(version 1.0 does not exist yet)
feature requests can be defined and watched here: https://github.com/courtiol/IsoriX/issues
the previous released introduced an error in how the variance of the assignment test is computed in the absence of calibration (with important consequence in terms of assignments). This is now fixed (#151).
the base package 'colourspace' is now suggested to avoid a note in R CMD check.
the function 'calibfit()' gains an argument method that allows for selecting one of four calibration methods ("wild", "lab", "desk", "desk_inverse"). This allows for users to use 1) calibration samples associated with unknown environmental isotopic values, 2) calibration samples associated with known environmental isotopic values, or 3 & 4) the intercept and slope of a calibration relationship computed by others (e.g. values found in a paper). Note: the desk* methods allow for the consideration of a fractionaction factor too (i.e. slope = 0). See
calibfit for details. (#20 & #142)
the function 'getelev()' has been completely rewriten so as to rely on the package **elevatr** to download elevation data. You should check '?getelev' for learning how to use the new version of the function, but we retained the core principle of the previous function so that old workflow will only require minor adjustements. The new version still saves a *.tif file on the disk, albeit uing a different file name to avoid (data) confusion. (#140 & #107)
the function 'isofind()' gains an argument 'neglect_covPredCalib' that allows for the computation of a covariance term that was so far neglected in IsoriX. See '?isofind' for details. (#143)
the function 'prepraster()' gains an argument 'values_to_zero' to turn a range of elevation values to zeros (nullify negative elevation values by default). This is particular useful because the new version of 'get_elev()' download an elevation raster that includes bathymetry.
new internal function '.invert_reg()' to invert regression (used for method "desk_inverse" in 'calibfit()')
when calling 'plot()' on an object created with 'calibfit()', the plotting function now returns the fitted values and CI for users to be able to make alternative plots (#44)
new argument 'xlim' for the plotting function for calibration fits
new argument 'line' for customising how to plot the regression line in calibration fits
the summary method for calibration fits now displays the residual variance
'calibfit()' performs more check on extrapolation (#119)
when using 'plot()' on an object of class ISOFIT, the x-axis for the plot showing the Matern correlation should have a range more adequate irrespective when autocorrelation is strong over short distances (#134)
documentation for '?plot()' now contains a description of what symbols mean in plots (#138)
when calling 'plot()' on an object created with 'isofind()', the plotting function now detects sample of size 1 and no longer displays "Group" in the title of the assignment plot even if 'who' = "group" (#120)
all functions accepting a 'data.frame' as input should also now be compatible when provided with a 'tibble' (#118)
typos have been corrected (#130)
default y-axis title changed to "Isotopic value in the environment" when plotting calibration fits to be flexible enough irrespective of the methods used in 'calibfit()'
the argument 'long_min', 'long_max', 'lat_min' & 'lat_max' function 'prepsources()' now have explicit default values and should no longer be missing.
the version of spaMM required by IsoriX has changed to 3.13 so as to benefit from a new extractor we rely on for the computation of the 4th variance term during assignment (#143)
the function depending on the package RandomFields are no longer available since that package has been (for now) retired by CRAN :-(
IsoriX should now work with tibbles as inputs (#118)
the printing method for the object of class ISOSCAPE was somehow not exported and thus not used (unreported issue)
plotting on a sphere ISOFIND objects did not work in some cases (#126)
argument ylim for the plotting function for calibration fits
it is now possible to calibrate data containing missing isotopic values
it is now possible to assign data containing missing isotopic values
the SpatialPolygons CountryBorders and OceanMask have been rebuilt for possibly improving the compatibility with new sp & rgdal
the website for WorlClim has now changed address, so links have been updated
rgdal is now listed as a suggested package
several weblinks had changed and have been updated
all old defunct functions have been removed from the package
fix issue #113: the plotting function was not working for isoscapes not stored in memory due to a wrong use of the quantile function. Many thanks to Dr. Gary Roemer and Amy Withers for reporting it!
the datasets used in Courtiol et al. 2019 are now provided
many useful functions from raster, rasterVis, lattice... are now re-exported so they can be used without attaching those packages
new option in plots that allows to map the isoscape onto a sphere
a new dataset PrecipBrickDE containing monthly precipitation amounts for Germany
an argument y_title for the plotting function for isoscapes to allow one to simply change the title
arguments xlab and ylab for the plotting function for calibration fits
new method points for plotting more than one calibration fit
the plotting function for assignments can now show the location of the assignment samples
the citations for the package have been updated!
