01-prac1.Rmd

# (PART) GIS tools {-}

# Geographic Information

## Learning outcomes

By the end of this practical you should be able to:

1.  Describe and explain GIS data formats and databases
1.  Source and pre-process spatial data
1.  Load and undertaken some basic manipulation of spatial data in QGIS and R
1.  Evaluate the (dis)advantages of each GIS you have used

## Homework

Outside of our scheduled sessions you should be doing around 12 hours of extra study per week. Feel free to follow your own GIS interests, but good places to start include the following:

::: {.infobox .assignment data-latex="{note}"}
**Exam**

Each week we will provide a short task to test your knowledge, these should be used to guide your study for the final exam.

The task is to join some non spatial data to some spatial data.

* Go to the [New Zealand spatial data portal](https://datafinder.stats.govt.nz/) and download the file Territorial Authority 2018 (generalised), [these are city or district councils](https://www.localcouncils.govt.nz/lgip.nsf/wpg_url/About-Local-Government-Local-Government-In-New-Zealand-Councils-roles-and-functions#TerritorialAuthorities(District%20and%20City%20Councils)). Make sure it's the Territorial Authority 2018 data **not** SA1. 

* Go to the [Stats NZ website](https://www.stats.govt.nz/information-releases/statistical-area-1-dataset-for-2018-census-updated-march-2020) and download the Statistical area 1 dataset for 2018 for the whole of New Zealand. Download the excel file this week, **not** the `.csv`.

* Unzip the downloaded census file and open  2018-SA1-dataset-individual-part-3a-total-NZ_updated_16-7-20, you will see a tab for Territorial authority. Join the 2018 paid employee field to the spatial data and make a basic map. Hint, you may need to make a new `.csv` file from the data.

:::

::: {.infobox .note data-latex="{note}"}
**Reading**

Each week i'll provide a core chapter to support the skills taught here. However, **you are free to read whatever interests you** and Adam and I strongly encourage this, so please don't be constrained to this recommendation. Consult the [reading list](https://rl.talis.com/3/ucl/lists/139FBAF8-DACD-60FB-8BDC-E9C9E09BA885.html?lang=en-GB&login=1), practical and lecture each week for more ideas.

This week:

-   [GIS and Cartography](https://www.sciencedirect.com/science/article/pii/B9780080449104000341) by Goodchild (2009)

-   [Chapter 2 "Geographic data in R"](https://geocompr.robinlovelace.net/spatial-class.html) from Geocomputation with R by Lovelace, Nowosad and Muenchow (2020).

-   [Opening practice: supporting reproducibility and critical spatial data science](https://link.springer.com/article/10.1007/s10109-020-00334-2#Sec3) by Brunsdon and Comber (2020)

**Watching**

-   [What is Spatial Data Science](https://www.youtube.com/watch?v=osAbJeTho5w) from some of our friends at CARTO
:::

## Recommended listening `r emo::ji("headphones")`

Some of these practicals are long, take regular breaks and have a listen to some of our fav tunes each week.

[Andy](https://open.spotify.com/album/1A3nVEWRJ8yvlPzawHI1pQ). Week 1. No other choice than Vampire Weekend. The band formed during college (University in the USA) and produced their first album of their own whilst working full time jobs! Incredible. The lead vocalist and guitarist Ezra Koenig was a school teacher in Brooklyn, New York. I've seen them a few times at Glastonbury and in London during which they gave the people what they wanted and took requests from their entire catalogue, really amazing musicians. Note some of the guest appearances in this album as they also might make an apperance later in the term!

[Adam](https://open.spotify.com/album/3LVWk5ZlGgFPvTUVSo8EKh?si=0tvDERBmRAOQgyRztZK-iw) OK, it's week 1, so I'm going in HARD. This week it's the raver's raver, the DJ's DJ. This man is a style icon and so drum & bass it hurts - it's Voltage! Here he is with his new album, Balance Over Symmetry. This will surprise a lot of people as it's an eclectic mix of DnB bangers through to some Burial-inspired minimal dubstep. Follow him on [Instagram](https://www.instagram.com/voltage_uk_/) and get your ears around his brilliant.

