Skip to contents

Creation of the InputsCrit object required to the ErrorCrit functions

Source: R/CreateInputsCrit.GRiwrmInputsModel.R, R/CreateInputsCrit.InputsModel.R, R/CreateInputsCrit.R
CreateInputsCrit.Rd

This function can be used either for a catchment (with an InputsModel object) or for a network (with a GRiwrmInputsModel object)

Usage

# S3 method for class 'GRiwrmInputsModel'
CreateInputsCrit(
  InputsModel,
  FUN_CRIT = ErrorCrit_KGE2,
  RunOptions,
  Obs,
  AprioriIds = getDefaultAprioriIds(InputsModel),
  k = 0.15,
  AprCelerity = 1,
  ...
)

# S3 method for class 'InputsModel'
CreateInputsCrit(InputsModel, FUN_CRIT, ...)

CreateInputsCrit(InputsModel, ...)

Arguments

InputsModel

object of class InputsModel or GRiwrmInputsModel. See CreateInputsModel

FUN_CRIT

[function (atomic or list)] error criterion function (e.g. airGR::ErrorCrit_RMSE, airGR::ErrorCrit_NSE)

RunOptions

object of class RunOptions or GRiwrmRunOptions, see CreateRunOptions

Obs

numeric, matrix or data.frame series of observed flows, see details

AprioriIds

(optional) named list or named vector whose entries are character vectors (ids of a priori sub-catchments) or numeric vectors (user-defined a priori parameters), used for the parameter regularization (see details)

k

(optional) numeric weight coefficient used in the parameter regularization (See airGR::CreateInputsCrit_Lavenne)

AprCelerity

(optional) numeric Default celerity used as a priori parameter for upstream catchments

...

arguments passed to airGR::CreateInputsCrit, see details

Value

Depending on the class of InputsModel argument (respectively InputsModel and GRiwrmInputsModel object), the returned value is respectively:

  • a InputsCrit object (See airGR::CreateInputsCrit)

  • a GRiwrmInputsCrit object which is a list of InputsCrit objects with one item per modeled sub-catchment

Details

See airGR::CreateInputsCrit documentation for a complete list of arguments.

Obs argument is equivalent to the same argument in airGR::CreateInputsCrit except that it must be a matrix or a data.frame if InputsModel is a GRiwrmInputsModel object. Then, each column of the matrix or data.frame represents the observations of one of the simulated node with the name of the columns representing the id of each node.

With a GRiwrmInputsModel object, all arguments are applied on each sub-catchments of the network.

Parameter regularization consists of defining a priori parameters which are used in a composed criterion based on the formula proposed by Lavenne et al. (2019) (See airGR::CreateInputsCrit_Lavenne). The parameter AprioriIds allows to define which neighbor sub-catchments are used for providing a priori parameters. Its format is as follows: AprioriIds <- list("Downstream sub-catchment 1" = c("A priori upstream sub-catchment 1", ...)) where the quoted strings are the ids of the sub-catchments. The nodes providing a priori parameters must be calibrated before the current one. The sequence order of calibration can be checked with getNodeRanking. If the latter is not adequate, this order can be forced by setting the node providing a priori parameters as donor of the current node in CreateGRiwrm. See vignettes for more details. The parameter AprCelerity is a default value used as a priori for the parameter 'Celerity' in case of an upstream catchment (without celerity parameter) is used as a priori catchment. In the calibration process, all a priori parameter sets are tested and the one getting the best ErrorCrit score is used for the parameter regularization. By default, the immediate upstream catchments are used as a priori catchments, as determined by getDefaultAprioriIds.

The entries of AprioriIds can also be numeric vectors of user-defined a priori parameters. In that case, the given parameters are used as the a priori parameters of the De Lavenne criterion of the named sub-catchment instead of the parameters of an upstream sub-catchment: AprioriIds <- list("Sub-catchment 1" = c("Upstream sub-catchment 1"), "Sub-catchment 2" = c(1, 50, -1, 3, 2)). A user-defined a priori parameter must have one value per model parameter of the sub-catchment (in the same order as the Param argument of airGR::RunModel: e.g. c(celerity, X1, X2, X3, X4) for a GR4J sub-catchment with celerity). User-defined a priori parameters are used as-is: no default celerity is added and no X4 rescaling is applied.

