Imperfect detection in plant populations can cause misestimates of demographic rates and missed population trends: The case for Astragalus microcymbus Barneby

Stage-structured population models are widely used for plant demographic studies to assess population dynamics. Over the last several decades, there have been advancements in mark-recapture methods in animal systems, but little to no use in plants because of the assumption that all individuals available for capture are seen. We examine population trends and demographic rates of a long-lived forb derived from conventional matrix population models compared to mark-recapture methods. We found a low probability of detection, primarily due to vegetative dormancy. While both models account for dormancy, only the mark-recapture method detected a population trend. Elasticity describes the effects of changes in vital rates on the population growth rate. Similar elasticity values are thought to buffer populations from environmental stochasticity. Vital rate estimates differed between methods, resulting in a higher growth rate with conventional methods but greater evenness among vital rate importance for the mark-recapture method.

Synthesis and applications: This study highlights that long-term demographic viability studies that have not addressed detectability can be revisited with minimal additional data to correct for imperfect detection. Plant studies have traditionally assumed perfect detection and may have missed trends, risk indicators, and conservation opportunities for rare plant populations.

Methods
Data were collected annually every July during the fruiting stage of Astragalus microcymbus since 1995. All plants found within four or five 3-meter radius circular plots at four sites were marked with 10-inch nails and numbered metal tags. Previously undocumented plants found within plots were assigned a new tag or a new decimal from an existing tag to track individuals with a unique tag ID over time. In the matrix population model analyses, individuals can transition among two aboveground stages (vegetative or reproductive) and one belowground stage (dormant). In the mark-recapture analyses, dormant individuals are treated as a non-detection.