## The Basics of Geographic Information

Geographic data, geospatial data or geographic information is data that identifies the location of features on Earth. There are two main types of data which are used in GIS applications to represent the real world. **Vectors** that are composed of points, lines and polygons and **rasters** that are grids of cells with individual values...

```{r echo=FALSE, out.width = "300pt", fig.align='center', cache=FALSE, fig.cap="Types of spatial data. Source: [Spatial data models](https://planet.uwc.ac.za/nisl/gis/web_page/page_15.htm)"}
knitr::include_graphics('prac1_images/rasvsvec.png') 
```

In the above example the features in the real world (e.g. lake, forest, marsh and grassland) have been represented by points, lines and polygons (vector) or discrete grid cells (raster) of a certain size (e.g. 1 x 1m) specifying land cover type.

### Data types in statistics

Before we go any further let's just quick go over the different types of data you might encounter

Continuous data can be measured on some sort of scale and can have any value on it such as height and weight.

Discrete data have finite values (meaning you can finish counting something). It is numeric and countable such as number of shoes or the number of computers within the classroom.

Foot length would be continuous data but shoe size would be discrete data.

```{r echo=FALSE, out.width = "450", fig.align='center', cache=FALSE, fig.cap="Continuous and discrete data. Source: [Allison Horst data science and stats illustrations](https://github.com/allisonhorst/stats-illustrations)"}
knitr::include_graphics('allisonhorst_images/continuous_discrete.png')
```

Nominal (also called categorical) data has labels without any quantitative value such as hair colour or type of animal. Think names or categories - there are no numbers here. 

Ordinal, similar to categorical but the data has an order or scale, for example if you have ever seen the chilli rating system on food labels or filled a happiness survey with a range between 1 and 10 --- that's ordinal. Here the order matters, but not the difference between them.

Binary data is that that can have only two possible outcomes, yes and no or shark and not shark.

```{r echo=FALSE, out.width = "550pt", fig.align='center', cache=FALSE, fig.cap="Nominal, ordinal and binary data. Source: [Allison Horst data science and stats illustrations](https://github.com/allisonhorst/stats-illustrations)"}
knitr::include_graphics('allisonhorst_images/nominal_ordinal_binary.png')
```

### Important GIS data formats

There are a number of commonly used geographic data formats that store vector and raster data that you will come across during this course and it's important to understand what they are, how they represent data and how you can use them.

#### Shapefiles

Perhaps the most commonly used GIS data format is the shapefile. Shapefiles were developed by [ESRI](http://www.esri.com/), one of the first and now certainly the largest commercial GIS company in the world. Despite being developed by a commercial company, they are mostly an open format and can be used (read and written) by a host of GIS Software applications.

A shapefile is actually a collection of files ---- at least three of which are needed for the shapefile to be displayed by GIS software. They are:

1.  `.shp` - the file which contains the feature geometry
2.  `.shx` - an index file which stores the position of the feature IDs in the `.shp` file
3.  `.dbf` - the file that stores all of the attribute information associated with the coordinates -- this might be the name of the shape or some other information associated with the feature
4.  `.prj` - the file which contains all of the coordinate system information (the location of the shape on Earth's surface). Data can be displayed without a projection, but the `.prj` file allows software to display the data correctly where data with different projections might be being used

On Twitter and want to see the love for shapefiles....have a look at [the shapefile account](https://twitter.com/shapefiIe)

#### GeoJSON

GeoJSON [Geospatial Data Interchange format for JavaScript Object Notation](http://geojson.org/) is becoming an increasingly popular spatial data format, particularly for web-based mapping as it is based on JavaScript Object Notation. Unlike a shapefile in a GeoJSON, the attributes, boundaries and projection information are all contained in the same file.

#### Shapefile and GeoJSON

We're now going to explore a shapefile (`.shp` ) and GeoJSON (`.geojson`) in action.

Go to: <http://geojson.io/#map=16/51.5247/-0.1339>

```{r echo=FALSE, out.width = "500pt", fig.align='center', cache=FALSE}
knitr::include_graphics('prac1_images/JSONwebsite.png') 
```

1.  Using the drawing tools to the right of the map window, create 3 objects: a point, line and a polygon as I have done above. Click on your polygon and colour it red and colour your point green
2.  Using the 'Save' option at the top of the map, save two copies of your new data -- one in `.geojson` format and one in `.shp` format
3.  Open your two newly saved files in a text editor such as notepad or notepad++. For the shapefile you might have to unzip the folder then open each file individually. What do you notice about the similarities or differences between the two ways that the data are encoded?

