CSIndicators (version 1.1.2)

MergeRefToExp: Merge a Reference To Experiments

Description

Some indicators are defined for specific temporal periods (e.g.: summer from June 21st to September 21st). If the initialization forecast date is later than the one required for the indicator (e.g.: July 1st), the user may want to merge past observations, or other references, to the forecast (or hindcast) to compute the indicator. If the forecast simulation doesn't cover the required period because it is initialized too early (e.g.: Initialization on November 1st the forecast covers until the beginning of June next year), a climatology (or other references) could be added at the end of the forecast lead time to cover the desired period (e.g.: until the end of summer).

Usage

MergeRefToExp(
  data1,
  data2,
  dates1 = NULL,
  dates2 = NULL,
  start1 = NULL,
  end1 = NULL,
  start2 = NULL,
  end2 = NULL,
  time_dim = "time",
  memb_dim = "member",
  ncores = NULL
)

Value

A multidimensional array created from the combination of 'data1' and 'data2'. The resulting array will contain the following dimensions: the original dimensions of the input data, which are common to both arrays and for the 'time_dim' dimension, the sum of the corresponding dimension of 'data1' and 'data2'. If 'memb_dim' is not null, regarding member dimension, two different situations can occur: (1) in the case that one of the arrays does not have member dimension or is equal to 1 and the other array has multiple member dimension, the result will contain the repeated values of the array one up to the lenght of member dimension of array two; (2) in the case that both arrays have member dimension and is greater than 1, all combinations of member dimension will be returned.

Arguments

data1

A multidimensional array with named dimensions. All dimensions must be equal to 'data2' dimensions except for the ones specified with 'memb_dim' and 'time_dim'.

data2

A multidimensional array of named dimensions matching the dimensions of parameter 'data1'. All dimensions must be equal to 'data1' except for the ones specified with 'memb_dim' and 'time_dim'.

dates1

A multidimensional array of dates with named dimensions matching the temporal dimensions of parameter 'data1'. The common dimensions must be equal to 'data1' dimensions.

dates2

A multidimensional array of dates with named dimensions matching the temporal dimensions on parameter 'data2'. The common dimensions must be equal to 'data2' dimensions.

start1

A list to define the initial date of the period to select from 'data1' by providing a list of two elements: the initial date of the period and the initial month of the period. The initial date of the period must be included in the 'dates1' array.

end1

A list to define the final date of the period to select from 'data1' by providing a list of two elements: the final day of the period and the final month of the period. The final date of the period must be included in the 'dates1' array.

start2

A list to define the initial date of the period to select from 'data2' by providing a list of two elements: the initial date of the period and the initial month of the period. The initial date of the period must be included in the 'dates2' array.

end2

A list to define the final date of the period to select from 'data2' by providing a list of two elements: the final day of the period and the final month of the period. The final date of the period must be included in the 'dates2' array.

time_dim

A character string indicating the name of the temporal dimension that will be used to combine the two arrays. By default, it is set to 'time'. Also, it will be used to subset the data in a requested period.

memb_dim

A character string indicating the name of the member dimension. If the 'data1' and 'data2' have no member dimension, set it as NULL. It is set as 'member' by default.

ncores

An integer indicating the number of cores to use in parallel computation.

Details

This function is created to merge observations and forecasts, known as the ‘blending’ strategy (see references). The basis for this strategy is that the predictions are progressively replaced with observational data as soon as they become available (i.e., when entering the indicator definition period). This key strategy aims to increase users’ confidence in the reformed predictions.

References

Chou, C., R. Marcos-Matamoros, L. Palma Garcia, N. Pérez-Zanón, M. Teixeira, S. Silva, N. Fontes, A. Graça, A. Dell'Aquila, S. Calmanti and N. González-Reviriego (2023). Advanced seasonal predictions for vine management based on bioclimatic indicators tailored to the wine sector. Climate Services, 30, 100343, tools:::Rd_expr_doi("10.1016/j.cliser.2023.100343").

Examples

data_dates <- c(seq(as.Date("01-07-1993", "%d-%m-%Y", tz = 'UTC'),
                   as.Date("01-12-1993","%d-%m-%Y", tz = 'UTC'), "day"),
               seq(as.Date("01-07-1994", "%d-%m-%Y", tz = 'UTC'),
                   as.Date("01-12-1994","%d-%m-%Y", tz = 'UTC'), "day"))
dim(data_dates) <- c(time = 154, sdate = 2)
ref_dates <- seq(as.Date("01-01-1993", "%d-%m-%Y", tz = 'UTC'),
                as.Date("01-12-1994","%d-%m-%Y", tz = 'UTC'), "day")
dim(ref_dates) <- c(time = 350, sdate = 2)
ref <- array(1001:1700, c(time = 350, sdate = 2))
data <- array(1:(2*154*2), c(time = 154, sdate = 2, member = 2))
new_data <- MergeRefToExp(data1 = ref, dates1 = ref_dates, start1 = list(21, 6),
                         end1 = list(30, 6), data2 = data, dates2 = data_dates,
                         start2 = list(1, 7), end = list(21, 9), 
                         time_dim = 'time')