Case study 4: The influence of range size on the relationship between species richness and body size

Mihai Valcu

2016-12-05

Appendix S5

Supporting information in Valcu, M., Dale, J., and Kempenaers, B. (2012). rangeMapper: a platform for the study of macroecology of life-history traits. Global Ecology and Biogeography 21, 945-951.

The example shown here is run on the wrens dataset which is part of the package. The wrens dataset has 84 species while the case study presented in the paper was run on 8434 bird species. Therefore both the settings and the results shown below are not identical with the results presented in Valcu et al 2012.

Project Set Up

require(rangeMapper)
breding_ranges = rgdal::readOGR(system.file(package = "rangeMapper",
     "extdata", "wrens", "vector_combined"), "wrens", verbose = FALSE)
data(wrens)
d = subset(wrens, select = c('sci_name', 'body_mass') )
con = ramp("wrens.sqlite", gridSize = 1, spdf = breding_ranges,
             biotab = d, ID = "sci_name",metadata = rangeTraits()['Area'],
             overwrite = TRUE)
## New session 2016-12-05 16:10:12
## PROJECT: wrens.sqlite 
## DIRECTORY: /tmp/RtmpLQb5kF
## Warning: Column name mismatch, columns will be matched by position. This
## warning may be converted to an error soon.
## Bounding box uploaded.PROJ4STRING set to +proj=longlat +datum=NAD83 +no_defs +ellps=GRS80 +towgs84=0,0,0
## Warning: Column names will be matched ignoring character case
## Grid size set to 1 map units.
## Canvas uploaded.
## Writing to project.
## 84 ranges updated; Elapsed time: 0.8 secs
## Extracting metadata...
## Table biotab saved as a  BIO_ table
## +______________+
## class            rangeMap
## Project_location /tmp/RtmpLQb5kF/wrens.sqlite
## Proj4            +proj=longlat +datum=NAD83 +no_defs +ellps=GRS80 +towgs84=0,0,0
## CellSize         1
## Extent           xmin=-164.8716,xmax=-34.60225,ymin=-55.72056,ymax=61.5632
## BIO_tables       biotab
## MAP_tables       species_richness
## METADATA_RANGES  Area
## +______________+

Range size classes

mt = RSQLite::dbReadTable(con, "metadata_ranges")

Q = quantile(log(mt$Area),  probs = seq(0.05, 1, 0.1) )
rangeA = data.frame(area = exp(Q), quant =  gsub("%", "", names(Q)) )

Run log10(median_body_mass) ~ sqrt(species_richness) regression for each range size interval

W = 4   # size of the moving window

output = vector(mode = "list", length = nrow(rangeA))
names(output) = rangeA$quant

for(i in seq(1:(nrow(rangeA) - W) ) ) {

  # Define  map subset
  area_subset =list(metadata_ranges = paste("Area between",  rangeA[i,"area"],
  "and", rangeA[i+W,"area"]))

  # Save map
  rangeMap.save(con, subset = area_subset , biotab = "biotab",
  biotrait = "body_mass", FUN = "median", tableName = "median_body_mass",
  overwrite = TRUE)

  # Fetch map
  m = rangeMap.fetch( con, c("species_richness", "median_body_mass"), spatial = FALSE )

  # Perform OLS regression
  # NOTE: In order to perform a spatial simultaneous autoregressive error
   # regression
  #  see 'errorsarlm ' function and the auxiliary neighbours list functions in
   # 'spdep' package.
  fm = lm( log10(median_body_mass) ~  sqrt(species_richness), m)

  output[[i]] =  fm

}

output = output[!sapply(output, is.null)]

Extract Regression Parameters and plot

X = lapply(output, function(x) data.frame(slope = coef(x)[2],
        ciu = confint(x)[2,1], cil = confint(x)[2,2]) )
X = do.call(rbind, X)
X$rangeSize = as.numeric(row.names(X))

# Plot

require(ggplot2)

ggplot(X, aes(x = rangeSize, y = slope)) +
    geom_errorbar(aes(ymin = cil, ymax = ciu), width= 0) +
    geom_line() +
    geom_point() +
    theme_bw()