#### Raster data

Most raster data is now provided in GeoTIFF (`.tiff`) format, which stands for Geostarionary Earth Orbit Tagged Image File. The GeoTIFF data format was created by NASA and is a standard public domain format. All necesary information to establish the location of the data on Earth's surface is embedded into the image. This includes: map projection, coordinate system, ellipsoid and datum type.

#### Other data formats

Aforementioned data types and formats are likely to be the ones you predominately encounter. However there are several more used within spatial analysis. These include:

**Vector**

-   GML (Geography Markup Language ---- gave birth to Keyhold Markup Language (KML))

**Raster**

-   Band SeQuential (BSQ) - technically a method for encoding data but commonly referred to as BSQ.
-   Hierarchical Data Format (HDF)
-   Arc Grid

There are normally valid reasons for storing data in one of these other formats. For example, BSQ are raster data with a separate text header file (`.hdr`) providing geographic spatial reference information. Earth observation data often monitors the electromagnetic spectrum in bands. Humans see in the visible range of the spectrum and our vision is composed of red, green and blue wavelengths. If we wanted to analyse just the red wavelength the BSQ format would let us *read in* only that data. In comparison a GeoTIFF might come with all the data 'packaged' in one file and when doing analysis over thousands of images would significantly slow things down. That said you can now often find GeoTIFFs separated in a similar format to BSQ and it's fairly straightforward to convert between raster formats.

#### Geodatabase

A geodatabase is a collection of geographic data held within a database. Geodatabases were developed by ESRI to overcome some of the limitations of shapefiles. They come in two main types: Personal (up to 1 TB) and File (limited to 250 - 500 MB), with Personal Geodatabases storing everything in a Microsoft Access database (`.mdb`) file and File Geodatabases offering more flexibility, storing everything as a series of folders in a file system. In the example below we can see that the FCC_Geodatabase (left hand pane) holds multiple points, lines, polygons, tables and raster layers in the contents tab.

```{r echo=FALSE, out.width = "500pt", fig.align='center', cache=FALSE}
knitr::include_graphics('prac1_images/geodatabase.png') 
```

#### GeoPackage

```{r echo=FALSE, out.width = "100pt", fig.align='center', cache=FALSE, fig.cap="GeoPacakge logo"}
knitr::include_graphics('prac1_images/geopkg.png')
```

A GeoPackage is an open, standards-based, platform-independent, portable, self-describing, compact format for transferring geospatial data. It stores spatial data layers (vector and raster) as a single file, and is based upon an SQLite database, a widely used relational database management system, permitting code based, reproducible and transparent workflows. As it stores data in a single file it is very easy to share, copy or move.

#### SpatiaLite

```{r echo=FALSE, out.width = "100pt", fig.align='center', cache=FALSE, fig.cap="SpatialLite logo"}
knitr::include_graphics('prac1_images/spatialite.png')
```

SpatialLite is an open-source library that extends SQLite core. Support is fairly limited and most software that supports SpatiaLite also supports GeoPackage, as they both build upon SQLite. It doesn't have any clear advantage over GeoPackage, however it is unable to support raster data.

#### PostGIS

```{r echo=FALSE, out.width = "100pt", fig.align='center', cache=FALSE, fig.cap="PostGIS logo"}
knitr::include_graphics('prac1_images/postGIS.jpg') 
```

PostGIS is an opensource database extender for PostrgeSQL. Essentially PostgreSQL is a database and PostGIS is an add on which permits spatial functions. The advantages of using PostGIS over a GeoPackage are that it allows users to access the data at the same time, can handle large data more efficiently and reduces processing time. In [this example](https://medium.com/@GispoLearning/learn-spatial-sql-and-master-geopackage-with-qgis-3-16b1e17f0291) calculating the number of bars per neighbourhood in Leon, Mexico the processing time reduced from 1.443 seconds (SQLite) to 0.08 seconds in PostGIS. However, data stored in PostGIS is much harder to share, move or copy.

#### What will I use

The variety of data formats can see a bit overwhelming. I still have to check how to load some of these data formats that I don't use frequently. But don't worry, most of the time you'll be using shapefiles, GeoPackages or raster data.

## General data flow

As Grolemund and Wickham state in [R for Data Science](https://r4ds.had.co.nz/)...

> "Data science is a huge field, and there's no way you can master it by reading a single book."

