R/hdf5_utils.R) that
supports both hdf5r (CRAN) and rhdf5 (Bioconductor) backends. Both
packages are now listed as suggested dependencies rather than hard
imports, and the backend is selected automatically at runtime based on
availability.caret package.
Cross-validation in estimate_calibration_error() is now
performed using base R, which reduces the package’s dependency
footprint. Note that Rsquared is no longer included in the
cross-validation output (R-squared is still reported from the model fit
itself).fit_carbon_regression() and
fit_water_regression() are now pre-allocated to the correct
size rather than a fixed 200,000 rows.convert_POSIXct_to_NEONhdf5_time() to use
format() instead of manual string construction with
individual lubridate accessor functions.setup_output_file,
write_carbon_*, write_water_*) now accept an
open file identifier, reducing redundant file open/close operations
during calibration output. The rhdf5::h5closeAll() calls in
the main calibration functions have been replaced with targeted
h5_close() calls.neonUtilities >= 2.3.0 (previously
>= 2.1.1).Substantial upgrades to package documentation.
The name “Bowling_2003” in methods and functions has been deprecated in favor of “gainoffset.” This change was made primarily to improve clarity of what the methods actually do. Code that uses the old “Bowling_2003” nomenclature will continue to work until at least version 0.8.0, but will issue a deprecation warning. # NEONiso 0.7.0
calibrate_carbon_bymonth and calibrate_water_bymonth have been removed. Use calibrate_carbon and calibrate_water instead.
terrestrial_relocatable_sites() has been renamed terrestrial_gradient_sites() to comply with a similar change made by NEON.
Patch to hard-coded 9 minute interval to comply with switch to 6 minute averaging interval for isoCo2 product, following ATM ISO TWG recommendations.