In USFWS/AKaerial: Analysis of Alaska Region Aerial Survey Data

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

library(AKaerial)

What is AKaerial?

AKaerial is a collection of R functions compiled as an R package that was initiated in 2017 to streamline the analysis and reporting of aerial survey data for the Division of Migratory Bird Management (MBM) in the Alaska Region of the US Fish and Wildlife Service. Pilots and observers in the Alaska Region have been collecting aerial survey data on waterfowl for over 50 years. In general, a pilot and observer fly low-level transects over a designated habitat-based study area (usually breeding grounds) and record observations of single drakes (or single birds for monomorphic species), breeding pairs, and flocks. Observations are summarized as densities, and densities are extrapolated to population indices. Resultant indices are sometimes used in harvest management decisions by both state and federal agencies. AKaerial was created not only to streamline these steps, but to provide a transparent, repeatable, and defensible alternative to past estimate-generating processes.

How does it work?

The path from data collection to report generation is summarized in 3 main steps in AKaerial:

Quality assurance / quality control (QAQC) of data files
Summarizing data files and generating estimates
Visualizing estimates in tables, figures, maps, and reports

1. QAQC

Data quality and consistency are paramount to creating a repeatable workflow. In an ideal setting, the data-generating process is driven by consistent protocol interpretation and data treatment throughout the life of the project. Unfortunately, the realities of a long-term survey program with hundreds of project-observer-year combinations can curtail any automated analysis, especially when standards of data management were simultaneously poorly developed and unenforced. AKaerial provides a suite of functions that work together to "greenlight" a standard aerial survey data file. The greenlight process results in 4 files:

The original, unchanged, raw transcribed data.
An .html report detailing known and potential issues.
The (if necessary, altered) archivable and "final" transcribed data file in .csv format.
An .html report detailing any components that were changed by AKaerial from raw to archivable data.

For a detailed description of QAQC in AKaerial, see the vignette GreenLighting Data, or GreenLight().

2. Generating Estimates

The primary objective of most aerial waterfowl surveys is to generate an index to a population of interest, generally to inform harvest or species recovery decision-making. A long-term survey will also consider trends in an index and growth or decline of an index over time as potential objectives. These indices and estimates are tied spatially and temporally to the spatial and temporal components of the survey (the "design"). The estimation process relies heavily on design decisions. Who conducted the survey? What were they surveying? Where was it flown? When did they fly? How were the observations recorded? In the absence of standardized protocols, many of these components end up being educated guesses. The primary components of the design that are required inputs in AKaerial are:

The spatial coverage of the area of inference, or study area, currently in .shp (ESRI shapefile) format, including any polygons that may represent habitat or other stratification schemes.
The spatial representation of the sample (transects) in the area of inference, numbered the same as the navigational file used by pilots and observers to generate transcribed data files, including the strip width (perpendicular distance from an observer to the edge of the sampled area).

The other required inputs for summarization or estimation are:

A spatially- and temporally-explicit (tied to the design files above), greenlight data file for each observer.
A list of the species of interest to generate estimates for (in the case where observations of secondary species are included in the data file).

AKaerial is currently coded to produce a design-based estimate. That means that we assume the design transect file was followed as closely as possible and can be used as a surrogate for the actual (naturally meandering) track file from the aircraft-mounted global positioning system. This does not preclude removal of transects (in whole or part) when equipment failure or weather-related factors prohibited full completion of the sampling design.

Basic spatial and observation summaries are provided as an additional data quality check once a spatial design is combined with an transcribed data file. We then use the classical ratio estimator described by Cochran (1977) modified for stratification and unequal transect lengths to summarize observations and design files as population indices.

For a detailed description of summaries and estimates in AKaerial, see the vignette Generating Estimates, or functions DataSelect and Densities.

Visualization

Visualizing data inputs and subsequent output can provide powerful insights on data quality, integrity, and general survey results. One of the most important features of AKaerial is its ability to generate on-demand tables, figures, maps, and reports of design files, observations, and estimates. AKaerial contains a series of functions that integrate Geographic Information Systems, design files, and transcribed observations to provide both publication-quality figures and interactive maps.

For a detailed description of summaries and estimates in AKaerial, see the vignette Visualization.