However, a nice place to start is looking at the typical workflow of a data science (or GIS) project which you will see throughout these practicals, which is summarised nicely in this diagram produced by [Dr. Julia Lowndes](https://twitter.com/juliesquid) adapted from Grolemund and Wickham.

```{r echo=FALSE, out.width = "600pt", fig.align='center', cache=FALSE, fig.cap="Updated from Grolemund & Wickham's classis R4DS schematic, envisioned by Dr. Julia Lowndes for her 2019 useR! keynote talk and illustrated by Allison Horst. Source: [Allison Horst data science and stats illustrations](https://github.com/allisonhorst/stats-illustrations)"}
knitr::include_graphics('allisonhorst_images/environmental_data_science_r4ds_general.png')
```

To begin you have to **import** your data (not necessarily environmental) into R or some other kind of GIS to actually be able to do any kind of analysis on it.

Once imported you might need to **tidy** the data. This really depends on what kind of data it is and we cover this later on in the course. However, putting all of your data into a consistent structure will be very beneficial when you have to do analysis on it --- as you can treat it all in the same way. Grolemund and Wickham state that data is tidy when "each column is a variable, and each row is an observation", we cover this more in next week in the [Tidying data] section.

When you have (or haven't) tidied data you then will most likely want to **transform** it. Grolemund and Wickham define this as "narrowing in on observations of interest (like all people in one city, or all data from the last year), creating new variables that are functions of existing variables (like computing speed from distance and time), and calculating a set of summary statistics (like counts or means)". However, from a GIS point of view I would also include putting all of your data into a similar projection, covered next week in [Changing projections] and any other basic process you might do before the core analysis. Arguably these processes could include things such as: clipping (cookie cutting your study area), buffering (making areas within a distance of a point) and intersecting (where two datasets overlap).

Tidying and transfrom = data wrangling. Remember from the introduction this could be 50-80% of a data science job!

```{r echo=FALSE, out.width = "450pt", fig.align='center', cache=FALSE, fig.cap="dplyr introduction graphic. Source: [Allison Horst data science and stats illustrations](https://github.com/allisonhorst/stats-illustrations)"}
knitr::include_graphics('allisonhorst_images/dplyr_wrangling.png')
```

After you have transformed the data the next best thing to do is **visualise** it --- even with some basic summary statistics. This simple step will often let you look at your data in a different way and select more appropraite analysis.

Next up is **modelling**. Personally within GIS i'd say a better term is processing as the very data itself is usually a computer model of reality. The modelling or processing section is where you conduct the core analysis (more than the basic analysis already mentioned) and try to provide an answer to your research question.

Fianlly you have to **communicate** your study and outputs, it doesn't matter how good your data wrangling, modelling or processing is, if your intended audience can't interpret it, well, it's pretty much useless.

::: {.infobox .tip data-latex="{note}"}
In a few weeks come back and revisit this data flow section to see how what you have learnt fits the framework presented.
:::

## Data

The volume of geographic information which is freely available for use in the UK is increasing exponentially and spatially referenced data can often be found in many different places. In this practical we're going to use data from the London data store --- a free and open data-sharing portal provided by the Greater London Authority (GLA), also known as City Hall that is the devolved regional governance body of London.

We are going to get spatial data of the London boroughs and join flytipping (the illegal deposit of waste, commonly on road verges) data that is provided as a `.csv` file. `.csv` stands for comma-separated values (CSV) --- it uses a comma to separate each value.

At the end of this document I'll also run through some common sources of data that will stand you in good stead (be advantageous) for the rest of the course.

### File paths

In your N drive: create a new folder called GIS and within this a sub folder called wk1. It is up to you how you organise your files. Make sure you change the file paths where appropriate to your own.

### Data download

Firstly we need to get a spatial outline of the London boroughs. The geographic boundaries that are used in the UK are a complex, often inter-related, but ever changing mass of areas. For anyone new to the UK (or indeed not a trained quantitative geographer), it can be quite a daunting task to attempt to understand all of the boundaries that are in use. Fortunately the Office for National Statistics (ONS) has an online beginners guide to UK geography. If you need more information on the vast array of different UK geographies, go and explore these resources:

-   <https://geoportal.statistics.gov.uk/datasets/c0db0e8c67d04935bcf1749ca6027fef/about>

Another [easy to read guide on census / administrative geography](https://www.towerhamlets.gov.uk/Documents/Borough_statistics/Research-tools-and-guidance/RB-Census2011-Census-Geography-Guide-2013-05.pdf) was produced by the London Borough of Tower Hamlets - skip to page 2 for a visual summary

Let's download some data..

