RESEARCH REPOSITORY.

Transcription

1 RESEARCH REPOSITORY This is the author s final version of the work, as accepted for publication following peer review but without the publisher s layout or pagination. The definitive version is available at: Short, J., O'Neill, S. and Richards, Jacqueline D. (2017) Irruption and collapse of a population of pale field-rat (Rattus tunneyi) at Heirisson Prong, Shark Bay, Western Australia. Australian Mammalogy, In press. Copyright: Australian Mammal Society It is posted here for your personal use. No further distribution is permitted.

2 Pale field rat on HP_review.doc 18/06/ Irruption and collapse of a population of pale field-rat Rattus tunneyi at Heirisson Prong, Shark Bay, Western Australia Jeff Short A, B, C, Sally O Neill B and Jacqueline D. Richards A A CSIRO Sustainable Ecosystems, Wembley, WA, B Faculty of Sustainability, Environmental and Life Science, Murdoch University, South Street, Murdoch, WA, C Current address: Wildlife Research and Management Pty Ltd, P.O. Box 1360, Kalamunda, WA, jeff@wildliferesearchmanagement.com.au Abstract. Pale field-rats have long disappeared from Australia s arid and semiarid zones, other than for some Pilbara islands and a single mainland population of indeterminate status and extent identified at Shark Bay in Hence, it was noteworthy when a field-rat was first caught at Heirisson Prong in 1994, 40 kilometres north-east of the previous location at Shark Bay. Further individuals were caught regularly from late The population peaked in July to October 2000 (with captures of c. 190 individuals per month) and had collapsed by July 2001 (with only the occasional animal caught thereafter). None were caught beyond 2006, despite regular trapping to This irruption and collapse was beyond the established range of the species and was in atypical habitat. Widespread trapping post-collapse suggested that the population inhabited few localised source areas and a broad area of sink habitat, with the latter only occupied after extraordinarily high rainfall events leading to higher grass cover. A return to dry years and the consequent loss of cover (aided by an abundant rabbit population) and strong growth in predator numbers (feral cats and small birds of prey) in response to the high number of field-rats appears to have facilitated the collapse. Additional keywords: eruption, outbreak, native rodent, source-sink, refuge, Mus, Pseudomys, temporal synchrony Introduction The pale field-rat or Tunney s rat Rattus tunneyi is one of seven species of native Rattus that occur on mainland Australia. It is a small and docile rat, with short tail, protruding eyes in a broad, rounded head, and often of light yellow-brown colour (Watts and Aslin 1981). It commonly weighs from g. Braithwaite and Griffiths (1989) reported the species widespread decline across the arid and semi-arid zones of Australia and expressed concern for the species in the wet-dry tropics. They highlighted the paradox of Rattus tunneyi whereby the species was a significant agricultural pest in some parts of its range (chiefly southeastern Queensland), but had almost entirely disappeared from other parts, prompting conservation concern. This disparity of fortune was reflected in a recent action plan (Woinarski et al. 2014), where the western sub-species R. t. tunneyi was considered

3 2 Short,O'Neill and Richards near threatened (approaches A2abce), but the eastern sub-species R. t. culmorum was unlisted. R. tunneyi is now predominantly a species of grassland habitat across tropical and sub-tropical Australia from Broome in the west to northern New South Wales in the south-east. However early specimens and skeletal remains from owl pellets suggest that it was formerly widespread through much of arid Australia, and in Western Australia its range extended to the mesic south-west (with a specimen from Moore River in 1843: Morris 2000). The causes of the species widespread decline across this region have not been formally identified, but is believed to be as a result of loss of its preferred creek-line habitat to grazing by European rabbits Oryctolagus cuniculus and domestic stock (Aplin et al. 2008). In this paper we provide information on the dynamics of R. tunneyi from a subpopulation at Shark Bay in Western Australia the only known surviving population on the mainland of arid and semi-arid Australia. A small relict population was identified at False Entrance on the western coast of Edel Land on the south-western margin of Shark Bay in December 1968 (Kitchener and Vicker 1981). This population was separated from the nearest other known mainland population, some 1600 km to the north-east. The species occurs on many Pilbara islands, such as Legendre, West Lewis, and Weld Islands (Woinarski et al. 2014), considerably further south, but has not been recorded on the adjacent mainland. While a representative of a typically resilient genus, the Shark Bay population may be at risk. R. tunneyi (mean adult weight of 90 g) falls within the critical weight range ( g mean adult body weight; Burbidge and McKenzie 1989) that characterises mammal species most at risk of decline and extinction. The species is a ground dwelling herbivore and the population at Shark Bay occurs within the arid and semiarid zones; both factors associated with vulnerability to decline (Burbidge and McKenzie 1989). This study documents the irruption and subsequent crash of a population of R. tunneyi on Heirisson Prong, presumably linked by dispersal to the False Entrance population. The dynamic changes in R. tunneyi numbers on Heirisson Prong are compared with those of other Australian rodents displaying eruptive behaviour and juxtaposed against a range of environmental and biotic factors in an effort to establish likely

