cbase-essd: Cloud-base Tools

#' Visualization of geographical distribution of cloud bases
#' 
#' @param df data.frame.  Result of \code{bases.cbase()}
#' @return NULL
#' @export
vis.geo <- function(df) {
    df %>%
        ## dplyr::mutate(lon = lon.caliop, lat = lat.caliop) %>%
        plotutils::discretize(lon, 180) %>%
        plotutils::discretize(lat, 90) %>%
        group_by(lon, lat, feature.qa.lowest.cloud, feature.above.surface) %>%
        summarize(cloud.base.altitude = median(cloud.base.altitude),
                  cloud.base.height = median(cloud.base.altitude - surface.elevation))  -> tmp

    tmp %>%
        filter(feature.above.surface %in% c("cloud", "clear air"),
               feature.qa.lowest.cloud == "high") %>%
        ggplot2::ggplot(ggplot2::aes(x = lon, y = lat, fill = cloud.base.height)) +
        ggplot2::geom_raster() +
        ggplot2::scale_fill_distiller("Base (km AGL)", palette = "Spectral",
                                      guide = ggplot2::guide_colorbar(## fill = ggplot2::guide_legend(##
                                          ## keywidth = 2, keyheight = 4,
                                          barwidth = 2, barheight = 16,
                                          title.position = "right",
                                          title.theme = ggplot2::element_text(angle = 270, vjust = 0.5, hjust = 0.5, size = 24))) +
        plotutils::geom_world_polygon() +
        plotutils::scale_x_geo(facet = TRUE) +
        plotutils::scale_y_geo() +
        ggplot2::facet_grid(feature.qa.lowest.cloud ~ feature.above.surface) +
        ggplot2::theme_bw(24)

    tmp %>%
        ## filter(feature.above.surface %in% c("cloud", "clear air"),
        ##        feature.qa.lowest.cloud == "high") %>%
        ggplot2::ggplot(ggplot2::aes(x = lon, y = lat, fill = cloud.base.height)) +
        ggplot2::geom_raster() +
        ggplot2::scale_fill_distiller("Base (km AGL)", palette = "Spectral") +
        plotutils::geom_world_polygon() +
        plotutils::scale_x_geo(facet = TRUE) +
        plotutils::scale_y_geo() +
        ggplot2::facet_grid(feature.qa.lowest.cloud ~ feature.above.surface) +
        ggplot2::theme_bw()

    NULL
}

#' Visualization of geographical distribution of cloud base heights
#' 
#' @param df data.frame.  Result of \code{bases.cbase()}
#' @return NULL
#' @export
vis.mean <- function(df, base_size, portrait = TRUE) {
    base.bins <- c(seq(250, 1050, 100), seq(1350, 2250, 300))
    df %>%
        ## dplyr::mutate(date = as.POSIXct("1993-01-01 00:00:00") + time,
        ##               season = plyr::laply(months(date), function(x)
        ##                   switch(x,
        ##                          "December" =, "January" = , "February" = "DJF",
        ##                          "March" =, "April" = , "May" = "MAM",
        ##                          "June" =, "July" = , "August" = "JJA",
        ##                          "September" =, "October" = , "November" = "SON")) %>%
        ##               factor(levels = c("DJF", "MAM", "JJA", "SON"), ordered = TRUE)) %>%
        plotutils::discretize(lon, seq(-180, 180, 5)) %>%
        plotutils::discretize(lat, seq(-90, 90, 5)) %>%
        dplyr::group_by(lon, lat, daynight #, season
                        ) %>%
        dplyr::summarize(n = n(),
                         error = sqrt(1 / sum(1 / pred.rmse ^ 2)),
                         base = sum(pred.ceilo / pred.rmse ^ 2) * error ^ 2) %>%
        plotutils::discretize(error, c(seq(5, 55, 10), 85, 105, 600)) %>%
        dplyr::mutate(error = factor(error, levels = stats::filter(c(seq(5, 55, 10), 85, 105, 600), c(0.5, 0.5)), ordered = TRUE)) %>%
        dplyr::mutate(base = pmax(400, pmin(1900, base))) %>%
        plotutils::discretize(base, base.bins) %>%
        dplyr::mutate(base = factor(base, levels = stats::filter(base.bins, rep(0.5, 2)), ordered = TRUE)) %>%
        ## ggplot2::ggplot(ggplot2::aes(x = lon, y = lat, fill = error)) +
        ggplot2::ggplot(ggplot2::aes(x = lon, y = lat, fill = base)) +
        ## ggplot2::ggplot(ggplot2::aes(x = lon, y = lat, fill = log10(n))) +
        ggplot2::geom_raster() +
        plotutils::geom_world_polygon() +
        plotutils::scale_x_geo(facet = !portrait) +
        plotutils::scale_y_geo() +
        ggplot2::coord_fixed(1) +
        ## ggplot2::scale_fill_distiller(palette = "Spectral") +
        # ggplot2::scale_fill_brewer(palette = "Spectral") +
        ggplot2::scale_fill_brewer("Mean cloud base height (m AGL)",
                                   palette = "Spectral",
                                   guide = ggplot2::guide_legend(direction = "horizontal",
                                                                 nrow = 1,
                                                                 keywidth = 2,
                                                                 title.vjust = 0.75,
                                                                 label.hjust = 0.5,
                                                                 label.position = "bottom")) +
        (if (portrait)
             ggplot2::facet_grid(daynight ~ .)
         else ggplot2::facet_grid(. ~ daynight)) +
        ggplot2::theme_bw(base_size) + ggplot2::theme(legend.position = "bottom")
    
