Download - WCS Fiji

Effects of alteration to 

catchments and streams on 

freshwater fish communities 

of Vanua Levu, Fiji 

Stacy Jupiter 1 , Aaron Jenkins 2 , Kini Koto 1,2 , John Ah Tong 3 , 

Toka Bwebe 3 , Akuila Cakacaka 1 , Sirilo Dulunaqio 1 , 

Margaret Fox 1 , Laisiasa Kuritani 4 , Sam Mario 3 , Waisea 

Naisilisili 1 , Yashika Nand 1 , Tukana A 3 , Rebeccca Weeks 1 , 

Naushad Yakub 1 

1 Wildlife Conservation Society Fiji Country Program 

2 Wetlands International‐Oceania 

3 Fiii Department of Fisheries 

4 Fiji Department of Forestry

© 2012 Wildlife Conservation Society 

This document should be cited as: 

Jupiter S, Jenkins A, Koto K, Ah Tong J, Bwebe T, Cakacaka A, Dulunaqio S, Fox M, Kuritani L, 

Mario S, Naisilisili N, Nand Y, Tukana A, Weeks R, Yakub N (2012) Effects of alteration to 

catchment and streams on freshwater fish communities of Vanua Levu, Fiji. Wildlife 

Conservation Society, Suva, Fiji, 17 pp.

Executive Summary 

In October and November 2010, staff from the Wildlife Conservation Society, Wetlands 

International‐Oceania, Department of Fisheries and Department of Forestry conducted riparian 

and stream surveys at 32 small stream sites in Wainunu, Kubulau, Macuata and Sasa districts of 

Bua and Macuata provinces on the island of Vanua Levu. The sites were chosen in areas of 

greater or less than 50% sub‐catchment forest cover and with intact or degraded riparian zones 

to assess the impact of catchment and stream alteration on in‐stream freshwater fish 

communities. We set out to address the question: “How does the size and composition of the 

riparian forest buffer strip in varying overall catchment cover conditions influence fish 

abundance, diversity, and water quality in the adjacent river?” 

We found that the tree community size structure of the riparian zone may only have marginal 

influence on in‐stream fish abundance. The factors that were most strongly related to fish 

presence/absence and abundance were: sub‐catchment forest cover; conductivity; and the 

presence of downstream overhanging culverts. 

Our prior research indicated that fish community composition is substantially affected when 

catchment forest cover falls below 50%. These findings were confirmed in our present study. 

This is likely due to increased sediment erosion from the adjacent lands into streambeds, which 

can impact feeding, breeding and resting habitat of Fiji’s native fish. Our present survey found 

elevated conductivity at the most degraded sites, which can be related to concentrations of 

suspended sediment and dissolved organic material. 

Secondly, we found reduced species richness and fish abundance at sites upstream from 

overhanging culverts, even in locations with high cover of primary forest and intact riparian 

zones. Overhanging culverts block upstream migrations of fish, and a large proportion of Fiji’s 

freshwater fish fauna make obligate migrations from the upper or mid‐reach of streams to the 

sea at some phase in their life cycles. Many species that were absent from the fish assemblages 

upstream from overhanging culverts are those with importance for subsistence or livelihoods 

for inland communities. Thus, there is a pressing need to think about improved culvert design 

to allow for safe fish passage. 

From our research, we have developed two important rule of thumb recommendations to 

guide local communities to manage their freshwater systems. First, communities should aspire 

to protect waterways in sub‐catchments with greater than 50% forest cover. It is much easier 

and more cost‐effective to protect existing intact landscapes than to attempt to restore them. 

Secondly, communities should preferentially prioritize freshwater streams for protection that 

are clear of downstream overhanging culverts. We have been using these guidelines to assist 

communities throughout Vanua Levu to designate or expand terrestrial and freshwater 

protected areas. 

1

Introduction 

Riparian habitats are critically important as habitat corridors for wildlife (Catteral 1993) and 

sources of organic detritus for downstream secondary production (Caraco and Cole 2004). They 

additionally provide water quality benefits through bank stabilization and sediment trapping 

(McKergow et al. 2003) and nutrient and chemical filtering (Hubbard and Lowrance 1994), 

therefore potentially protecting downstream reef systems from sediment and nutrient 

pollution. Because Fiji’s fishes are highly migratory (~99% of fishes found in freshwater make 

contact with saltwater during their life cycles; Jenkins et al. 2010), protection of riparian 

systems is also critical to protect biodiversity and important fisheries resources. 

There is a large body of evidence which suggests that broad‐scale catchment land‐clearing has 

both direct and indirect effects on tropical, native in‐stream community structure (Naiman and 

Decamps 1997). In Fiji, Jenkins et al. (2010) note a marked decline in fish diversity in mid‐ 

reaches of streams where catchment forest cover is reduced below 50%. These impacts are 

particularly pronounced in degraded catchments during the wet season, when seasonal flood 

pulses bring high volumes of sediments and associated pollutants into waterways (Jenkins and 

Jupiter 2011). Invertebrates are likely to be similarly affected: Haynes (1999) found consistently 

lower diversity in streams adjacent to logged areas over a three year study period. She 

hypothesized that the low abundance of neritid gastropods in streams of a logged catchment 

was due to sediment covering the periphyton on which they grazed (Haynes 1999). The decline 

of these prey species may strongly affect predator species such as gudgeons, which are 

important local freshwater fisheries resources in Fiji and feed preferentially on bottom‐dwelling 

invertebrates (Jenkins et al. 2010). 

