Access the full text.
Sign up today, get DeepDyve free for 14 days.
Loading next page...
/lp/wiley/assessing-the-effectiveness-of-long-term-monitoring-of-the-broad-TGaqHl38fP
- Publisher
- Wiley
- Copyright
- Copyright © 2023 Ecological Society of Australia and John Wiley & Sons Australia, Ltd
- ISSN
- 1442-7001
- eISSN
- 1442-8903
- DOI
- 10.1111/emr.12574
- Publisher site
- See Article on Publisher Site
Abstract
Implications for ManagersNineteen years of existing monitoring data are insufficient to determine whether the Barrington Tops National Park Broad‐toothed Rat population has increased, decreased, or remained constant.Scat surveys may provide a cost‐effective means of monitoring the Broad‐toothed Rat, but further work is needed to determine if the detection of scats within the Barrington Tops is near‐perfect and constant over space and time.If detection is confirmed as near‐perfect and has no spatio‐temporal variation under all conditions, then naïve occupancy is an appropriate method, otherwise, occupancy models provide an alternative.IntroductionBiodiversity monitoring is critical for ascertaining the status and trends of biodiversity, and the success of conservation actions (Lindenmayer & Likens 2009). Despite this, monitoring programs are often poorly implemented (Vos et al. 2000; Legg & Nagy 2006; Lindenmayer et al. 2010; Robinson et al. 2018) due to insufficient funding; short‐term planning; indecision over what to monitor; as well as inappropriate methodology and inadequate training (Lindenmayer et al. 2010; Legge & Fleming 2012). Likewise, there are many features common to good quality monitoring programs: well‐defined aims (Lindenmayer et al. 2010, 2012); potentially set using specific, measurable, assignable, realistic, and time‐related targets (Doran 1981); long‐term commitment and transparent data storage (Lindenmayer et al. 2010; Legge & Fleming 2012; Robinson et al. 2018); maintaining data collection standards (Lindenmayer et al. 2012) and incorporating a measure of error that reflects data accuracy (Prowse et al. 2021). Moreover, it is important to ensure a good experimental design that uses recent innovations and produces robust outcomes (Lindenmayer et al. 2010, 2012). Lindenmayer and Likens (2009) proposed a system of adaptive monitoring in which managers regularly repeat question setting, data collection, analysis, and interpretation; which allows them to repeatedly review their program and, if necessary, make appropriate alterations.To support an adaptive monitoring approach, we analysed a range of variables collected by the long‐term monitoring of the Broad‐toothed Rat (Mastacomys fuscus) in the Barrington Tops National Park (henceforth referred to as the Barrington Tops). The Broad‐toothed Rat is a native Australian rodent listed as Near Threatened by the International Union for Conservation of Nature (IUCN, https://www.iucnredlist.org/species/18563/22429430). During the Pleistocene glaciation, this species occurred throughout south‐east Australia (Cherubin et al. 2019; Eldridge et al. 2020). Today, the species is found in disconnected populations in eastern Australia between Tasmania and the Barrington Tops (Dickman & McKechnie 1985; Green 2003; Beilharz & Whisson 2016). This patchy distribution is due to strict habitat requirements: mean annual temperature <10°C and mean average rainfall >1000 mm (Green 2003).We examined patterns of population change across time using the range of monitoring variables collected within the long‐term monitoring of the Broad‐toothed Rat in the Barrington Tops. We aimed to determine whether these patterns could be predicted by climatic variables and whether patterns of change amongst variables were concordant. Specifically, we determined how naïve occupancy based on scat surveys and an index of relative abundance produced from trapping surveys changed over time, as well as with minimum temperature and precipitation. Naïve occupancy refers to a measure of occupancy that does not account for imperfect detection (Guillera‐Arroita et al. 2014). Second, we tested for a correlation between the index of relative abundance and latency to find scats. Finally, we re‐analysed the trapping data using a multi‐season occupancy model. Given the numerous threats faced by the Barrington Tops Broad‐toothed Rat, we hypothesised that the population has likely declined since 1999 (Green & Osborne 2003; Green et al. 2008; O'Brien et al. 2008). Additionally, given that the Broad‐toothed Rat requires cold temperatures and high precipitation (Green 2003; Shipway et al. 2020), we hypothesised that both temperature and precipitation would influence Broad‐toothed Rat populations. Following authors Hayward et al. (2015), Bled et al. (2013), Guillera‐Arroita et al. (2014), we refer to methods that account for imperfect detection as ‘robust’ methods.MethodsStudy areaThe Barrington Tops is located ~100 km north‐west of Newcastle in New South Wales (NSW) and together with the neighbouring Mount Royal National Park covers ~67,530 ha (Zoete 2000). The region is a Key Biodiversity Area with many species, such as the Broad‐toothed Rat, found only in high‐altitude regions (Key Biodiversity Areas Partnership 2022). The Barrington Tops were historically used as a meeting place for the Traditional Owners of the surrounding lowland area: the Worimi, Biripi, Gringai and Wonnarua Nations. The south‐east of Barrington Tops National Park is centred around the Barrington Tops plateau, ~1500 m above sea level, which has an average annual