Landscape metrics that incorporate species-specific characteristics with landscape analyses, i.e. ecologically-scaled landscape indices, are becoming an increasingly used approach to aid conservation planning. In an evaluation of bee habitat around agricultural sites, we previously used an ecologically-scaled landscape model to predict pollinator abundance. Our model results varied in their ability to capture observed abundance patterns. Results based on three main parameters for which values were derived by expert opinion: (1) proportion of suitable nesting habitat per land-use and land-cover type, (2) proportion of suitable foraging habitat per land-use and land-cover, and (3) expected foraging distances of bee taxa. Recognizing an inherent risk in relying solely on an expert-driven modeling approach, we present a Markov chain Monte Carlo (MCMC) parameterization routine for ecologically-scaled landscape models to estimate foraging, nesting, and floral values that yield model pollinator abundance scores that best match empirical observations of pollinator abundance at sampled farm sites. We validated this approach by partitioning datasets into two samples; one used to parameterize foraging, nesting, and floral values and one used validate the parameterization’s ability to predict pollinator abundance at novel sites.
Results/Conclusions
Using parameter values from expert opinion, the ecologically-scaled landscape model explained up to 35% of the variation in observed pollinator abundance. In contrast, the MCMC parameterization routine yielded model results that explained 55% and 80% of the variation in abundance scores. The fit of the MCMC approach improved as the number of samples within a study increased. For particular cover types, confidence intervals of estimates for forage and nesting quality decreased with increasing variation among sites in the amount of that habitat. Using the MCMC parameterization provides a method to refine parameter values based solely on expert opinion and a potentially powerful approach to link observed data with ecologically-scaled landscape indices that could improve conservation planning.