xclim 0.39.0

New features and enhancements

• The general xclim description and ReadMe have been updated to reflect recent enhancements. (GH/1185, PR/1209).

• Documentation now supports intersphinx mapping references within code examples via sphinx-codeautolink and copying of code blocks via sphinx-copybutton. (PR/1182).

• Log-logistic distribution added to stats.py for use with standardized_precipitation_index and standardized_precipitation_evapotranspiration_index. (GH/1141, PR/1183).

• New option in many indices allowing for resampling in given periods after run_length operations. (GH/505, GH/916, GH/917, PR/1161).

• New base indicator class for sdba: StatisticalPropertyMeasure, those measures that also reduce the time (as a property does). (PR/1198).

• xclim.core.calendar.common_calendar to find the best calendar to use when uniformizing an heterogeneous collection of data. (PR/1217).

• xclim.ensembles.create_ensemble now accepts calendar=None, and uses the above function to guess the best one. It also now accepts cal_kwargs to fine tune the calendar conversion. (GH/1190, PR/1217).

• New data check : xclim.core.datachecks.check_common_time that ensures all inputs of multivariate indicators have the same frequency (and the same time anchoring for daily and hourly data). (GH/1111, PR/1217).

New indicators

• New indices first_day_temperature_{above | below} and indicators xclim.indices.first_day_{tn | tg | tx}_{above | below}. These indices/indicators accept operator (op) keyword for finer threshold comparison controls. (GH/1175, PR/1186).

• New generic indice cumulative_difference for calculating difference between values and thresholds across time (e.g. temperature: degree-days, precipitation: moisture deficit), with or without resampling/accumulating by frequency. (PR/1202).

• New spatial sdba properties and measures : spatial_correlogram, scorr and first_eof. The later needs the optional dependency eofs. (PR/1198).

Breaking changes

• Indices that accept lat or lon coordinates in their call signatures will now use cf-xarray accessors to gather these variables in the event that they are not explicitly supplied. (PR/1180). This affects the following:

  • huglin_index, biologically_effective_degree_days, cool_night_index, latitude_temperature_index, water_budget, potential_evapotranspiration

• cool_night_index now optionally accepts lat: str = "north" | "south" for calculating CNI over DataArrays lacking a latitude coordinate. (PR/1180).

• The offset value in standardized_precipitation_evapotranspiration_index is changed to better reproduce results in the reference library monocongo/climate_indices. (GH/1141, PR/1183).

• The first_day_below and first_day_above indices are now deprecated in order to clearly communicate the variables they act upon (GH/1175, PR/1186). The suggested migrations are as follows:

  • xclim.indices.first_day_above -> xclim.indices.first_day_temperature_above

  • xclim.indices.first_day_below -> xclim.indices.first_day_temperature_below

• The first_day_below and first_day_above atmos indicators are now deprecated in order to clearly communicate the variables they act upon (GH/1175, PR/1186). The suggested migrations are as follows:

  • xclim.atmos.first_day_above -> xclim.indices.first_day_{tn | tg | tx}_above

  • xclim.atmos.first_day_below -> xclim.indices.first_day_{tn | tg | tx}_below

• The degree_days generic indice has been deprecated in favour of cumulative_difference that is not limited only to temperature variables (GH/1200, PR/1202). The indices for atmos.{heating | cooling | growing}_degree_days are now built from generic.cumulative_difference.

• Running pytest now requires the pytest-dist distributed testing dependency. This library has been added to the dev requirements and conda environment configuration. (PR/1203).

• Parameters reducer and window in xclim.indices.rle_statistics are now positional. (PR/1161).

• The relative_annual_cycle_amplitude and annual_cycle_amplitude have been rewritten to match the version defined in the VALUE project, outputs will change drastically (for the better) (PR/1198).

• English indicator metadata has been adjusted to remove frequencies from fields in the long_name of indicators. English indicators now have an explicit title and abstract. (GH/936, PR/1123).

• French indicator metadata translations are now more uniform and more closely follow agreed-upon grammar conventions, while also removing frequency fields in long_name_fr. (GH/936, PR/1123).

• The freshet_start indice is now deprecated in favour of first_day_temperature_above with thresh=’0 degC’, window=5. The freshet_start indicator is now based on first_day_temperature_above, but is otherwise unaffected. (GH/1195, PR/1196).

• Call signatures for several indices/indicators have been modified to optionally accept op for manually setting threshold comparison operators (GH/1194, PR/1197). The affected indices and indicators as follows:

  • hot_spell_max_length, hot_spell_frequency, cold_spell_days, cold_spell_frequency, heat_wave_index, warm_day_frequency (indice only), warm_night_frequency (indice only), dry_days, wetdays, wetdays_prop.

• Cleaner xclim.core.calendar.parse_offset : fails on invalid frequencies, return implicit anchors (YS -> JAN, Y -> DEC) and implicit is_start_anchored (D -> True). (GH/1213, , PR/1217).

Bug fixes

• The docstring of cool_night_index suggested that lat was an optional parameter. This has been corrected. (GH/1179, PR/1180).

• The mean_radiant_temperature indice was accessing hardcoded lat and lon coordinates from passed DataArrays. This now uses cf-xarray accessors. (PR/1180).

• Adopt (and adapt) unit registry declaration and preprocessors from cf-xarray to circumvent bugs caused by a refactor in pint 0.20. It also cleans the code a little bit. (GH/1211, PR/1212).

Internal changes

• The documentation build now relies on sphinx-codeautolink and sphinx-copybutton. (PR/1182).

• Many docstrings did not fully adhere to the numpy docstring format. Fields and entries for many classes and functions have been adjusted to adhere better. (PR/1182).

• The xdoctest namespace now provides access to session-scoped {variable}_dataset accessors, as well as a path_to_atmos_file object. These can be used for running doctests on all variables made in the pytest atmosds() fixture. (PR/1882).

• Upgrade CodeQL GitHub Action to v2. (GH/1188, PR/1189).

• New generic index first_day_threshold_reached is now used to compose all first_day_XYZ indices. (GH/1175, PR/1186).

• In order to reduce computation footprint, the GitHub CI full testing suite and doctests are now only run once a pull request has been reviewed and approved. The number of simultaneously triggered builds has also been reduced. (GH/1155, PR/1203).

• ReadTheDocs now only builds full documentation (including running notebooks) when pull requests are merged to the main branch. (GH/1155, PR/1203).

• xclim now leverages pytest-xdist to distribute tests among Python workers and significantly speed up the testing suite. (PR/1203).

• show_versions can now accept a list of dependencies so that other libraries can make use of this utility. (PR/1215).

• Pull Requests now are automatically tagged (CI, docs, indicators, and/or sdba) according to files modified using the GitHub Labeler Action. (PR/1214).

New features and enhancements

• Adjustment methods of SBCK are wrapped into xclim when that package is installed. (GH/1109, PR/1115).

  • Wrapped SBCK tests are also properly run in the tox testing ensemble. (PR/1119).

• Method FAO_PM98 (based on Penman-Monteith formula) to compute potential evapotranspiration. (PR/1122).

• New indices for droughts: SPI (standardized precipitations) and SPEI (standardized water budgets). (GH/131, PR/1096).

• Most numba functions of sdba.nbutils now use the “lazy” compilation mode. This significantly accelerates the import time of xclim. (GH/1135, PR/1167).

• Statistical properties and measures from xclim.sdba are now Indicator subclasses (PR/1149).

New indicators

• xclim now has the McArthur Forest Fire Danger Index and related indices under a new xclim.indices.fire module. These indices are also available as indicators. (GH/1152, PR/1159)

• Drought-related indicators: SPI (standardized precipitations) and SPEI (standardized water budgets). (GH/131, PR/1096).

• ensembles.create_ensembles now accepts a realizations argument to assign a coordinate to the “realization” axis. It also accepts a dictionary as input so that keys are used as that coordinate. (PR/1153).

• ensembles.ensemble_percentiles, ensembles.ensemble_mean_std_max_min and ensembles.change_significance now support weights (PR/1151).

• Many generic indicators that compare arrays or against thresholds or now accept an op keyword for specifying the logical comparison operation to use in their calculations (i.e. {“>”, “>=”, “<”, “<=, “!=”, “==”}). (GH/389, PR/1157).

  • In order to prevent user error, many of these generic indices now have a constrain variable that prevents calling an indice with an inappropriate comparison operator. (e.g. The following will raise an error: op=">", constrain=("<", "<=")). This behaviour has been added to indices accepting op where appropriate.

Breaking changes

• scipy has been pinned below version 1.9 until lmoments3 can be adapted to the new API. (GH/1142, PR/1143).

• xclim now requires xarray>=2022.06.0. (PR/1151).

• Documentation CI (ReadTheDocs) builds will now fail if there are any misconfigured pages, internal link/reference warnings, or broken external hyperlinks. (GH/1094, PR/1131, GH/1139, PR/1140, PR/1160).