4 Pale field-rats at Shark Bay, Western Australia linkages. Knowledge of factors likely impacting the Heirisson Prong sub-population and the spatial structure of the population are used to derive management recommendations to assist the persistence of the wider population on Edel Land. Methods Study site Heirisson Prong (26.06 o S, o E) is a long, narrow peninsula that juts into Shark Bay from the south (Fig. 1). Its northern tip was fenced in 1989 in an attempt to exclude foxes Vulpes vulpes and feral cats Felis catus from a 1200 ha area. Heirisson Prong and the more westerly peninsulas of Steep Point and Bellefin Prong make up an area known as Edel Land. They were, at the time of survey, part of the 805 km 2 Carrarang pastoral station. Carrarang was considered poorly watered from a pastoral perspective (Payne et al. 1987), and hence Steep Point, Bellefin Prong and the northern part of Heirisson Prong were subject to low levels of grazing by stock. Feral goats were a problem in the far west of the area, particularly along the Zutydorp cliffs to c when control was implemented. The pastoral history of Carrarang dates to 1873, with sheep numbers in the greater district peaking in the 1920s and crashing to half that peak in the late 1930s due to drought, overstocking, and consequent land degradation (Payne et al. 1987). Carrarang was reported to carry a mean of 9,300 sheep units over the period , with a maximum of 15,800 in The recommended sheep unit capacity was 8000 and the bulk of the station was considered to be in good range condition in the early 1980s (Payne et al. 1987). The three peninsulas are considered part of a common geomorphic district known as Coastal dunes. These consist of coastal dunes and undulating plains of shallow calcareous sand over limestone or calcrete. Heirisson Prong falls largely within the Edel land system (undulating sandy plains with minor dunes and limestone rises); whereas the peninsulas of Steep Point and Bellefin Prong fall within the Coast land system (large linear and reticulate coastal dunes, with minor limestone rises and steep coastal cliffs) (Payne et al. 1987). The Coast land system is considered highly susceptible to wind erosion if vegetation cover is depleted, and at least some parts consist of large blowouts and mobile dunes (Fig. 1). There is a strong west-east rainfall gradient of annual rainfall across Shark Bay with c. 300 mm falling on Dirk Hartog Island and c. 230 mm at Denham on Peron Peninsula.

5 4 Short,O'Neill and Richards The climate of Edel Land is considered to be dry warm Mediterranean, while that of Peron Peninsula is considered semi-desert Mediterranean (Payne et al. 1987). Heirisson Prong is largely separated from the western peninsulas by a series of ponds that impound water for a local salt harvesting operation established in the early 1960s (Fig. 1). These extend 23 km south of the open water of Useless Inlet and would formerly have been a shallow tidal estuary or saline flats. Connections across these ponds is by a series of man-made bars up to 5 km long. A 22-km flume carries saline water from impoundments on the western side of the Heirisson Prong peninsula to ponds on the eastern side. This forms a barrier to movement of mammals and on occasions field-rats have been caught in the strongly flowing brine and collected in a sieve at the end of the flume. Dune and sandplain habitats on Heirisson Prong are covered with sparse low shrubland or heath. Common shrub species include Pittosporum phylliraeoides, Acacia tetragonophylla, Acacia ligulata, Melaleuca huegelii, M. aff. cardiophylla, Heterodendrum oleifolium, Atriplex bunburyana, Acanthocarpus preissii, Exocarpus apyllus, Pimelea microcephala, Scaevola crassifolia, S. spinescens, S. tomentosa, Stylobasium spathulatum, Frankenia pauciflora, Brassica tournefortii and Thryptomene baeckeacea. Grasses appear rare on Heirisson Prong (other than in herbivore exclosures) due primarily to the abundance of the rabbits in the relative absence of mammalian predators. Rabbits are abundant in the sand dune and sand plain habitats of Heirisson Prong, and numbers fluctuate greatly with runs of high and low rainfall years (Short et al. 1997; Robley et al. 2002). Abundance at a similar coastal location to the north of Shark Bay were found to be linked to rainfall with high rabbit numbers associated with years of above-average rainfall (King et al. 1983). Rabbits preferentially feed on perennial grasses on Heirisson Prong (Robley et al. 2001) and have a major impact on their biomass. Species favoured by rabbits are Bromus arenarius, Eragrostis barrelieri, and Austrostipa elegantissima (Robley et al. 2001). Spinifex longifolius is common to the south of the fenced peninsula, but is now rare to its north, most likely due to grazing by rabbits. Rabbits also have a major impact on Acacia shrubs, particularly A. ligulata. They climb into the canopy during dry times, defoliating and often killing them. They also suppress any subsequent regeneration. This impact on regeneration of Acacia has been reported widely elsewhere (e.g. Lange and Graham 1983).

6 Pale field-rats at Shark Bay, Western Australia Small mammal species on Heirisson Prong included the introduced house mouse Mus musculus, native ash-grey mouse Pseudomys albocinereus, sandy inland mouse P. hermannsburgensis, little long-tailed dunnart Sminthopsis dolichura, and reintroduced greater stick-nest rat Leporillus conditor. Euros Macropus robustus were present throughout Edel Land, including a small population on Heirisson Prong north of the fence. Feral cats and foxes, both likely predators of R. tunneyi, are widespread throughout Edel Land. While they were controlled within the fenced area of Heirisson Prong, regular incursions and some breeding of feral cats occurred (Short and Turner 2005, Short 2016). Incursions by foxes also occurred, but were less frequent and typically short-lived. Indices of abundance for each species were available from regular spotlight counts. Sightings of owls, another likely predator of field-rats, were recorded also during these spotlight surveys. Sightings of small diurnal birds of prey (those other than eagles and osprey Pandion haliaetus) were recorded between February 2000 and October 2003 as part of another study. Another likely predator of field-rats and a very common species on Heirisson Prong was the sand goanna Varanus gouldii. This species was active each year during the warmer months from September through to April. King brown snakes Pseudechis australis were also present, although not commonly seen. A yard system of 17 ha enclosed by a predator and rabbit-proof fence was located within the fenced peninsula. It was used for initial establishment of threatened species and formed an effective predator refuge. It typically had fewer rabbits due to control actions and hence higher levels of vegetation cover, including perennial grasses, than areas outside. Other sites south of the barrier fence were trapped in an effort to establish the source of R. tunneyi on Heirisson Prong. These sites were chosen based primarily on similarities to the known habitat requirements of the species in northern Australia (particularly dense understorey with grass and/or sedge species) as well as accessibility to the existing track system. One site trapped to the south of the barrier fence was on the outskirts of the town of Useless Loop. Potable water for the town was derived from bore water treated using a reverse osmosis (RO) plant and associated cooling tower at the base of a tall dune approximately one kilometre south

