ARMI » Designing and Implementing a Monitoring Program

Designing and Implementing a Monitoring Program

Tyler and Bones on a boat.
T. Thigpen and B. Glorioso head out for a night of amphibian surveys in the Atchafalaya Basin LA.
Photo by: USGS

ARMI Philosophy on the Role of Monitoring: It's not a "Stand-alone" Activity. Our goal is to develop a monitoring program that considers how the data can be used to understand how amphibians respond to changes occurring on the landscape or in response to management even over long periods of time. These data can be combined with detailed experiments and habitat or species management, which allows ARMI scientists to learn how amphibians are doing in a more comprehensive manner.

Before starting a monitoring project, we answer the 3 fundamental questions of an informative monitoring program: Why are we monitoring; What are we monitoring; and How will we design the project?

Fundamental Questions of a Monitoring Program

  • 1. Why are we collecting monitoring data? To what extent do these data relate to the objectives of the management agency? Are we interested in understanding how the system changes over long periods of time, or over large areas?
  • 2. What state variables (e.g., population size, occupancy) and vital rates (e.g., survival) are most relevant to the characterization of the system state and objectives of the management agencies? Alternatively, what variables will allow us to best discriminate between the multiple hypotheses explaining the system state and vital rates? This approach allows us to design our monitoring so that the data we collect are most relevant to an informed conservation process
  • 3. How will we design the project to account for spatial variation and detection, which are necessary for the proper estimation of state variables? This relates to the specific species, habitats, or methods which are chosen to monitor.

Elements of Our Design

  • 1. Surveys are robust to the assumptions of the sampling and analysis methods used.
  • 2. Sampling locations are selected in a probabilistic manner, and the sampling frame is clearly identified in order to clearly draw conclusions about the area of inference.
  • 3. We estimate and incorporate the probability of detecting individuals, species, and occupied patches in our estimates.
  • 4. Using a model-based approach - We can use our monitoring data to discriminate among competing hypotheses/models that predict how amphibians respond to environmental covariates and management because we develop a set of competing models about how the system works in conjunction with our surveys.

Designing a Monitoring Program for Amphibians

Gyrinophilus porphyriticus danielsi, Whiteoak Blowhole Cave, Whiteoak sink, Great Smoky Mountains National Park

Designing a monitoring program about amphibians had a unique set of challenges. ARMI wanted a monitoring program that would provide information about the status of amphibians while simultaneously providing data on their ecology. There are approximately 287 species of amphibians in the United States whose ecology is as diverse as the habitat types they occupy. Further, amphibians have some characteristics that make them difficult to survey. For example: 1) some species are cryptic, even fossorial, and available for detection during short periods of time or in brief periods of the year; 2) several species are difficult to distinguish even when in hand, and 3) no salamanders call, and not all frogs and toads call.

The sampling technique, the design of the monitoring plan, and data interpretation should be congruent with the life history of the species being surveyed and the questions being asked. Examples of characteristics that make designing and interpreting survey data of amphibians difficult include: 1) many species are explosive breeders; 2) many species are "unavailable for sampling" during times when it is too cold, hot or dry for them to be active; 3) some species suffer local short-term extinctions and large fluctuations in population size as a normal part of their life history; 4) some sites (e.g., desert pools) do not exist every year; and 5) some amphibian habitat patches are obviously discrete (e.g., desert pools), while some are not (e.g., Atchafalaya Basin?).

A National Monitoring Program with a Modular Approach

Each of the ARMI Regional Principal Investigators (PI) is responsible for estimating the status of amphibians on a set of lands within a multi-state area. Each PI develops monitoring projects with its own goals and sampling methods. All of the PIs conduct monitoring on the state of a set of amphibian species on an ongoing basis to estimate the proportion of area that is occupied by the species of interest.

Even though we are a national program, we don't collect the same data across all of the regions for several reasons. Because we are using a model-based monitoring approach, each scientist can identify the factors believed to be primarily responsible for driving the populations in the areas to which the sampling is making inference. These factors (i.e.,covariates), are specific to the region, ecology, and history of the species being surveyed. For example, the impact of repeated fires may be an important covariate affecting populations of toads in the Intermountain West, but proximity to urbanization may be more influential to amphibians in the Northeastern U.S. However, there are 2 factors which are common to all of ARMI monitoring: 1) we use a probability based sampling design, and 2) we estimate detection probability.

Detection Probability

Savannah River National Wildlife Refuge, Dan Calhoun sampling, Kingfisher Pond, Jasper County, South Carolina

It is common for investigators to fail to locate every individual animal or every species during a survey. Unless the raw data, such as counts of individuals or species, and occupancy are adjusted for missed detections, the inferences drawn from the data can be misleading. The results will be biased towards those individuals or species which are easier to detect. This can result in biased conclusions and inappropriate management decisions.

However, it is also problematic to develop a single "correction factor" for missed detections and apply it to the species each time it is surveyed. Applying a fixed "correction factor" assumes that time, place, observer; or conditions which influence the probability of detecting individuals never changes, which is extremely unlikely. It is better to estimate detection directly. ARMI has shown how naive data (i.e., detection not estimated) can result in misleading conclusions in several studies (Grant et al. 2005, Mattfeld and Grant 2007).

The development of new quantitative tools has enabled ARMI scientists to increase the types of questions they can ask. Recent developments have allowed ARMI to handle data with missing observations, spatial correlations, and interactions among multiple species. By developing some of these advances and applying others as they become available, ARMI scientists are in the forefront of quantitative research on amphibians.


Links to several software programs available for download, including Program MARK and PRESENCE. These programs are used in the design of monitoring programs and to analyze the data.

Grant, E. H. C., R. E. Jung, J. D. Nichols, and J. E. Hines. 2005. Double-observer approach to estimating egg mass abundance of pool-breeding amphibians. Wetlands Ecology and Management 13:305-320.

Mattfeldt, S., and E. Grant. 2007. Are two methods better than one? Area constrained transects and leaf litterbags for sampling stream salamanders. Herpetological Review 38:43-45.

MacKenzie, D. I., J. D. Nichols, G. B. Lachman, S. Droege, J. A. Royle, and C. A. Langtimm. 2002. Estimating site occupancy rates when detection probabilities are less than one. Ecology 83: 2248-2255.

MacKenzie, D. I., J. D. Nichols, J. E. Hines, M. G. Knutson, and A. B. Franklin. 2003. Estimating site occupancy, colonization, and local extinction when a species is detected imperfectly. Ecology 84: 2200-2207.

MacKenzie, D.I., J.D. Nichols, J.A. Royle, K.H. Pollock, J.E. Hines and L.L. Bailey. 2005. Occupancy estimation and modeling: inferring patterns and dynamics of species occurrence Elsevier, San Diego, USA.

** This file type requires Adobe Reader or Google Chrome Browser (recommended) for viewing.