• Call signatures for generic indices have been reordered and/or modified to accept op, and optionally constrain, in many cases, and condition/conditional/operation has been renamed to op for consistency. (GH/389, PR/1157). The affected indices are as follows:

  • get_op, compare, threshold_count, get_daily_events, count_level_crossings, count_occurrences, first_occurrence, last_occurrence, spell_length, thresholded_statistics, temperature_sum, degree_days.

• All indices in xclim.indices.generic now use threshold in lieu of thresh for consistency. (PR/1157).

• Existing function xclim.indices.generic.compare can now be used to construct operations with op and constrain variables to allow for dynamic comparisons with user input handling. (GH/389, PR/1157).

• Two deprecated indices have been removed from xclim. (PR/1157):

  • xclim.indices._multivariate.daily_freezethaw_cycles -> Replaceable with the generic multiday_temperature_swing with thresh_tasmax=’0 degC’, thresh_tasmin=’0 degC’, window=1, and op=’sum’. The indicator version (xclim.atmos.daily_freezethaw_cycles) is unaffected.

  • xclim.indices.generic.select_time -> Was previously moved to xclim.core.calendar.

• The clix-meta indicator table parsing function (xclim.core.utils.adapt_clix_meta_yaml) has been adapted to support the new “op” operator handler. (PR/1157).

• Because they have been re-implemented as Indicator subclasses, statistical properties and measures of xclim.sdba no longer preserve attributes of their inputs by default. Use xclim.set_options(keep_attrs=True) to get the previous behaviour. (PR/1149).

• The xclim.indices.generic.extreme_temperature_range function has been fixed so it now does what its definition says. Results from xclim.indicators.cf.etr will change. (GH/1172, PR/1173).

• xclim now has a dedicated indices.fire submodule that houses all fire-related indices. The previous xclim.indices.fwi submodule is deprecated and will be removed in a future version. (GH/1152, PR/1159).

• The indicator xclim.indicators.atmos.fire_weather_indexes and indice xclim.indices.fire_weather_indexes have both been deprecated and renamed to cffwis_indices. Calls using the previous naming will be removed in a future version. (PR/1159).

• xclim now explicitly requires pybtex in order to generate documentation. (PR/1176).

Bug fixes

• Fixed saturation_vapor_pressure for temperatures in other units than Kelvins (also fixes relative_humidity_from_dewpoint). (GH/1125, PR/1127).

• Indicators that do not care about the input frequency of the data will not check the cell methods of their inputs. (PR/1128).

• Fixed the signature and docstring of heat_index by changing tasmax to tas. (GH/1126, PR/1128).

• Fixed a formatting issue with virtual indicator modules (_gen_returns_section) that was creating malformed Returns sections in sphinx-generated documentation. (PR/1131).

• Fix biological_effective_degree_days for non-scalar latitudes, when using method “gladstones”. (GH/1136, PR/1137).

• Fixed some extlink warnings found in sphinx and configured ReadTheDocs to use mamba as the dependency solver. (GH/1139, PR/1140).

• Fixed some broken hyperlinks to articles, users, and external documentation throughout the code base and jupyter notebooks. (PR/1160).

• Removed some artefact reference roles introduced in PR/1131 that were causing LaTeX builds of the documentation to fail. (GH/1154, PR/1156).

• Fix biological_effective_degree_days for non-scalar latitudes, when using method “gladstones”. (GH/1136, PR/1137).

• Fixed some extlink warnings found in sphinx and configured ReadTheDocs to use mamba as the dependency solver. (GH/1139, PR/1140).

• Fixed some broken hyperlinks to articles, users, and external documentation throughout the code base and jupyter notebooks. (PR/1160).

• Addressed a bug that was causing pylint to stackoverflow by removing it from the tox configuration. pylint should only be called from an active environment. (PR/1163)

• Fixed an issue with xclim.ensembles.kmeans_reduce_ensemble which caused it to fail when using dask arrays. (PR/1170).

• Addressed a bug that was causing pylint to stackoverflow by removing it from the tox configuration. pylint should only be called from an active environment. (PR/1163)

Internal changes

• Marked a test (test_release_notes_file_not_implemented) that can only pass when source files are available so that it can easily be skipped on conda-forge build tests. (GH/1116, PR/1117).

• Split a few YAML strings found in the virtual modules that regularly issued warnings on the code checking CI steps. (PR/1118).

• Function xclim.core.calendar.build_climatology_bounds now exposed via __all__. (PR/1146).

• Clarifications added to docstring of xclim.core.bootstrapping.bootstrap_func. (PR/1146).

• Bibliographic references for supporting scientific articles are now found in a bibtex file (docs/references.bib). These are now made available within the generated documentation using sphinxcontrib-bibtex. (GH/1094, PR/1131).

• Added information URLs to setup.py in order to showcase issue tracker and other sites on PyPI page (PR/1156).

• Configured the LaTeX build of the documentation to ignore the custom bibliographies, as they were redundant in the generated PDF. (PR/1158).

• Run length encoding (xclim.indices.run_length.rle) has been optimized. (GH/956, PR/1122).

• Added a sphinx-build -b linkcheck step to the tox-based “docs” build as well as to the ReadTheDocs configuration. (PR/1160).

• pylint is now setup to use a pylintrc file, allowing for more granular control of warnings and exceptions. Many errors are still present, so addressing them will need to occur gradually. (PR/1163).

• The generic indices count_level_crossings, count_occurrences, first_occurrence, and last_occurrence are now fully tested. (PR/1157).

• Adjusted the ANUCLIM indices by removing “ANUCLIM” from their titles, modifying their docstrings, and handling “op” input in a more user-friendly way. (GH/1055, PR/1169).

• Documentation for fire-based indices/indicators has been reorganized to reflect the new submodule structure. (PR/1159).

Announcements

• xclim is now compliant with PEP 563. Python3.10-style annotations are now permitted. (GH/1065, PR/1071).

• xclim is now fully compatible with xarray’s flox-enabled GroupBy and resample operations. (PR/1081).

• xclim now (properly) enforces docstring compliance checks using pydocstyle with modified numpy-style docstrings. Docstring errors will now cause build failures. See the pydocstyle documentation for more information. (PR/1074).

• xclim now uses GitHub Actions to manage patch version bumping. Merged Pull Requests that modify xclim code now trigger version-bumping automatically when pushed to the main development branch. Running $ bump2version patch within development branches is no longer necessary. (PR/1102).

New features and enhancements

• Add “Celsius” to aliases of “celsius” unit. (GH/1067, PR/1068).

• All indicators now have indexing enabled, except those computing statistics on spells. (GH/1069, PR/1070).

• A convenience function for returning the version numbers for relevant xclim dependencies (xclim.testing.show_versions) is now offered. (PR/1073).

  • A CLI version of this function is also available from the command line ($ xclim show_version_info). (PR/1073).

• New “keep_attrs” option to control the handling of the attributes within the indicators. (GH/1026, PR/1076).

• Added a notebook showcasing some simple examples of Spatial Analogues. (GH/585, PR/1075).

• create_ensembles now accepts a glob string to find datasets. (PR/1081).

• Improved percentile based indicators metadata with the window, threshold and climatology period used to compute percentiles. (GH/1047, PR/1050).

• New xclim.core.calendar.construct_offset, the inverse operation of parse_offset. (PR/1090).

• Rechunking operations in xclim.indices.run_length.rle are now synchronized with dask’s options. (PR/1090).

• A mention of the “missing” checks and options is added to the history attribute of indicators, where appropriate. (GH/1100, PR/1103).

Breaking changes

• xclim.atmos.water_budget has been separated into water_budget (calculated directly with ‘evspsblpot’) and water_budget_from_tas (original function). (PR/1086).

• Injected parameters in indicators are now left out of a function’s signature and will not be included in the history attribute. (PR/1086).

• The signature for the following Indicators have been modified (PR/1050):

  • cold_spell_duration_index, tg90p, tg10p, tx90p, tx10p, tn90p, tn10p, warm_spell_duration_index, days_over_precip_doy_thresh, days_over_precip_thresh, fraction_over_precip_doy_thresh, fraction_over_precip_thresh, cold_and_dry_days, warm_and_dry_days, warm_and_wet_days, cold_and_wet_days

• The parameter for percentile values is now named after the variable it is supposed to be computed upon. (PR/1050).

• pytest-runner has been removed as a dependency (it was never needed for xclim development). (PR/1074).

• xclim.testing._utils.py has been renamed to xclim.testing.utils.py for added documentation visibility. (PR/1074).

  • Some unused functions and classes (as_tuple, TestFile, TestDataSet) have been removed. (PR/1107).

New indicators

• universal_thermal_climate_index and mean_radiant_temperature for computing the universal thermal climate index from the near-surface temperature, relative humidity, near-surface windspeed and radiation. (GH/1060, PR/1062).

  • A new method ITS90 has also been added for calculating saturation water vapour pressure. (GH/1060, PR/1062).