A large question remains regarding the role and nature of riparian vegetation and processes in 

mitigating the effects of broader‐catchment land clearing, particularly in tropical systems. Some 

studies have found riparian width and past disturbance histories to be important factors in 

determining impacts from land‐clearing: for example, a study from Tasmania found an 80% 

reduction in macroinvertebrate abundance between control sites upstream of logging and 

downstream sites where riparian buffer width was less than 30 m (Davies and Nelson 1994). In 

a tropical example, Iwata et al. (2003) showed decreases in abundance and diversity of aquatic 

insects, shrimp, crabs, and benthic‐dwelling fish in relation to increases in fine sediments, 

eroded banks, and depositional habitat as a result of previous slash‐and‐burn clearing within 

riparian zones in Borneo. 

Local management can play a major role in halting these declines and, in one case, community‐ 

based management of catchment areas in Fiji was successful at preserving native fish diversity 

in freshwater systems. The small, coastal catchment of Macuata‐i‐wai, which is surrounded by 

heavily cultivated and degraded land, had much greater fish diversity than non‐managed 

catchments with comparable forest cover (Jenkins et al. in 2010). For two years prior to 

sampling, the community leaders had strictly enforced a ban on logging, fishing and waste 

disposal within the vicinity of the stream, which may have preserved the clean, rocky substrate 

preferred by species such as the endemic Stiphodon sp. 1 and the overhanging riparian 

2

vegetation which provides leaf litter detritus on which specialized detritivores feed (such as 

Ophiocara porocephala). Yet following the lifting of this ban, all benefits of protection were 

rapidly removed (Jenkins and Jupiter 2011), indicating that riparian and freshwater protection 

must be consistent, long‐term and enforced for it to be effective. 

However, an assumption is often made that the water quality and community benefits from 

preserved and restored riparian vegetation will be universally applicable. They typically only 

occur where the vegetation communities are proximate to pollution sources and when surface 

runoff moves slowly across the root zone (Norris 1993; Lowrance et al. 1997; Jupiter and 

Marion 2008). In the section of Kubulau District’s (Bua Province) ecosystem‐based management 

plan on best practices for management of freshwater habitats, there is a recommendation to 

“restore degraded river banks and riparian zones by planting native trees and shrubs” (WCS 

2009). Because this is a time consuming and expensive process, we need to first be confident 

that areas with more intact riparian zones will indeed support greater biological diversity in the 

context of broader land‐clearing for agricultural activities and logging. Secondly, we need to be 

able to prioritize where would be the best places along freshwater corridors to protect existing 

riparian habitat and restore degraded habitat. 

In this study, we collect a range of field data on in situ predictor variables (e.g. riparian 

vegetation composition, dissolved oxygen, conductivity, water temperature) and response 

variables (e.g. fish diversity and relative abundance) in order to better understand drivers of in‐ 

stream community composition. Secondly, we use the lessons from this current research and 

prior studies (Jenkins et al. 2010; Jenkins and Jupiter 2011) to set rules of thumb for identifying 

priorities for terrestrial and riparian protection across Fiji. 

Methods 

Study region and site selection 

Our study region focused on the districts of Wainunu and Kubulau (Bua Province) and Macuata 

and Sasa (Macuata Province) on the island of Vanua Levu, Fiji. For site selection, we used spatial 

layers of Fiji forest cover, roads, villages and catchments boundaries, and referenced imagery 

within Google Earth, to identify locations that met the following criteria in our stratified 

sampling design: 

Greater than 50% sub‐catchment forest cover Less than 50% sub‐catchment forest cover 

Greater than 30 meters 

riparian width 

Less than 30 meters 






We used this technique to select 32 sampling sites (8 sites for each treatment) and pre‐ 

uploaded the GPS coordinates to enable field location of the sampling areas. Upon arriving at 

the field sites in October and November 2010, the field team found that the actual width of the 

riparian zones in many cases was larger or smaller than anticipated, leading to an unbalanced 

site design (Figure 1, Appendix 1): 

3

Greater than 50% sub‐catchment forest cover Less than 50% sub‐catchment forest cover 


riparian width (n = 13) 







Figure 1. Location of riparian and instream field sites in Wainunu, Kubulau, Macuata and Sasa districts. 

It was particularly difficult for the field team to locate more than 2 sites in areas of substantially 

cleared catchments where riparian buffer zones had been maintained at widths greater than 30 

m. This type of site is extremely valuable as it simulates what restoration efforts might be able 

to provide. Unfortunately, both of the sites located were compromised by the presence of 

downstream overhanging culverts, which serve as barriers for upstream movements of most 

fish that cannot climb. Thus, our analysis focused on evaluating the biophysical determinants of 

similarities and differences between sites, using continuously distributed predictor variables 

(e.g., conductivity, water temperature, dissolved oxygen concentrations, mean stream depth, 

mean stream width), ranked qualitative measurements of stream characteristics (e.g., substrate 

type, overhanging canopy cover, number of root masses, number of undercuts), and categorical 

predictor variables (e.g., presence/absence of downstream overhanging culverts). 

