SNAMP PUB #8: Estimating population impacts via dynamic occupancy analysis of Before–After Control–Impact studies
Article Title: Estimating population impacts via dynamic occupancy analysis of Before–After Control–Impact studies.
Authors: Viorel D. Popescu, Perry de Valpine, Douglas Tempel, and M. Zachariah Peery.
- Using computer simulations, we assessed the ability of Before-After Control-Impact (BACI) studies to detect changes in wildlife presence or absence (i.e., occupancy) at locations where some type of experimental treatment is applied.
- We ran simulations under different scenarios, in which we varied factors such as sample size (number of sites surveyed), study design (how many years of pre- and post-treatment data were collected), and model parameters (e.g., local survival and colonization; defined below).
- We chose scenarios and model parameters that should be similar to those for the SNAMP Owl Team’s research, but our results apply to any species of interest.
- We found that the statistical power (defined below) to detect changes in occupancy during an 8-year study was low for all scenarios unless the treatment effects were large (at least a 50% decrease in local survival).
BACI studies are sometimes used to assess the potential effects of anthropogenic disturbance on ecological systems. In a BACI study, the ‘Impact’ sampling units are exposed to an experimental treatment (e.g., timber harvest), while the ‘Control’ sampling units are not exposed to the treatment. Data are collected from all sampling units ‘Before’ and ‘After’ the experimental treatments are applied to the ‘Impact’ units. In this way, any changes to the ‘Impact’ units that are due to the experimental treatment can be separated from changes that may occur naturally over time.
In an occupancy study, the sampling units are specific locations that are repeatedly surveyed over a number of years. From this survey data, researchers can estimate local survival and colonization (defined below) and assess what factors may affect changes in local survival and colonization. The SNAMP Owl Team is conducting a study that is a combination of BACI and occupancy study designs. Due to concerns that our sample size (i.e., number of owl territories) was too small to detect changes in owl occupancy in response to SPLAT treatments, we collaborated with two researchers from University of California, Berkeley (Dr. Viorel Popescu and Dr. Perry de Valpine) to assess the statistical power (defined below) of our study using computer simulations.
As expected, we found that the statistical power to detect post-treatment changes in local survival was greater when more sites were sampled, over a longer period of time (12 years vs. 8), and the species was more frequently detected during surveys. We also found that unbalanced study designs (those having a different number of Control and Impact sites, or a different number of years Before and After treatment) performed worse than balanced study designs. Our SNAMP occupancy study is an unbalanced design. In contrast, we found that implementing experimental treatments in different years at each site resulted in a modest improvement in power. Again, our SNAMP occupancy study consists of historical treatments that have occurred in different years at each owl territory. Finally, the power to detect 20% declines in local survival was low (< 0.5) under most scenarios, whereas sufficient power (≥ 0.7) was generally achieved when the declines in local survival were large (50% or 80%).
- Given the low power to detect small changes (20%) in post-treatment local survival, dynamic occupancy-BACI studies should be used when treatment impacts are expected to be large or when very large sample sizes can be obtained.
- At least 3 years of pre-treatment data are needed to accurately estimate changes in local survival, even when sample sizes are large.
- Dynamic occupancy-BACI studies perform well when treatments are implemented at different times (e.g., due to logistic, financial, or sociopolitical reasons).
- Based on these results, the SNAMP Owl Team decided to change its workplan and methods. The Owl Team will now pursue a 20-year, retrospective analysis of owl territories on both the Eldorado and Last Chance Study Areas in order to obtain an adequate sample size.
Full Reference: Popescu, de Valpine, Tempel, and Peery. 2012. Estimating population impacts via dynamic occupancy analysis of Before–After Control–Impact studies. Ecological Applications 22: 1389-1404.
The full paper is available here.
For more information about the SNAMP project and the owl team, please see: here.
Statistical power: The probability of detecting a treatment effect of a specified size during a research study, given the sample size for the study.
Local survival: The probability that an occupied location (i.e., site) will still be occupied by the target species during the next sampling period (typically the following year in most study designs).
Local colonization: The probability that an unoccupied site will be occupied by the target species during the next sampling period.