, Volume 166, Issue 1, pp 131–140

Population sinks resulting from degraded habitats of an obligate life-history pathway


    • Marine Ecology Research Group, School of Biological SciencesUniversity of Canterbury
  • David R. Schiel
    • Marine Ecology Research Group, School of Biological SciencesUniversity of Canterbury
Population ecology - Original Paper

DOI: 10.1007/s00442-010-1834-7

Cite this article as:
Hickford, M.J.H. & Schiel, D.R. Oecologia (2011) 166: 131. doi:10.1007/s00442-010-1834-7


Many species traverse multiple habitats across ecosystems to complete their life histories. Degradation of critical, life stage-specific habitats can therefore lead to population bottlenecks and demographic deficits in sub-populations. The riparian zone of waterways is one of the most impacted areas of the coastal zone because of urbanisation, deforestation, farming and livestock grazing. We hypothesised that sink populations can result from alterations of habitats critical to the early life stages of diadromous fish that use this zone, and tested this with field-based sampling and experiments. We found that for Galaxias maculatus, one of the most widely distributed fishes of the southern hemisphere, obligate riparian spawning habitat was very limited and highly vulnerable to disturbance across 14 rivers in New Zealand. Eggs were laid only during spring tides, in the highest tidally influenced vegetation of waterways. Egg survival increased to >90% when laid in three riparian plant species and where stem densities were great enough to prevent desiccation, compared to no survival where vegetation was comprised of other species or was less dense. Experimental exclusion of livestock, one of the major sources of riparian degradation in rural waterways, resulted in quick regeneration, a tenfold increase in egg laying by fish and a threefold increase in survival, compared to adjacent controls. Overall, there was an inverse relationship between river size and egg production. Some of the largest rivers had little or no spawning habitat and very little egg production, effectively becoming sink populations despite supporting large adult populations, whereas some of the smallest pristine streams produced millions of eggs. We demonstrate that even a wide-ranging species with many robust adult populations can be compromised if a stage-specific habitat required to complete a life history is degraded by localised or more diffuse impacts.


Attractive sinksDiadromous fishesGalaxias maculatusHabitat qualitySource-sink dynamics


Many animals traverse and use a wide range of habitats over their lifetimes. The types and quality of these habitats affect demographic rates, especially survival (Doak 1995), behavioural decisions by individuals on habitat choice, and dispersal among habitats (Delibes et al. 2001a), and these can have significant consequences on population structure and dynamics (Gundersen et al. 2001). One consequence is that source and sink populations develop, depending on whether or not local demographic characteristics and habitat quality can sustain net population growth (Pulliam 1988; Dias 1996). For sink populations, local reproduction is insufficient to balance mortality and populations only persist through immigration from more productive source populations (Pulliam 1988). Source–sink dynamics, therefore, can critically underpin the ecological and evolutionary feedbacks among populations (Kawecki and Holt 2002). The selection and use of poor habitats can be maladaptive and lead to ecological traps (Battin 2004), whereby a sink habitat is chosen over better alternatives, even when population densities are low (Gilroy and Sutherland 2007). These problems are exacerbated where habitats have become fragmented or degraded (Remeš 2000), effectively isolating populations or reducing the habitat choices available (Robinson et al. 1995), and potentially leading to localised extinctions in marginal or sparse habitats (Kawecki 2008). A critical component underpinning source–sink dynamics, therefore, is knowing what constitutes good and poor habitat, the temporal variability of the habitats (Johnson 2004), and how these habitats affect survival and consequent population dynamics.

