Choice of model and re-nesting probability function influences behaviour of avian seasonal productivity models and their demographic predictions

Abstract

Measuring seasonal productivity is difficult in multi-brooded species without labour-intensive ringing studies. Individual-based (IB) models have been used to estimate seasonal productivity with no direct knowledge of number of nesting attempts, but they are often based on simplified re-nesting probability (φ_R) step-functions instead of observed or more biologically plausible ones. We present a new, open-source IB seasonal productivity model parameterized from studies of Black Redstart Phoenicurus ochruros and Yellowhammer Emberiza citrinella. We examined how the φ_R function shape (empirical versus simplified) influenced (1) model performance, (2) re-nesting compensation and (3) population-level predictions of a simulated management intervention. Population-level predictions were made only for Yellowhammer as we had more detailed demographic data, such as survival rates, available. Pattern-oriented modelling revealed that IB models produced realistic within-population distributions of breeding parameters, and those specified with an observed or empirically derived φ_R function generally outperformed those specified with simpler step functions. Strength of re-nesting compensation differed depending on the φ_R function used. For Yellowhammers, type of φ_R function in IB models marginally influenced population-level predictions of a simulated management intervention (potential population growth rate increased between 23% and 29% relative to no management intervention). In contrast, a simple deterministic productivity model, which did not simulate re-nesting compensation, predicted a 41% increase in potential population growth. At a population level, choice of φ_R function may have less influence on IB model predictions, but choice of model itself (IB versus deterministic) may have substantial impact. We discuss how more biologically plausible φ_R functions might either be observed directly, derived from nest data, or estimated from proxy information such as moult or brood patch changes.