generateScenarios: Produces time series of hydroclimatic variables for an exposure space.

Description

generateScenarios produces time series of hydroclimatic variables using requested climate attributes that correspond to a target exposure space using a reference daily time series as an input.

Usage

generateScenarios(
  reference,
  expSpace,
  simLengthNyrs = NULL,
  numReplicates = 1,
  seedID = NULL,
  controlFile = NULL
)

Value

The function returns a list containing the time series data generated. The list can contain multiple replicates (named as Rep1, Rep2 etc.) equal to the numReplicates function argument. Each replicate can contain multiple targets (named as Target1, Target2 etc.) based on the specified exposure space (expSpace). The expSpace and controlFile are also returned as part of this output list.

Arguments

reference

data.frame or list; contains reference daily climate data.
For single site data, reference is a data.frame with columns named year, month, day, *variable_name1*, *variable_name2*. Note that the first three columns of the data.frame contain the year, month, and day of the data. The columns have to be named as specified. For multi-site data, reference is a list, with elements named year, month, day, *variable_name1*, *variable_name2*. List format is suitable for both single and multi-site data. Climate variables are specified as matrices, with columns for each site.
Use viewModels() to view the valid variable names. Please refer to data provided with the package that may be loaded using data(tankDat) and data(barossaDat) for examples of the expected format of single site and multi-site reference.

expSpace

a list; created using the function createExpSpace

simLengthNyrs

a number; a scalar that specifies the length in years of each generated scenario. This argument is used only with stochastic generation. If NULL (the default), the generated simulation will be as long as reference.

numReplicates

a number; a scalar that specific the number of stochastic replicates to be generated. The default is 1.

seedID

a number; a scalar that specifies the seed to be used for the first replicate. Subsequent replicates will use seeds incremented by one. If seedID is NULL (which is the default), the function will use a random seed for stochastic time series generation. The seed used will be specified in the output. This argument is intended for use in cases that aim to reproduce an existing simulation.

controlFile

a string; to specify the model/optimisation options used for simulating time series data. The valid values are:

NULL: the simulation uses the foreSIGHT default stochastic model settings.
"scaling": the simulation uses scaling (simple/seasonal) instead of a stochastic model. If all attributes in expSpace are annual totals/averages, then simple scaling is used. If seasonality ratio attributes are also included in expSpace, then seasonal scaling is used.
path to a JSON file: the JSON file contains advanced options specify the stochastic model and optimisation inputs. These options can be used to change stochastic model types, overwrite default model parameter bounds, change default optimisation arguments, and set penalty attributes to be used in optimisation. Please refer to the function writeControlFile in order to create an controlFile JSON file.

Examples

Run this code

# Example 1: Simple scaling
#-----------------------------------------------------------------------
attPerturb<-c("P_ann_tot_m","Temp_ann_avg_m")
attPerturbType = "regGrid"
attPerturbSamp = c(2, 2)
attPerturbMin = c(0.8, -1)
attPerturbMax = c(1.1, 1)
expSpace <- createExpSpace(attPerturb = attPerturb,
                           attPerturbSamp = attPerturbSamp,
                           attPerturbMin = attPerturbMin,
                           attPerturbMax = attPerturbMax,
                           attPerturbType = attPerturbType)
data(tankDat)
simScaling <- generateScenarios(reference = tank_obs,
                                expSpace = expSpace,
                                controlFile = "scaling")

# Example 2: Seasonal scaling
#-----------------------------------------------------------------------
attPerturb<-c("P_ann_tot_m","P_ann_seasRatio")
attPerturbType = "regGrid"
attPerturbSamp = c(2, 2)
attPerturbMin = c(0.8, 0.9)
attPerturbMax = c(1.1, 1.2)
expSpace <- createExpSpace(attPerturb = attPerturb,
                           attPerturbSamp = attPerturbSamp,
                           attPerturbMin = attPerturbMin,
                           attPerturbMax = attPerturbMax,
                           attPerturbType = attPerturbType)
data(tankDat)
seasScaling <- generateScenarios(reference = tank_obs,
                                expSpace = expSpace,
                                controlFile = "scaling")