Internal changes

• Typing syntax has been updated within pre-commit via isort. Pre-commit hooks now append from __future__ import annotations to all python module imports for backwards compatibility. (GH/1065, PR/1071)

• isort project configurations are now set in setup.cfg. (PR/1071).

• Many function docstrings, external target links, and internal section references have been adjusted to reduce warnings when building the docs. (PR/1074).

• Code snippets within documentation are now checked and reformatted to black conventions with blackdoc. A pre-commit hook is now in place to run these checks. (PR/1098).

• Test coverage statistic no longer includes coverage of the test files themselves. Coverage now reflects lines of usable code covered. (PR/1101).

• Reordered listed authors alphabetically. Promoted @bzah to core contributor. (PR/1105).

• Tests have been added for some functions in xclim.testing.utils.py; some previously uncaught bugs in list_input_variables, publish_release_notes, and show_versions have been patched. (GH/1078, PR/1107).

• A convenience command for installing xclim with key development branches of some dependencies has been added ($ make upstream). (GH/1088, PR/1092; amended in GH/1113, PR/1114).

  • This build configuration is also available in tox for local development purposes ($ tox -e pyXX-upstream).

Bug fixes

• Clean the bias_adjustement and history attributes created by xclim.sdba.adjust (e.g. when an argument is an xr.DataArray, only print the name instead of the whole array). (GH/1083, PR/1087).

• pydocstyle checks were silently failing in the pre-commit configuration due to a badly-formed regex. This has been adjusted. (PR/1074).

• adjust_doy_calendar was broken when the source or the target were seasonal. (GH/1097, GH/1091, PR/1099)

Bug fixes

• Invoking lazy_indexing twice in row (or more) using the same indexes (using dask) is now fixed. (GH/1048, PR/1049).

• Filtering out the nans before choosing the first and last values as fill_value in _interp_on_quantiles_1D. (GH/1056, PR/1057).

• Translations from virtual indicator modules do not override those of the base indicators anymore. (GH/1053, PR/1058).

• Fix mmday unit definition (factor 1000 error). (GH/1061, PR/1063).

New features and enhancements

• xclim.sdba.measures.rmse and xclim.sdba.measures.mae now use numpy instead of sklearn. This improves their performances when using dask. (PR/1051).

• Argument append_ends added to sdba.unpack_moving_yearly_window (PR/1059).

Internal changes

• Ipython was unpinned as version 8.2 fixed the previous issue. (GH/1005, PR/1064).

New indicators

• New indicator specific_humidity_from_dewpoint, computing specific humidity from the dewpoint temperature and air pressure. (GH/864, PR/1027)

New features and enhancements

• New spatial analogues method “szekely_rizzo” (PR/1033).

• Loess smoothing (and detrending) now skip NaN values, instead of propagating them. This can be controlled through the skipna argument. (PR/1030).

Bug fixes

• xclim.analog.spatial_analogs is now compatible with dask-backed DataArrays. (PR/1033).

• Parameter dmin added to spatial analog method “zech_aslan”, to avoid singularities on identical points. (PR/1033).

• xclim is now compatible with changes in xarray that enabled explicit indexing operations. (PR/1038, xarray PR).

Internal changes

• xclim now uses the check-json and pretty-format-json pre-commit checks to validate and format JSON files. (PR/1032).

• The few logging artifacts in the xclim.ensembles module have been replaced with warnings.warn calls or removed. (GH/1039, PR/1044).

Announcements

• xclim now officially supports Python3.10. (PR/1013).

Breaking changes

• The version pin for bottleneck (<1.4) has been lifted. (PR/1013).

• packaging has been removed from the xclim run dependencies. (PR/1013).

• Quantile mapping adjustment objects (EQM, DQM and QDM) and sdba.utils.equally_spaced_nodes will not add additional endpoints to the quantile range. With those endpoints, variables are capped to the reference’s range in the historical period, which can be dangerous with high variability in the extremes (ex: pr), especially if the reference doesn’t reproduce those extremes credibly. (GH/1015, PR/1016). To retrieve the same functionality as before use:

from xclim import sdba

# NQ is the the number of equally spaced nodes, the argument previously given to nquantiles directly.
EQM = sdba.EmpiricalQuantileMapping.train(
    ref, hist, nquantiles=sdba.equally_spaced_nodes(NQ, eps=1e-6), ...
)

• The “history” string attribute added by xclim has been modified for readability: (GH/963, PR/1018).

  • The trailing dot (.) was dropped.

  • None inputs are now printed as “None” (and not “<NoneType>”).

  • Arguments are now always shown as keyword-arguments. This mostly impacts sdba functions, as it was already the case for Indicators.

• The cell_methods string attribute appends only the operation from the indicator itself. In previous version, some indicators also appended the input data’s own cell_method. The clix-meta importer has been modified to follow the same convention. (GH/983, PR/1022)

New features and enhancements

• publish_release_notes now leverages much more regular expression logic for link translations to markdown. (PR/1023).

• Improve performances of percentile bootstrap algorithm by using xarray.map_block (GH/932, PR/1017).

Bug fixes

• Loading virtual python modules with build_indicator_module_from_yaml is now fixed on some systems where the current directory was not part of python’s path. Furthermore, paths of the python and json files can now be passed directly to the indices and translations arguments, respectively. (GH/1020, PR/1021).

Internal changes

• Due to an upstream bug in bottleneck’s support of virtualenv, tox builds for Python3.10 now depend on a patched fork of bottleneck. This workaround will be removed once the fix is merged upstream. (PR/1013, see: bottleneck PR/397).

  • This has been removed with the release of bottleneck version 1.3.4. (PR/1025).

• GitHub CI actions now use the deadsnakes python PPA Action for gathering the Python3.10 development headers. (PR/1013).

• The “is_dayofyear” attribute added by several indices is now a numpy.int32 instance, instead of python’s int. This ensures a THREDDS server can read it when the variable is saved to a netCDF file with xarray/netCDF4-python. (GH/980, PR/1019).

• The xclim git repository now offers Issue Forms for some general issue types.

Announcements

• xclim no longer supports Python3.7. Code conventions and new features for Python3.8 (PEP 569) are now accepted. (GH/966, PR/1000).

Breaking changes

• Python3.7 (PEP 537) support has been officially deprecated. Continuous integration testing is no longer run against this version of Python. (GH/966, PR/1000).

Bug fixes

• Adjusted behaviour in dataflags.ecad_compliant to remove data_vars of invalids checks that return None, causing issues with dask. (PR/1002).

• Temporarily pinned ipython below version 8.0 due to behaviour causing hangs in GitHub Actions and ReadTheDocs. (GH/1005, PR/1006).

• indices.stats methods where adapted to handle dask-backed arrays. (GH/1007, :pull:`1011).

• sdba.utils.interp_on_quantiles, with extrapolation='constant', now interpolates the limits of the interpolation along the time grouping index, fixing a issue with “time.month” grouping. (GH/1008, PR/1009).

Internal changes

• pre-commit now uses Black 22.1.0 with Python3.8 style conventions. Existing code has been adjusted. (PR/1000).

• tox builds for Python3.7 have been deprecated. (PR/1000).

• Docstrings and documentation has been adjusted for grammar and typos. (PR/1000).

• sdba.utils.extrapolate_qm has been removed, as announced for xclim 0.33. (PR/1009).

Announcements

• Deprecation: Release 0.33.0 of xclim will be the last version to explicitly support Python3.7 and xarray<0.21.0.

• xclim now requires yaml files to pass yamllint checks on Pull Requests. (PR/981).

• xclim now requires docstrings have valid ReStructuredText formatting to pass basic linting checks. (PR/993). Checks generally require:

  • Working hyperlinks and reference tags.

  • Valid content references (e.g. :py:func:).

  • Valid NumPy-formatted docstrings.

• The xclim developer community has now adopted the ‘Contributor Covenant’ Code of Conduct v2.1 (text). (GH/948, PR/996).

New indicators

• jetstream_metric_woollings indicator returns latitude and strength of jet-stream in u-wind field. (GH/923, PR/924).

New features and enhancements

• Features added and modified to allow proper multivariate adjustments. (PR/964).

  • Added xclim.sdba.processing.to_additive_space and xclim.sdba.processing.from_additive_space to transform “multiplicative” variables to the additive space. An example of multivariate adjustment using this technique was added to the “Advanced” sdba notebook.

  • xclim.sdba.processing.normalize now also returns the norm. xclim.sdba.processing.jitter was created by combining the “under” and “over” methods.

  • xclim.sdba.adjustment.PrincipalComponent was modified to have a simpler signature. The “full” method for finding the best PC orientation was added. (GH/697).

• New xclim.indices.stats.parametric_cdf function to facilitate the computation of return periods over DataArrays of statistical distribution parameters (GH/876, PR/984).

• Add copy parameter to percentile_doy to control if the array input can be dumped after computing percentiles (GH/932, PR/985).

• New improved algorithm for dry_spell_total_length, performing the temporal indexing at the right moment and with control on the aggregation operator (op) for determining the dry spells.