Riparian surveys 

4

Diameter at breast height (dbh) was measured from all trees with diameter greater than 10 cm 

within four replicate 30 m x 2 m belt transects running perpendicular to the stream bank at 

each site (Figure 2a). Dominant trees and other vegetation types were noted for each transect, 

as well as remarks about landscape characteristics. Vegetation was crosschecked against Keppel 

(2005) and Keppel and Ghazanfar (2006) to determine whether species were endemic, 

indigenous or introduced. 

a) b) 

Figure 2. Schematic representation of measurements taken for (a) riparian zone surveys and (b) stream 

biophysical characteristics surveys. RM = root mass, UC = undercut, yellow square = quadrat. 

Stream biophysical characteristics surveys 

At each site, we collected measurements of water quality variables (temperature, conductivity, 

and dissolved oxygen) and stream characteristics (stream width, stream depth, substrate 

coarseness, canopy cover, number of root masses, and number of undercuts) predicted to 

influence fish community assemblages (Figure 2b). We measured water quality variables with a 

hand‐held YSI multi‐meter before entering the water to minimize disturbance. We measured 

stream width across 5 replicate transects spread at a minimum of 20 m apart. For each 

transect, we estimated canopy cover (0‐20%, 20‐40%, 40‐60%, 60‐80%, 80‐100%) from the 

centre of the stream. We noted whether there were any root masses or undercuts present at 

each stream bank side along each transect. We measured stream depth from 5 quadrats spaced 

randomly across each transect. We ranked substrate coarseness into the following classes, 

evaluated for each quadrat: 1 – silt; 2 – sand; 3 – gravel; 4 – pebble; 5 – cobble; 6 – boulder; 7 – 

bedrock. 

In‐stream fish community surveys 

We surveyed fish species richness and abundance in streams at study sites in Wainunu, 

Kubulau, Macuata and Sasa districts described above. We systematically sampled fish 

communities using the exact methods of Jenkins et al. (2010), modified from field protocols of 

Parham (2005) and Fitzsimons et al. (2007). In brief, we used a variety of techniques to collect 

fauna from the streams to ensure comprehensive presence/absence assessment. These 

5

techniques included: electrofishing using a Smith‐Root (500 V, 10A) backpack unit; netting with 

gill nets (1 in mesh), large seine nets (0.4 cm 2 mesh), medium pole seine nets (1 mm 2 mesh) and 

small hand nets (1 mm 2 mesh); and observations by mask and snorkel. At each site, 4–6 

surveyors made collections from downstream to upstream for 1 h total. We fixed all specimens 

that could not be identified in the field in 10% formalin solution and transferred them to 70% 

ethanol solution after 1–2 weeks fixation for accurate taxonomic verification. 

Statistical analyses 

We first conducted RELATE tests comparing Bray‐Curtis resemblance matrices calculated for 

mean density of trees across dbh size classes distributions with Bray‐Curtis resemblance 

matrices (with dummy variable added due to sites with no fish collected) for fish 

presence/absence and abundance data. A RELATE test operates on the null assumption that 

there is no underlying relationship between the two sets of site‐based data being compared 

and is assessed based on the number of times the calculated rho (ρ) statistic exceeds that found 

in 95% of simulations based on 999 permutations of the data labels (Clarke and Gorley 2006). 

As per Jenkins and Jupiter (2011), we used the BIO‐ENV procedure within the BEST function of 

PRIMER version 6 software to evaluate potential stream biophysical correlates of fish 

presence/absence and abundance data (Clarke and Ainsworth 2003). We compared Euclidean 

distance similarity matrices of normalised stream biophysical variables plus presence of 

downstream overhanging culverts with Bray‐Curtis similarity matrices (with dummy variable 

added) for fish presence/absence and abundance over 999 permutations. The output statistic 

for the BIO‐ENV procedure is also rho (ρ). We used non‐metric dimensional scaling (nMDS) to 

ordinate fish presence/absence and abundance data and evaluated the significance of resulting 

clusters using cluster analysis with similarity profile (SIMPROF) tests (Clarke and Gorley 2006). 

Lastly, we conducted two‐factor permutational multivariate analysis of variance (PERMANOVA) 

analyses with 4,999 permutations using log10 Modified Gower resemblance matrices 

(Anderson 2001) of fish presence/absence and abundance data. We conducted separate two‐ 

factor PERMANOVA analyses with catchment forest cover class and presence of culverts as 

fixed factors and with riparian zone width class and presence of culverts as fixed factors as 

there was not enough replication to evaluate the interaction of catchment forest class and 

riparian width class due to difficulties finding sites in heavily cleared catchments with remnant 

riparian zones (as described above). 

Results 

Riparian zone communities 

Dominant vegetation is described for each district below. Unless otherwise indicated to be 

endemic or introduced, plants are of indigenous origin. 

