FatalityCMR is an R-package to fit a superpopulation capture-recapture model to carcass count data. It uses trial data to estimate et correct for detection and daily persistence probabilities. It has an ‘evidence of absence’ mode for the case when no or few carcasses are found. It allows users to distinguish between different turbine types (e.g. different vegetation cover under turbines, or different technical properties), as well between two carcass age-classes or states, with transition between those classes (e.g, fresh and old). A data simulation feature may be used at the planning stage to optimize sampling design. Resulting mortality estimates can be used 1) to quantify the required amount of compensation, 2) inform mortality projections for proposed development sites, 3) inform decisions about management of existing sites, and 4) improve the design of carcass search protocols and trial experiments.
Download here from USGS webpage
Pairs with this paper in J. of Applied Ecology
User’s manual here
Using the R-package nestAbund, nest searching and hatching/fledging success data can be used to estimate nest density and thereby make more precise inference about population-environment relationships, habitat quality, and habitat choice. In its current version nestAbund mostly works for ducks/geese nests but there are plans to expand it for use with passerines and near-passerines. The inference is based on the age of the nests at first detection, determined using the candling method, the age of the nestling, or other method. The software allows modeling separately daily nest survival and detection probabilities. Both continuous and binary individual (nest-level) covariates are allowed. By simultaneously estimating nest success and nest density, a more complete picture of bird population dynamics can be obtained, and more precise hypotheses can be tested that pertain to density-dependence and habitat choice.
Download here from USGS webpage
Pairs with this paper in Ecology
Surveys of animal population are affected by multiple sources of error a.k.a. sampling variance: 1) imperfect detection; 2) movements in and out of the survey area; 3) group-living which causes deviations from usual distributions; 4) variation among observers. Chamois is a software to address this issue of sampling variance in count data. By analyzing repeat count data, and allowing several options to estimate imperfect detection probability (distance sampling, time-to-detection, and multiple observers), it separates sampling variance from process variance, such as that caused by variation between sites and across time. The software also accomodates irregularly-shaped survey plots, making it especially designed for mountain ungulates. Its very basic algorithm is both its drawback (not very time efficient) and its advantage (quite robust and eventually always converges to a global minimum).
Here is the user’s manual.
Update: We will be looking for a Master’s student (M2 or Ingénieur) to finish this work. The project would be tailored to the student’s interest but would need to incorporate some form of benchmark including existing options and simpler alternatives to our software. Possibilities exist to work more on the connection with field work (in the French Alps) or to participate in further software development. Experience with R or a computer science background required.