airGRiwrm 0.8.1
Breaking changes:
- CreateRunOptions [Breaking change]: The default value of
Outputs_Sim has changed from "all" to a
minimal set
(c("DatesR", "WarmUpQsim", "Qsim", "Qsim_m3", "QsimDown", "Param", "StateEnd"))
in order to reduce the size of the GRiwrmOutputsModel object for large
networks. Use Outputs_Sim = "all" to restore the previous
behaviour (#141).
- Calibration [Breaking change]:
Calibration on a simple
InputsModel now requires a RunOptions argument,
inserted in the second position
(Calibration(InputsModel, RunOptions, CalibOptions, ...)).
Positional calls written as
Calibration(InputsModel, CalibOptions) are broken; named
calls and the calibration of network models (GRiwrmInputsModel) are
unchanged (#199).
- CreateInputsCrit [Breaking change]: the default error criterion
changed from
ErrorCrit_NSE to ErrorCrit_KGE2
and the default AprioriIds changed from NULL
to a computed a priori node
(getDefaultAprioriIds(InputsModel)); several a priori nodes
are now tested and the best one is kept. The numeric results of the
default criteria and of the calibration change (#197, #196).
- CreateInputsCrit [Breaking change]: the De Lavenne criterion now
raises an error if the X4 parameter of the a priori node is below 0.5,
and the a priori error criterion is now computed from the current
sub-catchment error criterion; the values of
InputsCritLavenne and the calibration results change
accordingly (#193).
- Calibration [Breaking change]: reservoirs now follow the
force-release/diversion logic during calibration by default
(
forceReservoirObs = TRUE in
RunModel/Calibration of a GRiwrmInputsModel,
with a calibration flag in the reservoir RunOptions); the simulated
reservoir behaviour during calibration can change (#199).
New features:
- CreateRunOptions: allow users to select which simulation outputs are
returned via the
Outputs_Sim argument, either as a
character vector applied to all nodes, or as a named list with per-node
selection (#141). CemaNeige nodes now also carry Precip in
their default Outputs_Sim: the default plot of a CemaNeige
node requires it in the OutputsModel to compute the
precipitation panel axis as soon as the per-layer solid precipitation
(Psol) is present (upstream airGR ylimP
limitation; the standalone airGR bug is still to be fixed on the airGR
side). Precip is a recorded meteo input, not a CemaNeige
Fortran output: the CemaNeige NOutputs and the model run
are unchanged (#178).
- CreateInputsCrit: allow user-defined a priori parameters for the De
Lavenne criterion: entries of
AprioriIds can now be numeric
vectors (one value per model parameter of the node,
e.g. c(celerity, X1, X2, X3, X4)) used as the a priori
parameters instead of the parameters of an upstream sub-catchment
(#61).
- plot.GRiwrmOutputsModel: display the meteorological time series
(precipitation and potential evapotranspiration) at the whole-catchment
scale: when available in the GRiwrmOutputsModel object, the
time series of each node are replaced on the fly by their area-weighted
mean over the catchment of the node and all its upstream nodes, so that
the displayed meteorology reflects the whole catchment integrated in the
simulated discharge (#178). For a CemaNeige node, the per-layer series
displayed by
airGR::plot.OutputsModel are covered as well:
the mean air temperature (Temp), the solid precipitation
(Psol) and the snow pack state (SnowPack) of
each elevation layer are replaced by their whole-catchment mean, as well
as the mean over the layers that airGR derives from them. The same
adjustment is available as a new exported function
adjustMeteoToCatchment(): it computes, for a node of a
GRiwrmOutputsModel object, the meteorological time series at
the whole-catchment scale (including the CemaNeige layer series), so
that users can compute statistics from the meteorological data of the
whole catchment instead of the local sub-basin data of the node
(#178).
- RunModel: a new S3 method
RunModel.GRiwrmOutputsModel
re-simulates a model from a previous GRiwrmOutputsModel object
(with modified inputs), and a previous OutputsModel can now be
merged with a new one (merge.OutputsModel,
merge.GRiwrmOutputsModel); the new exported function
getNextTimeSteps() helps to select the time steps of the
re-simulation (#173).
- RunModel.Supervisor: a new
Yinit argument sets the
initial Y state for a hot restart; the Yinit can be computed from a
previous GRiwrmOutputsModel (the warm-up outputs are kept in
the object and feed the supervised run) (#208, #209). The initial Y
state is now applied at every restart of the supervision (#210).
- CreateInputsModel: a new
FUN_REGUL argument allows
providing a user-defined regulation function (“local rule”) for a node;
the function is called with an extra named argument
e = environment() (contract to document on the user side)
(#29).
- CreateController: the variable to control/measure can now be
selected at each node location (reservoir volume, derived flows as
controlled/measured variables), and the
Y and
U locations accept NULL (#188, #174).