Wainunu. Wainunu District had the highest cover of primary forest near streams of any district 

surveyed. Dominant species in sites included: Atuna racemosa (makita); Gironniera celtidifolia 

(sisisi); Pometia pinnata (dawa); Ficus vitiensis (lolo; endemic); Cyathea spp. (balabala); 

Miscanthus floridulus (gasau); Inocarpus fagifer (ivi); Myristica castaneifolia (kaudamu; 

endemic); and Intsia bijuga (vesi). Other large trees noted included: Garcinia pseudoguttifera 

6

(bulu m); Bischofia javanica (koka); Dysoxylum lenticellare (malamala); Endospermum 

macrophyllum (kaukula; endemic); Serianthes melanesica (vaivai ni veikau); and Dillenia biflora 

(kuluva). There was additionally a sighting of the now rare indigenous hardwood Fagraea 

gracilipes (buabua) at site W3. Tree sizes were skewed towards saplings and on average, trees 

were present along each transect in every size class (Figure 3a,b). 

Kubulau. Survey sites in Kubulau were located in a mix of primary and secondary forest, as well 

as land previously cleared for plantations. At sites with recent or prior disturbance, there was a 

large cover of the vine Meremia peltata on the forest canopy or creeping across open space. 

Dominant species in sites with greater than 50% sub‐catchment forest cover included: P. 

pinnata (dawa); G. celtidifolia (sisisi); I. fagifer (ivi); Cyathea spp. (balabala); D. biflora (kuluva); 

G pseudoguttifera (bulu m); A. racemosa (makita); F. vitiensis (lolo; endemic); 

Dysoxylum richii (sasawira; endemic); Parinari insularum (sa); M. castaneifolia (kaudamu; 

endemic); Aleurites moluccana (sikea; aboriginal introduction); Macaranga harveyana (gadoa); 

I. bijuga (vesi); and Pagiantha thurstonii (tadalo; endemic). Other large trees noted included: 

Balaka seemannii (balaka; endemic); and Heritiera ornithocephala (rosarosa). More degraded 

landscapes included the trees, shrubs and grasses: Piper puberulum (yaqoyaqona vula); 

Geoniostoma vitiense (boiboida); Cynometra insularis (cibicibi; endemic); I. fagifer (ivi); F. 

vitiensis (lolo; endemic); M. harveyana (gadoa); Cocos nucifera (niu); Piper methisticum 

(yaqona); Alocasia sp. (via); Syzygium malaccense (kavika; aboriginal introduction); Hibiscus 

tiliaceus (vau); Spathodea campanulata (African tulip; introduced); Pandanus tectorius (vadra); 

P. pinnata (dawa); Theobroma cacao (cocoa); Leucena leucocephala (vaivai; introduced) and 

Canaga odorata (makosoi; possibly introduced from Hawaii). Like in Wainunu, tree sizes were 

skewed towards saplings and on average, trees were present along each transect in every size 

class (Figure 3c,d). 

Macuata. Riparian landscapes in Macuata were fairly degraded, with sparse canopy adjacent to 

grazing lands and/or dry forest. Species found included a combination of mesic forest trees and 

trees cultivated for fruits, such as: I. fagifer (ivi); Mangifera indica (maqo; early European 

introduction); S. malaccense (kavika; aboriginal introduction); C. nucifera (niu); D. richii 

(sasawira; endemic); Gyrocarpus americanus (wiriwiri); C. odorata (makosoi; possibly 

introduced from Hawaii); Psidium guajava (guava; aboriginal introduction); Decaspermum 

vitiensis (nuqanuqa; endemic); L. leucocephala (vaivai; introduced); A. moluccana (sikea; 

aboriginal introduction); Guioa sp. (drausasa); M. harveyana (gadoa); Casuarina equisetifoeia 

(nokonoko); Morinda citrifolia (kura); F. vitiensis (lolo; endemic); S. campanulata (African tulip; 

introduced); Pinus caribaea (pine; introduced); I. bijuga (vesi); P. insularum (sa); Premna 

protusa (yaro; endemic); and P. tectorius (vadra). Sites had reasonable density of saplings, but 

few trees with dbh between 50 to 100 cm, suggesting considerable past disturbance (Figure 

3e,f). 

Sasa. The Sasa sites were the most disturbed, with highly cleared, largely open canopied, grassy 

riparian zones. Most larger trees were left standing likely because of food or fiber resources 

that they produce. Species included: P. tectorius (vadra); P. guajava (guava; aboriginal 

introduction); Mangifera indica (maqo; early European introduction); Erythrina variegata 

7

(drala); P. caribaea (pine; introduced); S. campanulata (African tulip; introduced); C. odorata 

(makosoi; possibly introduced from Hawaii); and L. leucocephala (vaivai; introduced). Cassava, 

yams, eggplants and paragrass were found throughout transects, with other grasses and herbs. 

Sasa had the lowest tree density, with the fewest saplings and low or missing values from many 

dbh size classes (Figure 3g,h). 

