UsingRForSpatialAnalysis.Rmd

---
title: "Using R for Spatial Analysis"
author: "G Perkins"
date: "March 2, 2019"
output: ioslides_presentation
output:
  xaringan::moon_reader:
    css: ["default", "default-fonts", "css/my-theme.css"]
    chakra: libs/remark-latest.min.js
    lib_dir: libs
    nature:
      highlightStyle: github
      highlightLines: true
      countIncrementalSlides: false
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = FALSE)
```

## R Markdown

This is an R Markdown presentation. Markdown is a simple formatting syntax for authoring HTML, PDF, and MS Word documents. For more details on using R Markdown see <http://rmarkdown.rstudio.com>.

When you click the **Knit** button a document will be generated that includes both content as well as the output of any embedded R code chunks within the document.

## Slide with Bullets

- Bullet 1
- Bullet 2
- Bullet 3

## Slide with R Output

```{r cars, echo = TRUE}
summary(cars)
```

## Slide with Plot

```{r pressure}
plot(pressure)
```


---
title: "R for Geospatial Data Science"
subtitle: "Victoria GeoGeeks Meetup"
author: "Andy Teucher"
date: "2018/01/30"
output:
  xaringan::moon_reader:
    css: ["default", "default-fonts", "css/my-theme.css"]
    chakra: libs/remark-latest.min.js
    lib_dir: libs
    nature:
      highlightStyle: github
      highlightLines: true
      countIncrementalSlides: false
---

