Visualization improvements. PlotRobinson parameters and forecast ensemble methods

As previously mentioned, we will need to plot other methods than forecast ensemble mean.

I am already doing so by adding a few lines to the function plot_ensemble_mean.R so instead of:

  # Compute ensemble mean
  ensemble_mean <- s2dv::MeanDims(fcst$data, 'ensemble')

I added a parameter to the function called method, which is 'median' by default.

  # Compute ensemble mean or other
  if (!is.null(method)) {
    method <- tolower(method)
  }
  if (method == 'mean') {
    ensemble_mean <- s2dv::MeanDims(fcst$data, 'ensemble')
  } else if (method == 'iqr') {
    ensemble_mean <- Apply(fcst$data, target_dim = 'ensemble',
                          fun = function(x) {
                          IQR(x)})$output1
  } else {
    ensemble_mean <- Apply(fcst$data, target_dim = 'ensemble', 
                          fun = function(x) {
                          median(x)})$output1
  }

I haven't changed the name of the function or the object ensemble_mean but it could be an improvement as well.

On the other hand, I needed the lenged to be centered in zero and use the S2S4E color palette. So, I modified the code in line https://earth.bsc.es/gitlab/es/sunset/-/blob/master/modules/Visualization/R/plot_ensemble_mean.R#L80

        max_value <- max(abs(var_ens_mean), na.rm = T)
        brks <- pretty(c(-1 * max_value, max_value), n = 15, min.n = 8)

For output files and titles, I substituted 'mean' by method, e.g.:

      outfile <- paste0(outdir[[var]], "forecast_ensemble_", method,
                        "-", i_syear)

I also needed to configure the PlotRobinson parameters, so, I added the output_configuration

	  output_configuration <- output_conf$PlotRobinson$forecast_ensemble_mean
          base_args <- list(data = NULL, mask = NULL, dots = NULL,
                            lon = longitude, lat = latitude,
                            lon_dim = 'longitude', lat_dim = 'latitude',
                            target_proj = target_proj, legend = 's2dv',
                            style = 'point', brks = brks, cols = cols,
                            bar_extra_margin = c(3.5, 0, 3.5, 0),
                            point_size = "auto", title_size = 10,
                            dots_size = 0.2,
                            width = 8, height = 5,
                            units = units)
          base_args[names(output_configuration)] <- output_configuration

This development, to use the output_size.yml file is also needed in plot_metrics.R

In the case of plot_mostlikely, I removed some lines because my probabilities were already in an array:

REMOVED

 if (is.null(probabilities$probs_fcst$prob_b33) ||
      is.null(probabilities$probs_fcst$prob_33_to_66) ||
      is.null(probabilities$probs_fcst$prob_a66)) {
    stop("The forecast tercile probability bins are not present inside ",
         "'probabilities', the most likely tercile map cannot be plotted.")
  }

  probs_fcst <- abind(probabilities$probs_fcst$prob_b33,
                      probabilities$probs_fcst$prob_33_to_66,
                      probabilities$probs_fcst$prob_a66,
                      along = 0)
  names(dim(probs_fcst)) <- c("bin",
                              names(dim(probabilities$probs_fcst$prob_b33)))

ADDED:

probs_fcst <- probabilities$probs_fcst

Maybe we can add a check asking for the object class, if it is a list, the use the previous lines, otherwise look for 'bin' category.

Well, I have a mess in my branch, so, let me know how to proceed to add any of these developments to sunset.

Cheers,

Núria