Estimating density from presence/absence data in clustered populations

1. Inventories of plant populations are fundamental in ecological research and monitoring, butsuch surveys are often prone to field assessment errors. Presence/ absence (P/A) samplingmay have advantages over plant cover assessments for reducing such errors. However, thelinking between P/A data and plant density depends on model assumptions for plant spatialdistributions. Previous studies have shown, for example, how that plant density can beestimated under Poisson model assumptions on the plant locations. In this study, newmethods are developed and evaluated for linking P/A data with plant density assuming thatplants occur in clustered spatial patterns. 2. New theory was derived for estimating plant density under Neyman–Scott-type cluster models such as the Matérn and Thomas cluster processes. Suggested estimators, correspondingconfidence intervals and a proposed goodness-of-fit test were evaluated in a Monte Carlosimulation study assuming a Matérn cluster process. Furthermore, the estimators were applied to plant data from environmental monitoring in Sweden to demonstrate their empiricalapplication. 3. The simulation study showed that our methods work well for large enough sample sizes.The judgment of what is ’large enough’ is often difficult, but simulations indicate that asample size is large enough when the sampling distributions of the parameter estimators aresymmetric or mildly skewed. Bootstrap may be used to check whether this is true. Theempirical results suggest that the derived methodology may be useful for estimating densityof plants such as Leucanthemum vulgare and Scorzonera humilis. 4. By developing estimators of plant density from P/A data under realistic model assumptions about plants’ spatial distributions, P/A sampling will become a more useful tool forinventories of plant populations. Our new theory is an important step in this direction.
Session C9, part 2.
Statistical methods for ecological research on data from long term monitoring programs; financially supported by the Swedish Research Council