Interpolation and plotting of 2d/spatial timeseries data on an irregular grid with R

Question

(this is my first post so (i) I hope not to break too many rules and (ii) have to store example plots externally)

I would like to visualize irregular gridded timeseries data where the displayed parameter is also a function of a geographical measure like latitude or water depth. An example data file that contains the date (date), the geographical parameter water depth (dep) and the parameter of interest salinity (sal) and a preliminary scatterplot produced with ggplot2 are stored at our

owncloud

password: timeseries

The R-code for the ggplot plot is:

# Load required packages
library(ggplot2)
library(data.table)
library(colorRamps)
library(scales)

# Import spatial timeseries data
df      <- data.table(read.csv("timeseries_example.csv"))
df$date <- as.POSIXct(strptime(df$date, format="%m/%d/%Y", tz="GMT"))

# Scatterplot with color representing the z parameter

Fig <-
ggplot(data=df, aes(date, dep, col=Sal))+
  geom_point()+
  scale_y_reverse()+
  scale_colour_gradientn(colours = matlab.like2(7), oob=squish)

tiff("./example_timeseries_R_ggplot.tiff", width = 200, height = 100, 
  units =  'mm', res = 300, compression = 'lzw')
Fig
dev.off()

As the data are spaced irregular in space and time, plotting with ggplot's geom_tile() function requires interpolation.

The freely available software ocean data view ( ODV ) enables such interpolation and I would like to reproduce the ODV plot also stored at our owncloud (link above) with R.

As this problem is similar to previously solved issues, I tried to interpolate the parameter sal on a finer grid of date and dep with the package akima . However, this did not work with the x parameter being a POSIXct object.

Does anyone have a solution to this?

Answer 1

I've had good luck with the MBA package:

# Load required packages
library(ggplot2)
library(lubridate)
library(reshape2)
library(colorRamps)
library(scales)
library(MBA)

# Import spatial timeseries data
df      <- read.csv("timeseries_example.csv")
df$date <- as.POSIXct(strptime(df$date, format="%m/%d/%Y", tz="GMT"))
df$date <- decimal_date(df$date)

mba <- mba.surf(df[,c('date', 'dep', 'Sal')], 100, 100)
dimnames(mba$xyz.est$z) <- list(mba$xyz.est$x, mba$xyz.est$y)
df3 <- melt(mba$xyz.est$z, varnames = c('date', 'depth'), value.name = 'salinity')

Fig <-
  ggplot(data=df3, aes(date, depth))+
  geom_raster(aes(fill = salinity), interpolate = F, hjust = 0.5, vjust = 0.5) +
  geom_contour(aes(z = salinity)) + 
  geom_point(data = df, aes(date, dep), colour = 'white') +
  scale_y_reverse() +
  scale_fill_gradientn(colours = matlab.like2(7))
Fig

There are some anomalies that you may be able to clean up with the with the interpolation settings. I just used the default.