}

#' Visualization of geographical distribution of cloud base heights
#' (day-night differences)
#' 
#' @param df data.frame.  Result of \code{bases.cbase()}
#' @return NULL
#' @export
vis.diff <- function(df, base_size) {
    df %>%
        plotutils::discretize(lon, seq(-180, 180, 2)) %>%
        plotutils::discretize(lat, seq(-90, 90, 2)) %>%
        dplyr::group_by(lon, lat, daynight) %>%
        dplyr::summarize(n = n(),
                         error = sqrt(1 / sum(1 / pred.rmse ^ 2)),
                         base = sum(pred.ceilo / pred.rmse ^ 2) * error ^ 2) %>%
        dplyr::group_by(lon, lat) %>%
        dplyr::filter(n() == 2) %>%
        ## dplyr::summarize(diff = ifelse(n() == 2,
        ##                                base[daynight == "day"] - base[daynight == "night"],
        ##                                NA)) %>%
        dplyr::summarize(diff = base[daynight == "day"] - base[daynight == "night"]) %>%
        dplyr::mutate(diff = pmax(-1000, pmin(1000, diff))) %>%
        plotutils::discretize(diff, c(seq(-1100, 1100, 200))) %>%
        dplyr::mutate(diff = factor(diff, stats::filter(c(seq(-1100, 1100, 200)), rep(0.5, 2)), ordered = TRUE)) %>%
        ## ggplot2::ggplot(ggplot2::aes(x = lon, y = lat, fill = error)) +
        ggplot2::ggplot(ggplot2::aes(x = lon, y = lat, fill = diff)) +
        ## ggplot2::ggplot(ggplot2::aes(x = lon, y = lat, fill = log10(n))) +
        ggplot2::geom_raster() +
        plotutils::geom_world_polygon() +
        plotutils::scale_x_geo(facet = TRUE) +
        plotutils::scale_y_geo() +
        ggplot2::coord_fixed(1) +
        ## ggplot2::scale_fill_distiller(palette = "RdYlBu") +
        ggplot2::scale_fill_brewer(palette = "RdBu", direction = -1) +
        ## ggplot2::scale_fill_distiller("Mean cloud base height (m)",
        ##                            palette = "Spectral",
        ##                            guide = ggplot2::guide_legend(direction = "horizontal",
        ##                                                          nrow = 1,
        ##                                                          keywidth = 3,
        ##                                                          label.hjust = 0.5,
        ##                                                          label.position = "bottom")) +
        ggplot2::theme_bw(base_size) ## + ggplot2::theme(legend.position = "bottom")
    
}

#' Visualization of geographical distribution of cloud-base
#' uncertainties (quantiles)
#' 
#' @param df data.frame.  Result of \code{bases.cbase()}
#' @param base_size Font base size passed to ggplot2::theme_bw()
#' @param rmse.quantiles Quantiles of the uncertainty to be displayed
#' @return NULL
#' @export
vis.quantiles <- function(df, base_size,
                          rmse.quantiles = c(0.01, 0.1, 0.25, 0.5, 0.9)) {
    df %>%
        plotutils::discretize(lon, seq(-180, 180, 5)) %>%
        plotutils::discretize(lat, seq(-90, 90, 5)) %>%
        plyr::ddply(~ lon + lat + daynight, function(df) {
            quantile(df$pred.rmse, rmse.quantiles)
        }) %>%
        tidyr::gather(quantile, rmse, dplyr::matches("%")) %>%
        dplyr::mutate(quantile = gsub("%", "\\\\%", quantile)) %>%
##        dplyr::mutate(quantile = gsub("%", "th", quantile)) %>%
        dplyr::mutate(rmse = pmax(176, pmin(674, rmse))) %>%
        plotutils::discretize(rmse, seq(175, 675, 50)) %>%
        dplyr::mutate(rmse = factor(rmse)) %>%
        ggplot2::ggplot(ggplot2::aes(x = lon, y = lat, fill = rmse)) +
        ggplot2::geom_raster() +
        plotutils::geom_world_polygon() +
        plotutils::scale_x_geo(facet = TRUE) +
        plotutils::scale_y_geo() +
        ggplot2::coord_fixed(1) +
        ggplot2::scale_fill_brewer("Cloud base height RMSE (m)",
                                   palette = "RdYlGn",
                                   guide = ggplot2::guide_legend(direction = "horizontal",
                                                                 nrow = 1,
                                                                 keywidth = 2,
                                                                 title.vjust = 0.75,
                                                                 label.hjust = 0.5,
                                                                 label.position = "bottom")) +
        ggplot2::facet_grid(quantile ~ daynight) +
        ggplot2::theme_bw(base_size) + ggplot2::theme(legend.position = "bottom")
    