7 6 Short,O'Neill and Richards of the town. This leaked water into the surrounding area in the lee of the dune and allowed the growth of dense impenetrable shrubland to 3 m tall of Acacia ligulata and Alyogyne cuneiformis with areas of dense tall (to 1.5 m high) grass Sporobolus virginicus. The site was <2 ha in area and included a small dam of fresh water. The peninsulas of Steep Point and Bellefin Prong and the northern tip of Heirisson Prong were excised from Carrarang pastoral station and returned to the State Government in January 2008 (McCluskey 2008) to allow a tenure more in keeping with their role for conservation. These areas are proposed as the future Edel Land National Park. Trapping and animal measurement Rattus tunneyi were collected as by-catch in studies of the impact of feral cats on the small mammal and reptile faunas (Risbey et al. 2000; Richards and Short unpublished data) and the establishment of reintroduced threatened species (the marsupials Bettongia lesueur (Short et al. 1994; Short and Turner 2000) and Perameles bougainville (Richards and Short 2003; Short 2016) and the native rodents Pseudomys fieldi and L. conditor to Heirisson Prong. Data were derived from four sources: (i) trapping using grids of pitfall buckets of 20 l with low drift fences of flyscreen mesh (six grids of 16 buckets in two habitats: shrubland and heath) between February 1992 and May 2002; (ii) trapping of Heirisson Prong using small treadle-operated wire cages (Sheffield Wire, Welshpool, WA; 55 x 20 x 20 cm) set at 100 m intervals around some 40 km of track system at 3 to 6 month intervals between October 1994 and May 2013 (> 30,000 trap nights); (iii) trapping of the 17-hectare predator refuge using a combination of Elliott traps (30 x 10 x 10 cm) and small cage traps on a marked grid with trap spacing of 40 m between May 1994 and May 2005 (> 16,000 trap nights); and (iv) trapping of heath and shrubland habitat immediately beyond the predator refuge using Elliott traps and small cages (9 x 5 grid with trap spacing of 40 m) between June 1996 and November 2003 (> 2,500 trap nights). Cage traps were baited with rolled oats, peanut butter, and sardines, while Elliott traps were baited with either this mix or sunflower seeds. Traps were set just prior to sunset and checked at first light. When using Elliott traps, two traps were placed at each grid intersection. Some cage traps were typically set in conjunction with Elliott traps to reduce interference by burrowing bettongs. It is likely that pitfall buckets were only able to contain juvenile rats, so provide an indication of presence only.

8 Pale field-rats at Shark Bay, Western Australia Trapping specifically targeting R. tunneyi was conducted from 2001 to Fourteen sites were trapped on Heirisson Prong, south of the fence, in an effort to establish the possible source(s) of field-rats colonising the northern Prong (S. O Neill, unpublished data). Trapping took place between May 2001 and March Sites were trapped using 48 Elliott traps and 6 cage traps in a 40 x 50 m grid with a trap spacing of 10 m for 1 or 2 nights. Sites were located between 2 and 28 km south of the fence and were mainly located down the western flank of the peninsula (Fig. 2). Only one of these sites yielded field-rats. This area was trapped seven times between January 2002 and February 2004, typically using a grid of 9 x 9 Elliott traps covering 1.2 ha with traps opened for three consecutive nights. Measurements taken of R. tunneyi included location of capture, individual identification if marked, weight, sex, and hind foot length. Individuals were marked by ear punching. Not all individuals were measured on each trapping occasion. At times, particularly of high abundance, only total captures of R. tunneyi were noted to determine overall trap success. Reproductive status was not typically assessed. In some instances, hind foot measurements were taken, but not weights. These were subsequently estimated from a regression of weight (g) on hind foot length (mm) from all available data from Heirisson Prong. Condition indices were calculated using the method of Krebs and Singleton (1993). Data analysis Pearson correlation co-efficients were calculated for indices of abundance of R. tunneyi (trap success) and prior rainfall over various periods from 1 18 months, spotlight indices of rabbit and cat abundance, cat abundance with time lags of 2-7 months, spotlight counts of owls, and diurnal counts of small birds of prey. A paired t-test was used to compare capture rates inside and outside of the predator refuge matched for date of trapping. One way ANOVA with Fisher s least significant difference pair-wise comparison test was used to compare weight, hind foot and condition at different stages of the irruption. All analyses employed the software Minitab 15.