# Example 3: Stochastic simulation using foreSIGHT default settings
#----------------------------------------------------------------------
if (FALSE) {
# create an exposure space
attPerturb <- c("P_ann_tot_m", "P_ann_nWet_m", "P_ann_R10_m")
attHold <- c("P_Feb_tot_m", "P_SON_dyWet_m", "P_JJA_avgWSD_m", "P_MAM_tot_m",
"P_DJF_avgDSD_m", "Temp_ann_rng_m", "Temp_ann_avg_m")
attPerturbType = "regGrid"
attPerturbSamp = c(2, 1, 1)
attPerturbMin = c(0.8, 1, 1)
attPerturbMax = c(1.1, 1, 1)
expSpace <- createExpSpace(attPerturb = attPerturb,
                           attPerturbSamp = attPerturbSamp,
                           attPerturbMin = attPerturbMin,
                           attPerturbMax = attPerturbMax,
                           attPerturbType = attPerturbType,
                           attHold = attHold,
                           attTargetsFile = NULL)
# load example data available in foreSIGHT
data(tankDat)
# perform stochastic simulation
simStochastic <- generateScenarios(reference = tank_obs,
                                   expSpace = expSpace,
                                   simLengthNyrs = 30)
                                   }
# Example 4: Simple Scaling with multi-site data
#-----------------------------------------------------------------------
attPerturb <- c("P_ann_tot_m","P_ann_seasRatio")
attPerturbType = "regGrid"
attPerturbSamp = c(3, 3)
attPerturbMin = c(0.8, 1.2)
attPerturbMax = c(0.8, 1.2)
expSpace <- createExpSpace(attPerturb = attPerturb,
                           attPerturbSamp = attPerturbSamp,
                           attPerturbMin = attPerturbMin,
                           attPerturbMax = attPerturbMax,
                           attPerturbType = attPerturbType)
# load multi-site rainfall data
data(barossaDat)
# perform simple scaling
simScaling <- generateScenarios(reference = barossa_obs,
                                expSpace = expSpace,
                                controlFile = "scaling")

# Example 5: Multi-site stochastic simulation
#-----------------------------------------------------------------------
if (FALSE) {
attPerturb <- c("P_ann_tot_m")
attHold <- c("P_ann_wettest6monSeasRatio","P_ann_wettest6monPeakDay",
"P_ann_P99","P_ann_avgWSD_m", "P_ann_nWetT0.999_m")
attPerturbType = "regGrid"
# consider unperturbed climates in this example
attPerturbSamp = attPerturbMin = attPerturbMax = c(1)
expSpace <- createExpSpace(attPerturb = attPerturb,
                           attPerturbSamp = attPerturbSamp,
                           attPerturbMin = attPerturbMin,
                           attPerturbMax = attPerturbMax,
                           attPerturbType = attPerturbType,
                           attHold = attHold)
# load multi-site rainfall data
data(barossaDat)
# specify the penalty settings in a list
controlFileList <- list()
controlFileList[["penaltyAttributes"]] <- c("P_ann_tot_m",
"P_ann_wettest6monSeasRatio","P_ann_wettest6monPeakDay")
controlFileList[["penaltyWeights"]] <- c(0.5,0.5,0.5)
# specify the alternate model selections
controlFileList[["modelType"]] <- list()
controlFileList[["modelType"]][["P"]] <- "latent"
# specify model parameter selection
controlFileList[["modelParameterVariation"]] <- list()
controlFileList[["modelParameterVariation"]][["P"]] <- "harmonic"
# specify settings for multi-site model
controlFileList[["spatialOptions"]] <- list()
# specify spatial correlation perturbation factor
controlFileList[["spatialOptions"]][["spatCorFac"]] = 0.9
# write control file sttings to file
controlFileJSON <- jsonlite::toJSON(controlFileList, pretty = TRUE, auto_unbox = TRUE)
write(controlFileJSON, file = paste0(tempdir(), "controlFile.json"))
# run multi-site stochastic simulation - this will take a long time (e.g. hours)
sim <- generateScenarios(reference = barossa_obs, expSpace = expSpace,
                         controlFile = paste0(tempdir(), "controlFile.json"),seed=1)}

Run the code above in your browser using DataLab

Description

Usage

Value

Arguments

See Also

Examples