Why not use coord_proj("+proj=laea")?

[mdsumner]

A rhetorical question, but should be interesting to explore different approaches.

How would you make this map?

The chlorophyll-a data:

u <- "https://oceandata.sci.gsfc.nasa.gov/cgi/getfile/A20022132017243.L3m_MC_CHL_chl_ocx_9km.nc"
download.file(u, basename(u), mode = "wb")

The chlorophyll-a palette (breaks and colours):

pal <- palr::chlPal(palette = TRUE)

The projection and extents:

• outer boundary is -20 (20S)
• central longitude is 180, central latitude -90
• inner graticule is xx <- c(0, 90, 180, 270, 360); yy <- c(-60, -40)

"+proj=laea +lat_0=-90 +lon_0=180 +datum=WGS84 +ellps=WGS84 +no_defs +towgs84=0, 0, 0"

The fronts are from orsifronts::orsifronts:

subset(orsifronts::orsifronts, front %in% c("pf", "saf"))

[mdsumner]

(having some problems saving this)

[mdsumner]

for chl-a colours in gg, this almost works

# The breaks has a dummy 1000 in it, to capture any large values above the
# scaled maximum of 20 (it works for plotting rasters).

# In ggplot, you can use the gradientn to give it any sequence of colours, but
# you have to give it values scaled between 0,1. When I scaled pal$breaks that
# meant it spread out the range in the big gap between 20 and 1000, so I have to
# drop that. Here x is longitude, y is latitude, and "chl" is standard
# MODISA/SeaWiFS chlorophyll-a conc:
# 

## this is close
## chl a data frame of x, y, chl
names(chl) <- "chl"
chl[chl>20] <- 20
ggplot(as.data.frame(chl, xy = TRUE) %>% dplyr::filter(!is.na(chl)), aes(x, y)) + geom_raster(aes(fill = chl)) + scale_fill_gradientn(values = scales::rescale(head(pal$breaks, -1)), colours = pal$cols)

[mdsumner]

Primarily, to answer "why not ggdal::coord_proj"? This a raster, and coord_proj must assume long lat input, and there's no way to draw a projected grid like this, I believe the transformation happens after the geoms are set up (keen to be proved wrong!).