For most of the wide-ranging species that occur in marine environments, source–sink dynamics pose special problems and are poorly understood. In contrast to the terrestrial environments on which most source–sink models are based, marine environments have many species with highly dispersive larvae and specific habitat requirements during different phases of their life histories. In most marine fishes, for example, bi-partite life histories feature obligate development of eggs and larvae in the pelagic oceanic environment, usually well away from where juveniles or adults live (Kinlan et al. 2005). Also, specific spawning grounds or habitats for egg laying may be well removed from adult habitats (McKeown 1984). Consequently, the habitats required over an entire lifetime of a single fish may cover very large landscapes and hundreds or even thousands of kilometres (Metcalfe et al. 2002). Diadromous fishes are an extreme example of this type of life history, spread over widely separated marine and freshwater habitats. Sockeye salmon, for example, spawn in relatively small upstream habitat patches, develop in river and stream habitats, and then migrate to spend their adult life in large lakes or open marine waters (Schtickzelle and Quinn 2007). Degradation or reduction in spawning habitats can lead to poor habitat choices for salmon and result in sink populations with poor survival, even though habitats for later life stages are still intact (Nehlsen et al. 1991). For diadromous species, therefore, a reduction in the availability of, or access between, stage-specific habitat components may lead to demographic deficits in sub-populations (Lucas et al. 2009).

Here, we test this concept and the role of life stage-specific habitat in source–sink dynamics of the diadromous fish Galaxias maculatus (Jenyns 1842) in New Zealand. This is one of the most widely distributed freshwater fishes in the world, occurring across the southern hemisphere in New Zealand, Australia, South America, Lord Howe Island, the Chatham Islands, and the Falkland Islands. Wide-scale dispersal occurs during their marine phase of development but, unlike salmon, there is no evidence that G. maculatus are philopatric or that there is genetic differentiation of populations among rivers and regions in New Zealand (Waters et al. 2000).

Galaxias maculatus spawns in the tidal reaches of rivers, during autumn new and/or full moon spring tides when the tidal range is maximal (Benzie 1968; Taylor 2002). Eggs are deposited supratidally within inundated streamside vegetation (McDowall and Charteris 2006). The eggs adhere to the stems and aerial root-mat of riparian vegetation (Taylor 2002) and develop for around 28 days before being re-immersed during subsequent spring tides, hatching and being washed out to sea (Benzie 1968). After a 4- to 6-month pelagic development period (McDowall et al. 1994), post-larval juveniles (‘whitebait’) return to rivers by sensing freshwater plumes following rain events (McDowall and Eldon 1980). They are the basis of a culturally important whitebait fishery in New Zealand in which juveniles are netted as they migrate upstream (McDowall and Eldon 1980). Some 95% of the whitebait returning from offshore are G. maculatus (McDowall 1965; Rowe et al. 1992). This is one of the few fisheries worldwide based on capturing larvae and post-larval juveniles (cf. Romanelli et al. 2002; Gisbert and López 2008). Adults become reproductive in their first year and live for up to 3 years, but most do not survive breeding (McDowall 1968). Adults do not migrate between streams or rivers but spend their entire lives in the waterway to which they returned as juveniles.

Given this life history, the life cycle of G. maculatus depends on obligate, stage-dependent events in three distinct habitats: freshwater waterways as adults, riparian vegetation in tidally influenced river mouths for spawning, and the marine environment for larval development (non-diadromous populations of G. maculatus are rare; in some cases, land-locked individuals make a potamodromous migration between habitats to spawn; Pollard 1971; Chapman et al. 2006). Therefore, restriction or closure of one of these life-history gateways can greatly influence the ability of fish to complete their life history as they move between habitats. A time of great vulnerability is during spawning and egg development (Hickford et al. 2010), which occur in the near-coastal zone where land clearing, deforestation, farming and urbanisation have greatly affected stream and river habitats (Kennish 2002).