}

#' Visualization of geographical distribution of cloud-base
#' uncertainties (CDF)
#' 
#' @param df data.frame.  Result of \code{bases.cbase()}
#' @return NULL
#' @export
vis.cdf <- function(df, base_size) {
    doParallel::registerDoParallel(4)
    df %>%
        plotutils::discretize(lon, seq(-180, 180, 5)) %>%
        plotutils::discretize(lat, seq(-90, 90, 5)) %>%
        plyr::ddply(~ lon + lat + daynight, function(df) {
            cdf <- ecdf(df$pred.rmse)
            data.frame(rmse = seq(320,480, by = 40)) %>%
                dplyr::mutate(cdf = cdf(rmse),
                              ncdf = nrow(df) * cdf)
        }, .parallel = TRUE) %>%
        plotutils::discretize(cdf, c(0,0.1,0.2,0.4,0.6,0.8,1)) %>%
        plotutils::discretize(ncdf, c(0,10,50,100,200,500,1000)) %>%
        dplyr::mutate(cdf = factor(cdf),
                      ncdf = factor(ncdf)) %>%
        ggplot2::ggplot(ggplot2::aes(x = lon, y = lat, fill = cdf)) +
        ggplot2::geom_raster() +
        plotutils::geom_world_polygon() +
        plotutils::scale_x_geo(facet = TRUE) +
        plotutils::scale_y_geo() +
        ggplot2::coord_fixed(1) +
        ggplot2::scale_fill_brewer("CDF",
                                   direction = +1,
                                   palette = "Spectral",
                                   guide = ggplot2::guide_legend(direction = "horizontal",
                                                                 nrow = 1,
                                                                 keywidth = 3,
                                                                 title.vjust = 0.75,
                                                                 label.hjust = 0.5,
                                                                 label.position = "bottom")) +
        ggplot2::facet_grid(rmse ~ daynight) +
        ggplot2::theme_bw(base_size) + ggplot2::theme(legend.position = "bottom")
}

#' Visualization of geographical distribution of cloud-base
#' uncertainties (CDF)
#' 
#' @param df data.frame.  Result of \code{bases.cbase()}
#' @return NULL
#' @export
vis.n <- function(df, base_size) {
    doParallel::registerDoParallel(4)
    df %>%
        plotutils::discretize(lon, seq(-180, 180, 5)) %>%
        plotutils::discretize(lat, seq(-90, 90, 5)) %>%
        plyr::ddply(~ lon + lat + daynight, function(df) {
            cdf <- ecdf(df$pred.rmse)
            data.frame(rmse = seq(320,480, by = 40)) %>%
                dplyr::mutate(cdf = cdf(rmse),
                              ncdf = nrow(df) * cdf)
        }, .parallel = TRUE) %>%
        plotutils::discretize(cdf, c(0,0.05,0.1,0.15,0.2,0.4,0.6,0.8,1)) %>%
        plotutils::discretize(ncdf, c(0,10,50,100,200,500,1000)) %>%
        dplyr::mutate(cdf = factor(cdf),
                      ncdf = factor(ncdf)) %>%
        ggplot2::ggplot(ggplot2::aes(x = lon, y = lat, fill = ncdf)) +
        ggplot2::geom_raster() +
        plotutils::geom_world_polygon() +
        plotutils::scale_x_geo(facet = TRUE) +
        plotutils::scale_y_geo() +
        ggplot2::coord_fixed(1) +
        ggplot2::scale_fill_brewer(palette = "Reds", direction = +1) +
        ggplot2::facet_grid(rmse ~ daynight) +
        ggplot2::theme_bw(base_size) + ggplot2::theme(legend.position = "bottom")
}