TITLE: Shiny app to explore climate space of SEOSAW region
DATE: 2022-09-10
AUTHOR: John L. Godlee
====================================================================


I made an R Shiny web app to explore the climate space of the 
SEOSAW region. The app can be found here, on shinyapps.io.

  [here, on shinyapps.io]: 
https://johngodlee.shinyapps.io/climate_space/

A big part of getting the app to run smoothly was to pre-process 
the data sources so they could be loaded quickly from disk, 
subsetted quickly, and rendered quickly with ggplot(). I haven't 
styled the app much to make it look pretty, as it was more a 
learning experience on how code reactive objects in Shiny.

I loaded country outlines of Africa and the SEOSAW ecoregion from 
the {seosawr} R package, and simplified them using {rmapshaper}:

    africa <- seosawr::africa
    seosaw_region <- seosawr::seosaw_region
    seosaw_bbox <- seosawr::seosaw_bbox

    africa_simp <- ms_simplify(africa, 
      keep = 0.01, keep_shapes = FALSE) %>% 
      st_intersection(., seosaw_bbox)

    seosaw_region_simp <- ms_simplify(seosaw_region, 
      keep = 0.01, keep_shapes = FALSE)

I used climate data from WorldClim, which I downloaded at 10 minute 
spatial resolution using the {raster} package:

  [WorldClim]: https://www.worldclim.org/

    bioclim <- getData("worldclim", var = "bio", res = 10)

which returns a raster stack object. Then I cropped and masked the 
climate data with the SEOSAW region polygon:

    bioclim_crop <- mask(crop(bioclim, seosaw_region_simp), 
seosaw_region_simp)

and finally, extracted the values and coordinates of each raster 
cell for each raster layer, resulting in a large matrix, with cells 
for rows, and bioclim variables or coordinates as columns, which I 
saved as a .rds file.

    bioclim_val <- cbind(coordinates(bioclim_crop), 
values(bioclim_crop))
    bioclim_val_fil <- bioclim_val[
      !apply(bioclim_val, 1, function(x) { 
        all(is.na(x[!names(x) %in% c("x", "y")])) 
      }),
    ]
    saveRDS(bioclim_val_fil, "app/data/bioclim_val_fil.rds")

The app allows you to draw a rectangle around the climate space of 
interest using two bioclim variables which you can select from a 
dropdown list. This process uses the brush operator in the Shiny 
plotOutput() function. I subsetted the raster matrix to the values 
returned by input$brush using reactive() in the Shiny app.

    rasterMapInput <- reactive({
    val_sel <- val[,c("x", "y", input$xvar, input$yvar)]
    if (!is.null(input$brush)) {
      xmin <- input$brush$xmin
      xmax <- input$brush$xmax
      ymin <- input$brush$ymin
      ymax <- input$brush$ymax
      val_sel <- val_sel[
        val_sel[,3] > xmin &
          val_sel[,3] < xmax &
          val_sel[,4] > ymin & 
          val_sel[,4] < ymax, 
        c("x", "y")] 
    }
    as.data.frame(val_sel)
    })

Then I simply used ggplot() with rasterMapInput() as the data input 
to geom_tile() to map the climate space selected on the map of 
southern Africa.

    # Extract values from selected raster layers
    valInput <- reactive({
    as.data.frame(val[,c(input$xvar, input$yvar)])
    })

    output$plot1 <- renderPlot(
    ggplot() + 
      geom_bin2d(data = valInput(), 
        mapping = aes_string(x = names(valInput())[1], y = 
names(valInput())[2]),
          colour = bg_col, fill = bg_col, bins = 100) + 
      theme_classic() + 
      theme(legend.position = "none")
    )

    output$plot2 <- renderPlot(
    map_plot + 
      geom_tile(data = rasterMapInput(), 
        aes(x = x, y = y), 
        fill = bg_col) 
    )

  ![Screenshot of the app with a dry and wet region 
selected.](https://johngodlee.xyz/img_full/shiny_climate/scrot.png)