9 8 Short,O'Neill and Richards Results Population dynamics The first R. tunneyi caught on Heirisson Prong was in March 1994 (Fig. 3) during exploratory Elliott trapping to establish a release site for the Shark Bay mouse (Speldewinde 1996; cf. McKenzie et al. 2000). Another specimen was brought home by a domestic cat to a house in Useless Loop in June 1995 (B. Cane, pers. com.) and others were caught in town in June 1996, September 1998, June 2001 and January, April and October Carcases of several others were recovered from the northern end of the flume where it emptied into crystalliser ponds - in September and October 1995, in January 1996, and in September 1999 (3 records) (R. Gregory, pers. com.). Grids of pitfall traps were operated on Heirisson Prong north of the fence from February The first record of R. tunneyi in these traps was in July 1997 (Fig. 3) and one or two were caught regularly in most sessions from July 1998 through to February Field-rats caught in pitfall traps were typically juveniles ( = 22.9 g; n = 2) and it is likely that adults were able to either evade the pits or escape from them. The first R. tunneyi caught during regular cage trapping of the track system on Heirisson Prong was in May One or two individuals were caught in July and October 1996 and February 1997 (Fig. 3). However, in July 1997, six R. tunneyi were caught and they were caught on most trapping sessions thereafter through to October Captures were highly variable, ranging from a single capture through to 17 captures in January 1998 (1.8% trap success) and 12 captures in October 2000 (1.7%). Typically, a large number of traps would catch other species, limiting the number of traps available to catch field-rats. For example, in March 2000, 254 of 331 trap nights caught bettongs and bandicoots reducing the traps available to catch field-rats. R. tunneyi were first caught in the predator refuge in July 1997, when three were trapped (Fig. 4a). Numbers caught per trapping session remained relatively low (between one and nine captures) until March 1999 when they suddenly rapidly increased. They remained high (16 to 180 captures) before rapidly falling to two captures in July 2001 and a single capture in October Sixty field-rats (3.5 ha -1 ) were transferred from the predator refuge in August and September 1999 (Fig. 4a) to dense shrubland some 4 kilometres to the south-west to reduce potential competition with greater stick-nest rats shortly to be reintroduced to the predator refuge. Despite

10 Pale field-rats at Shark Bay, Western Australia this transfer, trap success continued to grow strongly in the refuge through the remainder of 1999 to >20%. In total, 375 of 505 captures of R. tunneyi north of the barrier fence (74%) were in the predator refuge. Shallow burrow systems were common within the predator refuge during the period of high field-rat numbers. The last records of the capture of a R. tunneyi on Heirisson Prong north of the barrier fence were in October 2005 (the first capture on standard trap lines in four years; 607 trap nights, 0.16% trap success; a male) and in November 2006 (434 trap nights, 0.23%; a single capture of a male). Hence, there appeared to be a dispersal phase characterised by low densities that extended from March 1994 to March 1999 (<5% trap success in predator refuge), followed by an irruptive phase from March 1999 to October 2000, with trap success rising to 24% in October This was followed by rapid decline in No R. tunneyi were caught on Heirisson Prong north of the barrier fence from October 2001 until single individuals were caught in October 2005 and November Only one site of 14 trapped south of the barrier fence yielded R. tunneyi. This was an area on the outskirts of the town of Useless Loop adjoining the reverse osmosis plant. R. tunneyi were present on all seven occasions this site was trapped from January 2002 to February Results for trap success are plotted in Figure 4a. Field-rats were also reported from town throughout 2002 (Fig. 3). Other small murids within the predator refuge showed highly variable patterns of abundance (Fig. 4b). Trap success over time was highly dynamic for the house mouse, with major peaks in May 1997 and May to October 1999 and lesser peaks in December 2001 and September The ash-grey mouse was in high numbers at the beginning of the survey period but collapsed in numbers thereafter and did not recover. In contrast, the sandy inland mouse occurred in low numbers at the start of survey and built in numbers through the survey period. The only species to show some temporal synchrony with field-rats was the house mouse in 1999, although that peak and collapse appeared to slightly proceed that for field-rats. Factors influencing population dynamics Key factors likely to affect the dynamics of R. tunneyi are plotted in Figure 5 and include rainfall, an index of rabbit numbers, trends in abundance of nocturnal and

11 10 Short,O'Neill and Richards diurnal birds of prey (total counts) and an index of abundance of feral cats, juxtaposed with changes in capture rates of R. tunneyi over time. Rainfall at Denham (22 km to the north-east) was below average for the years (178, 211, 171 mm respectively). This was followed by a period of above average rainfall in 1996 (303 mm) and (246, 259, and 384 mm) (Fig. 5a). High annual totals were largely driven by very high rainfall events in July 1996 (132 mm), July 1998 (121 mm), May 1999 (136 mm), and March 2000 (208 mm). Annual rainfalls in were all well below average (177, 160, 129 mm). Trap success of R. tunneyi on Heirisson Prong north of the barrier fence was significantly correlated with rainfall over the previous 12 months (r = 0.554; P < 0.001) and 18 months (r = 0.451; P = 0.001) (Table 1). Rainfall for the 12 month period prior to the October December 2000 peak in field-rat numbers was 350 mm, exceeding the 90 th percentile of annual rainfall (348 mm). There were no significant correlations between R. tunneyi numbers and rainfall over lesser time periods (1 9 months prior). The increase in capture rates of R. tunneyi (Fig. 5c) occurred during a period of sustained low rabbit numbers following several years after a major irruption of rabbits that peaked in late 1997 (Fig. 5b). High rabbit numbers had a major impact on the vegetation, likely reducing the amount of cover and food available to rodents. The spotlight index of rabbit abundance was not correlated with the abundance of R. tunneyi (Table 1). Owls (frogmouths, boobooks, and barn owls), small diurnal birds of prey (kestrels, kites, falcons, goshawks, and harriers) and feral cats all showed increases in number coinciding with the major peak in R. tunneyi numbers in July to October 2000 (Fig. 5d, e). Owls peaked in July 2000, while small birds of prey peaked in February to October Feral cats peaked in abundance in December 2000 and March 2001, 3-6 months after peak field-rat numbers (Fig. 5e). This was reflected in a significant correlation between abundance of R. tunneyi and the cat index lagged by 5 months (Table 1). Field-rat abundance was also significantly correlated with a spotlight count of owls to the south of the barrier fence (perhaps suggesting the eruption of field-rats extended to this area) and with diurnal counts of small birds of prey. There was a significant difference in capture success for R. tunneyi inside the predator refuge as compared with that immediately outside (t = 2.76; p = 0.025; Fig. 4a). Data

