Using Transfer Function Analysis to develop biologically and economically efficient restoration strategies

Rare species across taxonomic groups and biomes commonly suffer from multiple threats and require intensive restoration, including population reintroduction and threat control. Following reintroduction, it is necessary to identify what level of threat control is needed for species to persist over time. Population reintroduction and threat control are time intensive and costly. Thus, it is pragmatic to develop economically efficient restoration strategies. We combined transfer function analysis and economic cost analysis to evaluate the effects of biologically meaningful increases in demographic processes on the persistence of a reintroduced population of a Hawaii endemic long-lived shrub, Delissea waianaeensis. We show that an increase in fertility by 0.419 following the suppression of non-native rodents or an increase by 0.098 in seedling growth following the suppression of invasive molluscs would stabilize the population (i.e., λ = 1). Though a greater increase in fertility than seedling growth was needed for the reintroduced population to persist over time, increasing fertility by suppressing rodents was the most cost effective restoration strategy. Our study emphasizes the importance of considering the effects of large increases in plant vital rates in population projections and incorporating the economic cost of management actions in demographic models when developing restoration plans for endangered species.


A: Background information for Rattus rattus and non-native molluscs
Rattus rattus (black ship rat) is one of the most disruptive vertebrates to invade oceanic islands and often listed as a primary driver of population decline and extinction of native plants 1,2 . The estimated home range of R. rattus is 4 ha 3 . When foraging, R.
rattus are the most active in areas with thick understory vegetation cover 10-30 cm in height 3 . Rattus rattus dens are often below ground in soil and fractured rock substrate, under logs, in thick understory vegetation, and inside partially dead trees 3 . Thought R.
rattus are omnivores, seeds and fruits are the dominant portion of their diet 2 . Following consumption and digestion by R. rattus, small seeds (0.5-1.2 mm) remain intact and viable 4 . The relatively small size of D. waianaeensis seeds (1.0-1.2 mm) suggests that R. rattus do not alter seed viability of this taxon following consumption and digestion.
However, the large home range, den characteristics, and foraging behavior of R. rattus imply D. waianaeensis seeds consumed by R. rattus are deposited in unsuitable habitat for seedling establishment.
The establishment of non-native molluscs (Mollusca: Gastropoda) is often implicated in the decline of oceanic island species 5,6 . Molluscs are generalist herbivores, primarily consuming foliage on the forest floor. In Hawaii, a total of 12 different nonnative mollusc species have established throughout natural areas 7,8 . In mesic to wet forest communities in Hawaii, terrestrial molluscs significantly reduce the density of numerous native seedlings and thus, the suppression of non-native molluscs is often incorporated as part of the recovery strategy for endangered species 6,9 . Over the five-year study period, we did not observe mollusc herbivory damage on later life stages. Thus, for modeling purposes, we assumed mollusc herbivory only influenced seedling survival and growth.

B: Range of biologically meaningful perturbations in targeted vital rates
The range of biologically meaningful perturbations that we used for this study was determined using a combination of field experiments and the results of previous studies. To set a realistic range of increases in fertility ϕ m following the suppression of R. rattus, we quantified the percent of fruits consumed and the identity of the consumer using a modified version of the methods developed by Pender, et al. 10 . Given there are other potential frugivores that may consume D. waianaeensis fruits, including native and non-native birds, tracking tunnels were used to isolate the effects of R. rattus on fertility.
Specifically, we installed 24 tracking tunnels at equal distance along four transects (50 cm x 10 cm x 10 cm; Connovation Limited, Auckland, New Zealand), with tracking cards inserted (The Black Trakka Gotcha Traps LTD, Warkworth, New Zealand) ( Figure   S1.1). The four transects spanned the length of the D. waianaeensis population; capturing intrapopulation habitat heterogeneity. Each tracking tunnel was baiting with one mature fruit and checked at a two-day interval. On each visit, fruit consumption was recorded and the tracking tunnels were re-baited. The footprints left on the tracking cards each visit were used to identify the frugivore consuming D. waianaeensis fruits ( Figure S1.1).
In total, the tracking tunnels were baited five times during the 2015 fruiting season. In this field experiment, we found that the mean R. rattus frugivory rate was 83%. To mimic the effects of R. rattus control on fertility and population dynamics, we set the biologically feasible increases in ϕ m from the observed value to a 5.9 fold increase. The proportion of fruits consumed by R. rattus at the D. waianaeensis population was consistent with previous studies that have examined the effect of R. rattus on the fertility of a related taxon in the Campanulaceae group, Cyanea superba ssp. superba 10 . Though it is possible that rodent control increases the effects of other stressors on plant vital rates, this has not been documented at similar field sites in Hawaii following rodent suppression 10 .
To set a range of biologically meaningful increases in seedling growth ! s-si following the suppression of non-native molluscs, we used the results of a previous field experiment 9 . In this experiment, 200 seeds of three localized Hawaiian endemic species (Cyanea superba ssp. superba, Cyrtandra dentata, and Schiedea obovata) were sown on the top layer of soil in 12 plots, 15m x 15m in diameter. Six plots were treated with a molluscicide, Sluggo (Neudorff Co., Fresno, California), and the other six plots were left exposed to normal herbivory intensity at the field site. The density of seedlings in each plot was recorded on a weekly basis for six weeks. This study illustrated that seedling density between the mollusc suppression and mollusc control treatments were significantly different for localized endemic plants, by up to 33% 9 . The range of biologically feasible increases in ! s-si following the suppression of non-native molluscs that we used for our study was from the observed ! s-si value to ! s-si +.33 by a perturbation magnitude of .01.

C: Cost of management actions
To suppress R. rattus at the D. waianaeensis field site would require 20 Goodnature A24 self-resetting multi-species kill traps (Tyler Bogardus,  time needed to suppress R. rattus than to suppress molluscs is due, in part, to the shorter duration of time that R. rattus needs to be suppresses. While R. rattus only needs to be suppressed during the D. waianaeensis fruiting season, molluscs need to be suppressed year round. The fixed cost of equipment to suppress R. rattus was $2,500 (i.e., 20 Goodnature A24 rodent traps x $125 per trap). However, when amortized over the lifetime of the equipment (i.e., 10 years), the yearly equipment cost to suppress R. rattus was $250. Including labor costs, the total yearly cost to suppress R. rattus (i.e., yearly fixed cost of equipment and labor costs) was $1,805. For mollusc suppression, there is no upfront fixed cost of equipment and the total yearly labor cost was $4,562.