tile_selected_extremes <- as.character(list_tb_extremes[[3]]$df_extremes$tile_id)

  list_fig_filename <- vector("list",length=length(tile_selected_extremes))

    tile_selected <- tile_selected_extremes[i]

    #tb_subset$tile_id

    tb_tile_tmp <- subset(tb_subset,tb_subset$tile_id==tile_selected)

    tb_tile <- aggregate(rmse~date,tb_tile_tmp,min)

    #idx <- test2$date #transform this format...

    idx <- as.Date(strptime(tb_tile$date, "%Y%m%d"))   # interpolation date being processed

    d_z <- zoo(tb_tile,idx) #make a time series ...

    #add horizontal line...

    #plot(d_z[[metric_name]])

    fig_filename <- paste("Figure3a_time_series_extremes_tile_",model_name[j],"_",tile_selected,"_",

                      region_name,"_",out_suffix,".png",sep="")

    res_pix <- 960

    col_mfrow <- 1

    row_mfrow <- 1

    #only mod1 right now

    png(filename=fig_filename,

        width=col_mfrow*res_pix,height=row_mfrow*res_pix)

    title_str <- paste("Time series of",metric_name,"for",tile_selected,"and",region_name,sep=" ")

    plot(d_z$rmse,ylab=metric_name,xlab="dates")

    title(title_str)

    #length(unique(test$year_predicted))

    #unique(tb$year_predicted)

    dev.off()

    list_fig_filename[[i]] <- fig_filename

    ##Now plot the location

    #Take the max for now?

    val_selected <- max(d_z$rmse)

    d_z_selected <- d_z[d_z$rmse==val_selected,]

    date_selected <- d_z_selected$date

    #data_v_selected <- subset(tb_data_v,tb_data_v$date==date_selected)# & tb_data_v$tile_id==tile_selected)

    data_v_selected <- subset(tb_data_v,tb_data_v$date==date_selected & tb_data_v$tile_id==tile_selected)

    #tb_tile[,rmse==]

    data_s_selected <- subset(tb_data_s,tb_data_s$date==date_selected & tb_data_s$tile_id==tile_selected)

    fig_filename <- paste("Figure3b_data_stations_trainaing_testing_extremes_tile_",model_name[j],"_",tile_selected,"_",

                      region_name,"_",out_suffix,".png",sep="")

    coordinates(data_s_selected)<- c(data_s_selected$lon,data_s_selected$lat)

    coordinates(data_v_selected)<- c(data_v_selected$lon,data_v_selected$lat)

    res_pix <- 960

    col_mfrow <- 1

    row_mfrow <- 1

    #only mod1 right now

    png(filename=fig_filename,

        width=col_mfrow*res_pix,height=row_mfrow*res_pix)

    title_str <- paste("Data stations",metric_name,"for",tile_selected,"and",region_name,sep=" ")

    p_shp <- spplot(reg_layer,"ISO3" ,col.regions=NA, col="black") #ok problem solved!!

    #title("(a) Mean for 1 January")

    #tb_sorted$freq_extremes2 <- tb_sorted$freq_extremes/17

    p <- bubble(tb_sorted,"freq_extremes",main=paste("Extremes per tile and by ",model_name[j]," for ",

                                                                threshold_val[i]))

    plot(d_z$rmse,ylab=metric_name,xlab="dates")

    title(title_str)

    #length(unique(test$year_predicted))

    #unique(tb$year_predicted)

    dev.off()

  #Now plot the location

  tile_selected_extremes <- as.character(list_tb_extremes[[3]]$df_extremes$tile_id)

  list_fig_filename <- vector("list",length=length(tile_selected_extremes))