Biophysical characteristics of streams 

Wainunu and Kubulau had the highest mean canopy cover (80‐100%) over streams, with slightly 

greater width and depth of streams than in Macuata and Wainunu (Table 1). Wainunu streams 

had the coarsest substrate and lowest conductivity. Temperature was notably elevated and 

dissolved oxygen notably lower in Sasa streams, which had very open canopy cover (20‐40%). 

Macuata had the highest average number of root masses observed per site, while none of the 

districts had high mean numbers of undercuts along stream banks. 

Table 1. Mean site stream biophysical parameters for each district for: stream width (m); stream depth 

(m); ranked substrate coarseness; estimated canopy cover; number of undercuts; number of root 

masses; conductivity (µS cm ‐1 ), temperature (⁰C); and dissolved oxygen (DO, mg L ‐1 ). 

Stream width Stream depth Substrate Coarseness Canopy cover Undercuts Root masses Conductivity Temperature DO 

Wainunu 3.837 0.306 5.043 80‐100% 0.615 1.462 83.431 24.608 7.04 

Kubulau 2.744 0.306 2.112 80‐100% 0.667 2.556 122.479 24.856 6.076 

Macuata 2.479 0.25 2.479 40‐60% 0.167 3 370.2 24.633 6.375 

Sasa 1.439 0.157 1.439 20‐40% 0.25 0 217.85 28.575 0.987 

Factors driving fish community assemblages 

The RELATE tests maintained the null hypothesis that riparian tree size distribution is not 

related to in‐stream fish presence/absence or abundance, however the rho values were only 

barely non‐significant, particularly for fish abundance (abundance: ρ = 0.161, p = 0.051; 

presence/absence: ρ = 0.145, p = 0.072). No combination of stream biophysical variables 

significantly explained fish presence/absence distribution patterns in the BIO‐ENV analysis. The 

two factors with the strongest correlation (ρ = 0.202, p = 0.284) were conductivity and the 

presence of downstream overhanging culverts. The patterns in site‐level conductivity did, 

however, significantly relate to fish abundance (ρ = 0.311, p = 0.047). Cluster analyses of fish 

communities at the site level based on presence/absence and abundance data indicated that 

the sites can be separated into three significantly different groups (Figure 4): (1) sites (W21, K8, 

K29) with high species richness and abundance, despite lack of a 30 m riparian buffer zone, but 

without downstream overhanging culverts; (2) sites with extremely low species richness and 

abundance, containing only very hardy fish or no fish (M32, M31, S10, M15N) due to extreme 

environmental degradation and/or presence of a downstream overhanging culvert; and (3) 

everything else. Sites W21, K8 and K9 were the most speciose, with each site containing the 

following species that did not appear in any other sites: Ambassis miops; Kuhlia munda; and 

Microphis brachyurus (Appendix 2). Results from PERMANOVA show that catchment forest 

cover class and presence of downstream culverts both significantly influence site‐based fish 

presence/absence and abundance, however there was no significant interaction between them 

8

(Table 2). In the PERMANOVA analysis with riparian width class and presence of culverts, only 

culverts significantly influenced the fish community structures (Table 3). 

Figure 3. Size class distributions of mean tree density per transect (60 m 2 ) in diameter at breast height 

(dbh) and representative photograph of survey locations from (a) Wainunu, (b) Kubulau, (c) Macuata, 

and (d) Sasa districts. 

9

Figure 3. nMDS plots of (a) fish presence/absence and (b) fish abundance by site. Colours indicate: 

orange triangle – greater than 50% sub‐catchment forest cover and presence of downstream 

overhanging culvert; dark blue triangle – greater than 50% forest cover and no downstream culvert; light 

blue square – less than 50% forest cover and no downstream culvert; and yellow diamond – less than 

50% forest cover and presence of downstream culvert. Sites to the left of the plot have the most species 

and most numbers of fish. Dashed circles indicate clusters which are significantly different from one 

another. 

10

Table 2. PERMANOVA results with sub‐catchment forest cover class and presence of culvert as fixed 

factors for site‐pooled (a) fish presence/absence and (b) fish abundance. Significant values are in bold. 

(a) Fish Presence/Absence 

(b) Fish Abundance 

Source df SS MS Pseudo‐F P(perm) df SS MS Pseudo‐F P(perm) 

Catchment Forest Class 1 0.747 0.747 1.821 0.046 1 0.747 0.747 1.821 0.046 

Culvert Presence 1 0.836 0.836 2.036 0.022 1 0.836 0.836 2.036 0.020 

Forest x Culvert 1 0.359 0.359 0.874 0.587 1 0.359 0.359 0.874 0.576 

Residual 28 11.490 0.410 28 11.490 0.410 

Total 31 13.498 31 13.498 

Table 3. PERMANOVA results with riparian width class and presence of culvert as fixed factors for site‐ 

pooled (a) fish presence/absence and (b) fish abundance. Significant values are in bold. 

