Using Survival Analysis to Develop Models for Estimating
Size-at-Detection of Invasive Species under Surveillance
Abstract
Invasive species are non-native plants and animals that can
significantly harm societal and natural values such as agriculture,
social amenity, the environment and native ecosystems and species.
Efficient preparation for their incursion requires understanding their
potential impact, influenced by factors such as introduction pathways,
host material availability, climate suitability, and the value of
affected agriculture. A crucial factor that is specific to the incursion
and therefore unpredictable beforehand is the size of the outbreak at
the time of detection, which can curtail the range of management
options: if the invading population is small then eradication may be
affordable, whereas if it is large then eradication may be impossible.
We propose a statistical model for this random variable to aid
decision-support systems. We analyze the relationship between
surveillance and organism detection using survival analysis, treating
detection as analogous to a failure event. This approach links the
distribution of infestation size at detection with the probability of
detecting an incursion—specifically, the hazard function describing
the instantaneous detection rate. Under this survival model, we connect
the probability density function of infestation size at detection to the
hazard function. Moreover, we introduce an approximation using the
Weibull distribution to model the population size before pest detection.
This approximation holds when dealing with a small fixed number of traps
or a low probability of detection per trap. By assuming a relationship
between the invasive population size and the time it remains undetected,
we estimate the probability density function for the population’s
duration of occupancy. We develop a computer program to perform the
analysis, using the Mediterranean fruit fly as a case study to
demonstrate its application. We believe that representing the invasive
population size at detection provides valuable insights into control and
eradication strategies, potentially applicable to broader invasive
species management efforts.