• Added properties.py and measures.py in order to perform diagnostic tests of sdba (GH/424, PR/967).

• Update how percentile_doy rechunk the input data to preserve the initial chunk size. This should make the computation memory footprint more predictable (GH/932, PR/987).

Breaking changes

• To reduce import complexity, select_time has been refactored/moved from xclim.indices.generic to xclim.core.calendar. (GH/949, PR/969).

• The stacking dimension of xclim.sdba.stack_variables has been renamed to “multivar” to avoid name conflicts with the “variables” property of xarray Datasets. (PR/964).

• xclim now requires cf-xarray>=0.6.1. (GH/923, PR/924).

• xclim now requires statsmodels. (GH/424, PR/967).

Internal changes

• Added a CI hook in .pre-commit-config.yaml to perform automated pre-commit corrections with GitHub CI. (PR/965).

• Adjusted CI hooks to fail earlier if lint checks fail. (PR/972).

• TrainAdjust and Adjust object have a new skip_input_checks keyword arg to their train and adjust methods. When True, all unit-, calendar- and coordinate-related input checks are skipped. This is an ugly solution to disappearing attributes when using xr.map_blocks with dask. (PR/964).

• Some slow tests were marked slow to help speed up the standard test ensemble. (PR/969).

  • Tox testing ensemble now also reports slowest tests using the --durations flag.

• pint no longer emits warnings about redefined units when the logging module is loaded. (GH/990, PR/991).

• Added a CI step for cancelling running workflows in pull requests that receive multiple pushes. (PR/988).

Bug fixes

• Fix mistake in the units of spell_length_distribution. (GH/1003, PR/1004)

Announcements

• Code coverage (coverage/coveralls) is now a required CI check for merging Pull Requests. Requirements are now:

  • No individual run may report <80% code coverage.

  • Some drop in coverage is now tolerable, but runs cannot dip below -0.25% relative to the main branch.

New features and enhancements

• Added an optimized pathway for xclim.indices.run_length functions when window=1. (PR/911, GH/910).

• The data input frequency expected by Indicator is now in the src_freq attribute and is thus controllable by subclassing existing indicators. (GH/898, PR/927).

• New **indexer keyword args added to many indicators, it accepts the same arguments as xclim.indices.generic.select_time, which has been improved. Unless otherwise specified, the time selection is done before any computation. (PR/934, GH/899).

• Rewrite of xclim.sdba.ExtremeValues, now fixed with a correct algorithm. It has not been tested extensively and should be considered experimental. (PR/914, GH/789, GH/790).

• Added days_over_precip_doy_thresh and fraction_over_precip_doy_thresh indicators to distinguish between WMO and ECAD definition of the Rxxp and RxxpTot indices. (GH/931, PR/940).

• Update xclim.core.utils.nan_calc_percentiles to improve maintainability. (PR/942).

• Added heat_index indicator. Added heat_index indicator. This is similar to humidex but uses a different dew point as well as heat balance equations which account for variables other than vapor pressure. (GH/807) and (PR/915).

• Added alternative method for xclim.indices.potential_evapotranspiration based on mcguinnessbordne05 (from Tanguay et al. 2018). (PR/926, GH/925).

• Added snw_max and snw_max_doy indicators to compute the maximum snow amount and the day of year of the maximum snow amount respectively. (GH/776, PR/950).

• Added index for calculating ratio of convective to total precipitation. (GH/920, PR/921).

• Added wetdays_prop indicator to calculate the proportion of days in a period where the precipitation is greater than a threshold. (PR/919, GH/918).

Breaking changes

• Following version 1.9 of the CF Conventions, published in September 2021, the calendar name “gregorian” is deprecated. core.calendar.get_calendar will return “standard”, even if the underlying cftime objects still use “gregorian” (cftime <= 1.5.1). (PR/935).

• xclim.sdba.utils.extrapolate_qm is now deprecated and will be removed in version 0.33. (PR/941).

• Dependency pint minimum necessary version is now 0.10. (PR/959).

Internal changes

• Removed some logging configurations in xclim.core.dataflags that were polluting python’s main logging configuration. (PR/909).

• Synchronized logging formatters in xclim.ensembles and xclim.core.utils. (PR/909).

• Added a helper function for generating the release notes with dynamically-generated ReStructuredText or Markdown-formatted hyperlinks (PR/922, GH/907).

• Split of resampling-related functionality of Indicator into new ResamplingIndicator and ResamplingIndicatorWithIndexing subclasses. The use of new (private) methods makes it easier to inject functionality in indicator subclasses. (GH/867, PR/927, PR/934).

• French translation metadata fields are now cleaner and much more internally consistent, and many empty metadata fields (e.g. comment_fr) have been removed. (PR/930, GH/929).

• Adjustments to the tox builds so that slow tests are now run alongside standard tests (for more accurate coverage reporting). (PR/938).

• Use xarray.apply_ufunc to vectorize statistical functions. (PR/943).

• Refactor of xclim.sdba.utils.interp_on_quantiles so that it now handles the extrapolation directly and to better handle missing values. (PR/941).

• Updated heating_degree_days and fraction_over_precip_thresh documentations. (GH/952, PR/953).

• Added an intersphinx mapping to xarray. (PR/955).

• Added a CodeQL security analysis GitHub CI hook on push to master and on Friday nights. (PR/960).

Bug fixes

• Fix bugs in the cf_attrs and/or abstract of continuous_snow_cover_end and continuous_snow_cover_start. (PR/908).

• Remove unnecessary keep_attrs from resample call which would raise an error in futur Xarray version. (PR/937).

• Fixed a bug in the regex that parses usernames in the history. (PR/945).

• Fixed a bug in xclim.indices.generic.doymax and xclim.indices.generic.doymin that prevented the use of the functions on multidimensional data. (PR/950, GH/951).

• Skip all missing values in xclim.sdba.utils.interp_on_quantiles, drop them from both the old and new coordinates, as well as from the old values. (PR/941).

• “degrees_north” and “degrees_east” (and their variants) are now considered independent units, so that pint and xclim.core.units.ensure_cf_units don’t convert them to “deg”. (PR/959).

• Fixed a bug in xclim.core.dataflags that would misidentify the “extra” variable to be called when running multivariate checks. (PR/957, GH/861).

New indicators

• thawing_degree_days indicator returns degree-days above a default of thresh=”0 degC”. (PR/895, GH/887).

• freezing_degree_days indicator returns degree-days below a default of thresh=”0 degC”. (PR/895, GH/887).

• Several frost-free season calculations are now available as both indices and indicators. (PR/895, GH/887):

  • frost_free_season_start

  • frost_free_season_end

  • frost_free_season_length

• growing_season_start is now offered as an indice and as an indicator to complement other growing season-based indicators (threshold calculation with op=”>=”). (PR/895, GH/887).

New features and enhancements

• Improve cell_methods checking to search the wanted method within the whole string. (PR/866, GH/863).

• New align_on='random option for xclim.core.calendar.convert_calendar, for conversions involving ‘360_day’ calendars. (PR/875, GH/841).

• dry_spell_frequency now has a parameter op: {“sum”, “max”} to choose if the threshold is compared against the accumulated or maximal precipitation, over the given window. (PR/879).

• maximum_consecutive_frost_free_days is now checking that the minimum temperature is above or equal to the threshold ( instead of only above). (PR/883, GH/881).

• The ANUCLIM virtual module has been updated to accept weekly and monthly inputs and with improved metadata. (PR/885, GH/538)

• The sdba.loess algorithm has been optimized to run faster in all cases, with an even faster special case (equal_spacing=True) when the x coordinate is equally spaced. When activated, this special case might return results different from without, up to around 0.1%. (PR/865).

• Add support for group’s window and additional dimensions in LoessDetrend. Add new RollingMeanDetrend object. (PR/865).

• Missing value algorithms now try to infer the source timestep of the input data when it is not given. (PR/885).

• On indices, bootstrap parameter documentation has been updated to explain when and why it should be used. (PR/893, GH/846).

Breaking changes

• Major changes in the YAML schema for virtual submodules, now closer to how indicators are declared dynamically, see the doc for details. (PR/849, GH/848).

• Removed xclim.generic.daily_downsampler, as it served no purpose now that xarray’s resampling works with cftime (PR/888, GH/889).

• Refactor of xclim.core.calendar.parse_offset, output types were changed to useful ones (PR/885).

• Major changes on how parameters are passed to indicators. (PR/873):

  • Their signature is now consistent : input variables (DataArrays, optional or not) are positional or keyword arguments and all other parameters are keyword only. (GH/855, GH/857)

  • Some indicators have modified signatures because we now rename variables when wrapping generic indices. This is the case for the whole cf module, for example.

  • Indicator.parameters is now a property generated from Indicator._all_parameters, as the latter includes the injected parameters. The keys of the former are instances of new xclim.core.indicator.Parameter, and not dictionaries as before.

  • New Indicator.injected_parameters to see which compute function arguments will be injected at call time.

