diveMove (version 0.7-4)

austFilter: Filter satellite locations

Description

Apply a three stage algorithm to eliminate erroneous locations, based on the procedure outlined in Austin et al. (2003).

Usage

austFilter(time, lon, lat, id=gl(1, 1, length(time)),
           speed.thr, dist.thr, window=5)
grpSpeedFilter(x, speed.thr, window=5)
rmsDistFilter(x, speed.thr, window=5, dist.thr)

Arguments

time

POSIXct object with dates and times for each point.

lon

Numeric vectors of longitudes, in decimal degrees.

lat

Numeric vector of latitudes, in decimal degrees.

id

A factor grouping points in different categories (e.g. individuals).

speed.thr

Speed threshold (m/s) above which filter tests should fail any given point.

dist.thr

Distance threshold (km) above which the last filter test should fail any given point.

window

Integer indicating the size of the moving window over which tests should be carried out.

x

3-column matrix with column 1: POSIXct vector; column 2: numeric longitude vector; column 3: numeric latitude vector.

Value

• grpSpeedFilter returns a logical vector indicating those lines that passed the test.

  rmsDistFilter and austFilter return a matrix with 2 or 3 columns, respectively, of logical vectors with values TRUE for points that passed each stage. For the latter, lines that don't pass the first stage do not pass the other stages. The second and third columns returned by austFilter, as well as those returned by rmsDistFilter are independent of one another; i.e. lines that don't pass stage 2 do not necessarily not pass stage 3.

Warning

This function applies McConnell et al.'s filter as described in Austin et al. (2003), but other authors may have used the filter differently. Austin et al. (2003) have apparently applied the filter in a vectorized manner. It is not clear from the original paper whether the filter is applied iteratively or in a vectorized fashion, so authors may be using it inconsistently.

Details

These functions implement the location filtering procedure outlined in Austin et al. (2003). grpSpeedFilter and rmsDistFilter can be used to perform only the first stage or the second and third stages of the algorithm on their own, respectively. Alternatively, the three filters can be run sequentially using austFilter.

The first stage of the filter is an iterative process which tests every point, except the first and last two, for travel velocity relative to the preceeding/following two points. If all these four speeds are greater than the specified threshold, the point is marked as failing the first stage. In this case, the next point is tested, removing the failing point from the set of test points.

The second stage runs McConnell et al. (1992) algorithm, which tests all the points that passed the first stage, in the same manner as above. The root mean square of all four speeds is calculated, and if it is greater than the specified threshold, the point is marked as failing the second stage.

The third stage is run simultaneously with the second stage, but if the mean distance of all four pairs of points is greater than the specified threshold, then the point is marked as failing the third stage.

The speed and distance threshold should be obtained separately (see distSpeed).

References

McConnell BJ, Chambers C, Fedak MA. 1992. Foraging ecology of southern elephant seals in relation to bathymetry and productivity of the Southern Ocean. Antarctic Science 4:393-398.

Austin D, McMillan JI, Bowen D. 2003. A three-stage algorithm for filtering erroneous Argos satellite locations. Marine Mammal Science 19: 371-383.

See Also

distSpeed

Examples

locs <- readLocs(system.file(file.path("data", "sealLocs.csv"),
                             package="diveMove"), idCol=1, dateCol=2,
                  dtformat="%Y-%m-%d %H:%M:%S", classCol=3, lonCol=4,
                  latCol=5)
summary(locs)

ringy <- subset(locs, id == "ringy")

## Austin et al.'s group filter
grp <- grpSpeedFilter(ringy[, 3:5], speed.thr=0.9)

## McConnell et al.'s filter (root mean square test), and distance test
rms <- rmsDistFilter(ringy[, 3:5], speed.thr=0.9, dist.thr=100)

## All three tests in sequence (Austin et al. procedure)
austin <- austFilter(locs$time, locs$lon, locs$lat, id=locs$id,
                     speed.thr=0.9, dist.thr=100)