```{r setup, include=FALSE}
options(htmltools.dir.version = FALSE)
knitr::opts_chunk$set(warning = FALSE, 
                      message = FALSE, 
                      cache = TRUE, 
                      fig.align = 'center')
```


background-image: url(img/spreadsheet.png)

???

When Paul invited me I asked him how many people would be R users... and he said maybe none, possibly as many as three. So instead of spending a lot of time convincing you of R's spatial powers (which are many), I'm going to try to convince why using a tool like R is a good idea for any sort of analytical work, including spatial, which by the way R is REALLY good for.

I'm going to start off by describing a pretty common scenario, and then talk about how using **R** can help:

- You're doing a big project for a client. You have a four or five spatial files and a few csvs or Excel sheets containing data related to the spatial layers. 
- First you do a bunch of data cleaning in the spreadsheets - fixing date formats, unit conversions, transformations, unify inconsistent variable names, etc.

---

background-image: url(img/arcmap.jpg)

???

- You import the layers into a GIS, do a bunch of spatial operations - intersects, unions, buffers, spatial joins, etc.
- You import the tabular data and join it to the appropriate layers, and do more work.

---

background-image: url(img/arcmap-layout.jpg)

???

- Spend some time creating really nice looking map for final report

---

background-image: url(img/graph.png)

???

- Export the tabular data out so you can summarize it and make some nice graphs in Excel

---

class: middle, center

.pull-left[![](img/report.jpg)]
--

# "Thank you for your hard work, this is exactly what we were looking for!"

???

- Combine the results in a word document, convert it to a pdf and give to your client. They thank you for your great work, but say 'Oh, I think I gave you the wrong version of one of the spatial files, and that Excel spreadsheet has been updated.'

---
class: inverse

## "But, I think I gave you the wrong version of shapefile #2. Oh, and that Excel spreadsheet has been updated."

--

.center[![](img/leia-what.gif)]

---

background-image: url(img/full-cycle.png)

???

- Can you redo the analysis exactly the same way? How long will it take you?

---

background-image: url(img/Rlogo.png)

## Open source programming language
## Data analysis focus
## Cross-platform

???

# What is R?

- R is an open-source programming language for data analysis and visualization.
  - In many ways it is similar to other general purpose interpreted languages like Python, but unlike Python, R was originally created as a statistical programming language and so it is designed from the ground up around the manipulation, analysis, and visualization of data - it is now often considered the 'lingua franca' of data.
  - It also easy to make publication-quality graphics.
  - R is available for Windows, MacOS and most popular Linux distributions.
  - You can run it interactively by typing directly into the console, or from a sequence of commands written in text files, or "scripts". Scripts can be run against a dataset to produce a defined output, such as a chart or another dataset which contains the results of the analysis.


---

background-image: url(img/efficiency-car.jpg)

# Efficiency

???

So, what can **R** give you?

---

background-image: url(img/full-cycle.png)

???

The entire workflow can be done efficiently in **R**, including:

- reading in raw data from many different sources - spatial and non-spatial, 
- preparing the data for analysis, including easily integrating tabular and spatial data,
- performing spatial data manipulation and analysis,
- performing statistical analysis
- creating publication-quality graphs and figures.

---

background-image: url(img/R-flow.png)

???

The entire workflow can be done efficiently in **R**, including:

- reading in raw data from many different sources - spatial and non-spatial, 
- preparing the data for analysis, including easily integrating tabular and spatial data,
- performing spatial data manipulation and analysis,
- performing statistical analysis
- creating publication-quality graphs and figures.
- All your iteration happens in one place, and is recorded in scripts

---

background-image: url(img/photocopier.jpg)

# Reproducibility

???

Reproducibilty means:

- *You can easily recreate or update exactly what you did*: In the scenario I described, If you coded the entire above workflow in an R script, you can just change a few lines to change the input data and rerun the scripts.
- *You know what you did*: Doing your data preparation and analysis by coding it in **R** necessarily documents every step. The analysis and documentation are inextricably intertwined.
- *Others know what you did*: They can inspect and re-run your code to understand what you did, verify you've done it correctly, and build on your work.

---

background-image: url(img/k-car.jpg)

# Reliability

???

Using **R** can help you:

- *Minimize errors*: point and click and copy and paste operations are very error-prone, and it's hard to catch those errors
- And you can *find and correct errors* more easily: Changing a bit of code and re-running your script is much more efficient than doing it manually (without making another error).

---

background-image: url(img/swiss-army.jpg)

# Versatility

???

R has a huge array of add-on packages that extend the capabilities of R for performing diverse tasks, including:

- data manipulation, 
- data visualization, and
- almost every statistical method you can think of
- With a wide range of packages R also supports advanced geospatial statistics, modeling and visualization.
- Packages are generally published on CRAN - or the "Comprehensive R Archive Network", these packages are easily installable by a user from within R.

---

# Community

.img-right[![](img/R-community.jpg)]

.footnote[['Faces of R' by Maelle Salmon](http://www.masalmon.eu/2017/03/19/facesofr/)]

???

I think one of the best reasons to use **R** is the:

Fantastic online **R** Community

- **R** is hugely popular right now.
- If you have a question or are having a problem, 99% of the time you'll find that it has already been asked and answered online. And if it hasn't, people are almost always willing to help out.
- (rstats, community.rstudio.com, mailing lists, StackOverflow, RWeekly)

---

# Data Science
.center[![](img/data-science-cycle.png)]

.footnote[Wickham and Grolemund: [R for Data Science](http://r4ds.had.co.nz/)]

???

- hyped word, lots of definities out there
- To me - it's data analysis but inclusive of all steps of the data cycle, with a greater emphasis on programming.
- Spatial analysis can easily be fit into this framework, another side of the same coin.

---

# Geographical Data Science
.center[![](img/data-science-cycle.png)]

.footnote[Wickham and Grolemund: [R for Data Science](http://r4ds.had.co.nz/)]

???

- movement labelled Geographical Data Science (GDS). Combines ‘data science’ with GIS.
- focus on reproducibility and programming interface

---

### .center[Geographic Information Systems (GIS)</br>vs</br>Geographic Data Science (GDS)<sup>1</sup>]

.middle[
```{r echo=FALSE}
library(readr)
tbl_11.1 <- read_csv('Attribute,GIS,GDS
Home disciplines,Geography,"Geography, Computing, Statistics"
Software focus,Graphical User Interface,Code
Reproducibility,Minimal,Maximal')
knitr::kable(tbl_11.1, format = "html")
```
]

.footnote[
[1] Lovelace, Nowosad, and Muenchow: Geocomputation with R ([Table 1.1](https://geocompr.robinlovelace.net/intro.html#tab:gdsl))
]

???

Why R for spatial?

- Generally the same arguments as I made above for general data analysis.

- reading, writing and manipulating spatial data; making static and interactive maps; and spatial statistics and modelling

- Ability to seamlessly switch between spatial and non-spatial data processing, modeling and visualization tasks

---
background-image: url(img/vic-open-data.png)
---
background-image: url(img/dog.png)
---
```{r include=FALSE, eval=FALSE}
temp <- tempfile()
download.file("http://vicmap.victoria.ca/_GISData/DogsByNeighbourhoodSHP.zip", destfile = temp)
unzip(temp, exdir = "data/dogs")
```

```{r}
library(sf)
library(mapview)
dogs <- read_sf("data/dogs/DogsByNeighbourhood.shp")
mapview(dogs)
```

???

sf - new package that implements the simple features spec in R for representing spatial data
 - Binds to GDAL for reading and writing data, to GEOS for geometrical operations, and to Proj.4 for projection conversions and datum transformations
 - can read and write almost any format

---