precipitation of ~2000 mm (Zoete 2000). In lower altitude areas surrounding the plateau, the annual mean maximum temperature is ~23–24°C and the minimum is ~10–11°C. Temperatures on the plateau are usually around 5–7°C lower (Zoete 2000). There is a series of sub‐alpine sphagnum swamps scattered throughout the plateau where the Broad‐toothed Rat occurs (Dickman & McKechnie 1985; O'Brien et al. 2008; Fig. 1). The vegetation is dominated by Orites, Cyperus and Hakea species as well as Spiny‐head Mat‐rush (Lomandra longifolia), Restio stenocoleus, Sedges and Sphagnum.1FigureBroad‐toothed Rat distribution throughout the Barrington Tops National Park, Australia. Red dots represent sites used in the long‐term Broad‐toothed Rat monitoring program. Black lines represent the boundaries of the Barrington Tops National Park.These swamps support a functioning but fragile metapopulation of the Broad‐toothed Rat (O'Brien et al. 2008): namely, ‘a set of local populations within some larger area, where typically migration from one local patch to at least some other patches is possible’ (Hanski & Simberloff 1997) and the processes of colonisation and extinction operate independently within each local population. Local extinctions are more likely in isolated swamps, due to the larger distances individuals must travel to colonise and rescue these populations (O'Brien et al. 2008). The metapopulation is threatened by climate change, in addition to invasive fauna including Feral Horses (Equus caballus); Pigs (Sus scrofa); Cats (Felis catus); Red Foxes (Vulpes vulpes; Green & Osborne 1981; Green 2002; Milner et al. 2015). Invasive flora including Scotch Broom (Cytisus scoparius) and Yorkshire Fog (Holcus lanatus) is also an issue (O'Brien et al. 2008). Additionally, they may be threatened by competition with the Swamp Rat (Rattus lutreolus; Green et al. 2008).MonitoringThe Broad‐toothed Rat metapopulation in the Barrington Tops has been monitored by semi‐annual surveys between 1999 and 2018. Survey results have been documented in unpublished reports of the New South Wales National Parks and Wildlife Service (Green, K., unpubl. data, 2000; NPWS, unpubl. data, 2007; NPWS, unpubl. data, 2008; NPWS, unpubl. data, 2009; NPWS, unpubl. data, 2010; NPWS, unpubl. data, 2011; Schroder, M. and Fawcett, A. unpubl. data, 2018; Keating, J. unpubl. data, 2002; Keating, J. unpubl. data 2003; Green 2003), as well as by O'Brien et al. (2008). Monitoring was initiated as part of the NSW Fox Threat Abatement Plan (TAP) program (NPWS, unpubl. data, 2001) using the methodology and sites established by Green, K., unpubl. data (2000), implemented by local staff. At the time of his surveys in 1999, Green was the primary expert on this species, hence why his site selection and methods were adopted. The methods included scat surveys to determine naïve occupancy, and live trapping surveys to create an index of relative abundance. Monitoring sites were selected by managers of the species in Kosciusko National Park and similarity with the species' habitat in that region was used as a basis of site selection in the Barrington Tops. Although 27 sites were included in the program, not all sites were surveyed annually due to weather and access issues. Note there is an overlap between the sites used in this study and that of O'Brien et al. (2008), but we only analysed data from 1999, after the monitoring program was formally established (Supporting Information, Appendix A, Table A1).Scat surveys were conducted in 27 swamps at Barrington Tops. Broad‐toothed Rats produce 200 to 400 scats per day, which have a uniquely distinctive green, fibrous appearance when fresh, and are easy to distinguish from those of other species (Happold 1989; Schulz et al. 2019). This makes scat surveys an ideal way to determine Broad‐toothed Rat presence. Methods involved a varying number of observers spending 10 min searching for scats—if scats were found, a population was considered present at the site, and if no scats were found after 10 min, then the Broad‐toothed Rat was assumed absent. When conducting scat surveys, managers also recorded latency to first finding scats: that is, the time (in seconds) from commencing the search until the first scat was found. The age of scats was not recorded. Surveys were repeated annually but there were no repeat visits within years.Monitoring also involved live‐trapping surveys. Trapping was conducted in 20 sites: Barrington River (trapped from 2002–2004); Cronje Swamp (2004 and 2006); Edwards Creek (2004, 2006 and 2010); Edwards Swamp (2003, 2004, 2006 and 2010); Edwards Swamp North (2010); Edwards Swamp West (2010); Gloucester Falls Picnic Area (2009 and 2010); Junction Pools (1999, 2002 and 2004); Kerripit Swamp (2009 and 2010); Little Murray East (1999, 2002–2004 and 2006); Little Murray West (2004 and 2006); Nolans Swamp (2002–2004 and 2006–2018); Polblue Creek walkway (2004); Polblue Picnic Area (2004, 2006–2018); Polblue Swamp Campground (1999, 2004, 2006–2018); Saxby Creek (2004); Upper Barrington River (2004); Wallaby Hill (2004; 2006, 2007 and 2010), Watergauge Trail Creek (2002–2004 and 2006–2018) and Watergauge Trail Gate (1999, 2002–2004 and 2006–2017). One trapping session was conducted each year except for 2004 when there were two trapping sessions. There have been 214 captures since 1999 (mean of 2.10 per site each year ±0.33 standard error). Overall, there was a ratio of 2.09 Broad‐toothed Rat captures to every 1.79 Swamp Rat captures and no significant change in the number