writeShapefile producing invalid geometries

[Brittany B]

Hello! First of all, I want to thank you so much for incorporating the ability to implement hard boundaries within the akde function in this version of ctmm. Everything is working beautifully for a species of bass I am working with in a riverine environment, but when I export the 95% UD polygons I get invalid/self-intersecting geometries in the "high" estimate (see attached image). Any quick fixes for this issue?

Best,

Brittany

)

[Christen Fleming]

Hi Brittany,

Weird. When you plot the contours in ctmm, do they look messed up like this, or is it only after exporting? In either case, if you can send me a minimal working example (data + script), then I can take a look at it.

Best,

Chris

[Brittany B]

Sure thing--they look great when plotted in ctmm, and the errors only occur once exported. I imported the polygon boundary layer as a shapefile--can I send that over to you via e-mail? 

Here's data from one tag:

LotekID    Lat    Long    DateTime
58    27.40528    -81.15264    2021/02/08 14:40:00+00
58    27.41154    -81.14277    2021/02/18 14:48:00+00
58    27.41135    -81.14317    2021/03/04 00:00:00+00
58    27.41168    -81.14264    2021/03/08 16:07:00+00
58    27.41178    -81.14234    2021/03/15 16:01:00+00
58    27.41044    -81.1447    2021/03/22 15:59:59+00
58    27.41235    -81.14223    2021/03/30 15:23:00+00
58    27.41007    -81.14509    2021/04/05 15:57:00+00
58    27.4119    -81.14245    2021/04/23 15:30:00+00
58    27.41071    -81.1391    2021/06/03 11:31:00+00

Script:

locs_xy<-locs_df[c("Long","Lat")]
coordinates(locs_xy)<-~Long+Lat
proj4string(locs_xy)<-CRS("+init=epsg:4326")
ref<-CRS(SRS_string = "EPSG:32617")
locs_utm<-spTransform(locs_xy, CRSobj=ref)
locs_utm_spdf<-SpatialPointsDataFrame(locs_utm, data=locs_df)
locs_utm_spdf$Name<-as.factor(locs_utm_spdf$LotekID)
locs_utm_spdf$hdop<-1 #locations were typically exact with confirmed visual sighting

ref<-CRS("+init=epsg:32617")
GPS.HDOP<-as.data.frame(locs_utm_spdf$hdop)
hdop<-as.data.frame(locs_utm_spdf$GPS.HDOP)
names(GPS.HDOP)[1]<-"GPS.HDOP"
date<-as.POSIXct(locs_utm_spdf$DateTime)
xy<-as.data.frame(cbind(locs_utm_spdf$Long,locs_utm_spdf$Lat))
ids<-as.factor(locs_utm_spdf$Name)
trajectories<-as.ltraj(xy=xy, date=date, id=ids, typeII = T, infolocs=GPS.HDOP)

# Create Move object
locs_move<-move(trajectories)
proj4string(locs_move)<-"+proj=utm +zone=17 +datum=WGS84 +units=m +no_defs"

# Create telemetry object
locs_telem<-as.telemetry(locs_move, keep = FALSE, projection="+proj=utm +zone=17 +datum=WGS84 +units=m +no_defs")

## Setting the RMS UERE
uere(locs_telem)<-1
L58<-locs_telem$X58_58
SVF<-variogram(L58)
SVf_fit<-variogram.fit(SVF) #Interactive fit, make sure to check error box for irregular sampling
L58_sel<-ctmm.select(L58, CTMM=GUESS, IC="AICc", verbose=TRUE)
L58_fit<-ctmm.fit(L58, GUESS)
L58_akde<-akde(L58, L58_fit, SP=SP1_spdf, SP.in=T,weights=T, grid=2,debias = T, smooth=T)
writeShapefile(L58_akde, ".",level.UD=0.95, level=0.95, "L58", overwrite=T)

Thank you for looking into this!

Best,

Brittany

[Christen Fleming]

Yes, please send me everything via email.

Best,

Chris

[Christen Fleming]

Hi Brittany,

You are calculating an AKDE with a model from ctmm.fit rather than ctmm.select. The ctmm.fit model is far from the selected model and has extremely wide confidence intervals, which cause some coverage area polygons with <4 vertices, which sp doesn't like (I tried removing those, but sp just gave a different error instead). If you use the result of ctmm.select instead, it looks fine after exporting.

If this happens on a selected model where it matters, then let me know and I'll see if using sf instead will prevent this.

Best,

Chris