- CreateInputsModel:
ZInputs now accepts a list (coerced
to a numeric vector) (#206).
- RunModel_Reservoir: pass-through mode (“transparent” reservoir) with
Qrelease = NA and Vsim = 0 (#201).
- CreateInputsCrit: a new exported function
getDefaultAprioriIds(InputsModel) returns the default a
priori node ids (#197).
- CreateGRiwrm: gauged nodes inside an ungauged cluster are now
detected (#168).
- plot.OutputsModelReservoir and plot.GRiwrmOutputsModel: the observed
volume series (
Vobs) is now displayed in the plots
(#189).
- plot.GRiwrm: a new
server argument and the
MERMAID_URL environment variable provide an alternative to
the mermaid.ink service (#184).
Changes:
- The minimum required version of ‘airGR’ has been raised to 1.7.8:
CreateRunOptions reads the per-node
FeatFUN_MOD attribute for every node, and Reservoir nodes
require the fix of airGR issue #167 (both provided by the airGR 1.7.8
CRAN release, 2025-12-12). Users on airGR <= 1.7.6 now get a clear
dependency error at installation instead of a runtime crash in
CreateRunOptions().
- CreateRunOptions: the initial states (
IniStates) are
now taken into account and checked for Reservoir nodes (previously
ignored); new errors/warnings may be raised on configurations that were
previously accepted (#182).
- Calibration: the calibration of ungauged clusters now takes the
IniStates of the involved nodes into account, and a new
error is raised if a Reservoir node is in an ungauged cluster
(#182).
Bug fixes:
- RunModel: an explicit
Outputs_Sim selection now works
on semi-distributed networks: the outputs required to propagate the
simulated flows between nodes (in particular WarmUpQsim
when the warm-up period is non-zero) are now included automatically in
an explicit Outputs_Sim selection, instead of failing with
the error “Time series WarmUpQsim in ‘QcontribDown’ should have the same
length as ‘RunOptions$IndPeriod_WarmUp’” (#98).
- RunModel.Supervisor: crash with
RunModel_Lag on a node
(#191).
- RunModel.Supervisor: crash with Reservoir + Diversion nodes
(#190).
- RunModel.Supervisor: wrong detection of the warm-up period
(#174).
- RunModel.Supervision: crash at the hot restart (#176).
- RunModel.Supervision: missing attributes (class,
DatesR) for GR and OutputsModelReservoir
objects (#174, #177).
- RunModel.Supervision: wrong retrieval of the data for the direct
injections (#211).
- RunModel:
NA handling in IniStates
(#174).
- RunModel.GRiwrmOutputsModel: the original warm-up data are now kept
in the OutputsModel (#179).
- RunModel.OutputsModel: sign error of the Diversion flow (#173).
- RunModel_Lag (routing): crash at the
StateEnd
computation on very short simulations without a warm-up period
(#192).
- CreateInputsCrit: the De Lavenne criterion now updates the upstream
discharges before computing the criterion (#193).
- Calibration:
RunModel_ungauged is now allowed for the a
priori criterion of an ungauged cluster (#193).
- getDefaultAprioriIds: unique node ids guaranteed and bypass on
reservoirs (#196).
- transferGRparams: wrong computation of the X4 parameter transferred
from an upstream donor node; the transferred parameter values in
ungauged cluster calibrations change accordingly (#205).
- CreateCalibOptions: crash when all parameters are fixed (#183).
- Ungauged clusters: fixed parameters in a cluster, a gauged node in a
cluster without fixed parameters, the donor search with Diversion,
receivers removed from the cluster, and the gauged node detection were
fixed (#168, #167).
- CreateGRiwrm: wrong rank of nodes with an ungauged cluster and
reservoirs (#168).
- CreateController: direct injection in
Y with
var = "Qupstream" and the case of a missing Y
location (#188).
- CreateInputsModel:
PrecipScale is now defined at the
node level (#187) and Qrelease now accepts a matrix
(#200).
- plot.OutputsModelReservoir: the
which argument is now
handled when the plot is called from box (#203).
Documentation:
- New vignette V07: example combining tactical and operational
management (PID controller on a reservoir,
RunModel.GRiwrmOutputsModel) (#180).
- Vignette V04: reformatted for readability and the NA handling in the
controller measurements documented (#97).
- Documentation: the irstea.fr links replaced by the corresponding
forge.inrae.fr pages, the irrigation and reservoir vignette examples
revised, and the
CreateGRiwrm documentation updated
(#204).
- A documentation wave covered
Calibration,
CreateCalibOptions, CreateInputsModel,
RunModel.Supervisor, RunModel_Reservoir
(examples, target/observed release, PID supervisor examples),
transferGRparams and extractParam (#172,
#189).