  • See the pull request (PR/873) for all information.

• The call signature for huglin_index has been modified to reflect the correct variables used in its formula (tasmin -> tas; thresh_tasmin -> thresh). (PR/903, GH/902).

Internal changes

• Pull Request contributions now require hyperlinks to the issue and pull request pages on GitHub listed alongside changess in HISTORY.rst. (PR/860, GH/854).

• Updated the contribution guidelines to better give credit to contributors and more easily track changes. (PR/869, GH/868).

• Enabled coveralls code coverage reporting for GitHub CI. (PR/870).

• Added automated TestPyPI and PyPI-publishing workflows for GitHub CI. (PR/872).

• Changes on how indicators are constructed. (PR/873).

• Added missing algorithms tests for conversion from hourly to daily. (PR/888).

• Updated pre-commit hooks to use black v21.10.b0. (PR/896).

• Moved stack_variables, unstack_variables, construct_moving_yearly_window and unpack_moving_yearly_window from xclim.sdba.base to xclim.sdba.processing. They still are imported in xclim.sdba as before. (PR/892).

• Many improvements to the documentation. (PR/892, GH/880).

• Added regex replacement handling in setup.py to facilitate publishing contributor/contribution links on PyPI. (PR/906).

Bug fixes

• Fix a bug in bootstrapping where computation would fail when the dataset time coordinate is encoded using cftime.datetime. (PR/859).

• Fix a bug in build_indicator_module_from_yaml where bases classes (Daily, Hourly, etc) were not usable with the base field. (PR/885).

• percentile_doy alpha and beta parameters are now properly transmitted to bootstrap calls of this function. (PR/893, GH/846).

• When called with a 1D da and ND index, xclim.indices.run_length.lazy_indexing now drops the auxiliary coordinate corresponding to da’s index. This fixes a bug with ND data in xclim.indices.run_length.season. (PR/900).

• Fix name of heating degree days in French (“chauffe” -> “chauffage”). (PR/895).

• Corrected several French indicator translation description strings (bad usages of “.” in description and long_name fields). (PR/895).

• Fixed an error with the formula for huglin_index where tasmin was being used in the calculation instead of tas. (PR/903, GH/902).

New indicators

• climatological_mean_doy indice returns the mean and standard deviation across a climatology according to day-of-year (xarray.DataArray.groupby(“time.dayofyear”)). A moving window averaging of days can also be supplied (default:window=1).

• within_bnds_doy indice returns a boolean array indicating whether or not array’s values are within bounds for each day of the year.

• Added atmos.wet_precip_accumulation, an indicator accumulating precipitation over wet days.

• Module ICCLIM now includes PRCPTOT, which accumulates precipitation for days with precipitation above 1 mm/day.

New features and enhancements

• xclim.core.utils.nan_calc_percentiles now uses a custom algorithm instead of numpy.nanpercentiles to have more flexibility on the interpolation method. The performance is also improved.

• xclim.core.calendar.percentile_doy now uses the 8th method of Hyndman & Fan for linear interpolation (alpha = beta = 1/3). Previously, the function used Numpy’s percentile, which corresponds to the 7th method. This change is motivated by the fact that the 8th is recommended by Hyndman & Fay and it ensures consistency with other climate indices packages (climdex, icclim). Using alpha = beta = 1 restores the previous behaviour.

• xclim.core.utils._cal_perc is now only a proxy for xc.core.utils.nan_calc_percentiles with some axis moves.

• xclim now implements many data quality assurance flags (xclim.core.dataflags) for temperature and precipitation based on ICCLIM documentation guidelines. These checks include the following:

  • Temperature (variables: tas, tasmin, tasmax): tasmax_below_tasmin, tas_exceeds_tasmax, tas_below_tasmin, temperature_extremely_low (thresh=”-90 degC”), temperature_extremely_high (thresh=”60 degC”).

  • Precipitation-specific (variables: pr, prsn, ): negative_accumulation_values, very_large_precipitation_events (thresh=”300 mm d-1”).

  • Wind-specific (variables: sfcWind, wsgsmax/sfcWindMax): wind_values_outside_of_bounds

  • Generic: outside_n_standard_deviations_of_climatology, values_repeating_for_n_or_more_days, values_op_thresh_repeating_for_n_or_more_days, percentage_values_outside_of_bounds.

  These quality-assurance checks are selected according to CF-standard variable names, and can be triggered via xclim.core.dataflags.data_flags(xarray.DataArray, xarray.Dataset). These checks are separate from the Indicator-defined datachecks and must be launched manually. They’ll return an array of data_flags as boolean variables. If called with raise_flags=True, will raise an Exception with comments for each quality control check raised.

• A convenience function (xclim.core.dataflags.ecad_compliant) is also offered as a method for asserting that data adheres to all relevant ECAD/ICCLIM checks. For more information on usage, consult the docstring/documentation.

• A new utility “dataflags” is also available for performing fast quality control checks from the command-line (xclim dataflags –help). See the CLI documentation page for usage examples.

• Added missing typed call signatures, expected returns and docstrings for many xclim.core.calendar functions.

Breaking changes

• All “ANUCLIM” indices and indicators have lost their src_timestep argument. Most of them were not using it and now every function infers the frequency from the data directly. This may add stricter constraints on the time coordinate, the same as for xarray.infer_freq.

• Many functions found within xclim.core.cfchecks (generate_cfcheck and check_valid_*) have been removed as existing indicator CF-standard checks and data checks rendered them redundant/obsolete.

Bug fixes

• Fixes in sdba for (1) inputs with dimensions without coordinates, for (2) sdba.detrending.MeanDetrend and for (3) DetrendedQuantileMapping when used with dask’s distributed scheduler.

• Replaced instances of ‘◦’ (“White bullet”) with ‘°’ (“Degree Sign”) in icclim.yaml as it was causing issues for non-UTF8 environments.

• Addressed an edge case where test_sdba::test_standardize randomness could generate values that surpass the test error tolerance.

• Added a missing .txt file to the MANIFEST of the source distributable in order to be able to run all tests.

• xc.core.units.rate2amount is now exact when the sampling frequency is monthly, seasonal or yearly. Earlier, monthly and yearly data were computed using constant month and year length. End-of-period frequencies are also correctly understood (ex: “M” vs “MS”).

• In the potential_evapotranspiration indice, add abbreviated method names to docstring.

• Fixed an issue that prevented using the default group arg in adjustment objects.

• Fix bug in missing_wmo, where a period would be considered valid if all months met WMO criteria, but complete months in a year were missing. Now if any month does not meet criteria or is absent, the period will be considered missing.

• Fix bootstrapping with dask arrays. Dask does not support using loc with multiple indexes to set new values so a workaround was necessary.

• Fix bootstrapping when the bootstrapped year must be converted to a 366_day calendar.

• Virtual modules and translations now use ‘UTF-8’ by default when reading yaml or json file, instead of a machine-dependent encoding.

Internal Changes

• xclim code quality checks now use the newest black (v21.8-beta). Checks launched via tox and pre-commit now run formatting modifications over Jupyter notebooks found under docs.

Announcements

• It was found that the ExtremeValues adjustment algorithm was not as accurate and stable as first thought. It is now hidden from xclim.sdba but can still be accessed via xclim.sdba.adjustment, with a warning. Work on improving the algorithm is ongoing, and a better implementation will be in a future version.

• It was found that the add_dims argument of sdba.Grouper had some caveats throughout sdba. This argument is to be used with care before a careful analysis and more testing is done within xclim.

Breaking changes

• xclim has switched back to updating the history attribute (instead of xclim_history). This impacts all indicators, most ensemble functions, percentile_doy and sdba.processing (see below).

• Refactor of sdba.processing. Now all functions take one or more DataArrays as input, plus some parameters. And output one or more dataarrays (not Datasets). Units and metadata is handled. This impacts sdba.processing.adapt_freq especially.

• Add unit handling in sdba. Most parameters involving quantities are now expecting strings (and not numbers). Adjustment objects will ensure ref, hist and sim all have the same units (taking ref as reference).

• The Adjustment` classes of xclim.sdba have been refactored into 2 categories:

  • TrainAdjust objects (most of the algorithms), which are created and trained in the same call: obj = Adj.train(ref, hist, **kwargs). The .adjust step stays the same.

  • Adjust objects (only NpdfTransform), which are never initialized. Their adjust class method performs all the work in one call.

• snowfall_approximation used a “<” condition instead of “<=” to determine the snow fraction based on the freezing point temperature. The new version sticks to the convention used in the Canadian Land Surface Scheme (CLASS).

• Removed the “gis”, “docs”, “test” and “setup”`extra dependencies from ``setup.py`. The dev recipe now includes all tools needed for xclim’s development.

New features and enhancements

• snowfall_approximation has gained support for new estimation methods used in CLASS: ‘brown’ and ‘auer’.

• A ValidationError will be raised if temperature units are given as ‘deg C’, which is misinterpreted by pint.