Terrestrial development of eggs is very rare among fishes (Martin et al. 2004). In most cases, the benefits of aerial respiration and of reduced aquatic predation while emergent do not outweigh the physiological costs from desiccation and temperature stress (reviewed by DeMartini 1999). However, because G. maculatus eggs are deposited within riparian vegetation, it is likely that egg survival covaries with the composition of the vegetation and its ability to modify the sub-canopy physical environment (Geiger et al. 2003). Land use changes (Belsky et al. 1999), livestock grazing (Kauffman and Krueger 1984) and channelisation (Williamson et al. 1992) have caused significant modifications to the physical and biological character of riparian ecosystems. These changes have the potential to degrade the critical spawning habitat of G. maculatus by thinning or removing riparian vegetation or by limiting access to the riparian margin. Here, we test the hypothesis that sink populations result from spatially explicit alterations to habitats necessary for successful spawning and egg development of G. maculatus.

Materials and methods

Spawning habitat, egg laying and survival

Fourteen rivers on the west coast of New Zealand (Online Resource 1) were sampled to determine the extent and habitat characteristics of G. maculatus spawning sites. These rivers ranged in size from coastal streams (<50 km2 catchment) to major rivers (>6,000 km2 catchment). Fertilised G. maculatus eggs measure only 1.2 mm in diameter, and extensive preliminary searches showed that they occurred at the base of riparian vegetation along streams and rivers. Because spawning occurs on new and full moons when the tidal influence is greatest (Benzie 1968), the bank vegetation was searched for G. maculatus eggs shortly after peak spring tides (during April and May in 2006 and 2007). Initially, horizontal surveys were done along banks to determine the extent of egg deposition in relation to the tidal wedge. This confirmed that spawning and egg deposition occurred only in areas that are tidally influenced, but where low tide salinities were reduced to near freshwater levels (<2 practical salinity units). The upstream extent of the tidal saltwater wedge was then estimated in study rivers during spring tides using an array (n = 7) of Odyssey conductivity/temperature loggers (Dataflow Systems, Christchurch, New Zealand).

After each spawning event, the perimeters of all egg patches in each of the 14 rivers were traced with a real-time kinematic global positioning system (Trimble Navigation 4000SSI & 4400, Sunnyvale, USA). The area of these patches was summed to give the total area used for spawning, during each spawning event, in each river. To determine egg densities within these areas, eggs were counted in quadrats (100 × 100 mm) placed haphazardly inside multiple egg patches, and the species composition and height of vegetation above each quadrat were measured. The density of vegetation was estimated by counting the number of stems in three 50 × 10 mm transects placed across each quadrat. The overall composition of riparian vegetation was assessed in 1 × 1 m quadrats (n = 10 per location, per spawning event) placed randomly in the spawning area. Egg production in each river during each spawning event was estimated by multiplying the mean egg density in egg patches by the total area of the patches.

To test whether the survival of G. maculatus eggs covaries with the composition of riparian vegetation, the survival of developing eggs was determined at three locations (Online Resource 1). Two of these (Pigeon and Robinsons) were grazed by livestock, but small areas of riparian vegetation at each location were protected from grazing by bank topography. Live eggs were counted in marked quadrats (100 × 100 mm) immediately after deposition and again before hatching around 28 days later, and instantaneous survival rates were calculated. Dead or unfertilised eggs were readily distinguished because they changed from translucent to white in colour. The relationship between egg survival and the surrounding species composition and density of riparian vegetation was tested using data combined across all locations (n = 79).

Spawning habitat experiment

The effect of spawning habitat quality on G. maculatus egg production and survival was experimentally tested in 2007 and 2008. Plots of 5 × 2.5 m were established in a spawning area at a location (Goughs; Online Resource 1) that was heavily degraded by livestock grazing (one of the major influences on non-urban streams and rivers). Eight plots were fenced and protected from grazing, and eight were left open to grazing. After 468 days, live eggs were counted in random quadrats (100 × 100 mm, n = 80) placed inside and outside the protected areas immediately after deposition, and again before hatching around 28 days later, and instantaneous survival rates were calculated. We measured the species composition, height and density of vegetation, and the depth of the aerial root-mat. To determine environmental correlates around the eggs, ground-level temperature and relative humidity were measured around all egg patches using Hobo Pro v2 temperature/RH loggers (Onset Computer, Bourne, MA, USA).