-   Spatial Data

1.  To get the data go to the [London data store](https://data.london.gov.uk/)

2.  Search for Statistical GIS Boundary Files for London

3.  Download the statistical-gis-boundaries-london.zip

4.  Unzip the data and save it to your wk1 folder.

-   CSV data

1.  On the same website search for fly-tipping incidents

2.  Download the `.csv` file

### Data pre-processing

**Question** Open the `.csv` in Excel, what do you notice about how the data is stored?

**Answer** The year is a column and for each area the values are repeated for different years. In our analysis it is easier to have the different years as a column and populated for each area. So, we want to go from this...

```{r echo=FALSE, out.width = "600pt", fig.align='center', cache=FALSE}
knitr::include_graphics('prac1_images/csv_original.png')
```

To this...

```{r echo=FALSE, out.width = "600pt", fig.align='center', cache=FALSE}
knitr::include_graphics('prac1_images/csv_pivot.png')
```

::: {.infobox .tip data-latex="{note}"}
In previous years there have errors in this `.csv`, for example there used to be year value of 2017-2018 that was incorrect. Whilst these have been resolved, it's always important to check your data. 
:::

As we are going to use this dataset in QGIS and R I've done it in Excel using a pivot table. In future we'll use R to automate tasks like this.

1.  Go to Insert \> PivotTable

2.  Select the original table and create a PivotTable in a new worksheet

3.  The PivotTable Field box will appear, experiment with the different fields in each of the areas

I've used the following:

```{r echo=FALSE, out.width = "400pt", fig.align='center', cache=FALSE}
knitr::include_graphics('prac1_images/pivot_example.png')
```

Note how I've altered the total_action_taken to the sum of... as the original was displaying incorrectly, to do so:

1.  Click on drop down button for total_action_taken \> Value Field Settings \> select Sum

It's important to think about what data we actually need in the next step and it's good practice to avoid data redundancy where possible.

**Spoiler** The spatial data we have downloaded already contains borough name, so we don't need it twice. However, we do need a field to link the two datasets on. You could use borough name, but when using text fields sometimes input variations can affect joins. For example, if you had the University of Manchester in one dataset and Manchester University in another the join would fail. Consequently it's usually best to join datasets on a column that has a code or number.

Now save the Excel sheet that contains the pivot table as a new `.csv`. Make sure that the first row of data holds the column titles. Remove all empty rows.

When saving the file also avoid any special characters (e.g. -) and spaces, use an underscore instead of spaces.

::: {.infobox .warning data-latex="{note}"}
**Warning** Spatial software does not like file names with spaces or special characters.
:::

Now it's time to load, inspect and do some basic manipulation of this data. As mentioned in the lecture there are several GIS software 'types', here we will repeat the same process across QGIS and R.

### QGIS

#### Load data

1.  Search for and open QGIS
2.  Click on the open data source manager. Just above the word **browser** in the top left of the screen

```{r echo=FALSE, out.width = "700pt", fig.align='center', cache=FALSE}
knitr::include_graphics('prac1_images/qgis_datamanager.png') 
```

1.  Navigate to the `London_Borough_Excluding_MHW.shp` and add it, you then have to close the data source manager.

You can right click on the layer to view the attribute table, however the `.csv` file must be loaded nto QGIS in order to join it to a shapefile.

1.  Open the data source manager and select Delimited Text
2.  Navigate to our `.csv` file and provide a suitable layer name
3.  Under Record and Fields Options make sure the number of header lines to discard is 0 and the First record has field names box is selected (this is assuming you left a title for each column in your `.csv`)
4.  Under Geometry Definition select No geometry (attribute table only)

```{r echo=FALSE, out.width = "700pt", fig.align='center', cache=FALSE}
knitr::include_graphics('prac1_images/csv_datamanager.png') 
```

Does the sample data seem right?

1.  If so, click add then close

::: {.infobox .warning data-latex="{note}"}
**Note** Your `.csv` might have a extra header present (a row before Row Labels). You can either: (a) close this and go and delete this in the `.csv` now and reload it or (b) change the number of header lines to discard to 1 in the Record and Fields Options in the screenshot above.
:::

#### Join data

1.  Right click on the London boroughs layer \> Properties \> Joins
2.  Click the plus button at the bottom of the box
3.  Complete the dialogue box

```{r echo=FALSE, out.width = "500pt", fig.align='center', cache=FALSE}
knitr::include_graphics('prac1_images/qgis_join.png') 
```

#### Export data

Now let's export to a GeoPackage.