of Swamp Rat captures over time (linear regression p‐value = 0.839, r2 = 0.00, n = 102). No Broad‐toothed Rats were captured from 2016 until the end of the project in 2018. In most years, individuals were microchipped so they could be identified if recaptured, thus potentially allowing for capture‐recapture analysis. Capture histories were not recorded systematically and we were only able to recover them for 10 years from hand‐written notes taken in the field. Mean trap effort was 467 trap nights per site each year (±41.6 SE), with a maximum of 2700 (Nolans Swamp was trapped for nine nights with 300 traps in 2002) and a minimum of 20 (Polblue Picnic Area and Wallaby Hill were both only trapped for one night with 20 traps in 2004). There were no recaptures between years while within years, there was a maximum of eight and a mean of 0.48 recaptures across all sites (±0.124 standard error). Unfortunately, there were too few recaptures for capture‐recapture analysis; hence, we used an index of capture success (captures per 100 trap nights) as an unvalidated index of relative abundance. The exact location of trap placement within each site was not recorded but the same areas of habitat were trapped most years.IndicesWe modelled two indices, naïve occupancy based on scats and relative abundance based on the number of captures of new individuals per 100 trap nights, to determine whether they changed over time, and in relation to precipitation and annual average minimum temperature. Given both methods are used by managers to monitor occupancy and abundance respectively, we assumed that they would be correlated if they were accurate, but if they were not correlated then one or both methods are inappropriate for monitoring purposes. Climate data were taken from WorldClim historical monthly weather data, CRU‐TS 4.03 (Harris et al. 2014) downscaled with WorldClim 2.1 (Fick & Hijmans 2017), with a spatial resolution of 2.5 min (~21 km2). This resolution is not fine enough to detect major variation between sites, instead, the main focus is examining temporal trends. For analysis of the index of relative abundance, we used the climate data from the month and year in which trapping was conducted. Unfortunately, the month in which scat surveys were conducted was not recorded, so we used the average of each month in the year the surveys were conducted instead. To obtain these values, we used the ‘raster’ package in R (Hijmans 2022) to combine data into one stack, which we then averaged. We also normalised precipitation and temperature values using the min‐max method: this transforms the minimum value to zero, the maximum to one and all others to values between zero and one, thus ensuring the coefficients of each variable are comparable.We modelled naïve occupancy using a generalised linear model with a binomial distribution and a logit link function with all combinations of two‐way interactions between variables. To model the index of relative abundance, we used a linear model with all combinations of two‐way interactions between variables. We also included trap effort as a confounding factor in this model. For both regressions, we only included sites that were surveyed in the analysis. This meant that while the naïve occupancy model included 27 sites the relative abundance model only included data for the 20 sites with trapping data. We then ranked all possible model combinations by Akaike's Information Criterion corrected for small sample size (AICc; Akaike 1974; Burnham & Anderson 2004). To obtain coefficient estimates, we averaged all models for which the AICc value was <2.0 from the best model (i.e. ΔAICc <2). Finally, we used a Pearson correlation test to assess whether there was a relationship between the two unvalidated indices of abundance: latency to find scats and relative abundance.Occupancy modelThe lack of repeat visits within each year's scat surveys made it impossible to conduct an occupancy model using scat data. Hence, we ran an occupancy model using the data from trapping surveys. We treated each trap night as a repeat visit and each 4‐to‐5‐night survey trip (one trip each year) as a season. We were only able to obtain capture history data from four sites (Polblue Swamp Campground; Nolans; Watergauge Trail Gate and Polblue Picnic Area) from 2008 to 2017 (with only half the 2011 data). We included five variables for occupancy: temperature and precipitation from the month in which the trapping was conducted, site, year, as well as swamp rat and house mouse presence at each site in the same year. Climate data were collected as described above and were again normalised. We also included trap effort as a detection variable. We conducted the multi‐season occupancy modelling using the R package ‘unmarked’ (Fiske & Chandler 2011). We ranked all possible model combinations using AICc. As the best model (lowest AICc value) was the only model with ΔAICc <2.0, we obtained coefficient estimates from this model rather than model averaging. The ‘unmarked’ package also provides estimates of occupancy, detection as well as seasonal colonisation and extinction probability when conducting a multi‐season model (Fiske & Chandler 2011). Finally, we used the goodness‐of‐fit model proposed by MacKenzie and Bailey (2004) with 10,000 simulations to assess the model fit.ResultsIndicesBased on naïve occupancy, 53.4% of sites were occupied (Supporting Information, Appendix A, Table A1). The best‐fitting model of naïve occupancy included precipitation (Akaike's weight w = 