• Functions computing run lengths (sequences of consecutive “True” values) now take the index argument. Possible values are first and last, indicating which item in the run should be used to index the run length. The default is set to “first”, preserving the current behavior.

• New sdba_encode_cf option to workaround a cftime/xarray performance issue when using dask.

New indicators

• effective_growing_degree_days indice returns growing degree days using dynamic start and end dates for the growing season (based on Bootsma et al. (2005)). This has also been wrapped as an indicator.

• qian_weighted_mean_average (based on Qian et al. (2010)) is offered as an alternate method for determining the start date using a weighted 5-day average (method="qian"). Can also be used directly as an indice.

• cold_and_dry_days indicator returns the number of days where the mean daily temperature is below the 25th percentile and the mean daily precipitation is below the 25th percentile over period. Added as CD indicator to ICCLIM module.

• warm_and_dry_days indicator returns the number of days where the mean daily temperature is above the 75th percentile and the mean daily precipitation is below the 25th percentile over period. Added as WD indicator to ICCLIM module.

• warm_and_wet_days indicator returns the number of days where the mean daily temperature is above the 75th percentile and the mean daily precipitation is above the 75th percentile over period. Added as WW indicator to ICCLIM module.

• cold_and_wet_days indicator returns the number of days where the mean daily temperature is below the 25th percentile and the mean daily precipitation is above the 75th percentile over period. Added as CW indicator to ICCLIM module.

• calm_days indicator returns the number of days where surface wind speed is below threshold.

• windy_days indicator returns the number of days where surface wind speed is above threshold.

Bug fixes

• Various bug fixes in bootstrapping:

  • in percentile_bootstrap decorator, fix the popping of bootstrap argument to propagate in to the function call.

  • in bootstrap_func, fix some issues with the resampling frequency which was not working when anchored.

• Made argument thresh of sdba.LOCI required, as not giving it raised an error. Made defaults explicit in the adjustments docstrings.

• Fixes in sdba.processing.adapt_freq and sdba.nbutils.vecquantiles when handling all-nan slices.

• Dimensions in a grouper’s add_dims are now taken into consideration in function wrapped with map_blocks/groups. This feature is still not fully tested throughout sdba though, so use with caution.

• Better dtype preservation throughout sdba.

• “constant” extrapolation in the quantile mappings’ adjustment is now padding values just above and under the target’s max and min, instead of ±np.inf.

• Fixes in sdba.LOCI for the case where a grouping with additionnal dimensions is used.

Internal Changes

• The behaviour of xclim.testing._utils.getfile was adjusted to launch file download requests for web-hosted md5 files for every call to compare against local test data. This was done to validate that locally-stored test data is identical to test data available online, without resorting to git-based actions. This approach may eventually be revised/optimized in the future.

Announcements

• The xclim binary package available on conda-forge will no longer supply clisops by default. Installation of clisops must be performed explicitly to preserve subsetting and bias correction capabilities.

New indicators

• snow_depth indicator returns the mean snow depth over period. Added as SD to ICCLIM module.

Internal Changes

• Minor modifications to many function call signatures (type hinting) and docstrings (numpy docstring compliance).

New features and enhancements

• Automatic load of translations on import and possibility to pass translations for virtual modules.

• New xclim.testing.list_datasets function listing all available test datasets in repo xclim-testdata.

• spatial_analogs accepts multi-indexes as the dist_dim parameter and will work with candidates and target arrays of different lengths.

• humidex can be computed using relative humidity instead of dewpoint temperature.

• New sdba.construct_moving_yearly_window and sdba.unpack_moving_yearly_window for moving window adjustments.

• New sdba.adjustment.NpdfTransform which is an adaptation of Alex Cannon’s version of Pitié’s N-dimensional probability density function transform. Uses new sdba.utils.rand_rot_matrix. Experimental, subject to changes.

• New sdba.processing.standardize, .unstandardize and .reordering. All of them, tools needed to replicate Cannon’s MBCn algorithm.

• New sdba.processing.escore, backed by sdba.nbutils._escore to evaluate the performance of the N pdf transform.

• New function xclim.indices.clausius_clapeyron_scaled_precipitation can be used to scale precipitation according to changes in mean temperature.

• Percentile based indices gained a bootstrap argument that applies a bootstrapping algorithm to reduce biases on exceedance frequencies computed over in base and out of base periods. Experimental, subject to changes.

• Added a .zenodo.json file for collecting and maintaining author order and tracking ORCIDs.

Bug fixes

• Various bug fixes in sdba :

  • in QDM.adjust, fix bug occurring with coords of ‘object’ dtype and interp='nearest'.

  • in nbutils.quantiles, fix dtype bug when using float32 data.

  • raise a proper error when ref and hist have a different calendar for map_blocks-backed adjustments.

Breaking changes

• spatial_analogs does not support sequence of dist_dim anymore. Users are responsible for stacking dimensions prior to calling spatial_analogs.

New indicators

• biologically_effective_degree_days (with method="gladstones") indice computes degree-days between two specific dates, with a capped daily max value as well as latitude and temperature range swing as modifying coefficients (based on Gladstones, J. (1992)). This has also been wrapped as an indicator.

• An alternative implementation of biologically_effective_degree_days (with method="icclim", based on ICCLIM formula) ignores latitude and temperature range swing modifiers and uses an alternate end_date. Wrapped and available as an ICCLIM indicator.

• cool_night_index indice returns the mean minimum temperature in September (lat >= 0 deg N) or March (lat < 0 deg N), based on Tonietto & Carbonneau, 2004 (10.1016/j.agrformet.2003.06.001). Also available as an indicator (see indices Notes section on indicator usage recommendations).

• latitude_temperature_index indice computes LTI values based on mean temperature of warmest month and a parameterizable latitude coefficient (default: lat_factor=75) based on Jackson & Cherry, 1988, and Kenny & Shao, 1992 (10.1080/00221589.1992.11516243). This has also been wrapped as an indicator.

• huglin_index indice computes Huglin Heliothermal Index (HI) values based on growing degrees and a latitude-influenced coefficient for day-length (based on Huglin. (1978)). The indice supports several methods of estimating the latitude coefficient:

  • method="smoothed": Marks latitudes between -40 N and 40 N with k=1, and linearly increases to k=1.06 at |lat|==50.

  • method="icclim": Uses a stepwise function based on the the original method as presented by Huglin (1978). Identical to the ICCLIM implementation.

  • method="jones": Uses a more robust calculation for calculating day-lengths, based on Hall & Jones (2010). This method is now also available for biologically_effective_degree_days.

• The generic indice day_length, used for calculating approximate daily day-length in hours per day or, given start_date and end_date, the total aggregated day-hours over period. Uses axial tilt, start and end dates, calendar, and approximate date of northern hemisphere summer solstice, based on Hall & Jones (2010).

Internal Changes

• aggregate_between_dates (introduced in v0.27.0) now accepts DayOfYear-like strings for supplying start and end dates (e.g. start="02-01", end="10-31").

• The indicator call sequence now considers “variable” the inputs annoted so. Dropped the nvar attribute.

• Default cfcheck is now to check metadata according to the variable name, using CMIP6 names in xclim/data/variable.yml.

• Indicator.missing defaults to “skip” if freq is absent from the list of parameters.

• Minor modifications to the GitHub Pull Requests template.

• Simplification of some yaml elements for virtual modules.

• Allow injecting freq without the missing checks failing.

New features and enhancements

• Rewrite of nearly all adjustment methods in sdba, with use of xr.map_blocks to improve scalability with dask. Rewrite of some parts of the algorithms with numba-accelerated code.

• “GFWED” specifics for fire weather computation implemented back into the FWI module. Outputs are within 3% of GFWED data.

• Addition of the run_length_ufunc option to control which run length algorithm gets run. Defaults stay the same (automatic switch dependent of the input array : the 1D version is used with non-dask arrays with less than 9000 points per slice).

• Indicator modules built from YAML can now use custom indices. A mapping or module of them can be given to build_indicator_module_from_yaml with the indices keyword.

• Virtual submodules now include an iter_indicators function to iterate over the pairs of names and indicator objects in that module.

• The indicator string formatter now accepts a “r” modifier which passes the raw strings instead of the adjective version.

• Addition of the sdba_extra_output option to adds extra diagnostic variables to the outputs of Adjustment objects. Implementation of sim_q in QuantileDeltaMapping and nclusters in ExtremeValues.

Breaking changes

• The tropical_nights indice is being deprecated in favour of tn_days_above with thresh="20 degC". The indicator remains valid, now wrapping this new indice.

• Results of sdba.Grouper.apply for Grouper``s without a group (ex: ``Grouper('time')) will contain a group singleton dimension.

• The daily_freezethaw_cycles indice is being deprecated in favour of multiday_temperature_swing with temp thresholds at 0 degC and window=1, op="sum". The indicator remains valid, now wrapping this new indice.