Statistical analyses

Differences in the vertical range and upper limit of the spawning area were tested with two-way ANOVA for the effects of river and month, nested within river. Time factors in this and subsequent analyses were treated as random. To examine the association between vegetation type and spawning, a replicated G test of goodness of fit (Sokal and Rohlf 1981) was done on the percentage cover of four riparian plant taxa in the spawning area at four locations and the percentage of eggs found under the same four vegetation types. Differences in the height and density of a dominant species [Schedonorus phoenix (Scop.) Holub] were tested with ANOVA for the effects of location. Egg survival was tested with ANOVA for the effects of vegetation type. Quantile regression analysis tested the relationship between the mix of riparian vegetation and egg survival for the 50–90% quantiles of the probability distribution, and least squares regression tested the relationship between vegetation density and egg survival.

A correlation analysis was done on data from the 14 west coast rivers to examine the association between a river’s catchment area and its Galaxias maculatus spawning area. Differences in the size of spawning areas were tested with two-way ANOVA for the effects of river and year. Differences in the density of eggs within spawning areas were tested with a three-way ANOVA for the effects of river, year and month (nested within year). Differences in egg production were tested with a two-way ANOVA for the effects of river and year.

To examine the effects of experimental removal of livestock grazing from spawning areas, ANOVAs tested differences in the height of vegetation, depth of aerial root mat, vegetation density, density and survival of eggs, and ground-level temperature and relative humidity between treatments.

Tukey’s HSD post hoc tests were used when ANOVA indicated significant effects. Unless otherwise specified, a significance level of α = 0.05 was used. Data were log- or log (x + 1)-transformed when necessary to satisfy the assumption of homogeneity of variance, as detected by Cochran’s tests. If transformation did not alleviate heterogeneous variance, critical α was reduced to α = 0.01 to account for an increased probability of Type I error (Underwood 1997). Quantile regression analysis was done with R software (version 2.9.1). All other tests were done with Statistica 8 (StatSoft).


The spawning habitat of G. maculatus was very limited horizontally and vertically. All eggs were found <320 m from the upstream extent of the saltwater wedge with most occurring within 150 m. Eggs were always found on the banks within a narrow vertical band, which averaged 0.15 m in width (Fig. 1). The upper limit of the egg band was a function of local tidal amplitude and river flow during the preceding spring tides and of the longitudinal river profile. The upper limit of the egg band varied between rivers (F3,185 = 399.25, P < 0.001) and between cohorts (within rivers; F10,185 = 6.92, P < 0.001), but successive cohorts in the same rivers (Punakaiki in April–May 2006 and Robinsons in March–April 2008) were only about 100 mm different in elevation. The lower limit of the egg band varied slightly because some eggs were washed down the river bank by remaining spring tides and riverine flooding. The vertical extent of the egg band was small and varied between rivers (F3,185 = 38.31, P < 0.001), with the overall average band at Punakaiki (0.26 m ± 0.01) being broader than that in the other rivers. Within each river, however, the extent of the egg band was very consistent between cohorts (within and between years; F10,185 = 1.56, n.s.). The projection of a vertical range onto a sloping river bank is aspect-dependent, but for a 45° slope, an egg band with a 0.15 m elevation range equates to only 0.21 m of bank surface.
Fig. 1

Elevation above mean sea level of upper and lower margins (mean ± SE) of the Galaxias maculatus egg band at two locations on the a east coast and b west coast of New Zealand. Data measured 3–4 times at each site between April 2006 and May 2008