12 Pale field-rats at Shark Bay, Western Australia are compared across nine time periods between July 1997 and November Mean capture success was 8.9% inside the refuge; 2.8% outside. The difference likely reflects the impact of cat predation outside the predator refuge in combination with less cover due to higher rabbit numbers. Biology Captures of R. tunneyi across all trap types (excluding trapping at the reverse osmosis plant) gave a male-dominant sex ratio (1.31M: 1F, n = 386). Weights ranged from 11 to 147 g, with a mean of 79.8 g (n = 371). Male weights were significantly greater than that of females (85.3 g cf g; F1, 351 = 16.93; P < 0.001). If weight at sexual maturity is taken as 60 g for females and 90 g for males (Braithwaite and Griffiths 1996), then 82.2% of females were adults and 43.9% of males. If the threshold for adulthood in males is lowered to 84 g (based on a hind foot measurement of 27.5 mm (Taylor and Horner 1973) and a regression of weight v. hind foot for Heirisson Prong data) then 56.6% of males were adults. The proportion of juveniles (M < 84 g; F < 60 g) in the population was 60% in 1996 and 1997 (n = 3 and 15), declining to 44% in 1998 (n = 34), 22% in 1999 (n = 196) and 17% in 2000 (n = 108) as population numbers rose, suggesting emigration of juveniles at higher densities. The sex ratio of R. tunneyi caught beyond the predator refuge had a strong male bias at 2.18M: 1F (n = 108) and a slightly lower mean weight (76.7 g; n = 92) compared to within the refuge (1.09M: 1F; 80.8 g (n = 279)). The sex ratios were significantly different (χ 2 1= 8.48; P = 0.004), while the difference in weights with location approached significance (F1, 351 = 3.73; P = 0.054). There was no significant interaction of location with gender for weight (P = 0.272). The proportion of juveniles (M < 84 g; F < 60 g) in the trappable population beyond the predator refuge was 0.45; compared with 0.28 in the predator refuge. Condition of R. tunneyi were calculated from the regression of weight on hind foot for all available captures on Heirisson Prong (weight (g) = *hind foot (mm); (r 2 = 23.8%; F1, 330 = , P< 0.05)). The condition of R. tunneyi was significantly better during the irruption than in the periods before and after the irruption (Table 2). This was driven largely by a change in mean weight.

13 12 Short,O'Neill and Richards Another significant difference evident across the irruption was in sex ratio. The preirruptive phase was characterised by a high proportion of males relative to other phases (Table 2). In contrast to that of Heirisson Prong, the sex ratio of R. tunneyi at the habitat refuge south of the barrier fence (reverse osmosis plant) was 0.92M: 1F (n = 71) and the proportion of juveniles captured was just 7.1%. Discussion Population dynamics and biology of pale field-rats on Heirisson Prong This study has documented an ephemeral population of R. tunneyi on Heirisson Prong, presumably linked by occasional dispersal from a small habitat refuge on Heirisson Prong south of the barrier fence and other populations on the two most westerly peninsulas of Edel Land (S. O Neill, unpublished data). The dense vegetation of the habitat adjoining the reverse osmosis plant at Useless Loop was perceived to be a key habitat refuge and source habitat (sensu Pulliam 1988), likely supplying field-rats to other sites on Heirisson Prong. Similarly, R. tunneyi occupied sites on Edel Land that were perceived to be of high quality and likely to be source sites (S. O Neill, unpublished data). The bulk of Heirisson Prong was perceived to be a sink habitat, suitable for occupation only in favourable seasons. R. tunneyi were first detected on Heirisson Prong in 1994 and appeared to peak in the area north of the barrier fence in October 2000 before rapidly collapsing. Early records and the subsequent irruption appeared to be triggered by high annual rainfall in the years 1996, 1998, 1999, and 2000 (Fig. 5a). These years typically included at least one exceptionally high rainfall pulse of > 120 mm per month, greater than four times the long-term median monthly rainfall for that month. The irruption of R. tunneyi on Heirisson Prong coincided also with years when rabbit numbers remained low (Fig. 5b). Rabbits peaked at very high levels in 1997, but remained low thereafter until spring The combined impacts of high rainfall and low rabbit numbers were high vegetation density, including perennial grasses. In a 2.5 year study on Heirisson Prong, Robley et al. (2002) found a 5-fold variation in percentage plant cover. Cover peaked shortly after a wet winter in 1996 (when winter rainfall was double the long-term mean) and declined thereafter through 1997 and the first half of 1998, a period of below average rainfall and high rabbit numbers.

