You can get some good information from the BBS site on routes, etc. but you may need to convert it to a more friendly format (i.e. a CSV file). Once you have your dataframes linked up, you can spatially subset based on some of the data that has already been assigned per BBS route…

 BBS <- BBS[BBS$BCR %in% c(1,2), ]  

This one may not be as step-by-step, because the GIS layer of BBS routes I have may not have been published yet. Nonetheless, here’s what I did.

1. Cleaning up the spatial data and splitting routes
2. Assigning segment order once the routes are split
3. Summarizing spatial covariates by segment
4. Determining which segments can be considered the same
5. Importing BBS and spatial data into R at the segment level

The final step looks similar to the route-level tutorial, but with pieceID being the variable to match on. Set “bird” to the AOU code of your species of interest.

dataperspecies <- subset(BBS[which(BBS$Aou == bird),]) 
vars$abundance <- dataperspecies$count[match(vars$pieceID, dataperspecies$pieceID)]
vars[is.na(vars)] <- 0

assess <- data.table(vars)
whattodrop <- assess[,list(abundance=sum(abundance)),by='rteno']
droprtes <- whattodrop$rteno[whattodrop$abundance==0]
perspp <- vars[!(vars$rteno %in% droprtes),]
perspp$Year <- as.integer(perspp$Year)
perspp$yearfix <- perspp$Year
perspp <- perspp[perspp$RunType > 0,]
#vars <- vars[vars$RPID < 102]
perspp <- perspp[c("Year","MN","firstyear","rteno","ObsN","abundance")]

Here is the full script: bbs_10stop_species

The BBS was designed to capture avian populations in natural areas, so if the land surrounding a portion of the route was developed, historically the route path was altered and given a new number. The problem is, that new numerical designation was not standardized (though often, it’s 100 + the original route number).

In wanting to maintain spatio-temporal continuity, I thought that the portions of the routes that were still spatially congruent should be considered the same segment. This serves my random effect of location, such that an observation isn’t considered to come from a different location just because the route number is different (if the “old” and “new” segments spatially coincide). I use “coincide” loosely, in that I considered original and +100 route segments to be the same if they intersected. You can measure the degree of overlap and incorporate that if you choose. I made a shape file of intersecting duplicated segments, and imported it into R to adjust the location designations.

repeats <- read.dbf("intersecting_duplicated_segment.dbf")
repeats$new_rteno <- repeats$rteno - 100
repeats$new_route <- paste(repeats$new_rteno,repeats$segment,sep="_")

Now, the newer segments have the same numbering as the original routes. I match the new segments with the original route numbers to the BBS data, and then replace the new route segment numbers with the original numbering.

BBS$new_route<-repeats[match(BBS$route_piece, repeats$route),"new_route"]
BBS$new_rteno<-repeats[match(BBS$route_piece, repeats$route),"new_rteno"]
BBS[!is.na(new_route)]$route_piece <- BBS[!is.na(new_route)]$new_route
BBS[!is.na(new_route)]$rteno <- as.integer(BBS[!is.na(new_route)]$new_rteno)

The new route segments that intersect the original routes now have the old route numbering, and thus counts from these segments appear in the code to to have come from the same routes as the original.

Here’s the full script: bbs_10stop

1. Break up the accounted for routes (or maybe even raw data set?) into “good” and “bad” routes depending on if the start is nearest a segment for a matching route
2. Work with “good starts” and “good routes” (the easiest!)
   1. identify the segments closest to the start point as the first segment of the route
      1. run near on the start points to find out which segment in the master file is closest
      2. use the near_FID to join with the FID of the route segment layer
      3. export the resultant start segments into their own layer (number them 1)
      4. remove join
   2. delete the start segments from the master file of all the broken up segments of the route (select by location: identical to)
   3. use select by location (touches the boundary of) to find the next closest line segment to the start in the master file, and number them 2
   4. repeat 2-4 for the remaining segments, and number them accordingly
3. Work with bad starts: once the good routes are out of the data set, is the bad point now closest to a remaining matching segment?
   1. break up those newly matched lines
   2. put leftover points into yet another file
4. Look at remaining points to see if any more lines can be put to use

Given the numbering scheme recommended in the steps above, your segments should be numbered 1-5 on each route. I buffered those segments and averaged tree cover within  each segment buffer. In order to match my segment numbering with the rest of my R code, I renamed the segments.

vcf_subroute <- read.dbf("landcover/bbs_segment_buffers.dbf")
vcf_subroute$interval <- paste("count",vcf_subroute$segment,"0",sep="")

The remainder of the code is similar to that described in the route-level MODIS post, just with the added level of the interval.

vcf_subroute <- vcf_subroute[,c("rteno","interval",grep("[[:digit:]]", names(vcf_subroute), perl=TRUE, value=TRUE))]
vcf_subroute <- melt(vcf_subroute,id=c("rteno","interval"))
vcf_subroute$variable <- as.character(vcf_subroute$variable)
vcf_subroute$Year <- str_extract(vcf_subroute$variable,"[[:digit:]]+")
vcf_subroute$var <- sapply(strsplit(vcf_subroute$variable, "[[:digit:]]+"),"[[",2)
vcf_subroute$variable <- NULL
vcf_subroute <- dcast(vcf_subroute,rteno + interval ~ var,fun.aggregate=mean)
vcf_subroute$rteno <- as.character(vcf_subroute$rteno)

The full script, to this point: bbs10vcf