1.00), temperature (w = 0.95), year (w = 0.97) and an interaction between temperature and year (w = 0.91)—these were also the most important variables (Table 1; Supporting Information, Appendix A, Table A2). Model averaging showed that the impact of precipitation on naïve occupancy had confidence intervals that overlapped with zero, meaning there is no evidence that precipitation explains variation in occupancy (coefficient estimates = −21.49 ± 79.49 standard error; Table 1). As the temperature increased, naïve occupancy decreased (coefficient estimate = −682.33 ± 240.99; Table 1). The year had a slightly negative effect on naïve occupancy (coefficient estimate = −0.08 ± 0.04; Table 1). There was also an interactive effect of temperature and year (coefficient estimate = 0.34 ± 0.12). The interactive effect between precipitation and year had confidence intervals that overlapped with zero, providing no evidence of an effect of precipitation on occupancy between years (coefficient estimate = 0.05 ± 0.07; Table 1). The same was true for the interactive effect between precipitation and temperature (coefficient estimate = 1.98 ± 2.37).1TableEstimated covariate effects of factors that impact the naïve occupancy of the Broad‐toothed Rat in the Barrington Tops National ParkCoefficient estimateStandard errorp‐valueWeightIntercept160.0987.170.07NAPrecipitation−21.4979.490.791.00Temperature−682.33240.990.010.95Year−0.080.040.070.97Temperature*Year0.340.120.010.91Precipitation* Temperature1.982.370.410.31Precipitation*Year0.050.070.490.31Covariate effects were produced by averaging all models with a delta value <2, weights were the sum of the weights of all possible models.The best‐fitting model of the index of relative abundance included precipitation (w = 0.63) only (Table 2; Supporting Information, Appendix A, Table A3). The most important variables included precipitation, trap effort (w = 0.55) and year (w = 0.51; Table 2). Model averaging showed that neither precipitation (coefficient estimate = 27.58 ± 99.93), nor year (coefficient estimate = 0.00 ± 0.01) had any impact on the index of relative abundance due to the large confidence intervals (Table 2). Additionally, trap effort had no impact on relative abundance (coefficient estimate = −0.00 ± 0.00; Table 2).2TableEstimated covariate effects of factors that impact the relative abundance of the Broad‐toothed Rat in the Barrington Tops National ParkCoefficient estimateStandard errorp‐valueWeightIntercept−1.2389.810.99NAPrecipitation27.5899.930.780.63Temperature103.27195.940.600.38Year0.000.060.980.51Temperature*Year−0.160.110.140.12Precipitation*Temperature−0.652.970.830.07Precipitation*Year−0.080.100.430.11Trap effort−0.000.000.120.56Covariate effects were produced by averaging all models with a delta value <2, weights were the sum of the weights of all possible models.We found no evidence of a relationship between latency to find scats and relative abundance (correlation coefficient = −0.07, t = −0.54, n = 64, p = 0.59).Occupancy modelThe best‐fitting model for Broad‐toothed Rat occupancy was the null model, meaning no variables impacted occupancy or detection (AICc = 123.58; Supporting Information, Appendix A, Table A3). Based on this model, Broad‐toothed Rat occupancy was 0.78 ± 0.23 and detectability was 0.51 ± 0.06. Seasonal colonisation was 0.36 ± 0.13 and extinction 0.44 ± 0.13 (Supporting Information, Appendix A, Table A3). The goodness‐of‐fit model test showed the model provided a sufficient fit (p‐value = 0.17; c‐hat = 1.1; total chi‐square = 169.86).DiscussionThe 19‐year monitoring program for the Broad‐toothed Rat in Barrington Tops National Park is unfit for purpose, being unable to identify population declines, range reductions or the drivers of change. The range of variables revealed inconsistent patterns of change across time and was associated with different predictors. Naïve occupancy decreased very slightly over time, while there was no trend in relative abundance. Robust occupancy also revealed no trend over time. The seasonal extinction rate (0.44 ± 0.13 per site) exceeded the seasonal colonisation rate (0.36 ± 0.13), but the confidence intervals around these estimates were too large to predict with certainty whether the population has declined. As expected, naïve occupancy decreased with increasing temperature, but contrary to expectations, temperature had no impact on relative abundance or robust occupancy. Also contrary to expectations, we found no evidence that precipitation explained naïve occupancy, relative abundance or robust occupancy.Current understanding predicts a poor outlook for the Barrington Tops Broad‐toothed Rat population for several reasons. First, the Broad‐toothed Rat is continually threatened by climate change, and this is especially true for the climatically marginal Barrington Tops metapopulation (Green et al. 2008). Indeed, Green et al. (2008) showed that continued climate change will lead to a decline in the quality of most outlying areas of Broad‐toothed Rat habitat. Notwithstanding this, we found inconsistent evidence of temperature affecting annual population parameters. Second, Swamp Rats, which may out‐compete the Broad‐toothed Rat, were found in Barrington Tops for the first time in 1999, the same year that the monitoring began (Green & Osborne 2003; Green et al. 2008) and if their numbers have increased since then Broad‐toothed Rats may have declined. Finally, O'Brien et al. (2008) predicted that limited dispersal and colonisation, caused by invasive species