Eggs were preferentially laid beneath particular types of vegetation and not in proportion to the vegetation’s abundance (Fig. 2; GP = 250.78, P < 0.001). However, the preferred vegetation varied between locations (GH = 90.83, P < 0.001). For example, in two small, relatively pristine rivers with no livestock grazing or bank modifications in the spawning areas (Orowaiti and Punakaiki), >99% of all eggs were found in S. phoenix, Agrostis stolonifera L., or Juncus edgariae L.A.S. Johnson and K.L. Wilson, although together they comprised 63–89% of the riparian vegetation (Fig. 2a, b). At Fox, which had twice the catchment size of Orowaiti, but was degraded with major bank modifications for flood control and roading around the spawning area, 85% of all eggs were found in one species, J. edgariae (Fig. 2c), despite it comprising only 24% of the riparian vegetation. At Little Wanganui, where the spawning area was grazed by livestock, eggs were spread relatively evenly across the three preferred vegetation types, but they were under-represented in the remaining vegetation (G = 65.87, P < 0.001; Fig. 2d).
Fig. 2

Percentage cover (mean + SE) of vegetation (filled bars) and percentage of eggs found in clumps dominated (>50%) by each vegetation taxa (open bars) in the Galaxias maculatus spawning zone at four locations along the west coast of New Zealand in April and May, 2006 and 2007. a Orowaiti, b Punakaiki, c Fox and d Little Wanganui. Vegetation was Schedonorus phoenix and Agrostis stolonifera (introduced), Juncus edgariae (native) or other. Sample sizes for vegetation were n = 40 at all locations, and for eggs were n = 100 at Orowaiti, n = 74 at Punakaiki, n = 44 at Fox and n = 40 at Little Wanganui

The stem density and grass height of the dominant species, S. phoenix, varied by river. Riparian vegetation on the relatively pristine rivers (Orowaiti and Punakaiki) had greater densities and greater height than the more impacted rivers (Fig. 3). This had effects on egg survival, which depended on the type of vegetation in which eggs were deposited (F4,69 = 4.89, P < 0.01). Eggs in clumps of vegetation dominated by S. phoenix, A. stolonifera or J. edgariae had far greater survival than those in other types of vegetation (Fig. 4a). Eggs in other vegetation types averaged only 3.4% survival. The survival of eggs increased exponentially as the vegetation became more dominated by S. phoenix, A. stolonifera, or J. edgariae (F1,72 = 43.23, P < 0.001; Fig. 4b). When the cover of these species was near 100%, egg survival over the 28-day development period averaged >30%, but when the cover was below 50%, less than 7% of eggs survived. The central tendency and variance of survival changed with changes in the aggregate cover of the three species. Clearly, survival of eggs was affected by more than just the composition of the riparian vegetation. S. phoenix, A. stolonifera and J. edgariae generally have high numbers of stems at ground level, and the survival of eggs increased as the vegetation they were developing in became more dense (F1,72 = 76.41, P < 0.001; Fig. 4c).
Fig. 3

Stem density (open bars) and plant height (filled bars) of introduced Schedonorus phoenix plants (mean + SE) in the Galaxias maculatus spawning zone in April and May, 2006 and 2007 at Orowaiti (n = 39), Punakaiki (n = 70), Fox (n = 40) and Little Wanganui (n = 15)

Fig. 4

a Survival (after 28 days) of developing Galaxias maculatus eggs in clumps of vegetation dominated (i.e. >50% cover) by Schedonorus phoenix, Agrostis stolonifera, Juncus edgariae, mixed clumps of these species (Mix) or other species (Other) at three locations (see below). Letters indicate homogeneous groups based on Tukey HSD with P < 0.05). b Relationship between survival of eggs in riparian vegetation and summed proportion of S. phoenix, A. stolonifera and J. edgariae in the vegetation at Pigeon (open symbols), Te Kawa (closed symbols) and Robinsons (shaded symbols). Also shown are 0.90, 0.75 and 0.50 quantile estimates (dotted lines) and least squares regression estimate of mean function (solid line). c Relationship between survival of the same eggs and stem density of the riparian vegetation