• CMIP6 variable names have been adopted whenever possible in xclim. Changes are:

  • swe is now snw (snw is the snow amount [kg / m²] and swe the liquid water equivalent thickness [m])

  • rh is now hurs

  • dtas is now tdps

  • ws (in FWI) is now sfcWind

  • sic is now siconc

  • area (of sea ice indicators) is now areacello

  • Indicators RH and RH_FROMDEWPOINT have be renamed to HURS and HURS_FROMDEWPOINT. These are changes in the _identifiers_, the python names (relative_humidity[...]) are unchanged.

New indicators

• atmos.corn_heat_units computes the daily temperature-based index for corn growth.

• New indices and indicators for tx_days_below, tg_days_above, tg_days_below, and tn_days_above.

• atmos.humidex returns the Canadian humidex, an indicator of perceived temperature account for relative humidity.

• multiday_temperature_swing indice for returning general statistics based on spells of doubly-thresholded temperatures (Tmin < T1, Tmax > T2).

• New indicators atmos.freezethaw_frequency, atmos.freezethaw_spell_mean_length, atmos.freezethaw_spell_max_length for statistics of Tmin < 0 degC and Tmax > 0 deg C days now available (wrapped from multiday_temperature_swing).

• atmos.wind_chill_index computes the daily wind chill index. The default is similar to what Environment and Climate Change Canada does, options are tunable to get the version of the National Weather Service.

Internal Changes

• run_length.rle_statistics now accepts a window argument.

• Common arguments to the op parameter now have better adjective and noun formattings.

• Added and adjusted typing in call signatures and docstrings, with grammar fixes, for many xclim.indices operations.

• Added internal function aggregate_between_dates for array aggregation operations using xarray datetime arrays with start and end DayOfYear values.

Announcements

• xclim no longer supports Python3.6. Code conventions and new features from Python3.7 (PEP 537 Features) are now accepted.

New features and enhancements

• core.calendar.doy_to_days_since and days_since_to_doy to allow meaningful statistics on doy data.

• New bias second-order adjustment method “ExtremeValues”, intended for re-adjusting extreme precipitation values.

• Virtual indicators modules can now be built from YAML files.

• Indicators can now be built from dictionaries.

• New generic indices, implementation of clix-meta’s index functions.

• On-the-fly generation of climate and forecasting convention (CF) checks with xc.core.cfchecks.generate_cfcheck, for a few known variables only.

• New xc.indices.run_length.rle_statistics for min, max, mean, std (etc) statistics on run lengths.

• New virtual submodule cf, with CF standard indices defined in clix-meta.

• Indices returning day-of-year data add two new attributes to the output: is_dayofyear (=1) and calendar.

Breaking changes

• xclim now requires xarray>=0.17.

• Virtual submodules icclim and anuclim are not available at the top level anymore (only through xclim.indicators).

• Virtual submodules icclim and anuclim now provide Indicators and not indices.

• Spatial analog methods “KLDIV” and “Nearest Neighbor” now require scipy>=1.6.0.

Bug fixes

• from_string object creation in sdba has been removed. Now replaced with use of a new dependency, jsonpickle.

Internal Changes

• pre-commit linting checks now run formatting hook black==21.4b2.

• Code cleaning (more accurate call signatures, more use of https links, docstring updates, and typo fixes).

Announcements

• Deprecation: Release 0.25.0 of xclim will be the last version to explicitly support Python3.6 and xarray<0.17.0.

New indicators

• land.winter_storm computes days with snow accumulation over threshold.

• land.blowing_snow computes days with both snow accumulation over last days and high wind speeds.

• land.snow_melt_we_max computes the maximum snow melt over n days, and land.melt_and_precip_max the maximum combined snow melt and precipitation.

• snd_max_doy returns the day of the year where snow depth reaches its maximum value.

• atmos.high_precip_low_temp returns days with freezing rain conditions (low temperature and precipitations).

• land.snow_cover_duration computes the number of days snow depth exceeds some minimal threshold.

• land.continuous_snow_cover_start and land.continuous_snow_cover_end identify the day of the year when snow depth crosses a threshold for a given period of time.

• days_with_snow, counts days with snow between low and high thresholds, e.g. days with high amount of snow (indice and indicator available).

• fire_season, creates a fire season mask from temperature and, optionally, snow depth time-series.

New features and enhancements

• generic.count_domain counts values within low and high thresholds.

• run_length.season returns a dataset storing the start, end and length of a season.

• Fire Weather indices now support dask-backed data.

• Objects from the xclim.sdba submodule can be created from their string repr or from the dataset they created.

• Fire Weather Index submodule replicates the R code of cffdrs, including fire season determination and overwintering of the drought_code.

• New run_bounds and keep_longest_run utilities in xclim.indices.run_length.

• New bias-adjustment method: PrincipalComponent (based on Hnilica et al. 2017 https://doi.org/10.1002/joc.4890).

Internal changes

• Small changes in the output of indices.run_length.rle.

New indicators

• days_over_precip_thresh, fraction_over_precip_thresh, liquid_precip_ratio, warm_spell_duration_index, all from eponymous indices.

• maximum_consecutive_warm_days from indice maximum_consecutive_tx_days.

Breaking changes

• Numerous changes to xclim.core.calendar.percentile_doy:

  • per now accepts a sequence as well as a scalar and as such the output has a percentiles axis.

  • per argument is now expected to between 0-100 (not 0-1).

  • input data must have a daily (or coarser) time frequency.

• Change in unit handling paradigm for indices, which as a result will lead to some indices returning values with different units. Note that related Indicator objects remain unchanged and will return units consistent with CF Convention. If you are concerned with code stability, please use Indicator objects. The change was necessary to resolve inconsistencies with xarray’s keep_attrs=True context.

  • Indice functions now return output units that preserve consistency with input units. That is, feeding inputs in Celsius will yield outputs in Celsius instead of casting to Kelvin. In all cases the dimensionality is preserved.

  • Indice functions now accept non-daily data, but daily frequency is assumed by default if the frequency cannot be inferred.

• Removed the explicitly-installed netCDF4 python library from the base installation, as this is never explicitly used (now only installed in the docs recipe for sdba documented example).

• Removed xclim.core.checks, which was deprecated since v0.18.

New features and enhancements

• Indicator now have docstrings generated from their metadata.

• Units and fixed choices set are parsed from indice docstrings into Indicator.parameters.

• Units of indices using the declare_units decorator are stored in indice.in_units and indice.out_units.

• Changes to Indicator.format and Indicator.json to ensure the resulting json really is serializable.

Internal changes

• Leave missing_options undefined in land.fit indicator to allow control via set_options.

• Modified xclim.core.calendar.percentile_doy to improve performance.

• New xclim.core.calendar.compare_offsets for comparing offset strings.

• New xclim.indices.generic.get_op to retrieve a function from a string representation of that operator.

• The CI pipeline has been migrated from Travis CI to GitHub Actions. All stages are still built using tox.

• Indice functions must always set the units (the declare_units decorator does no check anymore).

• New xclim.core.units.rate2amout to convert rates like precipitation to amounts.

• xclim.core.units.pint2cfunits now removes ‘ * ‘ symbols and changes Δ° to delta_deg.

• New xclim.core.units.to_agg_units and xclim.core.units.infer_sampling_units for unit handling involving aggregation operations along the time dimension.

• Added an indicators API page to the docs and links to there from the Climate Indicators page.

Bug fixes

• The unit handling change resolved a bug that prevented the use of xr.set_options(keep_attrs=True) with indices.

Breaking changes

• Renamed indicator atmos.degree_days_depassment_date to atmos.degree_days_exceedance_date.

• In degree_days_exceedance_date : renamed argument start_date to after_date.

• Added cfchecks for Pr+Tas-based indicators.

• Refactored test suite to now be available as part of the standard library installation (xclim.testing.tests).

• Running pytest with xdoctest now requires the rootdir to point at tests location (pytest –rootdir xclim/testing/tests/ –xdoctest xclim).

• Development checks now require working jupyter notebooks (assessed via the pytest –nbval command).

New indicators

• rain_approximation and snowfall_approximation for computing prlp and prsn from pr and tas (or tasmin or tasmax) according to some threshold and method.

• solid_precip_accumulation and liquid_precip_accumulation now accept a thresh parameter to control the binary snow/rain temperature threshold.

• first_snowfall and last_snowfall to compute the date of first/last snowfall exceeding a threshold in a period.

New features and enhancements

• New kind entry in the parameters property of indicators, differentiating between [optional] variables and parameters.

• The git pre-commit hooks (pre-commit run –all) now clean the jupyter notebooks with nbstripout call.

Bug fixes

• Fixed a bug in indices.run_length.lazy_indexing that occurred with 1D coords and 0D indexes when using the dask backend.

• Fixed a bug with default frequency handling affecting fit indicator.

• Set missing method to ‘skip’ for freq_analysis indicator.

• Fixed a bug in ensembles._ens_align_datasets that occurred when inputs are .nc filepaths but files lack a time dimension.

