1. Within Caughley’s (1994) declining population paradigm for conservation, we develop a realistic method of detecting population changes using floor and ceiling population models.
2. A theoretical framework for detecting declining populations is presented and applied to four ungulate species from Kruger National Park, South Africa.
3. Census data are fitted to a non-linear autoregressive population model. Based on this model two auxiliary models, called the floor and ceiling models, are derived. These models predict the upper and lower threshold densities (or abundance values) for the non-linear population model.
4. Threshold floor and ceiling estimates can be used as a benchmark for monitoring population status. In particular, when abundance values (densities) tend to fall below the floor estimate, this may merit concern that a population is declining.
5. In conjunction with appropriate response by management, the presence of such an early warning system may prevent a population from becoming low in numbers, possibly making it less susceptible to stochastic processes and a less likely candidate for local extinction.