14 Pale field-rats at Shark Bay, Western Australia The rapid collapse of the R. tunneyi population on Heirisson Prong north of the barrier fence (to a single capture in October 2001) occurred equally inside and outside the predator refuge, suggesting that it was primarily driven by a decline in available food or food quality over late summer and autumn This was borne out by the poor condition of field-rats in the post-irruption phase. The habitat at this time was perceived as largely unsuitable for R. tunneyi, due to decreasing cover and lack of green, growing grasses in the understory. Essentially, this was a sink habitat suitable only in years of favourable rainfall coinciding with low rabbit numbers. Feral cats were present on Heirisson Prong growing in numbers in spring 1999 and peaking (as assessed by spotlight index) in December 2000 closely matching the growth in the field-rat population (and a near-synchronous peak in Mus). They are likely to have hastened the subsequent crash in the field-rat population, but were not perceived to be primarily responsible as R. tunneyi within the predator refuge were not subject to this predation. In contrast, the population south of the fence in the vicinity of Useless Loop s reverse osmosis plant, persisted through to at least This was core refuge habitat possessing dense vegetation, both thick shrubland and dense high grass, providing cover from predation, available free-water, and extensive areas of grass exposed to leaking water providing year-round green feed. Rabbits were typically abundant in and around this site and cat sightings were relatively frequent, but did not apparently impact on persistence of field-rats at the site due to the dense cover. Only limited information is available on the age of this population. R. tunneyi were caught nearby in Useless Loop in 1995, 1996 and 1998, and likely originated from the population adjoining the reverse osmosis plant. Carcases of several other R. tunneyi were recovered from the northern end of the flume where it empties into crystalliser ponds in 1995, 1996, and in 1999, suggesting a source of animals further to the south. However, it is unknown whether this was from a population on the southern part of Heirisson Prong or from the more westerly peninsulas of Bellefin Prong and Steep Point/False Entrance via one of the man-made bars crossing the salt ponds. There appeared to be major differences in the makeup and dynamics of populations in the two refuges and in the wider population. The habitat refuge consisted largely of adults and an approximate equal sex ratio of males and females. Similarly, the

15 14 Short,O'Neill and Richards population in the predator refuge had a sex ratio near parity and juveniles were a lower proportion of the trappable population. In contrast, captures of R. tunneyi caught beyond either refuge were typically male biased and made up of a high proportion of juveniles. Hence there appeared to be stable adult populations in the two refuges, largely exporting their young to less favourable habitats beyond. The habitat refuge at the reverse osmosis plant was subject to constant resource inputs over time and likely maintained a relatively stable population. In contrast, shelter and food in the predator refuge were influenced by rainfall and varied sharply over time. This was reflected in the change in body condition and number of R. tunneyi across the period of the irruption and subsequent collapse. Irruptions of Rattus and other murids Rodent irruptions or plagues have long been a feature of arid Australia (Newsome and Corbett 1975) and have been documented for a range of rodent species following exceptional rainfall (e.g. Newsome and Corbett 1975 for M. musculus, Rattus villossisimus and Notomys alexis; Predavec and Dickman 1994 for R. villosissimus; Predavec 1994 for P. hermannsburgensis and N. alexis; Dickman et al for N. alexis, P. hermannsburgensis, P. desertor and M. musculus and Greenville et al for R. villosisimus). Rodents typically spiked in abundance at from 2-9 months following significant rainfall. Finlayson (1935, 1939) linked such increases to vegetative flushes caused by droughtbreaking rains and flooding river channels. This was reiterated by Newsome and Corbett (1975), who attributed rodent plagues to great and continuing rains locally, or flood waters hundreds of kilometres away producing spectacular flushes of herbage greatly increasing food supply for rodents. Source habitats (where rats were present during non-irruptive periods) for R. villosissimus were identified as boredrains, reedy springs and other mesic sites (Parker 1973, Watts and Aslin 1974). Such source habitats may be man-made: such as the feed shed on the cattle station at Brunette Downs in the Northern Territory, which appeared to be a significant refuge supplying abundant food and water to a large population of R. villosissimus (Carstairs 1976). Hence the source (localised mesic refuges) and trigger (exceptional rainfall) for irruption of the R. tunneyi population on Heirisson Prong appears similar in kind to that of R. villosissimus.

16 Pale field-rats at Shark Bay, Western Australia Korpimäki et al. (2004) contended that high rates of population increase in small mammals, leading to either cycles or outbreaks (irruptions), were driven largely by improvements in survival rather than in reproduction. In the case of R. tunneyi on Heirisson Prong, high inputs of rainfall likely increase the level of cover in sink habitat reducing the vulnerability of field-rats to predation, but also greatly increasing the spatial extent and amount of food resources. Aplin et al. (2008) suggested that R. tunneyi may have exhibited extreme fluctuations in abundance in response to climate variability in the drier part of its range, similar to that documented for R. villosissimus. The study on Heirisson Prong bears this out. Braithwaite and Griffiths (1996) considered R. tunneyi to be a poor coloniser relative to many other rodent species, although with a high reproductive rate (for example, ten teats and able to breed all year round in riparian habitat). Watts and Aslin (1981) reported the species to have an intermediate breeding potential; less than R. sordidus, R. colletti and R. villosissimus, but greater than R. leucopus. R. tunneyi is reported to have a litter size of 2-11 in the wild (Watts and Aslin 1981: 238) and in captivity can have litters in quick succession. Young attain sexual maturity as early as five weeks of age. The collapse of a rodent population following irruption is typically attributed to either depletion of food resources, to predation, or to a combination of the two. The poorer condition of R. tunneyi on Heirisson Prong during the decline phase suggests a food shortage related either to overshooting their food supply or to declining seasonal conditions impacting on food quality and supply as a key cause. This is consistent with Pavey et al. (2008) s conclusion that the decline of rodents after irruption is in part due to starvation resulting from declines in food availability, but at variance with that of Korpimäki et al. (2004) who saw food shortage acting primarily to stop population increases as the population approached peak numbers, rather than being responsible for the subsequent decline. Predators, including feral cats, are reported to increase in number during plagues of R. villossismus (Carstairs 1974, Newsome and Corbett 1975) and to speed population decline following irruptions (Newsome and Corbett 1975). Recent experimental work in tropical savanna has demonstrated the vulnerability of isolated concentrations of R. villosissimus to predation by feral cats (Frank et al. 2014). Korpimäki et al. (2004) viewed predation as primarily responsible for declines of small mammals following