References

De Lavenne, A., Andréassian, V., Thirel, G., Ramos, M.-H., Perrin, C., 2019. A Regularization Approach to Improve the Sequential Calibration of a Semidistributed Hydrological Model. Water Resources Research 55, 8821–8839. doi:10.1029/2018WR024266

See also

CreateGRiwrm(), CreateInputsModel.GRiwrm(), CreateRunOptions(), CreateCalibOptions(), Calibration()

Examples

# Loading catchment data
data(Severn)

# Creating catchment network
nodes <- Severn$BasinsInfo[, c("gauge_id", "downstream_id",
                               "distance_downstream", "area")]
nodes$model <- "RunModel_GR4J"
rename_columns <- list(id = "gauge_id",
                       down = "downstream_id",
                       length = "distance_downstream")
griwrm <- CreateGRiwrm(nodes, rename_columns)

# Preparation of InputsModel object
BasinsObs <- Severn$BasinsObs
DatesR <- BasinsObs[[1]]$DatesR
PrecipTot <- cbind(sapply(BasinsObs, function(x) {x$precipitation}))
PotEvapTot <- cbind(sapply(BasinsObs, function(x) {x$peti}))
Qobs <- cbind(sapply(BasinsObs, function(x) {x$discharge_spec}))
Precip <- ConvertMeteoSD(griwrm, PrecipTot)
PotEvap <- ConvertMeteoSD(griwrm, PotEvapTot)
InputsModel <- CreateInputsModel(griwrm, DatesR, Precip, PotEvap)
#> CreateInputsModel.GRiwrm: Processing sub-basin 54095...
#> CreateInputsModel.GRiwrm: Processing sub-basin 54002...
#> CreateInputsModel.GRiwrm: Processing sub-basin 54029...
#> CreateInputsModel.GRiwrm: Processing sub-basin 54001...
#> CreateInputsModel.GRiwrm: Processing sub-basin 54032...
#> CreateInputsModel.GRiwrm: Processing sub-basin 54057...

# Calibration period selection
# Set aside warm-up period and use the rest for calibration
IndPeriod_Run <- seq(
  which(InputsModel[[1]]$DatesR ==
        (InputsModel[[1]]$DatesR[1] + 365 * 24 * 60 * 60)),
  length(InputsModel[[1]]$DatesR)
)
IndPeriod_WarmUp <- seq(1, IndPeriod_Run[1] - 1)

# Preparation of RunOptions object
RunOptions <- CreateRunOptions(
  InputsModel,
  IndPeriod_WarmUp = IndPeriod_WarmUp,
  IndPeriod_Run = IndPeriod_Run
)

# De Lavenne criterion with the default node-based a priori parameters
InputsCrit <- CreateInputsCrit(
  InputsModel = InputsModel,
  FUN_CRIT = ErrorCrit_KGE2,
  RunOptions = RunOptions,
  Obs = Qobs[IndPeriod_Run, ],
  transfo = "sqrt"
)

# De Lavenne criterion with a user-defined a priori parameter for node
# "54057" instead of the parameters of an upstream sub-catchment:
# one value per model parameter of the node (e.g. for a GR4J sub-catchment
# with celerity: celerity, X1, X2, X3, X4)
InputsCrit <- CreateInputsCrit(
  InputsModel = InputsModel,
  FUN_CRIT = ErrorCrit_KGE2,
  RunOptions = RunOptions,
  Obs = Qobs[IndPeriod_Run, ],
  AprioriIds = list(
    "54057" = c(1, 50, -1, 3, 2),
    "54032" = "54001",
    "54001" = "54095"
  ),
  transfo = "sqrt"
)