Internal changes:
- Examples: the closed-loop Supervisor block of the shared Reservoir
example (used by the
RunModel_Reservoir and
plot.OutputsModelReservoir manual pages) is now wrapped in
\donttest — it calls R at every time step and made the
examples exceed the CRAN 5-second limit, which led to the rejection of
the 0.8.0 submission by the CRAN pretest (2026-11-06). The code stays
visible in the manual and is still executed by the package CI and by
R CMD check --run-donttest.
airGRiwrm 0.7.0
Breaking changes:
- CreateSupervisor [Breaking change]: Only allow to apply a command on
DirectInjection and Diversion nodes (#101)
- CreateInputsModel [Breaking change]: do not allow to use
Qobs on nodes other than Direct Injection and Diversion
(#99)
New features:
- Handle simple in-line reservoir inside the model (#51)
- Use of non gauged stations in the network (#42)
- RunModel.Supervision: handle NA values on injected flows (#94)
- Add checks on Supervisor functions (#102)
- Get
plot.Qm3s available as a function (#104)
- Implementation of “Diversion” node (#95)
- CreateInputsModel: downgrade error on “
Qobs only for
nodes Direct Injection and Diversion” to warning (#109)
- Diversion: Remove error “The downstream node of a Diversion node
must be different than the downstream node of the node is attached to”
(#111)
- Ungauged node: allow to define donor manually (#129)
- Ungauged node - CreateGRiwrm: Search donor through Diversion if not
available by natural network (#132)
- Highlight the water deficit at a node due to too much withdrawals
(#144)
- Integration of reservoirs as nodes in the network (#90)
- Handle direct withdrawal in reservoirs (#147)
- plot.GRiwrm: use web service instead of DiagrammeR::mermaid
(#150)
- plot.GRiwrm: sketch ungauged cluster with subgraphs (#151)
- Implementation of non gauged station with donor other than a
downstream gauged station (#92)
- CreateInputsCrit: allow a priori node not only at upstream
(#156)
- Allow Diversion on Reservoir (#146)
- CreateInputsModel: deprecate
Qobs parameter and use
Qinf instead (#120)
- CreateInputsModel: Specify the error message: “‘Qobs’ column names
must be included in ’id’s of the GRiwrm object” (#152)
plot.GRiwrmOutputsModel: handle other units than mm /
time step (#105)- Feature request: function for getting parameters from
GRiwrmOutputsCalib (#86)
- Speed up RunModel_Lag for RunModel.Supervisor (#164)
- plot.GRiwrm: allow to use optional parameters of
mermaid function (#160)
- Ungauged node cluster: improve outsider gauged node detection
(#168)
Bug fixes:
- Wrong file name for the “Get started” vignette (#79)
- CreateInputsCrit: wrong id in error message (#85)
- ConvertMeteoSD: crash with upstream nodes with no area (#89)
- Regularisation: check if current node and a priori node use the same
model (#93)
- Regularisation: taking into account X4 transformation (#88)
- RunModel.Supervisor doesn’t work with Diversion node (#106)
- Allow to use node with model RunModel_Lag (#107)
- Calibration does not work when upstream catchments are both gauged
and ungauged (#108)
- Ungauged node: incorrect definition of donor with Reservoir and bug
with Diversion nodes (#110)
- Ungauged node: crash with upstream Diversion node (#113)
- Ungauged node: crash on diversion to node outside the sub-network
(#112)
- Ungauged node: difference of ErrorCrit between Calibration and
RunModel (#115)
- getSD_Ids crashes on upstream nodes with Diversion (#116)
- CreateSupervisor: wrong definition of allowed nodes for command
(#117)
- Calibration: Diversion is not handled on upstream nodes (#122)
- Calibration: crash with ungauged node and multiple Diversions
(#123)
- CreateInputsModel: wrong area of sub-basin with upstream Reservoir,
Lag or Direct Injection nodes (#124)
- RunModel.Supervisor: error with Diversion node (#126)
- Ungauged node: Diversions are not handled correctly in Calibration
(#127)
- Calibration: crash with a diverted ungauged node (#128)
- Ungauged nodes: X4 transformation not handled with CemaNeige models
(#135)
- Crash with Hysteresis in CemaNeige (#134)
- Ungauged nodes: crash with a reservoir and several upstream nodes
(#136)
- Ungauged node: Diversion to Reservoir crashes Calibration
(#130)
- Wrong calibration node order with multiple ungauged node clusters
(#149)
- CreateGRiwrm: several Diversions on the same node do not raise error
(#125)
- Reservoir output plot color palette issue (#154)
- Wrong sorting for calibration of ungauged nodes (#155)
- Donor defined on gauged model node turns the node into ungauged at
Calibration (#157)
