My issue is that I am working on a really extensive dataset and that my tracking data encompass several UTM zones (12 to be specific: 32-33-34-35-36-37-38-39-40-41-42-43).

Snip of my data:

> head(springsub) id DateTime lat lon Altitude Sex Age Tag distance time_interval velocity velocity_kmh 1 N34234@1BF1A215 2022-03-15 10:42:00 45.50620 12.57837 5 1 6 GPS-GSM_TECHNOSMART 0.00000 NA NA NA 2 N34234@1BF1A215 2022-03-16 06:42:00 48.62932 20.51655 274 1 6 GPS-GSM_TECHNOSMART 695286.88086 72000 9.656762234 34.76434404 3 N34234@1BF1A215 2022-03-16 08:42:00 48.62761 20.51720 262 1 6 GPS-GSM_TECHNOSMART 195.38842 7200 0.027137281 0.09769421 4 N34234@1BF1A215 2022-03-16 10:42:00 48.62785 20.51662 271 1 6 GPS-GSM_TECHNOSMART 50.60765 7200 0.007028840 0.02530382 5 N34234@1BF1A215 2022-03-16 12:41:00 48.62737 20.51678 270 1 6 GPS-GSM_TECHNOSMART 54.77180 7140 0.007671121 0.02761603 6 N34234@1BF1A215 2022-03-16 18:42:00 49.30252 20.92450 914 1 6 GPS-GSM_TECHNOSMART 80799.38930 21660 3.730350383 13.42926138 utm_zone 1 33 2 34 3 34 4 34 5 34 6 34

Now many tracking data analysis workflows use Easting and Northing in UTM coordinates instead of longitude and latitude (which is what I have for now).

I tried projecting my tracks in UTM but obviously they get extremely distorted (when projecting around a “central” UTM zone).

# Project to UTM llcoord <- st_as_sf(springsub[, c("lon", "lat")], coords = c("lon", "lat"), crs = CRS("+proj=longlat +datum=WGS84")) utmcoord <- st_transform(llcoord, crs = CRS("+proj=utm +zone=35 +datum=WGS84")) # add Easting and Northing to data (in km) data[, c("x", "y")] <- st_coordinates(utmcoord)/1000

What should I do?
Splitting by UTM zones as in Converting latitude and longitude data to UTM with points from multiple UTM zones in R doesn’t sound like a good idea considering that each track encompasses several zones and represents an individual movement.

Should I use a Lambert projection or an Equal Earth projection Projecting long/lat with multiple UTM zones instead?

Would that mean still using Easting and Norting metric information suitable for (as an example) a Hidden Markov Model analysis?

Thank you