1.  Right click on the London boroughs layer \> Export \> Save Feature As
2.  Select the GeoPackage format and complete the File name (the saved file name for the GeoPackage) and the Layer name (the name for this layer within the GeoPackage). Recall that a GeoPackage can store many data layers as a single file
3.  The new layer will be added to the map, so you can remove the old one (Right click on the layer \> Remove Layer). Make sure you remove the right one!

```{r echo=FALSE, out.width = "500pt", fig.align='center', cache=FALSE}
knitr::include_graphics('prac1_images/make_geopkg.png') 
```

We've now made a GeoPackage that we can connect our map project to.

1.  Under **Browser**, Right click on GeoPackage \> New Connection \> Navigate to your GeoPackage
2.  Click the down arrow left to GeoPackage and you should see the one you just navigated to. Click the down arrow on the database and you can see your layer.
3.  Now we are going to import our `.csv` into our GeoPackage. To do so go: Database \> DB Manager
4.  Select your GeoPackage in the left hand pane
5.  Import layer/file then select the `.csv`. Click Ok.

Again, remember that the `.csv` in the Layers tab (bottom left) is the original. Remove it, then from the GeoPackage just click, hold and drag the `.csv` into the layer pane.

Now let's make a quick thematic map...

1.  Right click on your London boroughs layer \> Properties \> Symbology
2.  Select categorised and choose a data column and color ramp \> click Classify

We've selected categorised (think categories) as our flytipping data column types are most likely text (otherwise know as string). You can check by opening the attribute table and hovering over the column name...

```{r echo=FALSE, out.width = "400pt", fig.align='center', cache=FALSE}
knitr::include_graphics('prac1_images/QGIS_column_string.png') 
```

Alternatively we can covert the column type to numeric and selected graduated (for numeric columns). We will convert between text and numeric columns in future weeks. 

1.  Save your QGIS project

You should have produced something like this:

```{r echo=FALSE, out.width = "800pt", fig.align='center', cache=FALSE}
knitr::include_graphics('prac1_images/QGIS_final.png') 
```

::: {.infobox .tip data-latex="{note}"}
This a very basic map and we've just used raw counts to show how to join and map data. You should (almost) never do this, as counts can be misleading. This is because the counts are collected over areas of unequal size. It's best to divide the values by some sort of denominator (e.g. area of borough, population) to normalise the data giving something like occurrence of flytipping per km2 enabling comparison between the boroughs. We cover this in more detail before reading week.
:::

**We haven't talked about the Coordiante Reference System (CRS) (or Spatial Reference System (SRS)) of our map document**

A coordinate reference system is a series of parameters that define the coordinate system. Within GIS we use geographic or projected coordinate systems. The former uses a three-dimensional spherical surface to define locations of Earth, whereas the latter is defined on a flat, two-dimensional surface giving it constant lengths, angles and areas. We cover this in much more detail in subsequent weeks.

QGIS defaults to the Coordinate Reference System (CRS) WGS 84, or known by its European Petroleum Survey Group (EPSG) code 4326. However, when you add your first layer it will default the CRS of the layer. We'll go into the background of EPSG next time.

You can change the CRS by going Project \> Properties and selecting CRS in the left hand pane.

```{r echo=FALSE, out.width = "800pt", fig.align='center', cache=FALSE}
knitr::include_graphics('prac1_images/QGIS_CRS.png') 
```

::: {.infobox .tip data-latex="{note}"}
You can also access the project CRS by just clicking on the EPSG code in the bottom right of your QGIS window, in the screenshot above click EPSG:27700.
:::

::: {.infobox .warning data-latex="{note}"}
**Note** As we are going to open the same files in difference GIS systems it is important to close the software before moving on. If you don't then files can be *locked* and unreadable as they are still considered to be in use.
:::

### R

#### Introduction

R is both a programming language and software environment, originally designed for statistical computing and graphics. R's great strength is that it is open-source, can be used on any computer operating system and free for anyone to use and contribute to. Because of this, it is rapidly becoming the statistical language of choice for many academics and has a huge user community with people constantly contributing new packages to carry out all manner of statistical, graphical and importantly for us, geographical tasks.

The purpose of this practical is just to demonstrate data loading and manipulation in different software. The next practical will provide much more detail on R, so don't worry about that.