and unfavourable habitat in more isolated swamps, would lead to future declines in the Barrington Tops metapopulation. Despite this strong expectation, the present study provided little support for population decline. We suggest this might be due to methodological constraints.While naïve occupancy decreased over time as expected, the change was minimal. One possible explanation for this unexpected outcome is that the specific scat survey methodology used to calculated naïve occupancy underestimated population decline. Broad‐toothed Rat scats decay slowly, with 7% of Broad‐toothed Rat scats still visible after 2.5 years (Happold 1989). Even though it is possible to distinguish between fresh and old scats (Cherubin et al. 2019; Shipway et al. 2020), scat age was not recorded in the Barrington Tops scat surveys. This means the surveys are not just examining current occupancy, but also to some extent occupancy over the last 2.5 years. This could limit their potential to detect population change in a timely manner. Similarly, the trapping method may have also limited the potential for relative abundance and robust occupancy to detect population change. Extreme habitat degradation meant that non‐random trap placement was the only way to ensure future trapping success. This may have also resulted in a reduced ability to detect localised population declines caused by the same habitat degradation. With these issues in mind, it seems possible that rather than the population being stable, the methods have failed to detect population change over time.The effect of climate predictors was also unexpected based on the current literature. Indeed, it is well documented that the Broad‐toothed Rat requires cold temperatures and high precipitation (Green 2003; Green et al. 2008; Shipway et al. 2020). While naïve occupancy decreased with increasing temperature as expected, temperature had no impact on either relative abundance or robust occupancy (Supporting Information, Appendix A, Table A3). This again could be due to non‐random trap placement leading to a limited capacity of relative abundance and robust occupancy to detect population change. It is also possible that precipitation did not vary enough across the monitoring sites to reveal the influence of this climatic variable on Broad‐toothed Rat populations within the Barington Tops. This seems probable given that Barrington Tops has an average annual precipitation of ~2000 mm (Zoete 2000), which falls well within the mean average rainfall of >1000 mm required by the Broad‐toothed Rat (Green 2003). This means that when considering the impact of climate on the Barrington Tops Broad‐toothed Rat metapopulation, temperature is likely a more limiting factor than precipitation. Further research is needed to confirm this, especially given the large monthly variation in precipitation: the precipitation data used in our relative abundance analysis ranged from 49.1 to 230.7 mm (±48.9 mm standard deviation).To properly consider conclusions about population change, it is important to evaluate methods used to monitor the Barrington Tops Broad‐toothed Rat and to consider the potential impact of imperfect detection. Indices, including naïve occupancy and relative abundance, are methods that do not account for imperfect detection. These methods are advantageous in that they are fast and easy to replicate (Stander 1998; Blaum et al. 2008; Funston et al. 2010). At the same time, concerns have been raised about the accuracy of these methods: indices have the same assumptions as robust methods (i.e. methods that account for imperfect detection assume closed populations and independent detection events; Charbonnel et al. 2014) plus the assumption that they relate in some linear fashion to actual abundance or occupancy (Anderson 2003; Hayward & Marlow 2014; Gopalaswamy et al. 2015; Hayward et al. 2015; Fancourt 2016; Dröge et al. 2020). For this assumption to be true, there needs to be a constant detection rate (Anderson 2003; Fancourt 2016; Dröge et al. 2020), which is generally unlikely given the large number of spatio‐temporal factors that influence detectability (Anderson 2001; Hayward & Marlow 2014; Fancourt 2016).Previous research suggests a near‐perfect and constant detection rate of Broad‐toothed Rat scats. Shipway et al. (2020) found that with six randomly placed transects, the detection probability in their occupancy model was >0.99. Additionally, Green and Osborne (2003) found that when conducting non‐random 15‐min scat searches in the Snowy Mountains, scats were found within 5 min at 98% of the sites where scats were detected. Based on this, the authors concluded that the risk of failing to detect scats when they were present was very low. Given this, it seems likely that imperfect detection does not have a significant impact on Broad‐toothed Rat scat surveys, meaning both naïve and robust analyses of these surveys should produce the same results (Guillera‐Arroita et al. 2014) and that naïve occupancy based on scat surveys is a valid method.Nevertheless, it will be important to test the assumption of perfect detection, as well as whether spatially and temporally consistent scat detectability holds within the Barrington Tops. This is especially important to consider in the context of long‐term monitoring. Shipway et al. (2020) were able to assume constant detection of scats within and between sites in their study. While this may be the case for their relatively short‐term study, it does not necessarily hold true for the long‐term