-lived adults that showed a positive response to periphyton biomass and nutrients, other shredders have short-lived adults. Therefore, the negative response of shredders to periphyton and nutrients were mainly determined by nutrient effects acting on the food source. The positive link between collectors and periphyton biomass and total nitrogen might be due to the fact that collectors feed on both streambed alage and other organic matter. Therefore, they can take advantage of a high abundance of streambed algae at high nutrient levels. However, we found that collectors being plurivoltine increased their abundance with periphyton biomass and total nitrogen, while those being univoltine decreased (Fig. 3 and 4). These different responses might be due to univoltine trait being negatively linked with periphyton biomass and nutrients, but the association with plurivoltine trait was positive. Collectors having different life history traits, therefore, reacted differently to periphyton biomass and nutrients. Chapter 4. Nutrients affect invertebrate traits 57 Figure 3. The proportion of collectors having plur