R for Antarctic and Southern Ocean science

class: center, middle, inverse, title-slide

# R for Antarctic and Southern Ocean science
### Michael Sumner, Ben Raymond, Anton Van de Putte
### MEASO18

---

class: middle
background-image: url(extra/Rlogo.png)
background-size: auto 65%
background-position: 50% 50%
  
# R for Antarctic and Southern Ocean Science
  
- context and rOpenSci (Ben)
- modern state of R (Mike)
- some of our stuff (Ben, Mike)
- the biodiversity.aq Antarctic virtual research environment (Anton)
- general chat

---

# rOpenSci <img style="background-color: white; padding:10px; float:right;clear:none; margin:20px;" width="30%" src="extra/ropensci_logo.png" alt="ropensci logo" />
 
rOpenSci: R for reproducible science, open data & software

https://ropensci.org/community/
  
Tutorials, use cases, discussion forum, open calls, physical meetings

### For R developers/regular users:
  
- slack: sign up at http://slack.ropensci.org/
- package onboarding

SCAR/rOpenSci joint initiative to build the Antarctic R community: see the #antarctic channel on slack.

Note that the slack channel tends to be fairly technical, and probably most suited for regular R users and package developers. Mechanisms for broader engagement (discussion fora, how-to's etc) will happen as this initiative progresses.

---

# Antarctic/SO R packages
  
See https://github.com/SCAR/antarctic-r-packages

- taxonomy
- mapping
- environmental data access
- ocean colour processing
- biodiversity, diet data
- animal tracking

What packages do we need? What tasks/analyses should be easier than they currently are?
  
SCAR's github: https://github.com/SCAR

---
class: middle
background-image: url(extra/Rlogo.png)
background-size: auto 65%
background-position: 50% 50%

# Modern state of

---
# The tidyverse

https://www.tidyverse.org

A family of modern, fast, consistent packages.

The single best guide  http://r4ds.had.co.nz/  (free online!)

Key packages

* **dplyr** - data frame creation, manipulation, subsetting, and merging as`filter`, `select`, `mutate`, `rename`, `join`, consistent with database workflows
* **readr**, **readxl** - read data frames from text and Excel files
* **purrr** - functional programming, tools needed for unstructured, list data
* **ggplot2** - grammar of graphics, consistent visual exploration and plot making from data frame data

These packages work extremely well together.

---

# Key spatial R

* sp
* raster
* sf
* rgdal

These work together, somewhat.

Use google, stackoverflow, gis.stackexchange.com.

Twitter has a friendly and helpful R community: #rstats

---
class: middle
background-image: url(extra/Rlogo.png)
background-size: auto 65%
background-position: 50% 50%

# Some of our stuff

---

# Local techniques - a toolbox approach

We're creating R packages and collating related resources.

SOKI.aq overview of our local tools and approaches:

http://soki.aq/display/SET/1.+General+Strategy

* [angstroms](https://CRAN.R-project.org/package=angstroms) - bending raster for working with ROMS
* [rbgm](https://CRAN.R-project.org/package=rbgm) - working with Atlantis BGM Box Geometry Models

---

# A toolbox approach: blueant and raadtools

https://github.com/AustralianAntarcticDivision/blueant

- pre-configured definitions for environmental/satellite data sets
- functions for maintaining local copies of those data

```r
library(blueant)

## define where we want to store our data
my_data_dir <- "c:/my/big/data/collection"
cf <- bb_config(local_file_root=my_data_dir)

## add a few data sets to our configuration
my_sources <- c("NSIDC SMMR-SSM/I Nasateam sea ice concentration",
 "IBCSO bathymetry")

cf <- cf %>% bb_add(sources(my_sources))

## go!
bb_sync(cf)
```

---

# A toolbox approach: blueant and raadtools

https://github.com/AustralianAntarcticDivision/blueant

- *anyone* can use this to maintain their own copy of these data

- we use this to maintain a central data collection for a community of researchers

- with a scheduled task to keep it updated

- together with `raadtools` and virtual machine capability provides us with a consistent, community-level framework

- separation of data-getting (blueant) with data-access (raadtools)

---

# A toolbox approach: blueant and raadtools

- the `raadtools` package complements `blueant` 
- it provides functions for reading and manipulating environmental data in a simple and consistent way

https://github.com/AustralianAntarcticDivision/raadtools

```r
library(raadtools)
?readice
?readsst
?readssh
?readcurr
?readtopo

Functions that work on the blueant data collection:

http://australianantarcticdivision.github.io/raadtools/reference/index.html
```

---

# A toolbox approach: blueant and raadtools

```r
(ice <- readice(latest = TRUE))
```

```
## class       : RasterStack 
## dimensions  : 332, 316, 104912, 1  (nrow, ncol, ncell, nlayers)
## resolution  : 25000, 25000  (x, y)
## extent      : -3950000, 3950000, -3950000, 4350000  (xmin, xmax, ymin, ymax)
## coord. ref. : +proj=stere +lat_0=-90 +lat_ts=-70 +lon_0=0 +k=1 +x_0=0 +y_0=0 +a=6378273 +b=6356889.449 +units=m +no_defs 
## names       : nt_20180409_f18_nrt_s.bin 
## min values  :                         0 
## max values  :                       100 
## time        : 2018-04-09
```
---

