Threat maps have been commonly used to prioritize game and conservation management decisions. A common approach is to overlay problematic species distribution with that of the conservation target to identify areas of high conservation value. However, a number of different methods and data types have arisen that have been used as inputs into threat map modelling. These various methods have arisen due to individuals and agencies monetary or personnel constraints. In this study, we identified the two most common methods of obtaining feral pig threat data to test the correlative strength between data types in the creation of threat distribution maps.
(i) Create threat maps from two of the most common methods of obtaining threat data for feral pigs
(ii) Quantify correlative strength between both data sets in the creation of threat maps
(1) Random Stratified Sampling
(2) Camera Detection data
(3) Disturbance Survey Data
(4) Predictive Threat Map Modelling
(5) Linear Regression Analysis
(1) Sites were chosen at random using a random stratified sampling design by elevation across the island of Oʻahu.
(2) Six game cameras were deployed at each random site in a rectangular pattern with 50m intervals between each camera. Cameras were deployed in a manner that maximized probability of detection within a 10 meter radius of randomly pre-selected GPS points.
(3) At each camera location, signs of pig disturbance were recorded in four 10x10 meter quadrats over a standardized two minute search period per quadrat. In each quadrat we recorded the presence or absence of old and new signs of tracks, scat, digging, and vegetation damage.
(4) Both data types (camera detections & disturbance) were then used as inputs into a zero-inflated negative binomial model to ascertain the distribution of these proxy for impact.
(5) Linear regressions were used to determine directionality and correlative strength between the two forms of count data collected for this study
Figure 1. The likely impact of pigs on Oʻahu as predicted by the average number of camera detections per camera per site. Areas of highest impact are indicated in green. Red circles depict the location of survey sites.
Figure 2. The likely impact of pigs on Oʻahu as predicted by total observed sign per site. Areas of highest impact are indicated in green. Red circles depict the locations of survey sites.
Figure 3. Regression analysis between observed and predicted values from model fit. Observed data has high deviance and weak directional correlation. Predicted values from zero-inflated model fit show strong directional correlation with low deviance implying that outputs from both data sets will inform synonymous management decisions
(i) All forms of observed data exhibited weak directional correlations with high deviance (Figure 3).
(ii) Much of the deviance of the observed data was explained by fitting the data to zero-inflated models (Figure 3).
(iii) After model fit, all predicted data exhibited significant strong directional correlations with low deviance (all sign: R2 = 0.7946, all track: R2 = 0.8027, all dig: R2 = 0.8007).
- Both camera detections and sign data are good predictors of feral pig threat distribution
- Both data types are significantly correlated after model fit suggesting that resulting threat maps will inform synonymous management decisions
- In Hawai‘i, the environmental parameters account for a majority of the deviance between data types (may not be applicable elsewhere).
- Since data types are interchangeable in creating threat maps each method may be prioritized by the individual/agencies which are most suitable to their budget and personnel.
Mahalo to Shaya Hornarvar at the Division of Forestry and Wildlife for making this study possible and to the hunting community for their feedback on our resulting threat maps.