- Calibration: crash when transferring from upstream donor to upstream
receiver (#158)
- plot.GRiwrm crashes with a single node (#153)
- CreateInputsModel: Don’t allow ungauged donor (#131)
- Rename internal function
RunModel.SD (#166)
- Calibration: crash with downstream receiver node with upstream
reservoir (#167)
Documentation:
- Improve Supervisor documentation (#100)
- Update package citation reference (#119)
- Add a main page in the package documentation (#77)
- Improve RunModel documentation (#133)
Internal changes:
- Check failed on release version (#78)
- Transfer documentation to github pages (#148)
- Publish documentation package on github with github workflow
(#162)
- CI: add check with version “dev” of airGR (#84)
airGRiwrm 0.6.2
- airGR changes cause failed check (#139)
airGRiwrm 0.6.1
Changes:
- Simplify the use of airGR in airGRiwrm (#63)
CreateInputsModel: Make Qobs parameter
optional (#60)- airGR compatibility: change on LengthHydro unit (#32)
CreateInputsCrit: Change obs parameter characteristics
(#38)- Update URLs in the DESCRIPTION file (#45)
- Use S3 plot method for GRiwrm class objects (#26)
- Rename function GRiwrm to CreateGRiwrm (#46)
CreateInputsCrit: transfo is mandatory for
parameter regularization (#56)
New features:
plot.Qm3s: customize legend position (#75)- Regularisation: Add default value for parameter Celerity (#58)
- Add network consistency checks in
GRiwrm (#36)
- Handle CemaNeige compatibility (#52)
- Use S3 plot method for GRiwrmOutputsModel class objects (#26)
- Handling correctly initial conditions (#48)
- Calibration with parameter regularization (#54)
Bug fixes:
- Results differences between versions on vignette V04 (#75)
plot.GRiwrm not working in gitlab-ci (#74)plot.GRiwrm: mermaid code is displayed with the diagram
(#73)CreateGRiwrm crashes when keeping all columns and
rename some (#64)- Breaking change in airGR in issue HYCAR-Hydro/airgr#137 (#62)
- Review documentation for publication on CRAN (#43)
- Vignettes: working directory instability (#35)
- airGR compatibility: debugged version of RunModel_Lag (#33)
CreateInputsModel: Error when using data.frame for Qobs
(#37)RunModel.Supervisor: Error in ctrlr\(U[seq.int(length(sv\)ts.index), i]
(#39)- Supervisor: measurement on network downstream node returns
NULL (#40)
RunModel: Suspected bug on
OutputsModel$Qsim in the training example (#41)- Test fail after airGR update on outputting warm-up Qsim (#50)
- Wrong Qobs use in Lavenne function criteria (#57)
Internal changes:
- Prepare the package for v0.6.x CRAN submission (#71)
- Clone on github master and dev branches of the repository (#68)
- Update airGR dependency to CRAN v1.7.0 (#69)
- Add an airGR galaxy tab on the website? (#49)
- CI: dependency issues with Latex in Check as CRAN (#53)
- Remove dependency to R > 3.5 (#59)
- pkgdown: wrong documentation for methods CreateRunOptions and
CreateCalibOptions (#65)
- Review documentation for publication on CRAN (#43)
- Push roxygen outputs on the repository (#34)
- Generation of the https://airgriwrm.g-eau.net site documentation
(#44)
- Automatically update website from package repository (#47)
airGRiwrm 0.5.0
(Release date: 2021-03-07)
New features:
- Feedback control (#19)
- RunModel of GRiwrm networks: add a data.frame of simulated flows in
OutputsModel (#30)
- Plot simulated flows of all the nodes in m3/s (#31)
Changes:
- RunModel: Uncoupling of hydrological and hydraulic models (#28)
airGRiwrm 0.4.0
(Release date: 2020-12-28)
New features:
- Convert basin meteorological data to sub-basin level (#21)
Changes:
- Clarify dependency with ‘DiagrammeR’ package (#24)
Minor changes:
- Replace vignette examples on Seine River by a fake example from data
provided by airGR (#13)
Bug fixes:
- Impossibility to inject flow associated to an area (#23)
- Error in the area used for the sub basins (#22)
airGRiwrm 0.3.1
(Release date: 2020-08-07)
New features:
- Calibration of influenced semi-distributed model (#11)
airGRiwrm 0.3.0
(Release date: 2020-08-07)
New features:
- Add node of type “direct flow” in order to inject or withdraw flows
into the model (#5)
airGRiwrm 0.2.1
(Release date: 2020-06-11)
Changes:
- Remove Gits class object and use CreateInputsModel directly
(#7)
- Remove Girop class object and integrate hydrological model and area
in Ginet (#9)
- Rename “Ginet” class object to “Griwrm” (#10)
airGRiwrm 0.2.0
(Release date: 2020-06-06)
New features:
- Calibration of semi-distributed model (#3)
airGRiwrm 0.1.0
(Release date: 2020-05-25)
New features:
- Database structuring (#1)
- Scheduling airGR model runs (#2)