Across 200 km of coastline, there was a wide disparity, and an inverse relationship, between the sizes of spawning areas and the sizes of river catchments (r12 = −0.10, n.s.; Fig. 5a). Many of the small coastal rivers had significantly larger spawning areas than the major rivers (F13,28 = 4.18, P < 0.01; Tukey HSD, P < 0.01). For example, the Buller River, one of New Zealand’s largest with a catchment area of 6,380 km2, had only about 2 m2 of spawning area, whereas several of the smallest streams with catchments of 50–80 km2 had spawning areas >20 m2. The major exception to the trend was the Karamea River (1,212 km2 catchment) which had the second greatest spawning area of all the rivers sampled in 2006. Across these rivers, the size of spawning areas did not differ significantly between years (F1,28 = 2.88, n.s.), but the spawning areas of two rivers (Karamea and Orowaiti) were subsequently impacted by heavy sedimentation after floods, thereby reducing spawning areas to near-zero after 2007. Consequently, no G. maculatus eggs were found at Karamea in 2007 or in subsequent years, and the spawning area at Orowaiti decreased from 30 m2 to <5 m2.
Fig. 5

Characteristics of Galaxias maculatus spawning sites in 14 rivers on the west coast of New Zealand in 2006 and 2007. a The catchment (open bars) and spawning area (mean + SE, filled bars), and b egg density (mean + SE, open bars) and production (mean + SE, filled bars) in rivers ranked by catchment area

The spawning areas in most small rivers contained higher densities of eggs than those in the larger rivers. For example, the four largest rivers (other than Karamea) had egg densities <1 per cm2, whereas the four smallest streams had densities >4 per cm2. Overall, egg densities differed between rivers (F11,558 = 7.93, P < 0.001; Fig. 5b), but not between years (F1,558 = 1.36, n.s.) or months (within years) (F2,558 = 0.37, n.s.). Egg production per spawning event ranged from around 1.7 million eggs per spawning event in the Orowaiti to <3,000 eggs in the Pororari and Buller Rivers, to none in the third and fourth largest rivers (Fig. 5b). Because smaller waterways had more spawning habitat, high densities of eggs translated into greater production than in most of the larger rivers.

Multiple effects on riparian vegetation and developing eggs were evident following experimental removal of grazing by livestock from spawning areas. After 468 days, the vegetation in ungrazed plots was on average twice as high (F1,75 = 13.95, P < 0.001; Fig. 6a), the aerial root mat was 14 mm deeper (F1,75 = 24.79, P < 0.001; Fig. 6b) and vegetation density increased by 40% (F1,75 = 24.97, P < 0.001; Fig. 6c) relative to grazed controls. The consequence of this was that egg densities in ungrazed plots were 10 times greater (F1,75 = 13.49, P < 0.001; Fig. 6d) and egg survival (%) was three times greater (F1,36 = 10.43, P < 0.01; Fig. 6e) than in grazed controls. The riparian vegetation in ungrazed plots buffered temperature and humidity fluctuations more than did the shorter and less dense vegetation in grazed plots. The average daily temperature range beneath vegetation in ungrazed plots was 2.3°C less than in grazed plots (F1,87 = 4.29, P < 0.05), but the maximum temperature was often >5°C cooler than in grazed plots. The humidity range beneath vegetation in ungrazed plots was 8.2% less than in grazed plots (F1,87 = 7.74, P < 0.01). Furthermore, there were several extreme events in the grazed plots where the humidity was <60%. The combined effect was that many dead and undeveloped eggs were found in grazed plots, but not in ungrazed plots, at the end of the 28-day development cycle.
Fig. 6

Differences (mean + SE) in a vegetation height, b root mat depth, c stem density, d egg abundance and e survival of eggs (after 28 days) between plots that were grazed (n = 40) and protected from grazing for 468 days (n = 40) at the Galaxias maculatus spawning site at Goughs Bay, Banks Peninsula on the east coast of New Zealand