many objects have been renamed to prepare the release of the version 1.0
the vignettes have now been moved to a bookdown. To access the documentation you should now visit: https://bookdown.org/content/782/
all arguments 'bla.bla' have been renamed 'bla_bla'
the plotting function for calibfit gains an argument "..." for more control
a ploting method for rasterLayer has been included for conveniance
the function relevate is now called prepraster
the function prepdata is now called prepsources
in several functions the argument elevation.raster has been renamed as raster
in several functions the argument xxx.data has been renamed as data
the file storing the internal functions is now called zzz.R
the dontrun and donttest have been replaced by comments due to new R CMD check flags
the function downloadfile is now exported
large temporary objects are now deleted within isofind to limit memory usage
the package is now being tested using testthat, but tests will be implemented in the future
a lot of the internal code as been rewriten to comply more closely to the IsoriX coding style
the list of suggested packages has been revised and rgdal removed as it caused (again) problems with Travis CI
following a change in spaMM predict.HLfit, the prediction are now being made by chunck of 1000 points instead of 150. This should lead to a tiny gain in performance
the function isoscape was performing predictions twice every 150 (or now 1000) locations, this was not influencing the isoscapes produced, but this has now been corrected
the function prepraster now produces an raster stored in memory if it is possible. This should prevent bugs that appears when using loaded rasters that were previously saved (the temporary link to the hard drive location is no longer correct in this case).
the function .objective_fn_calib has been moved within the function calibfit as it is not used elsewhere
the function calibfit as been prepared for a possible activation of a random effect for species ID in the future. But whether it would make sense or not remains to be determined.
the function .Fisher_method now directly computes the exponential of the log pv if only one value is provided. This leads to much faster assignment in the case of a single observation.
the plotting function for calibration fit was displaying CI based on variance instead of SD
the function getprecip and prepcipitate were not handling paths manualy defined properly
the plotting functions were crashing in case of no variation in the landscape
the plotting functions were crashing when called on multiple-raster objects not stored 'inMemory'
the plotting function for fitted model was not displaying one plot in RStudio when called on objects of class MULTIISOFIT
this is a minor update necessary to maintain compatibility with spaMM 2.4
the syntax for the extraction of correlation terms of spaMM objects has changed
the calibration step is now optional, allowing for users to use an isoscape directly fitted on tissues instead of precipitation water
the function queryGNIP has been renamed and is now called prepdata, this function can also handle other datasets than GNIP
the function relevate has been modified to make crop possible around the pacific meridian -180/180 (but several issues remain to handle extra plot layers automatically)
an additional options as been added to prevent prompting during examples
new internal function .converts_months_to_numbers
the maximum duration of running time for examples can now be controlled using IsoriX.options(example_maxtime = XX)
due to new GNIP policies, we no longer provide the GNIP dataset for the entire World, but only a subset containing data for Germany (users should thus compile their precipitatin data themselves from the 'wiser' plateform provided by GNIP; see vignette Workflow)
it is now possible to control the colours and labels for the levels of isotopes or p-values in plots
for plotting, it is no longer needed to load the ocean mask and country borders (it now happens automatically)
the function relevate now allows for a cropping larger than the extent of the weather stations by means of the argument margin_pct
it is now possible to create the so-called annual averaged precipitation isoscapes!
queryGNIP can now split the dataset per month or year at each location during the aggregation
new function prepcipitate to prepare the precipitation brick
new function getprecip to download monthly precipitation rasters from WorldClim
new function isomultifit fitting isoscapes per strata (month, year, or any "split")
new function isomultiscape building isoscapes averaged across strata
new function create_aliens simulating of organism data
the inputs for filtering data by month or year using queryGNIP have changed
the default fixed effect structure for the mean model is isofit has changed
the namespace is now generated with Roxygen2
the datasets are now 'lazy-loaded'
new vignette for coding conventions
changed some object names following our coding convention (more to come)
the package could not be detached and reloaded
the citation was not correct
the path in getelev was breaking in some cases
the title of the assignment plot was missing when a single individual was plotted
new vignette explaining how to export spatial objects to GIS
the file GNIPdata has been updated and now contain 2014 data
names of all functions and objects have been refactored to remove upper cases
links to our GitHub directory have been added
new function downloadfile to download non standard elevation raster or any other file
function getelev can perform MD5 sum checks if the package 'tools' is installed
function getelev can display additional information during download if verbose > 1
the column animalID in the assignment dataset can now handle names with spaces
added Codecov to track test coverage for the package
the modification of the option set_ll_warn from the 'sp' package has been moved to onLoad (instead of onAttached) and the original state is now restored while unloading 'IsoriX'
the Earth distance method has been moved to the package 'spaMM'
function getelev lost its 'address' argument as downloadfile should now be used to download non-standard elevation rasters
some typo fixed in documentation files
RandomFields moved to suggest
.Rd files for documentation are now generated with Roxygen2
queryGNIP is now provided with a single month argument specifying the months to select
this was the first version of IsoriX submitted to CRAN