Added regression coverage for SIPNET event-JSON segmentation and segmented restart chaining (#4021).
write.config.SIPNET now maps the leafNResorptionFrac trait to its v2 parameter, so PFT supplied values reach the param file instead of silently keeping the template default. Completes the nitrogen cycle trait mappings alongside the existing leafOnReallocFrac. plantStorageNInit is deliberately not trait mapped; it is an initial condition, set from the IC layer like the other Init parameters.
For SIPNET v2, write.config.SIPNET now sets plantStorageNInit to the N needed for one carbon-limited leaf flush, computed from the initialized wood pool, instead of leaving the template value. With the nitrogen cycle on, a zero initial storage N suppressed the first leaf-on event entirely for vegetation initialized with standing wood (e.g. established orchards). Sipnet takes the smaller of its carbon and nitrogen limiters, so sizing N to the carbon-capped transfer removes the suppression without adding surplus N. A plantStorageNInit supplied through the IC argument still takes precedence, and zero-wood (annual) initializations still get 0.
More flexible handling of Sipnet “restart” files (checkpoints containing full model state at the end of the run):
New setting settings$model$copy.restart is a logical runtime flag controlling whether to copy each run’s restart.out to the output directory at job completion or leave it in the run directory.
New function collect_restarts works after the run and copies all the restart.outs from an entire workflow run into a single directory, for e.g. initializing follow-up workflows (#4110).
The collect_restarts() function and the copy.restart flag work independently of each other; point collect_restarts() at whichever of run/ or out/ contains the restart files.
Improvements to the job.sh written by write_segmented_configs:
Now places README.txt, segments.csv, and the full log files from each segment, into the outdir (as was already done for one-segment runs).
Now respects settings$model$delete.raw by deleting sipnet.out from segment dirs as well as the from the job outdir
No longer calls model2netcdf.SIPNET() inside each segment dir (it was redundant with the whole-job netcdf output).
Runs host setup and teardown steps once per job instead of once per segment. This reduces duplicate computation and allows using these steps for whole-rundir cleanups, e.g. settings$model$postrun = “rm -rf @RUNDIR@”. This may be especially useful to keep file counts reasonable in workflows with many crop changes (and hence many segment directories per site).
Fixed incorrect run paths in the settings$host block passed to single segments
model2netcdf.SIPNET takes LAI from sipnet.out if present (which requires Sipnet > v2.2). If it is not present, LAI is calculated as plantLeafC / leafCSpWt as previously.
split_inputs.SIPNET now avoids internal time format conversions, giving a substantial speedup and reduced memory use when processing multi-year files.
model2netcdf.SIPNET now detects the number of timesteps per day by taking the maximum count across all days in the first simulation year, rather than reading only from day 1. This prevents a factor-of-N error in flux unit conversions when the first day of output is partial (fewer timesteps than a complete day) (#3624, #3989).
Fixed a unit error in model2netcdf.SIPNET’s calculation of GWBI (kgC/m2/sec) from woodCreation (actually gC/m2/timestep, was being treated as gC/m2/day).
model2netcdf.SIPNET now takes NPP directly from sipnet.out rather than repeat Sipnet’s internal calculation (as GPP - Ra) with identical results.
Updated README with a more complete model description and instructions for installing SIPNET (#3705)
Removed tests/Rcheck_reference.log, which was used to ignore historic check messages that have now been fixed.
Initial support for SIPNET v2.0, whose features include simplified input files, simulation of management events, tracking of N cycle components, flooded soils, and anaerobic CH4 generation.
Breaking: met2model.SIPNET now writes 12-column clim files (as expected by Sipnet >= v2.0) by default. To get the previously standard 14-column output, set clim_format_version = "v1".
NITROGEN_CYCLE, LITTER_POOL, and ANAEROBIC enabled by default for SIPNET v2 runs.
model2netcdf.SIPNET now handles both V1 and v2 output header format (v2 has no Notes line; SIPNET #267).