(a) Fish Presence/Absence 

(b) Fish Abundance 

Source df SS MS Pseudo‐F P(perm) df SS MS Pseudo‐F P(perm) 

Riparian Width Class 1 0.719 0.719 1.719 0.065 1 0.719 0.719 1.719 0.065 

Culvert Presence 1 0.928 0.928 2.218 0.013 1 0.928 0.928 2.218 0.012 

Riparian x Culvert 1 0.174 0.174 0.416 0.965 1 0.174 0.174 0.416 0.969 

Residual 28 11.718 0.419 28 11.718 0.419 

Total 31 13.498 31 13.498 

Discussion 

The major goal of this research was to address the question: “How does the size and 

composition of the riparian forest buffer strip in varying overall catchment cover conditions 

influence fish abundance, diversity, and water quality in the adjacent river?” We sought this 

information specifically to inform recommendations for riparian zone protection, restoration 

and freshwater management. However, while we were able to demonstrate several factors that 

influence freshwater fish community composition that confirm and build on prior research 

(Jenkins et al. 2010; Jenkins and Jupiter 2011), we found it difficult to specifically link conditions 

in the riparian zone alone to fish assemblage characteristics, although they are likely to play a 

contributing role. 

There is a large body of literature that indicates that fragmentation and degradation within the 

riparian zone can affect biological communities and abiotic conditions within catchments and 

adjacent streams (Davies and Nelson 1994; Machtans et al. 1996; Debinski and Holt 2000; 

Heartsill‐Scalley and Aide 2003; Iwata et al. 2003). When we assessed the condition of riparian 

communities in relation to fish community variables based on tree size structure, which can 

give a good indication of disturbance to the forest community, we found that tree size structure 

at our sites surveyed may only have a marginal impact on fish community assemblages. 

There are a few reasons that could explain this phenomenon. First, we found that downstream 

overhanging culverts exerted strong influence on structuring fish communities because they 

11

epresent a huge barrier to upstream migration. Culverts, dams and other natural barriers (e.g. 

waterfalls) have been previously shown to interrupt migration of diadromous fishes (Holmquist 

et al. 1998; Fitzsimons et al. 2005; Greathouse et al. 2006; Hein et al. 2011). This is particularly 

problematic in Fiji where freshwater ichthyofaunal communities are dominated by 

amphidromous fish that make obligate migrations downstream as larvae and upstream as post‐ 

larvae (Jenkins et al. 2010). Jenkins et al. (2010) postulated that overhanging culverts may 

strongly influence mid‐reach fish assemblages, but they could not detect significant 

contribution of maximum downstream slope on fish species richness likely due to the 

coarseness of the slope data used. However, in our current survey at sites with good catchment 

and riparian forest cover where downstream overhanging culverts were documented (e.g. sites 

in Wainunu and Kubulau such as W18, W28, K7 and K), species richness and abundance were 

substantially lower than would have been otherwise predicted. Fish communities only included 

climbing species (Anguila marmorata); hardy species that may have been trapped and survived 

upstream (Eleotris fusca; Giurus margaritacea) and one of Fiji’s two endemic freshwater 

residents (Redigobius leveri). We, therefore, are looking to start conversations with Fiji 

Government about best practices for constructing culverts and retrofitting existing culverts with 

fish passageways (e.g. Kapitzke 2010) 

A second possible reason for the lack of strong influence of riparian and stream habitat in 

structuring fish assemblages may be due to lower overall diversity from reduced number of 

microhabitats in small stream systems compared with larger stream systems previously 

surveyed (Jenkins et al. 2010; Jenkins and Jupiter 2011). Niche partitioning through habitat and 

food specialization are important determinants of freshwater stream communities (Ross 1986). 

Our data indicated very few undercuts and root masses at any of the sites, which are important 

microhabitat features for many fish species (Pusey et al. 2004). Greater replication at the site 

level in future studies may help to uncover more of these important features. 

Despite these findings, the condition of the riparian zone and overall condition of adjacent 

landscape may still indirectly influence assemblages. Iwata et al. (2003) found strong 

relationships between the degree of riparian disturbance and regeneration in tropical 

landscapes and in‐stream depositional characteristics, with more eroded soil found in streams 

adjacent to disturbed areas. The most degraded riparian sites we surveyed in Macuata and Sasa 

districts had extremely elevated conductivity, which can increase with high sediment loads or 

high concentrations of dissolved organic material. Sediment may impact feeding, breeding and 

resting ability of many species of Fiji’s freshwater fish (Jenkins et al. 2010). Conductivity on its 

own at least partially explained fish abundance patterns across all sites surveyed. An important 

corollary to the Iwata et al. (2003) study was that in‐stream habitats adjacent to riparian areas 

well in progress of regeneration still had not recovered even one to two decades after 

agricultural activities ceased. 

With higher replication of survey sites, particularly sites with low sub‐catchment forest cover 

and intact riparian zones of at least 30 m width, we might be able to tease out impacts 

attributable to riparian zone condition apart from general catchment condition. Our finding 

that loss of sub‐catchment forest cover has strong impact on in‐stream communities echoes 

12

our prior work from Fiji that severe catchment degradation can result in near complete to 

complete loss of freshwater fish assemblages (Jenkins et al. 2010), and that these losses are 

pronounced in degraded catchments during the wet season (Jenkins and Jupiter 2011). Thus, 

our findings further give weight to our rule of thumb recommendation to protect waterways in 

sub‐catchments with greater than 50% forest cover. From this study, we add a second rule of 

thumb recommendation to preferentially prioritize freshwater streams for protection that are 

clear of downstream overhanging culverts. We have been using these guidelines to assist 

communities to designate or expand terrestrial and freshwater protected areas. To date, 

communities of Kubulau, Wainunu, Nadi and Solevu districts of Bua Province and Wailevu 

District of Cakaudrove Province have found the data and the rules of thumb useful to 

comprehend the threats to their freshwater resources and assist in selection of areas for 

management. 