monitoring of Barrington Tops. There are numerous factors that could have impacted scat detection including the number and experience of observers, time of day, vegetation density and rainfall. It is also possible that scat detection could vary between the Snowy Mountains and Barrington Tops: Green and Osborne (2003) as well as Green, K., unpubl. data (2000) noted that it generally took longer to detect scats in Barrington Tops compared to the Snowy Mountains. While Green, K., unpubl. data (2000) concluded that this indicated a relatively low abundance of Broad‐toothed Rats in Barrington Tops, without considering the possibility of differences in the detection probability between both sites, it is impossible to be sure to what extent this result is due to variation in abundance compared to variation in scat detection. There is already evidence that scat detection has some spatial variation. Milner et al. (2015) noted low detection of Broad‐toothed Rat scats in low‐spreading heath compared to more open areas. While we cannot exclude that detection is extremely high and that detection is constant between Broad‐toothed Rat sites, we argue that it will be an important facet of adaptive monitoring to test these assumptions over a long‐time frame and within Barrington Tops (Lindenmayer & Likens 2009). Until this time, we cannot be sure that naïve occupancy estimated via scat surveys is a valid method.When the assumption of perfect detection is violated, robust methods such as occupancy models provide a useful alternative approach. Occupancy models account for imperfect detection and allow detection to vary depending on specific factors (MacKenzie 2005; MacKenzie et al. 2006; Bailey et al. 2014; Rushing et al. 2019). For occupancy models to do this, factors that impact detection need to be included in models. When an occupancy model fails to include factors that impact detection then, just like indices, the model assumes constant detection (MacKenzie et al. 2006; Bailey et al. 2014). With this in mind, we think it probable that to some extent our occupancy model assumes constant detection. This is because it is likely that factors other than trap effort, such as trap placement and the number of subsequent trap nights, would have influenced detection in trapping surveys. Therefore, to some extent, our occupancy model is analysing detection probability in addition to occupancy. Given this limitation, the results of our occupancy model should be examined with caution. Our occupancy model returned no evidence that any of the variables considered here explained variation in occupancy or detection. As such, the analysis does not allow us to identify potential causes of population change. Even so, our analysis provides a useful illustration of how occupancy modelling can be used. While occupancy models built from trapping data may be unrealistic in practice because trapping has a low detection rate and is very time‐consuming, the method could be easily applied to data from scat surveys where detection is found to be imperfect.Our findings highlight the ongoing need for effective and continuous biodiversity monitoring for the Barrington Tops Broad‐toothed Rat metapopulation. We reviewed 19 years of data to understand how the Broad‐toothed Rat population is changing over time and whether climate is having an influence on its trajectory. We cannot confirm whether the population has increased, decreased or remained stable. While it seems that temperature is a more limiting factor for this population than precipitation, we draw this conclusion with caution given several potential methodological issues in both the surveys and analyses. These survey limitations are discussed here with an adaptive response in mind. The impact of detection in scat surveys is particularly important to consider. While scat surveys have been established to study the Broad‐toothed Rat effectively in other areas of its geographic distribution (Green & Osborne 2003; Schulz et al. 2019; Shipway et al. 2020), there is a need to record the age of scats in future monitoring and also to determine whether detection is near‐perfect and consistent across space and time within the Barrington Tops. If perfect and constant detection are confirmed, then naïve occupancy is an appropriate method, although this validation is required for all conditions and years. If these attributes of detection are not confirmed, then we suggest that it would be wise to consider using robust methods in future analyses. Many of the problems with this monitoring program are common to long‐term monitoring. Specifically, managers have many demands on their time meaning monitoring activities may be missed or done in an incomplete way. Moreover, staff turnover makes it difficult to maintain a consistent methodology, which is important for the integrity of long‐term data and the ability of a program to detect population trends (Lindenmayer et al. 2020). At the same time, monitoring can benefit from the inclusion of new technology and methods. The challenge faced by long‐term programs is how to incorporate new methods while also maintaining the long‐term integrity of the program (Lindenmayer et al. 2020). This is incredibly important as the effectiveness of a monitoring program underpins an institution's ability to collect good quality data and therefore make well‐informed management decisions.AcknowledgementsCharlotte Alley is supported by an Australian Government Research Training Program (RTP) Scholarship and the Holsworth