Removed 13 obsolete v1 parameters from the v2 parameter template template.param_v2 (microbeInit, qualityLeaf, etc.) and added 16 new nitrogen cycle, anaerobic, and methane parameters.
Added nitrogen cycle (mineral_N, soil_organic_N, litter_N, N2O, N leaching, N fixation, N uptake) and methane (CH4) output conversion to NetCDF in model2netcdf.SIPNET.
Fixed crash when litterWater column absent in v2 output (LITTER_WATER removed in v2).
write.config.SIPNET now validates runtime flag dependencies (NITROGEN_CYCLE requires LITTER_POOL and ANAEROBIC) and guards v1 only parameters (litterWHC, litWaterDrainRate, litterWFracInit, microbeInit, m_ballBerry) from being set when using v2 templates.
Removed workarounds for column naming bugs in output from long-outdated legacy Sipnet version sipnet.unk.
Added trait to parameter mappings for nitrogen cycle parameters in write.config.SIPNET, including tissue C:N ratios, N volatilization/leaching/resorption, fixation, and methane rates.
segment_dataframe() now returns an empty dataframe when date filtering removes all crop-cycle segments, instead of a single row with NA columns that caused downstream segment config errors (#4007).
PEcAn.SIPNET 1.10.0
Added
write.events.SIPNET() generates SIPNET events.in files from an events.json file (#3623).
met2model.SIPNET now accepts argument var.names, listing which variables should be extracted from the file. If not provided, it extracts all variables in the file (#3563).
Removed
The sipnet2datetime function is no longer used anywhere and therefore has been removed (#3622).
Changed
Breaking: Renamed the setting used to pass soil and hydrology parameters. write.config.SIPNET previously read these from settings$run$inputs$soilinitcond, now settings$run$inputs$soil_physics to better reflect that these are state factors applicable to the whole run rather than initial conditions (Quianyu Li, #3406).
model2netcdf.SIPNET no longer writes separate <year>.nc.var files for every year of output. Use PEcAn.utils::nc_write_varfiles() to create these as needed (#3611).
Restart and met2model functions now print less to the console unless verbose = TRUE (#3544, #3563).
Fixed
write.config.SIPNET now checks more carefully whether an optional variable exists in an initial condition file before trying to read it, therefore printing fewer messages about (expectedly) missing variables (#3545).
When passed a vector of multiple input paths, write.config.SIPNET was choosing one at random; it now throws an error (Blesson Thomas, #3298). Note that a single input path per call has always been the intended usage; being passed many was a second bug in PEcAn.uncertainty that is also now fixed.
model2netcdf.SIPNET no longer assumes a constant value of pecan_start_doy across years, which lead to incorrect calculations of sub_dates and sub_dates_cf at year boundaries (@DongchenZ, #3622).
When phenology inputs contain missing values, write.config.SIPNET now tries to use an average across years for that site before falling back to fixed cross-site defaults (Quianyu Li, #3680).
write.config.SIPNET now adjusts soil water capacity to match the specified soil depth (#3634).
Fixed unit errors in write.config.SIPNET calculation of leafCSpWt and Amax (#3608, #3664).
PEcAn.SIPNET 1.9.0
License change
PEcAn.SIPNET is now distributed under the BSD three-clause license instead of the NCSA Open Source license.
write.config.SIPNET() now consults PFT trait definitions to decide whether to initialize LAI in the leaf-on or leaf-off state, instead of the previous hard-coded defaults (#3419). Specifically:
a PFT with fracLeafFall > 0.5 will be treated as deciduous (previously hardcoded to deciduous for anything other than boreal conifers)
deciduous PFTs will get laiInit=0 if the simulation start date is not between leafOnDay and leafOffDay (previously hardcoded to May through September)
The generated Sipnet run script (job.sh) now works correctly, including across machines, when met/input/output files are specified relative to the working directory (#3418). Absolute paths continue to work as always.
PEcAn.SIPNET 1.8.0
Support for all Sipnet variables in read_restart and write_restart, for integration with state data assimilation workflows