The statistical methods will not be detailed in this document but should soon be available as a vignette, a publication, or both (we are currently working on it).
Isofit:The function Isofit fits a geostatistical model, which approximates the relationship between the topographic features of a location and its isotopic signature (see Isofit for details). The model fits observations of isotopic delta values at several geographic locations (hereafter, called sources). One common type of sources used in ecology is the delta values for deuterium in precipitation water collected at weather stations, but one may also use measurements performed on sedentary organisms. In either case, the accuracy of the isoscape (and thereby the accuracy of assignments) increases with the number and spatial coverage of the sources. Because fitting a geostatistical model may take several hours for large datasets, we have stored an already fitted model for users willing to explore our package (see GNIP_Model).
RElevate:Building isoscapes and assigning organisms to their origin requires an adequate elevation raster, i.e. a matrix representing altitude data on a spatial grid. The function RElevate allows restricting the extent of the raster to the area covered by isoscape data (in order to avoid extrapolation) and to reduce the resolution of the original elevation raster (in order to speed up computation in all following steps). Note that aggregating the raster may lead to different results for the assignment, because the elevation of raster cells changes depending on the aggregation function, which in turn will affect model predictions.
We will soon provide a link to download an elevation raster for the entire world at a resolution of one altitude per square-km, and other rasters may be used. We have also stored a low resolution raster in our package (see elevationrastersmall) for users to try things out, but we do not encourage its use for real application.
Isoscape:The function Isoscape generates the isoscape: it uses the fitted geostatistical model to predict the isotopic values for each raster cell defined by the elevation raster. Our package allows the production of fine-tuned isoscape figures (using the function plot.isoscape). Alternatively, the isoscape rasters can be exported as ascii raster and edited in any Geographic Information System (GIS) software (see Isoscape for details).
Calibfit:In most cases, organisms are of another kind than the sources used to build the isoscape (e.g. the isoscape is built on precipitation isotopic values and organisms are not water drops, i.e. the deuterium values of their keratin structures were modulated by their distinct physiology). In this situation, one must use sedentary organisms to study the relationship between the isotopic values in organisms and that of their environment. The function Calibfit fits a statistical model on such a calibration dataset.
If the isoscape is directly built from isotopic values of organisms, there is no need to fit a calibration model.
Isorix:The function Isorix tests for each location across the isoscape if it presents a similar isotopic signature than the unknown origin of a given individual(s). This assignment procedure considered the uncertainty stemming from the model fits (geostatistical model and calibration model). The function plot.isorix then draw such assignment by plotting the most likely origin with the prediction region around it. When several organisms are being assigned, both individuals assignments and a single assignment for the whole group can be performed.
### A simple workflow for IsoriX
### is provided in a vignette:
### vignette("Workflow", "IsoriX")
Run the code above in your browser using DataLab