17 16 Short,O'Neill and Richards outbreaks or peaks in cycles. Feral cats, in combination with native predators, appear to have played a role in declines on Heirisson Prong. The population of feral cats north of the barrier fence escaped control (Short and Turner 2005) and peaked contemporaneously with that of M. musculus and R. tunneyi, likely hastening their decline. An indication of the impact of feral cats on R. tunneyi on Heirisson Prong can be gauged from the more than three-fold difference in capture success on trapping grids inside the predator refuge to those immediately outside. There was some indication also of a localised increase in both nocturnal and diurnal birds of prey at times when R. tunneyi were peaking, with observations of nocturnal birds of prey particularly located at the predator refuge. The sand goanna, also, was conspicuous and abundant on Heirisson Prong and likely to have been a significant predator of R. tunneyi and other small mammals. Both birds and reptiles, unlike feral cats, would have acted on R. tunneyi both inside and outside the predator refuge. Distribution of pale field-rats Rattus tunneyi were recorded as present at Edel Land in the early 1970s (Kitchener and Vicker 1981), but as locally extinct on Peron Peninsula and the mainland to the east of Shark Bay (McKenzie et al. 2000). No sub-fossils were found on the islands of Dirk Hartog, Bernier, or Dorre (Baynes 1990, 2006, 2007). Sub-fossils of field-rats were found on Peron Peninsula in numerous small coastal caves and dune blowouts (Baynes 2007) and this species was regarded as forming part of the original mammal fauna (i.e. immediately pre-european) of this eastern part of Shark Bay. One adult female R. tunneyi was trapped in dense Spinifex longifolius on a sand ridge near Steep Point on Edel Land during widespread trapping using both pitfall traps and Elliott traps in 1989 (Sanders and Harold 1990). They trapped widely at 17 sites across the Steep Point, Bellefin, and Heirisson peninsulas for this single capture. In a later survey, two R. tunneyi were trapped with a single R. rattus at False Entrance (407 trap nights with Elliott traps) in March 1996 (A. Robley pers. com. 1996). Similar trapping (545 trap nights) at the same time on the southern part of Heirisson Prong along the salt flume yielded no Rattus. R. tunneyi were trapped also between 2002 and 2004 at numerous sites on the Steep Point and Bellefin peninsulas (S. O Neill, unpubl. data). These were in localised areas linked to fresh water seepage from tall dunes and were perceived to be important refuge sites for R. tunneyi.

18 Pale field-rats at Shark Bay, Western Australia The population of R. tunneyi on Edel Land occurs as a significant outlier to any other current mainland populations of the species. The nearest extant mainland population is 1600 km to the north in the Kimberley. R. tunneyi were previously distributed through much of continental Australia, including the south-west and arid regions of Western Australia (Braithwaite and Griffiths 1996). Fossil records are widespread in central Western Australia (see Fig. 1 of Braithwaite and Griffiths (1996); also McKenzie et al. (2000)). The species is considered extinct in the arid Northern Territory (that area receiving < 600 mm of annual rainfall) and in South Australia (Cole and Woinarski 2000; Robinson et al. 2000). However, the species still persists in the more mesic parts of its range. R. tunneyi has been trapped in a variety of mostly mesic or dense grassy habitats in the Mitchell Plateau, Kimberley in the wet-dry tropics, variously described as plateau swamp, riparian situation fringing a creek with low woodland and dense tall grass, mixed eucalypt open forest over tall grass, and woodlands with dense grasses to 2 m high in understorey (Bradley et al. 1987). Concern has been expressed for the species in Kakadu National Park in the Northern Territory, where capture rates suggested a 500-fold decline across two studies conducted in the late 1980s and early 1990s (Braithwaite and Griffiths 1996). They attributed this decline to a series of dry years that resulted in the depletion of aquifers and the consequent impact on seasonal creeks that formed important habitat for the species. Creeks dried up early in the dry season of each year in the early 1990s, in contrast to earlier years when waterholes were permanent. Refuge habitat: requirements, threatening processes, and conservation The importance of localised mesic refuge sites for the species is likely linked to their food requirements. R. tunneyi appear to specialise on nutritious food sources highly dependent on the availability of moisture (Braithwaite and Griffith 1996). Hence their preference for riparian habitat. They are known to favour stems and leaves of monocotyledonous plants (Watts 1977), although they are not adapted to a highly abrasive herbivorous diet (Braithwaite and Griffiths 1996). In the tropical savanna they favoured a small number of grasses, Sorghum and Alloteropsis. Watts and Aslin (1981: 20) considered they fed on a substantial amount of grass, eating mainly grass stems, seeds, and roots and were essentially herbivores. This was true also for Shark

