Skip to content

OpenGHG Inversions v0.2

Latest
Latest
Compare
Choose a tag to compare
@brendan-m-murphy brendan-m-murphy released this 19 Aug 17:10
· 66 commits to devel since this release
v0.2
71f7624

Changes for 0.2:

  • Added option to pass "mean" and "stdev" as parameters for lognormal BC prior #PR 190

  • Pinned numpy to version < 2.0 since PyTensor hasn't updated to numpy >= 2.0 #PR 148

  • Updated filtering to handle case inlet == "multiple". #PR 189

  • Added option to store merged data in a zarr ZipStore, which is essentially just a zipped zarr store. This should reduce the number of files created when saving merged data. #PR 185

  • Fixed issue where missing footprints times were dropped from basis function calculations. #PR 186

  • Made format for filtering in ini file allow for missing sites. Made inlet, instrument, fp_height, obs_data_level, and met_model
    accept a single string in the ini file, which will be converted to a list of the correct length. #PR 182. Bug fix: #PR 188

  • Added code to look for older flux data if none is found between start and end dates #PR 177

  • Moved code related to basis functions from utils.py to basis submodule #PR 162

  • Fixed bug in filtering function and updated tests to cover all filters #PR 179

  • Updated all docstrings (various PRs)

  • Cleaned up utils.py: adding typing, and updated docstrings #PR 158

  • Refactored filters.py so filter functions aren't nested inside filtering. Added code to keep track of filter functions. Updated docstrings. #PR 163

  • Replaced utils.combine_datasets with (nearly) equivalent function from openghg.analyse._scenario. There is currently a thin wrapper to make sure that the second
    dataset is loaded into memory, since this change is only on the devel branch of OpenGHG #PR 160

  • Moved basis and related functions from utils.py to basis._functions.py to make more consistent #PR 162

  • Moved filters from utils.py to new submodule filters.py #PR 159

  • Removed site_info.json and species_info.json and replaced with calls to functions in openghg.util, which pull the same info from openghg_defs. #PR 152

  • Removed unused functions from convert.py and updated docstrings. #PR 151

  • Added new option for computing min. model error based on percentiles. #PR 142

  • Update the docstrings of openghg_inversions.basis and openghg_inversions.array_ops #PR 150

  • Fixed "add averaging" functional, which adds the variability of obs over a resampling period to the measurement error (repeatability). This closes Issue #42 . #PR 144

  • Add option to pass the filters as dictionary (with the sites as keys). #PR 135

  • fixed issue with missing obs due to dropping NaNs from other variables in fp_data (e.g. wind_speed, etc). #PR 132

  • added option no_model_error to run inversions without model error (i.e. no min. model error and no pollution event scaling). #PR 131

  • added work-around for error in post-processing caused by the order of the flux dimensions deviating from 'lat', 'lon', 'time'. #PR 128

  • removed julian2time function from convert.py because it used code that was deprecated by matplotlib. This function is still available at github.com/ACRG-Bristol/acrg/acrg/time/convert.py. #PR 129

  • met_model is now used by data_processing_surface_notracer; it is an optional argument, passed as a list with the same length as the number of sites. #PR 125

  • Added option to pass "mean" and "stdev" to lognormal xpriors. Additionally, if reparameterise_log_normal = True is added to an ini file, then the
    log normal prior will be sampled by transforming samples from standard normal random variable to samples from the appropriate log normal distribution. #PR 107

  • Updated pblh filter to work with new variable names in footprints. #PR 101

  • NaNs are filled before converting to numpy and passing data to the inversion. This partly addresses Issue#97. #PR 101

  • add option to calculate an estimate of the minimum model error on the fly #PR 101

  • added documentation, including a "getting started" tutorial, as well as expanding the README file, and updating the example ini files. #PR 101

  • added land/sea mask file needed for weighted basis functions, and updated code to retrieve it #PR 101

  • restored option to save raw trace from inversion. #PR 101

  • added option to use Numpyro nuts sampler. #PR 101

  • fix for uncaught error when a filter removes all data from a site. The PBLH filter was also modified to return a value in all cases. #PR 105

  • unpinned OpenGHG (from v0.6.2) and made changes for compatibility with OpenGHG v0.8, which uses zarr as a backend. CI was updated to test against OpenGHG versions 0.7.1, 0.8, and the devel branch. Merged data has been changed from pickle files to either zarr or netCDF (if zarr is not available). #PR 92

  • updates to hbmcmc_post_process.py, including changes to site_info.json and species_info.json to remove dependencies on ACRG paths; updates to documentation; changed fluxmean to variable with default fluxmode; fixed bug in set_cmap, which would fail for datasets with many NaNs; no updates to DIC. #PR 88

Assets 2
Loading