Evidence for source–sink dynamics, even in well-studied species, is generally inferential because of the difficulty in acquiring accurate, unbiased, site-specific demographic data, particularly relating to births and migration (Pulliam 1988, 1996). For semelparous salmonids and G. maculatus, evidence for riverine source and sink populations is less onerous to achieve because there is no post-spawning emigration from rivers. A true riverine sink population will have highly compromised natality and few early life stage fish entering the marine realm. However, diadromous fishes pose other problems in source–sink dynamics because they are prone to multiple disturbances across ecosystem boundaries (McDowall 1993). For example, recent declines in some salmon populations along north-western America seem to be the result of oceanic influences on development (Wells et al. 2008) exacerbated by intensive fishing, altered waterways and degraded spawning habitats (Nehlsen et al. 1991). An interaction of multiple stressors can therefore affect a diadromous fish species by impeding or blocking sequential life-history pathways.

In our study, many of the rivers, especially the large ones, are effectively sink populations. We showed that egg production by G. maculatus has been severely compromised by reductions in the extent and quality of obligate spawning habitat. Stream bank topography, and the species composition, density and height of riparian vegetation greatly affected the density and survival of eggs. Several stressors, including channelisation, flooding, urbanisation and livestock grazing, are implicated in the reduction and deterioration of required riparian habitats. It is unlikely, however, that the intensity of these stressors is equal across different-sized waterways. Generally, larger rivers have been more severely modified by human uses. Frequent flooding in larger rivers has necessitated channelisation and coastal urban areas are often centred on large waterways. We showed that remediation of some of the stressors of riparian spawning habitats resulted in increased egg production and survival in localised areas. The cumulative effect of these stressors, however, is that many rivers have sink populations of G. maculatus that have little or no effective way of producing offspring for subsequent generations. There was an inverse correlation between the sizes of rivers and egg production, with most production limited to smaller rivers and streams. However, because large freshets and plumes produced by the major rivers are the largest cues for late-stage larvae to return from their oceanic larval pools (McDowall and Eldon 1980), it is the large rivers that attract the vast numbers of small fish that support the whitebait fishery. Taken together, the life history of G. maculatus encompasses a series of filters or gateways that can greatly affect population dynamics and feedbacks (Fig. 7), discussed below.
Fig. 7

Model of life stage critical habitats for Galaxias maculatus

Environmental gateways

G. maculatus larvae may disperse considerable distances from shore during their 100- to 200-day oceanic larval development (McDowall et al. 1994; Hickford and Schiel 2003). Nothing is known of mortality during oceanic development, but it is likely that most larvae die before they can complete this phase and return to coastal waters (Fig. 7). As juvenile G. maculatus return to freshwater from the oceanic environment, however, a significant source of mortality is whitebait fishing, which can leave relatively narrow channels available for upstream migration, even in the major waterways. For example, it is believed that the annual whitebait catch on the West Coast, New Zealand, peaked at 322,000 kg in 1955 (McDowall and Eldon 1980). With an average individual of 51 mm (fork length) and weight of 0.45 g, the whitebait fishery removes an enormous number of juvenile fish from the population.

There is no evidence of G. maculatus moving between rivers once they have entered a waterway and moved upstream, so the presence and quality of spawning habitat in each particular stream or river determines future egg production of that waterway (Fig. 7). Although some spawning may occur in marginal habitats, it is clear from this and other studies (Hickford et al. 2010) that egg survival in sub-optimal habitat is extremely low. Furthermore, it appears that G. maculatus repeatedly use the same spawning sites, which would further exacerbate the effects of localised impacts by drawing spawning aggregations back to degraded areas even when suitable riparian habitat exists nearby.

