Introduction
Climate change and habitat destruction has led to an increased species population decline over the last 100 years (Pachauri, et al., 2015), leading to 20% of vertebrates classified as threatened and 52 species moving one category closer to extinction each year on the IUCN Red List (Hoffman, et al., 2010). These anthropogenic problems have led to interventions to halt this decline but have had mixed success. The problem is that methods are based on intuition rather than primary evidence (Sutherland, et al., 2004; Ferraro & Pattanayak, 2006), making processes tougher and expensive and giving sub-optimal outcomes (Sutherland, et al., 2004).
Conservation needs to be based on evidence from past interventions to identify efficient methods. This will allow for a stretched budget to be better used (James, et al., 1999) and increased funding after demonstrating the effectiveness of these evidence-based methods (Sutherland, et al., 2004). This review of interventions will give an insight to which methods are effective and what confounding variables affect them.
We expect mainland species to be more effective in re-establishment because they have lower turnover rates (MacArthur & Wilson, 1967), a greater probability of success with increased ranges due to reduced intraspecific competition and an increase in success with increased initial population size from reduced genetic problems (Frankham, 1995). We expect that all intervention methods will have some level of success but captive breeding may be one of the most successful (Spooner, et al., 2015).
Word count – 200
Methods
Data Collection
Data was collected from conservation intervention studies researched by 30 people. Only interventions that had a physical, ecological, biological, political, economic, cultural and/or legal variable altered to halt the further decline of a population were included. Cases that conserved multiple species, whole ecosystems or their services were omitted. Both successful and failed interventions were collected to increase statistical power.
For every case study, data on the species’ identity, habitat and range, initial and ending population size, the cause of decline and the intervention details including dates and intervention type were collected.
Analysis
Some raw data was altered to ease analysis (see Table 1.) and studies that had no change in population size after intervention were categorised as failures. The natural range sizes were categorised as either: smaller than the Isle of Wight, Ireland, Australia or on a continental scale. Each study was given a binary value (0=no, 1=yes) for whether an intervention method was used.
A generalised linear model with binomial family errors was carried out using R statistical software, first considering each confounding variable to see which had a significant effect on success or failure and then each intervention type was added to this previous model in turn.
Table 1 – The problems encountered in the raw data and the alterations made before analysis.
Data Type Problem
Alterations made
Unknown data/information
Discarded from relevant analysis
Reducing 4 class categories to 3
Grouped reptiles and amphibians together
Inequalities in population size estimate
<X =X ; >X X*1.1
Range of body size given
Mid-point used as estimate
Word count – 198
Results
Confounding Variables
Although interventions involving birds seemed to be most successful (GLM, z=1.113, d.f.=28, P=0.266), followed by mammals (GLM, z=0.523, d.f.=28, P=0.601) and then reptiles/amphibians (GLM, z=-0.444, d.f.=28, P=0.657), class had no significant effect on the success of an intervention.
Interventions on mainland species seemed to be less successful (GLM, z=-0.935, d.f.=29, P=0.350) than those on islands (GLM, z=1.266, d.f.=29, P=0.206) but neither was significant.
When species’ body size (log) increased, it may have caused an increase in the probability of intervention success but this was not significant (GLM, z=1.182, d.f.=29, P=0.237).
The natural range size of the species had no significant effect on the likelihood that an intervention would be successful. Ranges up to the size of Ireland (GLM, z=0.143, d.f.=27, P=0.887), the Isle of Wight (GLM, z=0.000, d.f=27, P=1.000) nor Australia (GLM, z=0.900, d.f.=27, P=0.368) had any effect on whether an intervention was successful.
The initial population size (log) was the only confounding variable that had any significant effect on the probability of success for an intervention. This effect was that as initial population size grew, the likelihood of a successful intervention declined (GLM, z=-2.399, d.f.=29, P=0.0164).
Intervention Type
When added to the model including the significant effect of initial population size, neither captive breeding (GLM, z=0.965, d.f.=17, P=0.3347), translocation/relocation (GLM, z=0.701, d.f.=17, P=0.4832) nor habitat based strategies (GLM, z=0.943, d.f.=17, P=0.3454) had any significant effect on the probability of successfully halting the decline of a population. All three have shown evidence for increasing the likelihood of success but none are significant.
Intervention via disease treatment (GLM, z=-0.072, d.f.=17, P=0.266), controlling other animal taxa (GLM, z=-1.501, d.f.=17, P=0.1334) or by controlling humans (GLM, z=-0.735, d.f.=17, P=0.4626) seemed to decrease the probability of successful population growth, but this effect was not significant either.
Word count – 290
Discussion – 298
Main Findings
Most surprising were the effects of class, mainland/island, and range size. Past meta-analyses have shown higher rates of success involving reptiles/amphibians than mammals/birds. (Spooner, et al., 2015). Mainland species should be more successful than island species due to the lower turnover rate (MacArthur & Wilson, 1967) but evidence suggests the opposite. Species with larger ranges should be more successful because of reduced intraspecific competition but this was not shown.
Initial population size had a significant effect. You would expect larger populations to have greater success due to larger gene pools, reducing genetic defects like inbreeding depression (Frankham, 1995). We found the opposite. This may be because smaller populations are prioritised and to prevent extinction more effort is used, increasing the probability of success.
It was unusual to find no significant effects of intervention methods. It was expected that interventions would be effective but results highlight the issue with not using primary evidence (Sutherland, et al., 2004; Ferraro & Pattanayak, 2006). Past reviews have shown ranges in success, with ex-situ conservation having 85% success but habitat management only 60% (Spooner, et al., 2015).
Error and Bias
It is possible that errors arose from collection bias, publication bias or collection error. Data was collected non-randomly, and so a selection bias may have arisen from choosing more appealing taxa (Clark & May, 2002). Publication bias may have affected results with concerns that researchers only endeavour for publication if their results are significant and positive, which may overestimate success (Fanelli, 2010). Collection errors may have led to incorrect results with many cases not having their intervention added into the database through human error.
Future Work
To fully review conservation intervention methods, large-scale meta-analyses from a range of taxa and areas is needed (Spooner, et al., 2015). One idea is to establish a four-tier rule evaluating interventions (Ferraro & Pattanayak, 2006). If written into policy, data could be entered into a database to guide which interventions are best for certain situations (Spooner, et al., 2015), allowing for more efficient use of funding and saving more species from extinction.
Word count – 298