Wildlife Research Endowment. Matt Hayward is supported by ARC LP200100261. The long‐term monitoring program described here was conducted by New South Wales National Parks and Wildlife Services with whom some of our authors are affiliated: Peter Beard, Adam Fawcett and Geoffrey James. We declare this as a possible conflict of interest. Many people have worked hard to collect data for this monitoring program: M. Schroder; K. Green; J. Keating; N. Hunter; C. O'Brien; G. James and A. Fawcett, in addition to other staff and volunteers. The Barrington Tops were traditionally used as a meeting place for several surrounding peoples: the Worimi, Biripi, Gringai and Wonnarua. We acknowledge the Traditional Owners of the land and pay our respects to their Elders, past, present and emerging. Open access publishing facilitated by The University of Newcastle, as part of the Wiley ‐ The University of Newcastle agreement via the Council of Australian University Librarians.Conflict of interestSome of our authors are affiliated with New South Wales National Parks and Wildlife Services and we declare this as a possible conflict of interest.ReferencesAkaike H. (1974) A new look at the statistical model identification. In: Selected Papers of Hirotugu Akaike. (eds E. Parzen, K. Tanabe and G. Kitagawa) pp. 215–222. Springer, New York, NY.Anderson D. R. (2001) The need to get the basics right in wildlife field studies. Wildlife Society Bulletin 29, 1294–1297.Anderson D. R. (2003) Response to Engeman: index values rarely constitute reliable information. Wildlife Society Bulletin 31, 288–291.Bailey L. L., MacKenzie D. L. and Nichols J. D. (2014) Advances and applications of occupancy models. Methods in Ecology and Evolution 5, 1269–1279.Beilharz L. V. and Whisson D. A. (2016) Habitat selection by two sympatric rodent species in an alpine resort. Australian Journal of Zoology 64, 327–334.Blaum N., Engeman R. M., Wasiolka B. and Rossmanith E. (2008) Indexing small mammalian carnivores in the southern Kalahari, South Africa. Wildlife Research 35, 72–79.Bled F., Nichols J. D. and Altwegg R. (2013) Dynamic occupancy models for analyzing species' range dynamics across large geographic scales. Ecology and Evolution 3, 4896–4909.Burnham K. P. and Anderson D. R. (2004) Model Selection and Multimodel Inference, 2nd edn. Springer, New York, NY.Charbonnel A., D'Amico F., Besnard A. et al. (2014) Spatial replicates as an alternative to temporal replicates for occupancy modelling when surveys are based on linear features of the landscape. Journal of Applied Ecology 51, 1425–1433.Cherubin R. C., Venn S. E., Driscoll D. A., Doherty T. S. and Ritchie E. G. (2019) Feral horse impacts on threatened plants and animals in subalpine and montane environments in Victoria, Australia. Ecological Management & Restoration 20, 47–56.Dickman C. R. and McKechnie C. A. (1985) A survey of the mammals of mount royal and Barrington tops, New South Wales. Australian Zoologist 21, 531–543.Doran G. T. (1981) There's a S.M.a.R.T. way to write Management's goals and objectives. Management Review 70, 35–36.Dröge E., Creel S., Becker M. S., Loveridge A. J., Sousa L. L. and Macdonald D. W. (2020) Assessing the performance of index calibration survey methods to monitor populations of wide‐ranging low‐density carnivores. Ecology and Evolution 10, 3276–3292.Eldridge D. J., Ding J. and Travers S. K. (2020) Feral horse activity reduces environmental quality in ecosystems globally. Biological Conservation 241, 108367.Fancourt B. A. (2016) Avoiding the subject: the implications of avoidance behaviour for detecting predators. Behavioural Ecology and Sociobiology 70, 1535–1546.Fick S. E. and Hijmans R. J. (2017) WorldClim 2: new 1km spatial resolution climate surfaces for global land areas. International Journal of Climatology 37, 4302–4315.Fiske I. J. and Chandler R. B. (2011) Unmarked: an R package for fitting hierarchical models of wildlife occurrence and abundance. Journal of Statistical Software 43, 1–23.Funston P. J., Frank L., Stephens T. et al. (2010) Substrate and species constraints on the use of track incidences to estimate African large carnivore abundance. Journal of Zoology 281, 56–65.Gopalaswamy A. M., Delampady M., Karanth U., Kumar N. S. and Macdonald D. M. (2015) An examination of index‐calibration experiments: counting tigers at macroecological scales. Methods in Ecology and Evolution 6, 1067–1068.Green K. (2002) Selective predation on the broad‐toothed rat, Mastacomys fuscus (Rodentia Muridae), by the introduced red fox, Vulpes vulpes (carnivora: Canidae), in the Snowy Mountains, Australia. Austral Ecology 27, 353–359.Green K. (2003) Altitudinal and temporal differences in the food of foxes (Vulpes vulpes) at alpine and subalpine altitudes in the Snowy Mountains. Wildlife Research 30, 245–253.Green K. and Osborne W. S. (1981) The diet of foxes, Vulpes vulpes (L.), in relation to abundance of prey above the winter snowline in New South Wales. Australian Wildlife Research 8, 349–360.Green K. and Osborne W. S. (2003) The distribution and status of the broad‐toothed rat Mastacomys fuscus (Rodentia: Muridae) in New South Wales and the Australian Capital Territory. Australian Zoologist 32, 229–237.Green K., Stein J. A. and Driessen M. M. (2008) The projected distributions of Mastacomys fuscus and Rattus lutreolus in South‐Eastern Australia under a scenario of climate change: potential for increased competition? Wildlife Research 35, 113–119.Guillera‐Arroita G., Lahoz‐Monfort J. J., MacKenzie D. I., Wintle B. A. and