Attractive sinks

Because the density and extent of G. maculatus egg production in major waterways was generally very low in comparison to smaller, more pristine streams, and the major rivers produce the most extensive low-salinity plumes that attract whitebait navigation from the sea to river mouths (Grimes and Kingsford 1996), it appears that the most of the larger waterways are “deceptive sources” (Delibes et al. 2001b) where they have little chance of breeding successfully. There are no cues for juvenile G. maculatus entering a large river as to the likely availability of riparian spawning habitat when they mature 6–18 months later. In fact, given the temporal and spatial vulnerability of spawning habitat, it is possible that the spawning habitat may disappear after juveniles enter a river. Juvenile G. maculatus will enter these attractive sinks, showing maladaptive behaviour, if they perceive them to be sources.

The whitebait fishery is not a good indication of potential breeding populations if fishery-induced mortality, or mortality prior to maturation, is overwhelming. Escapement rates of juveniles in the whitebait fishery range from 81 to 91% (Allibone et al. 1999) with only around 20% of these surviving to spawn (Richardson et al. 2000), but the huge numbers of juveniles entering major rivers (McDowall and Eldon 1980) should still ensure a large potential breeding population. Despite this, however, we found little evidence of egg-laying in the Buller and none in the Mokihinui, even though there is evidence of very large numbers of G. maculatus juveniles entering both rivers: the whitebait catch in the Mokihinui in 2007 was estimated at c. 9,000 kg (Bush 2008); the Buller is deemed of national importance because of its whitebait fishery (Anon 2004). These and other large West Coast rivers with little spawning habitat are therefore major attractive sinks for returning juveniles (Fig. 7). Because spawning habitat is so restricted, often only a few square metres, there is a great potential for disruption of life cycles because small-scale localised impacts and large-scale disturbances can effectively create a sink by impeding or closing a critical life-history gateway (Fig. 7). It appears likely, therefore, that source populations are mainly smaller rivers and streams that have remained relatively pristine along their seaward riparian margins.

The dichotomy between the size of waterways and the number of G. maculatus eggs they produce, together with the return of large numbers of whitebait to rivers where there is little or no evidence of spawning, casts further doubt on the ability of G. maculatus to home to their natal river. We found no evidence in the largest rivers of the enormous number of eggs that would be necessary, after planktonic development and its associated high mortality, to produce the vast numbers of whitebait that enter these large rivers.

Among waterways, the overall G. maculatus population along a coastline is a combination of cumulative within-stream spawning success coupled with oceanic development, which undoubtedly vary among years. The extent to which the greatly different returns of the whitebait fishery over several decades (McDowall and Eldon 1980) is due to oceanic conditions or reduced spawning habitat is unknown. However, any recent declines in whitebait catches have coincided with large increases in intensified farming, urban development and channelisation that have negatively affected the coastal riparian margins of many streams and rivers. It is highly likely that reduced spawning will contribute substantially to lower fishery returns, and that this results primarily from a severe population bottleneck due to spawning habitat restrictions in most streams and rivers.

The cumulative effects of degraded habitats are known for a wide range of taxa across most ecosystems (Kappel 2005). If habitats are to be restored with the aim of rejuvenating populations, it is crucial to identify the key linkages and feedbacks between life histories and the environment. This study demonstrates that even a wide-ranging species with many robust adult populations can be compromised if a relatively small, stage-dependent habitat is required to complete a life history. This knowledge provides the necessary underpinnings for targeted rehabilitation and restoration.


We thank K. O’Connell, D. Taylor, K. Seaward, M. Møhl, S. Lilley and the Marine Ecology Research Group for assistance, Canterbury University for logistic support, and R. Warner for reviewing the manuscript. Thanks to R.M. McDowall and an anonymous reviewer for helpful comments. Thanks to the New Zealand Foundation for Research, Science and Technology (UOCX0502) and the A.W. Mellon Foundation of New York for funding.

Supplementary material

442_2010_1834_MOESM1_ESM.pdf (118 kb)
Supplementary material 1 (PDF 118 kb)

Copyright information

© Springer-Verlag 2010