Sign In
Plains Fish Optimal Sampling Protocol
Plains Fish Optimal Sampling Protocol

Led By

Ryan Fitzpatrick

Study Area

Plains streams of eastern Colorado in the South Platte and Arkansas River basins

Project Status

Ongoing (this site selection tool will be used on an annual basis to guide sampling efforts)

Research Objectives

To guide biologists to the most efficient sampling locations to reduce uncertainty given logistical and financial constraints.


Dr. Kristin Broms and Dr. Mevin Hooten, Colorado Cooperative Fish and Wildlife Research Unit, Colorado State University

Project Description

Due to logistical, financial, and time constraints on staff, it is important that field activities are conducted as efficiently as possible and result in data that are statistically rigorous and defensible. This is especially essential when there are several species of interest. This project provides a site selection tool for eastern plains native fishes that is adaptable to changing management priorities, and can be accomplished within the logistical parameters set by CPW staff. The site selection process consists of five steps (Figure 1):

  1. The data. The data provide structure for the model, the desired inference, and the design criterion. Defining the data includes setting the boundaries, scale, and resolution of the area of inference, checking and cleaning the data that have been collected, and obtaining potential covariates.
  2. The model. The model structure and output make the inference associated with the monitoring efforts explicit and concrete.
  3. The design criterion. The design criterion is a formal connection between monitoring and the model, and is the quantity of interest about which improved inference is desired. It is a single statistic that summarizes the uncertainty associated with the study and is used to compare the efficiencies of sampling at various sets of locations in the future.
  4. Selecting sites for future sampling. This step involves finding the set of sites that minimizes the design criterion, which in this case is uncertainty regarding native species’ distributions.
  5. Collect more data and repeat. After future sites are selected and sampled, the model is re-fit with the new data and modified as necessary. Because the design criterion is based on the parameter estimates or posterior distributions, the next set of optimal sites will change with the newly fitted model. It is this responsiveness to new data that makes the procedure ideal for optimal long-term monitoring.

This protocol results in a sampling design that is statistically rigorous and biologist friendly. Biologists tell the model how many sites they are able to sample, and the model optimizes on those constraints. This protocol is optimal in that it optimizes on one metric—uncertainty.

Uncertainty across the species and weights selected according to management priorities. The protocol is adaptive in that it incorporates new data learning—as management objectives change, this protocol can change with them. This procedure has been used by biologists for the previous three field seasons, and is scheduled to be an ongoing, annual site selection tool.

Associated Publications

Broms, K. M., M. B. Hooten, and R. M. Fitzpatrick. 2015. Accounting for imperfect detection in Hill numbers for biodiversity studies. Methods in Ecology and Evolution 6:99–108.

Broms, K. M., M. B. Hooten, and R. M. Fitzpatrick. 2016. Model selection and assessment for multi-species occupancy models. Ecology 97:1759–1770.

Broms, K. M. and R. M. Fitzpatrick. 2016. A procedure manual for optimal adaptive sampling design for multiple species of plains fishes in eastern Colorado.