# A toolbox approach: blueant and raadtools

```r
(ssh <- readssh("2017-04-10", xylim = extent(120, 150, -70, -30)))
```

```
## class       : RasterBrick 
## dimensions  : 160, 120, 19200, 1  (nrow, ncol, ncell, nlayers)
## resolution  : 0.25, 0.25  (x, y)
## extent      : 120, 150, -70, -30  (xmin, xmax, ymin, ymax)
## coord. ref. : +proj=longlat +rf=298.257 +a=6378136.3 
## data source : in memory
## names       : Absolute.dynamic.topography 
## min values  :                      -1.296 
## max values  :                       0.955 
## time        : 2017-04-10
```

---

# Converting between projections

Raster can reproject between grids to match disparate data, like sea ice
in a polar projection, sea surface temperature in longitude/latitude.

```r
sst <- readsst("2017-04-10", xylim = extent(-180, 180, -90, 0))
polar_sst <- projectRaster(sst, ice)
plot(polar_sst, col = palr::sstPal(64))
```

---
# Converting between projections

Or in the opposite direction.

```r
longlat_ice <- projectRaster(ice, sst)
plot(longlat_ice)
```

---
class: middle
background-image: url(extra/Rlogo.png)
background-size: auto 65%
background-position: 50% 50%

# More emerging R stuff

---
# New spatial framework

The "simple features"" package `sf` provides very fast and more consistent spatial data support, for map data, points, lines, polygons.

* a large shapefile?  try `sf::read_sf`. 
* convert to old style `as(x, "Spatial")` for use with `raster`, `spdep` ...
* `sf` support in `ggplot2` is coming

Still in development, but package `ggplot2` from Github has very good plotting for `sf`. Use `devtools::install_github("tidyverse/ggplot2"))`

---
# geom_sf

```r
## a map data set prepared earlier, the CCAMLR statistical areas
ccamlr <- readRDS("/home/shared/data/assessment/sectors/ccamlr_statareas.rds")
prj <- raster::projection(ccamlr)
library(ggplot2)
library(sf)
library(raster)
ice <- raadtools::readice(latest = TRUE)[[1L]]
ccamlr <- st_transform(st_as_sf(ccamlr), projection(ice))

## convert gridded sea ice to data frame
names(ice) <- "iceconc"
iced <- as.data.frame(ice, xy = TRUE)
```

(See plot on next slide)
---

```r
ggplot(ccamlr) + geom_raster(data = iced, aes(x, y, fill = iceconc)) + 
  geom_sf(fill = NA, aes(colour = name))
```

---
#  fasterize

Merging vector (polygons) and raster (gridded) data can be difficult!  Fasterize is super fast for this.

Being able to generate a gridded version of a polygon data set can be extremely powerful

- classifying a gridded data set
- looking up values per polygon

```r
library(fasterize)
ccamlrgrid <- fasterize(ccamlr, ice, field = "name")
```

(Plot on next slide)

---
# fasterize

```r
plot(ccamlrgrid, col = sample(rainbow(nrow(ccamlr)+1)[-1]))
```

---
# mapview, mapedit, plotly

Create interactive maps and even edit map data.

* [mapview](https://r-spatial.github.io/mapview/)
* [mapedit](https://CRAN.R-project.org/package=mapedit/)
* [plotly](https://CRAN.R-project.org/package=plotly)

---
class: middle
background-image: url(extra/Rlogo.png)
background-size: auto 65%
background-position: 50% 50%

# The Antarctic component of EU-LifeWatch

---
class: middle
background-image: url(extra/avdp1.png)
background-size: 100% auto
background-position: center center

---
class: middle
background-image: url(extra/avdp2.png)
background-size: 100% auto
background-position: center center

---
# For more info

http://marine.lifewatch.eu/

http://lifewatch.be/

http://www.lifewatch.eu

---
class: middle
background-image: url(extra/avdp4.png)
background-size: 100% auto
background-position: center center

---
class: middle
background-image: url(extra/avdp5.png)
background-size: 100% auto
background-position: center center

---
class: middle
background-image: url(extra/avdp6.png)
background-size: 100% auto
background-position: center center

---
class: middle
background-image: url(extra/avdp7.png)
background-size: 100% auto
background-position: center top

# R packages

[solong](https://github.com/SCAR/solong)

Allometric equations that relate the body size of Southern Ocean taxa to their body part measurements

[sohungry](https://github.com/SCAR/sohungry)

Access to the SCAR Southern Ocean Diet and Energetics Database

[antanym](https://github.com/SCAR/antanym)

Antarctic geographic place name information.