References 

Anderson MJ (2001) A new method for non‐parametric multivariate analysis of variance. 

Australian Ecology 26:32–46 

Caraco NF, Cole J (2004) When terrestrial organic matter is sent down the river: the importance 

of allochthonous carbon inputs to the metabolism of lakes and rivers. In: Polis GA, Power 

ME, Huxel GR (eds) Food Webs at the Landscape Level. The University of Chicago Press, 

Chicago, USA, pp 301‐316 

Catteral CP (1993) The importance of riparian zones to terrestrial wildlife. In: Bunn SE, Pusey BJ, 

Price P (eds) Ecology and Management of Riparian Zones in Australia. Land and Water 

Resources Research and Development Corporation Occasional Paper Series No: 05/93, 

Canberra, Australia, pp 41‐52 

Clarke KR, Ainsworth M (1993) A method of linking multivariate community structure to 

environmental variables. Marine Ecology Progress Series 92:205–219 

Clarke KR, Gorley RN (2006) PRIMER v6: User Manual/Tutorial. PRIMER‐E, Plymouth, 190 pp 

Davies PE, Nelson M (1994) Relationships between riparian buffer widths and the effects of 

logging on stream habitat, invertebrate community composition and fish abundance. 

Marine and Freshwater Research 45:1289‐1305 

Debinski DM, Holt RD (2000) A survey and overview of habitat fragmentation experiments. 

Conservation Biology 14:342–355. 

Fitzsimons JM, Parham JE, Benson LK, McRae MG, Nishimoto RT (2005) Biological assessment of 

Kahana stream, island of O’ahu, Hawai’i: an application of PABITRA survey methods. Pacific 

Science 59: 273–281. 

Fitzsimons JM, Nishimoto RT, Parham JE (2007) Stream ecosystems. In: Mueller‐Dombois D, 

Bridges KW, Deahler C (eds) Biodiversity assessment of tropical island ecosystems, PABITRA 

manual for interactive ecology and management. University of Hawaii, Honolulu, pp 105– 

138 

Greathouse EA, Pringle CM, McDowell WH, Holmquist JG (2006) Indirect upstream effects of 

dams: consequences of migratory consumer extirpation in Puerto Rico. Ecological 

Applications 16: 339–352 

Haynes A (1999) The long term effect of forest logging on the macroinvertebrates in a Fijian 

stream. Hydrobiologia 405:79‐85 

13

Heartsill‐Scalley T, Aide TM (2003) Riparian vegetation and stream condition in a tropical 

agriculture‐secondary forest mosaic. Ecological Applications 13:225‐234 

Hein CL, Pike AS, Blanco JF, Covich AP, Scatena FN, Hawkins CP, Crowl TA (2011) Effects of 

coupled natural and anthropogenic factors on the community structure of diadromous fish 

and shrimp species in tropical island streams. Freshwater Biology 56:1002‐1015 

Holmquist JG, Schmidt‐Gengenbach JM, Yoshioka BB (1998) High dams and marine‐freshwater 

linkages: effects on native and introduced fauna in the Caribbean. Conservation Biology 12: 

621–630 

Hubbard RK, Lowrance RR (1994) Riparian forest buffer system research at the coastal plain 

experiment station, Tifton, GA. Water, Air, and Soil Pollution 77:409‐432 

Iwata T, Nakano S, Inoue M (2003) Impacts of past riparian deforestation on stream 

communities in a tropical rain forest in Borneo. Ecological Applications 13:461‐473 

Jenkins AP, Jupiter SD, Qauqau I, Atherton J (2010) The importance of ecosystem‐based 

management for conserving migratory pathways on tropical high islands: A case study from 

Fiji. Aquatic Conservation: Marine and Freshwater Ecosystems 20:224‐238 

Jenkins AP, Jupiter SD (2011) Spatial and seasonal patterns in freshwater ichthyofaunal 

communities of a tropical high island in Fiji. Environmental Biology of Fishes 91:261‐274 

Jupiter SD, Marion GS (2008) Changes in forest area along stream networks in an agricultural 

catchment of the Great Barrier Reef lagoon. Environmental Management 42:66‐79 

Kapitzke R (2010) Culvert Fishway Planning and Design Guidelines Part B – Fish Migration and 

Fish Species Movement Behaviour. James Cook University School of Engineering and 

Physical Sciences, Townsville, Australia, 28 pp 

Keppel G (2005) Summary report on forests of the mataqali Nadicake Kilaka, Kubulau District, 

Bua, Vanua Levu. Prepared for the Wildlife Conservation Society Fiji Program, Suva, Fiji, 16 

pp 

Keppel G, Ghazanfar SA (2006) Trees of Fiji: A guide to 100 rainforest trees. Secretariat of the 

Pacific Community (SPC) and Deutsche Gesellschaft für Technische Zusammenarbeit (GTZ), 

Suva, Fiji, 341 pp 

Lowrance RR, Altier LS, Newbold JD, Schnabel RR, Groffman PM, Denver JM (1997) Water 

quality functions of riparian forest buffer systems in the Chesapeake Bay Watershed. 