```{r}
library(ggplot2)
ggplot() + 
  geom_sf(data = dogs, aes(fill = DogLicence)) + 
  scale_fill_viridis_c() + theme_minimal()
```

???

Plotting - built-in in sf, even better in ggplot2

---

background-image: url(img/hydrant.png)

---

```{r include=FALSE, eval=FALSE}
temp <- tempfile()
download.file("http://vicmap.victoria.ca/_GISData/HydrantsSHP.zip", destfile = temp)
unzip(temp, exdir = "data/hydrants")
```


```{r}
hydrants <- read_sf("data/hydrants/Hydrants.shp")
ggplot() + 
  geom_sf(data = hydrants) + 
  theme_minimal()
```

---

```{r}
hydrants <- st_intersection(hydrants, st_geometry(dogs))
ggplot() + 
  geom_sf(data = dogs, aes(fill = DogLicence)) + 
  geom_sf(data = hydrants, colour = "white") + 
  scale_fill_viridis_c() + theme_minimal()
```

???

### geometry generating logical operators

* `st_union`: union of several geometries
* `st_intersection`: intersection of pairs of geometries
* `st_difference`: difference between pairs of geometries
* `st_sym_difference`: symmetric difference (`xor`)

---

```{r eval=require(dplyr)}
dogs_and_hydrants <- st_join(dogs, hydrants) %>% 
  group_by(Neighbourh) %>% 
  summarize(licences = max(DogLicence), 
            n_hydrants = length(unique(ID)), 
            hydrants_per_dog = n_hydrants / licences)
```

```{r echo=FALSE}
knitr::kable(st_set_geometry(dogs_and_hydrants, NULL), format = "html")
```

???

### logical binary geometry predicates

* `st_intersects`: touch or overlap
* `st_disjoint`: !intersects
* `st_touches`: touch
* `st_crosses`: cross (don't touch)
* `st_within`: within
* `st_contains`: contains
* `st_overlaps`: overlaps
* `st_covers`: cover
* `st_covered_by`: covered by
* `st_equals`: equals
* `st_equals_exact`: equals, with some fuzz

### higher-level operations: summarise, interpolate, aggregate, st_join

* `aggregate` and `summarise` use `st_union` (by default) to group feature geometries
* `st_interpolate_aw`: area-weighted interpolation, uses `st_intersection` to interpolate or redistribute attribute values, based on area of overlap:
* `st_join` uses one of the logical binary geometry predicates (default: `st_intersects`) to join records in table pairs

### manipulating geometries

* `st_line_merge`: merges lines
* `st_segmentize`: adds points to straight lines
* `st_voronoi`: creates voronoi tesselation
* `st_centroid`: gives centroid of geometry
* `st_convex_hull`: creates convex hull of set of points
* `st_triangulate`: triangulates set of points (not constrained)
* `st_polygonize`: creates polygon from lines that form a closed ring
* `st_simplify`: simplifies lines by removing vertices
* `st_split`: split a polygon given line geometry
* `st_buffer`: compute a buffer around this geometry/each geometry
* `st_make_valid`: tries to make an invalid geometry valid (requires lwgeom)
* `st_boundary`: return the boundary of a geometry

---

```{r}
ggplot() + 
  geom_sf(data = dogs_and_hydrants, aes(fill = hydrants_per_dog)) + 
  scale_fill_viridis_c() + theme_minimal()
```

---

```{r fig.width=8}
ggplot(dogs_and_hydrants, aes(x = Neighbourh, y = hydrants_per_dog)) + 
  geom_col(aes(fill = licences)) + scale_fill_viridis_c() + 
  coord_flip() + theme_minimal()
```


---
.pull-left[
## Me:

Twitter: @andyteucher

GitHub: @ateucher
]

.pull-right[
## Resources:

http://fg2re.sellorm.com/

http://r4ds.had.co.nz/

https://geocompr.robinlovelace.net/

http://r-spatial.org/
]

.footnote[Source code for this presentation available [here](https://github.com/ateucher/vic-geogeeks-2018-01-30)

Made with [xaringan](https://slides.yihui.name/xaringan/)
]
---

```{r eval=require(leaflet)}
leaflet() %>%
  addProviderTiles("Stamen.TonerLite") %>% 
  addAwesomeMarkers(lng = -123.369407, lat = 48.428881, 
                    label = paste0("<h3>", emo::ji("beer"), "</h3>"), 
                    labelOptions = labelOptions(noHide = TRUE))
```
© 2019 GitHub