Internal changes

• core.cfchecks.check_valid now accepts a sequence of strings as its expected argument.

• Clean up in the tests to speed up testing. Addition of a marker to include “slow” tests when desired (-m slow).

• Fixes in the tests to support sklearn>=0.24, clisops>=0.5 and build xarray@master against python 3.7.

• Moved the testing suite to within xclim and simplified tox to manage its own tempdir.

• Indicator class now has a default_freq method.

Breaking changes

• Statistical functions (frequency_analysis, fa, fit, parametric_quantile) are now solely accessible via indices.stats.

New indicators

• atmos.degree_days_depassment_date, the day of year when the degree days sum exceeds a threshold.

New features and enhancements

• Added unique titles to atmos calculations employing wrapped_partials.

• xclim.core.calendar.convert_calendar now accepts a missing argument.

• Added xclim.core.calendar.date_range and xclim.core.calendar.date_range_like wrapping pandas’ date_range and xarray’s cftime_range.

• xclim.core.calendar.get_calendar now accepts many different types of data, including datetime object directly.

• New module xclim.analog and method xclim.analog.spatial_analogs to compute spatial analogs.

• Indicators can now accept dataset in their new ds call argument. Variable arguments (that use the DataArray annotation) can now be given with strings that correspond to variable names in the dataset, and default to their own name.

• Clarification to frequency_analysis notebook.

• Now officially supporting PEP596 (Python3.9).

• New methods xclim.ensembles.change_significance and xclim.ensembles.knutti_sedlacek to qualify climate change agreement among members of an ensemble.

Bug fixes

• Fixed bug that prevented the use of xclim.core.missing.MissingBase and subclasses with an indexer and a cftime datetime coordinate.

• Fixed issues with metadata handling in statistical indices.

• Various small fixes to the documentation (re-establishment of some internally and externally linked documents).

Internal changes

• Passing align_on to xclim.core.calendar.convert_calendar without using ‘360_day’ calendars will not raise a warning anymore.

• Added formatting utilities for metadata attributes (update_cell_methods, prefix_attrs and unprefix_attrs).

• xclim/ensembles.py moved to xclim/ensembles/*.py, splitting stats/creation, reduction and robustness methods.

• With the help of the mypy library, added several typing fixes to better identify inputs/outputs, and reduce object type mutations.

• Fixed some doctests in ensembles and set_options.

• clisops v0.4.0+ is now an optional requirements for non-Windows builds.

• New xclim.core.units.str2pint method to convert quantity strings to quantity objects. Main improvement is to make “3 degC days” a valid string that converts to “3 K days”.

Breaking changes

• Statistical functions (frequency_analysis, fa, fit, parametric_quantile) moved from indices.generic to indices.stats to make them more visible.

New indicators

New features and enhancements

• New xclim.testing.open_dataset method to read data from the remote testdata repo.

• Added a notebook, ensembles-advanced.ipynb, to the documentation detailing ensemble reduction techniques and showing how to make use of built-in figure-generating commands.

• Added a notebook, frequency_analysis.ipynb, with examples showcasing frequency analysis capabilities.

Bug fixes

• Fixed a bug in the attributes of frost_season_length.

• indices.run_length methods using dates now respect the array’s calendar.

• Worked around an xarray bug in sdba.QuantileDeltaMapping when multidimensional arrays are used with linear or cubic interpolation.

Internal changes

Breaking changes

• xclim.subset has been deprecated and now relies on clisops to perform specialized spatio-temporal subsetting. Install with pip install xclim[gis] in order to retain the same functionality.

• The python library pandoc is no longer listed as a docs build requirement. Documentation still requires a current version of pandoc binaries installed at system-level.

• ANUCLIM indices have seen their input_freq parameter renamed to src_timestep for clarity.

• A clean-up and harmonization of the indicators metadata has changed some of the indicator identifiers, long_names, abstracts and titles. xclim.atmos.drought_code and fire_weather_indexes now have indentifiers “dc” and “fwi” (lowercase version of the previous identifiers).

• xc.indices.run_length.run_length_with_dates becomes xc.indices.run_length.season_length. Its argument date is now optional and the default changes from “07-01” to None.

• xc.indices.consecutive_frost_days becomes xc.indices.maximum_consecutive_frost_days.

• Changed the history indicator output attribute to xclim_history in order to respect CF conventions.

New indicators

• atmos.max_pr_intensity acting on hourly data.

• atmos.wind_vector_from_speed, also the wind_speed_from_vector now also returns the “wind from direction”.

• Richards-Baker flow flashiness indicator (xclim.land.rb_flashiness_index).

• atmos.max_daily_temperature_range.

• atmos.cold_spell_frequency.

• atmos.tg_min and atmos.tg_max.

• atmos.frost_season_length, atmos.first_day_above. Also, atmos.consecutive_frost_days now takes a thresh argument (default : 0 degC).

New features and enhancements

• sdba.loess submodule implementing LOESS smoothing tools used in sdba.detrending.LoessDetrend.

• xclim now depends on clisops for subsetting, offloading several heavy GIS dependencies. This improves maintainability and reduces the size of a “vanilla” xclim installation considerably.

• New generic.parametric_quantile function taking parameters estimated by generic.fit as an input.

• Add support for using probability weighted moments method in generic.fit function. Requires the lmoments3 package, which is not included in dependencies because it is unmaintained. Install manually if needed.

• Implemented _fit_start utility function providing initial conditions for statistical distribution parameters estimation, reducing the likelihood of poor fits.

• Added support for indicators based on hourly (1H) inputs, and a first hourly indicator called max_pr_intensity returning hourly precipitation intensity.

• Indicator instances can be retrieved through their class with the get_instance() class method. This allows the use of xclim.core.indicator.registry as an instance registry.

• Indicators now have a realm attribute. It must be given when creating indicators outside xclim.

• Better docstring parsing for indicators: parameters description, annotation and default value are accessible in the json output and Indicator.parameters.

• New command line interface xclim for simple indicator computing tasks.

• New sdba.processing.jitter_over_thresh for variables with a upper bound.

• Added op parameter to xclim.indices.daily_temperature_range to allow resample reduce operations other than mean

• core.formatting.AttrFormatter (and thus, locale dictionaries) can now use glob-like pattern for matching values to translate.

Bug fixes

The ICCLIM module was identified as icclim in the documentation but the module available under ICCLIM. Now icclim == ICCLIM and ICCLIM will be deprecated in a future release.

Internal changes

• xclim.subset now attempts to load and expose the functions of clisops.core.subset. This is an API workaround preserving backwards compatibility.

• Code styling now conforms to the latest release of black (v0.20.8).

• New IndicatorRegistrar class that takes care of adding indicator classes and instances to the appropriate registries. Indicator now inherits from it.

Breaking changes

• Refactoring of the Indicator class. The cfprobe method has been renamed to cfcheck and the validate method has been renamed to datacheck. More importantly, instantiating Indicator creates a new subclass on the fly and stores it in a registry, allowing users to subclass existing indicators easily. The algorithm for missing values is identified by its registered name, e.g. “any”, “pct”, etc, along with its missing_options.

• xclim now requires xarray >= 0.16, ensuring that xclim.sdba is fully functional.

• The dev requirements now include xdoctest – a rewrite of the standard library module, doctest.

• xclim.core.locales.get_local_attrs now uses the indicator’s class name instead of the indicator itself and no longer accepts the fill_missing keyword. Behaviour is now the same as passing False.

• Indicator.cf_attrs is now a list of dictionaries. Indicator.json puts all the metadata attributes in the key “outputs” (a list of dicts). All variable metadata (names in Indicator._cf_names) might be strings or lists of strings when accessed as object attributes.

• Passing doctests are now strictly enforced as a build requirement in the Travis CI testing ensemble.

New features and enhancements

• New ensembles.kkz_reduce_ensemble method to select subsets of an ensemble based on the KKZ algorithm.

• Create new Indicator Daily, Daily2D subclasses for indicators using daily input data.

• The Indicator class now supports outputting multiple indices for the same inputs.

• xclim.core.units.declare_units now works with indices outputting multiple DataArrays.

• Doctests now make use of the xdoctest_namespace in order to more easily access modules and testdata.

Bug fixes

• Fix generic.fit dimension ordering. This caused errors when “time” was not the first dimension in a DataArray.

Internal changes

• datachecks.check_daily now uses xr.infer_freq.

• Indicator subclasses Tas, Tasmin, Tasmax, Pr and Streamflow now inherit from Daily.

• Indicator subclasses TasminTasmax and PrTas now inherit from Daily2D.

• Docstring style now enforced using the pydocstyle with numpy docstring conventions.

• Doctests are now performed for all docstring Examples using xdoctest. Failing examples must be explicitly skipped otherwise build will now fail.

• Indicator methods update_attrs and format are now classmethods, attrs to update must be passed.

• Indicators definitions without an accompanying translation (presently French) will cause build failures.

• Major refactoring of the internal machinery of Indicator to support multiple outputs.