3

I have a series of coordinates for fish caught from a boat at different datetimes and different trips. How do I determine whether the coordinates of a fish are likely to be incorrect (e.g. due to transcription error) based on time since last fish caught within that same trip and an assumed boat speed (say 10km/hour).

Here is a simple example dataset with 2 trips and two fish per trip.

library(sf)
library(ggplot2)

datetime <- ymd_hms('2017-05-13 14:00:00', tz = "Etc/GMT+8")
df <- data_frame(DateTimeCapture = c(datetime, datetime + minutes(35), datetime + days(2), 
                                     datetime + days(2) + minutes(20)),
                 Trip = c('1', '1', '2', '2'),
                 Order = c(1, 2, 1, 2),
                 X = c(648635, 648700, 647778, 658889),
                 Y = c(5853151, 5853200, 5854292, 5870000))

# if you prefer to work in sf
df_sf <-  st_as_sf(df, coords = c('X', 'Y'), crs = 32610)

# quick plot
ggplot() + 
  geom_point(data = df, aes(x = X, y = Y, color = Trip)) 

enter image description here

The distance between the two fish in the second trip is 19km:

st_distance(df_sf[3:4, ])
Units: m
         [,1]     [,2]
[1,]     0.00 19240.47
[2,] 19240.47     0.00

It is unlikely that a boat could travel 19km in 20 minutes. Thus this should be flagged as a possible error.

My preference is for solutions using sf, but may also accept solutions using sp. It has to be r-based solution.

  • looks like you're using library(lubridate) & library(dplyr) too? – jbosq Mar 30 '18 at 19:30
1

There is undoubtedly an sf solution that is more elegant and better accounts for the CRS, but here are my ugly 4 lines.

Step 1: Get the elapsed time between trips, and convert that to a max distance

fill <- NA
transform(df, elapsedTime = c(fill, diff(as.numeric(DateTimeCapture))) * c(fill, 1))

This works on the example 2 catches per trip, but here are alternatives when that varies. Use the same line to convert time to distance (10km/hr -> 2.77778m/sec)

df2 <- transform(df, maxDist = c(fill, (diff(as.numeric(DateTimeCapture)))*2.77778) * c(fill, 1))

Step 2: Determine catch distance and if that is greater than max distance

c^2 = a^2 + b^2, or distance = sqrt(sum((vector1-vector2)^2))

df3 <- transform(df2, dist = c(fill, sqrt(diff(X)^2 + diff(Y)^2)) * c(fill, 1))
#determine if the distance is greater than max dist
df3$dist>df3$maxDist

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.