McCarthy M. A. (2014) Ignoring imperfect detection in biological surveys is dangerous: a response to “fitting and interpreting occupancy models”. PLoS One 9, e99571.Hanski I. A. and Simberloff D. (1997) The metapopulation approach, its history, conceptual domain, and application to conservation. In: Metapopulation Biology: Ecology, Genetics and Evolution. (eds I. A. Hanski and M. E. Gilpin) pp. 5–26. Academic Press, San Diego.Happold D. (1989) The value of faecal pellets for ascertaining the presence of Mastacomys fuscus in field surveys. Victorian Naturalist 106, 41–43.Harris I., Jones P. D., Osborn T. J. and Lister D. H. (2014) Updated high‐resolution grids of monthly climatic observations ‐ the CRU TS3.10 dataset. International Journal of Climatology 34, 623–642.Hayward M. W. and Marlow N. (2014) Will dingoes really conserve wildlife and can our methods tell? Journal of Applied Ecology 51, 835–838.Hayward M. W., Botani L., Burrows N. et al. (2015) Ecologists need robust survey designs, sampling and analytical methods. Journal of Applied Ecology 52, 286–290.Hijmans R. J. (2022) raster: geographic data analysis and modelling. Available from URL: https://CRAN.R‐project.org/package=raster.Key Biodiversity Areas Partnership. (2022) Key Biodiversity Areas factsheet: Barrington Tops and Gloucester Tops. Extracted from the World Database of Key Biodiversity Areas. Developed by the Key Biodiversity Areas Partnership: BirdLife International, IUCN, American Bird Conservancy, Amphibian Survival Alliance, Conservation International, Critical Ecosystem Partnership Fund, Global Environment Facility, Re: wild, NatureServe, Rainforest Trust, Royal Society for the Protection of Birds, World Wildlife Fund and Wildlife Conservation Society. Downloaded from http://www.keybiodiversityareas.org/ on 29/11/2022.Legg C. J. and Nagy L. (2006) Why most conservation monitoring is, but need not be, a waste of time. Journal of Environmental Management 78, 194–199.Legge S. and Fleming A. (2012) Measuring and reporting on conservation management outcomes. In: Biodiversity Monitoring in Australia. (eds D. Lindenmayer and P. Gibbons) pp. 127–132. CSIRO Publishing, Collingwood, Victoria.Lindenmayer D. B. and Likens G. E. (2009) Adaptive monitoring: a new paradigm for long‐term research and monitoring. Trends in Ecology and Evolution 24, 482–426.Lindenmayer D. B., Likens G. E., Krebs C. J. and Hobbs R. J. (2010) Improved probability of detection of ecological “surprises”. Proceedings of the National Academy of Sciences of the United States of America 107, 21957–21962.Lindenmayer D., Gibbons P., Bourke M. et al. (2012) Improving biodiversity monitoring. Australian Journal of Ecology 37, 285–294.Lindenmayer D. B., Woinarski J., Legge S. et al. (2020) A checklist of attributes for effective monitoring of threatened species and threatened ecosystems. Journal of Environmental Management 262, 110312.MacKenzie D. I. (2005) What are the issues with presence‐absence data for wildlife managers. Journal of Wildlife Management 69, 849–860.MacKenzie D. I. and Bailey L. L. (2004) Assessing the fit of site‐occupancy models. Journal of Agricultural, Biological, and Environmental Statistics 9, 300–318.MacKenzie D. I., Nichols J. D., Royle A., Pollock K. H., Bailey L. L. and Hines J. E. (2006) Occupancy Estimation and Modeling: Inferring Patterns and Dynamics of Species Occurrence. Academic, New York, NY.Milner R. N. C., Starrs D., Hayes G. and Evans M. C. (2015) Distribution and habitat preference of the broad‐toothed rat (Mastacomys fuscus) in the Australian Capital Territory, Australia. Australian Mammalogy 37, 125–131.O'Brien C. M., Crowther M. S., Dickman C. R. and Keating J. (2008) Metapopulation dynamics and threatened species management: why does the broad‐toothed rat (Mastacomys fuscus) persist? Biological Conservation 141, 1962–1971.Prowse T. A. A., O'Connor P. J., Collard S. J., Peters K. J. and Possingham H. P. (2021) Optimising monitoring for trend detection after 16 years of woodland bird surveys. Journal of Applied Ecology 58, 1090–1100.Robinson N. M., Scheele B. C., Legge S. et al. (2018) How to ensure threatened species monitoring leads to threatened species conservation. Ecological Management & Restoration 19, 222–229.Rushing C. S., Royle J. A., Ziolkowski D. J. and Pardieck K. L. (2019) Modelling spatially and temporally complex range dynamics when detection is imperfect. Scientific Reports 9, 12805.Schulz M., Schroder M. and Green K. (2019) The occurrence of the broad‐toothed rat Mastacomys fuscus in relation to feral horse impacts. Ecological Management and Restoration 20, 31–36.Shipway S., Rowe K. M. C. and Rowe K. C. (2020) Persistence of the broad‐toothed rat (Mastacomys fuscus) across Victoria is correlated with climate and elevation. Wildlife Research 47, 267–278.Stander P. E. (1998) Spoor counts as indices of large carnivore populations: the relationship between spoor frequency, sampling effort and true density. Journal of Applied Ecology 35, 378–385.Vos P., Meelis E. and Ter Keurs W. J. (2000) A framework for the Design of Ecological Monitoring Programs as a tool for environmental and nature management. Environmental Monitoring Assessment 61, 317–344.Zoete T. (2000) Vegetation survey of the Barrington tops and mount Royal National Parks for use in fire management. Cunninghamia 6, 511–578.
Journal
Ecological Management & Restoration – Wiley
Published: Jan 1, 2023
Keywords: index of abundance; mastacomys fuscus; naïve occupancy; occupancy model; population change; scat detection