I have survey data, with recordings every 1 NM along parallel transects perpendicular to the coast. For each record I have information such as latitude and longitude, speed, bearing. Along the transects the speed is around 10 knots. I also have points along the inter-transects (where speed might be different, and bearing definitely is) and I have points outside the transects if a trawl was carried out.

What I would like to do is grouping all the points belonging to the same transects (e.g. see figure):

[

As you can see in the figure, this does not really work because whenever there is an inter-transect (like transect nr 5), this will be grouped together with the transect itself. Also, in transect 13, since the distance for some reason is slightly larger then 1 NM between 2 subsequent records, the points are separated into 2 transects (you can see the slightly different color).

Dataframe example shown here:

 |year  |datetime             |xkm   |ykm  |logdiff |time_diff |distance |bearing |speed |
 |<dbl> |<dttm>               |<dbl> |<dbl>|   <dbl>| <drtn>   |    <dbl>|   <dbl>| <dbl>|
 |------|----------           |------|-----|--------|------  --|---------|--------|------|
 |2023  |2023-09-26 15:03:00  |221.  |1606.|       1| 360 secs |        1|   -1.58|    10|
 |2023  |2023-09-26 15:08:00  |223.  |1606.|       1| 300 secs |        1|   -1.58|    12|
 |2023  |2023-09-26 15:14:00  |225.  |1606.|       1| 360 secs |        1|   -1.58|    10|
 |2023  |2023-09-26 15:19:00  |227.  |1606.|       1| 300 secs |        1|   -1.58|    12|
 |2023  |2023-09-26 15:25:00  |229.  |1606.|       1| 360 secs |        1|   -1.84|    10|
 |2023  |2023-09-26 15:30:00  |231.  |1606.|       1| 300 secs |        1|   -1.85|    12|
 |2023  |2023-09-26 15:36:00  |233.  |1606.|       1| 360 secs |        1|   -1.58|    10|
 |2023  |2023-09-26 15:41:00  |234.  |1605.|       1| 300 secs |        1|   -1.85|    12|
 |2023  |2023-09-26 15:47:00  |236.  |1605.|       1| 360 secs |        1|   -1.58|    10|
 |2023  |2023-09-26 15:52:00  |238.  |1605.|       1| 300 secs |        1|   -1.58|    12|

What could be the best approach to tackle that in R? I thought about some unsupervised machine learning algorithm, like clustering using dbscan, but I’m not sure I’m using it properly. Beside distance between points, I would like to use other parameters to classify whether a point belongs to a transect or not (such as bearing and speed for example).

My attempt:

# Prepare data for clustering
clustering_data <- df %>% select(year, speed, bearing, xkm, ykm)

# Apply DBSCAN clustering
set.seed(123)
db <- dbscan(clustering_data, eps = 1.8, minPts = 5)

Any suggestion?

Thanks,

Val