Environmental Management 21:687‐712 

Machtans CS, Villard MA, Hannon SJ (1996) Use of riparian buffer strips as movement corridors 

by forest birds. Conservation Biology 10:1366–1379. 

McKergow LA, Weaver DM, Prosser IP, Grayson RB, Reed AEG (2003) Before and after riparian 

management: sediment and nutrient exports from a small agricultural catchment, Western 

Australia. Journal of Hydrology 270:253‐272 

Naiman RJ, Decamps H (1997) The ecology of interfaces: riparian zones. Annual Review of 

Ecology and Systematics 28:621‐658 

Norris V (1993) The use of buffer zones to protect water quality: a review. Water Resources 

Management 7:257‐272 

Parham JE (2005) Survey techniques for freshwater streams on oceanic islands: important 

design considerations for the PABITRA project. Pacific Science 59:283–291 

Pusey BJ, Kennard MJ, Arthington AH (eds) (2004) Freshwater fishes of north‐eastern Australia. 

CSIRO, Canberra, Australia, 700 pp 

14

Ross ST (1986) Resource partitioning in fish assemblages: A review of field studies. Copeia 

1986:352‐388 

WCS (2009) Ecosystem‐Based Management Plan: Kubulau District, Vanua Levu, Fiji. Wildlife 

Conservation Society, Suva, Fiji, 121 pp 

15

Appendix 1. Site location information, including: district; survey date; latitude and longitude; 

whether downstream overhanging culverts were present; percent sub‐catchment forest cover 

(greater than or less than 50%); and riparian zone width (greater than or less than 30 m). 

Colours indicate site categories: green ‐ greater than 50% forest cover and greater than 30 m 

riparian zone; blue – greater than 50% forest cover and less than 30 m riparian zone; purple – 

less than 50% forest cover and greater than 30 m riparian zone; brown – less than 50% forest 

cover and less than 30 m riparian zone. 

Site Code District Survey Date Lat Lon Culverts Catchment Forest Riparian width 

W1 Wainunu 18‐10‐2010 178.92564 E 16.86486 S Yes >50 >30 

W18 Wainunu 18‐10‐2010 178.92468 E 16.86502 S Yes >50 >30 

W23 Wainunu 19‐10‐2010 178.95524 E 16.89368 S No >50 >30 

W5 Wainunu 19‐10‐2010 178.95535 E 16.89136 S No >50 >30 

W21 Wainunu 20‐10‐2010 178.94571 E 16.88878 S No >50 50 >30 

W19 Wainunu 21‐10‐2010 178. 94078 E 16.87970 S No >50 >30 

W3 Wainunu 21‐10‐2010 178.93813 E 16.87899 S No >50 >30 

WN3 Wainunu 21‐10‐2010 178.88951 E 16.82419 S No >50 50 >30 

W2 Wainunu 22‐10‐2010 178.92590 E 16.88123 S No >50 >30 

W4 Wainunu 25‐10‐2010 178.95535 E 16.89134 S No 30 

K9 Kubulau 28‐10‐2010 178.98396 E 16.90012 S No >50 50 >30 

K30 Kubulau 30‐10‐2010 179.01674 E 16.87512 S Yes

Appendix 2. Freshwater fish species presence (dark grey) for each site in each of the four districts surveyed. 

Wainunu Kubulau Sasa Macuata 

Species 

Ambassis miops 

Anguilla marmorata 

Anguilla megastoma 

Anguilla obscura 

Apogon laterallis 

Awaous guamensis 

Butis butis 

Eleotris fusca 

Eleotris melanosoma 

Giurus margaritacea 

Glossogobius sp. 

Gymnothorax polyuranodon 

Hypseleotris guentheri 

Kuhlia marginata 

Kuhlia munda 

Kuhlia rupestris 

Liza vagiensis 

Lutjanus argentimaculatus 

Microphis brachyurus 

Moringua macrochir 

Ophiocara porocephala 

Periopthalmus kalolo 

Redigobius bikolanus 

Redigobius leveri 

Scatophagus argus 

Schismatogobius vitiensis 

Sicyopterus lagocephalus 

Sicyopus zosterophorum 

Siganus vermiculatus 

Sphraena flavicauda 

Stenogobius sp. 

Stiphodon rutilaureus 

Stiphodon sp. 

Yirrkala sp. 

Zenarchopterus dispar 

W1 W18 W23 W5 W21 W20N W19 W3 WN3 W28 W2 W4 W22 K24 K9 K25 K8 K K26K7K30K29S16S11S12NS10NM17NM15NM31M32M14M15 17

Download - WCS Fiji

Create successful ePaper yourself

Delete template?

Save as template?