Trade‐offs associated with occupancy models in a multispecies grassland plant population study

Abstract

Abstract Occupancy models have been underused in multispecies plant studies. We evaluated the effectiveness of occupancy models to document impacts of management on grassland plant species populations in a pilot study. The study was designed to answer three questions: (1) From a list of 44 plant species, how many will have sufficient frequency to be analyzed? (2) Within a single site, which factor (sampling year or unit) influences occupancy and detection of each plant species? (3) How does the number of plots and visits affect the number of species with sufficient frequency for analysis? In 2012 and 2013, occupancy and detection probabilities of 44 plant species were assessed using permanently marked plots on two different grassland sites. Within each site, a fire and grazing managed unit and a burn‐only control unit were used to test the efficacy of occupancy modeling for multiple plant species and study design. Four of the 44 species had sufficiently high known presence (π) on both sites and years to be analyzed. There was little evidence to support that occupancy was different between years, despite very different weather patterns. There was support that occupancy was different in the burned and grazed unit compared with the control for Rhus copallinum L. Detection was highest (≥0.8) for R. copallinum, while the other species analyzed varied. In 2012, eight species had sufficient π for analysis. In 2013, 12 species had sufficient π for analysis. Occupancy models can be useful for investigating multiple plant species; however, a priori investigation of species' distribution is necessary to ensure species can be analyzed. Detection of plants was never 1, and therefore, detection should continue to be incorporated in plant population models, regardless of modeling technique.

Keywords

• detection probability
• grassland
• management
• multispecies occupancy
• occupancy
• patch‐burn grazing