19 18 Short,O'Neill and Richards Bay, where monocotyledons made up a large proportion of the diet at all times examined (S. O Neill unpublished data). R. villosissimus was only able to survive for 7-22 days without green vegetation or water (Baverstock 1976) and it is likely that R. tunneyi may have similar or more demanding water requirements. Braithwaite and Muller (1997) emphasised the role of exotic herbivores (rabbits, sheep and cattle) in impacting refuge habitat. In addition to reducing the availability of cover and food, hard-hooved domestic stock, including feral goats, may impact R. tunneyi through the trampling of their shallow burrows. Feral and domestic goats are the major factors likely to impact refuge soaks in Edel Land. Rabbits, while present at many sites where R. tunneyi were trapped on Bellefin Prong and Steep Point, were far less abundant than on Heirisson Prong (S. O Neill, pers. obs.). However, rabbits fed on grasses on Heirisson Prong and had a major impact on their density, affecting the temporary suitability of sink habitat also. In addition, the reduction in vegetation cover by exotic herbivores is likely to facilitate higher levels of predation from both cursorial and avian predators, impacting field-rats in both refuge (source) and sink habitat. Pulliam (1988) highlighted the conservation consequences of populations with sourcesink dynamics. Much of the population might occur in sink habitats of broad area, dependent on a relatively small source population in a different habitat. This might lead to the conclusion that the destruction of the latter habitat (by virtue of its small size) would be inconsequential for the persistence of the species. This situation appears to reflect the status of R. tunneyi on Heirisson Prong. The habitat refuge at the reverse osmosis plant was surveyed for Rattus in 2014 (camera traps, analysis of cat scats and owl pellets: Palmer and Morris 2014) and trapped in June 2015, without detecting or capturing any field-rats. It is likely that the resident population may have been poisoned during intensive rabbit control immediately around the town of Useless Loop over the previous several years using bait stations of oats coated with the poison pindone. R. tunneyi is likely to have a low tolerance to pindone (another native R. fuscipes has an LD50 of 2-8 mg/kg: Twigg et al. 1999) and has a low tolerance to the other major poison 1080 used for rabbit control (LD50 of 3-5 mg kg -1 : Twigg et al. 2003). This highlights a management dilemma, in that while rabbits are likely to be a key adverse factor for persistence of R. tunneyi, management of rabbits by poisoning may have greater adverse impacts.

20 Pale field-rats at Shark Bay, Western Australia The current status of the species on the more westerly peninsulas of Edel Land is unknown. Further survey is required to clarify their status, identify key habitat refuges, and document changes in distribution and abundance in response to rainfall. Management should be directed at maintaining high vegetation cover by excluding goats and reducing the incidence and extent of fire. Acknowledgements We thank Bruce Turner for assistance with data collection, Bryan Cane for observations of domestic cat predation on Rattus tunneyi as well as his role in community leadership at Useless Loop, community members from Useless Loop, Shark Bay Salt Joint Venture (now Shark Bay Resources) for its logistic support, and Earthwatch volunteers and Earthwatch Australia for field assistance. Keith Morris, Bruce Turner and two anonymous referees commented on an earlier draft of this manuscript. References Aplin, K. P., Braithwaite, R. W., and Baverstock, P. R. (2008). Pale Field-rat Rattus tunneyi (Thomas, 1904). In 'The Mammals of Australia'. (Eds. S. Van Dyck and R. Strahan.) Pp (Reed New Holland: Australia.) Baverstock, P. R. (1976). Water balance and kidney function in four species of Rattus from ecologically diverse environments. Australian Journal of Zoology 24, Baynes, A. (1990). The mammals of Shark Bay, Western Australia. In 'Research in Shark Bay. Report of the France-Australe Bicentenary Expedition Committee'. (Eds. P. F Berry, S. D. Bradshaw, and B. R. Wilson.) Pp (Western Australian Museum: Perth, Western Australia.) Baynes, A. (2006). The original mammal fauna of Dirk Hartog Island: results from field work in Unpublished report to the Department of Environment and Conservation, Perth, WA. 6 pp. Baynes, A. (2007). Final report on an investigation of the original non-volant mammal fauna of Nanga Station, Shark Bay. Unpublished report to the Department of Environment and Conservation, Perth, WA. 9 pp. Bradley, A. J., Kemper, C. M., Kitchener, D. J., Humphreys, W. F., How, R. A. (1987). Small mammals of the Mitchell Plateau area, Kimberley, Western Australia. Australian Wildlife Research 14, Braithwaite, R. W. and Griffiths, A. D. (1996). The paradox of Rattus tunneyi: endangerment of a native pest. Wildlife Research 23, Braithwaite, R. W. and Muller, W. (1997). Rainfall, groundwater and refuges: predicting extinctions of Australian tropical mammal species. Australian Journal of Ecology 22, Burbidge, A.A. and McKenzie, N.L. (1989). Patterns in the modern decline of Western Australia's vertebrate fauna: causes and conservation implications. Biological Conservation 50, Carstairs, J. L. (1974). The distribution of Rattus villosissimus (Waite) during plague and non-plague years. Australian Wildlife Research 1, Carstairs, J. L. (1976). Population dynamics and movements of Rattus villosissimus (Waite) during the plague at Brunette Downs, N.T. Australian Wildlife Research 3, 1-9. Cole, J. R. and Woinarski, J. C. Z. (2000). Rodents of the arid Northern Territory: conservation status and distribution. Wildlife Research 27,