In RStudio go:

1.  File \> New File \> R Script

You should be able to see these quadrants (without the code):

```{r echo=FALSE, out.width = "800pt", fig.align='center', cache=FALSE}
knitr::include_graphics('prac1_images/RStudio.png') 
```

Below are bits of code, to start we will work using the **console**. So just copy the bits of code into the console window, changing the file names to where your data is stored. Then at the end of this section I'll show you how to make a script.

R works on packages that are collections of functions and data. Packages are bits of code that extend R beyond the basic statistical functionality it was originally designed for. For spatial data, they enable R to process spatial data formats, carry out analysis tasks and create some of the maps that follow. Most packages are easily recognisable through a hexagon (hex) logo, or sticker. 

```{r echo=FALSE, out.width = "500px", fig.align='center', cache=FALSE, fig.cap="R package hex sticker wall. Source: [Mitchell O'Hara-Wild arranging Hex stickers in R](https://www.mitchelloharawild.com/blog/hexwall/)"}
knitr::include_graphics('prac1_images/hex_stickers.png') 
```

Bascially, without packages, R would be very limited. With packages, you can do virtually anything! One of the issues you will come across is that packages are being continually developed and updated and unless you keep your version of R updated and your packages updated, there may be some functions and options not available to you. This can be a problem, particularly with University installations which (at best) may only get updated once a year. Therefore, apologies in advance if things don't work as intended, but there is often a work around.

For this practical we will need the ones listed in the code chunk below. Whilst we've installed them (with the code below), we haven't yet loaded them. It's best practice to do all this at the start of your code, however, for demonstration purposes I'll load each one as we need it.

```{r, message=FALSE, warning=FALSE, eval=FALSE, cache=FALSE}
install.packages(c("sf", "tmap", "tmaptools", "RSQLite", "tidyverse"), 
                 repos = "https://www.stats.bris.ac.uk/R/")
```

1.  We can also install and load packages through RStudio. If you can see the package listed in the package tab it is installed and a tick means it is loaded.

```{r echo=FALSE, out.width = "500px", fig.align='center', cache=FALSE}
knitr::include_graphics('prac2_images/r_packages.png') 
```

Packages we've installed with our code:

-   `sf`: simple features, standard way to encode spatial vector data
-   `tmap`: layer-based and easy approach to make thematic maps
-   `tmaptools`: set of tools for reading and processing spatial data
-   `RSQLite`: embeds the SQLite database engine in R

Here, `repos` stands for repository where we will download the packages from. Once you have set it, you shouldn't need to specify it again. Whilst we've installed the packages we haven't loaded them...this is done through using `library()` function.

```{r echo=FALSE, out.width = "600pt", fig.align='center', cache=FALSE, fig.cap="Installing vs loading (with library) packages. Source: [Dianne Cook](https://twitter.com/visnut)"}
knitr::include_graphics('prac1_images/install_library.jfif') 
```

#### Load data

Load the `sf` package so we can read our shapefile in --- remember to change to filepath to your shapefile.

```{r echo=FALSE, out.width = "450pt", fig.align='center', cache=FALSE, fig.cap="sf introduction graphic. Source: [Allison Horst data science and stats illustrations](https://github.com/allisonhorst/stats-illustrations)"}
knitr::include_graphics('allisonhorst_images/sf.png', )
```

For this first practical we'll just leave the data where it is, next week we'll show you how to use projects to make calling data much easier.

> **Note** by default in R, the file path should be defined with `/` but on a windows file system it is defined with `\` (e.g. when looking at files on your computer they will be folder\file). Using `\\` instead allows R to read the path correctly -- alternatively, just use `/`

```{r, message=FALSE, warning=FALSE, cache=FALSE}
library(sf)
# change this to your file path!!!
shape <- st_read("C:/Users/Andy/OneDrive - University College London/Teaching/CASA0005/2020_2021/CASA0005repo/prac1_data/statistical-gis-boundaries-london/ESRI/London_Borough_Excluding_MHW.shp")
```

> **Note** If you are working on the UCL RStudio Server and have a shapefile (along with all the other relevant files required) you will just need to use...

```{r, message=FALSE, warning=FALSE, cache=FALSE, eval=FALSE}
shape <- st_read("London_Borough_Excluding_MHW.shp")
```

As RStudio cloud knows you have setup a project and to look for files within it...we cover this very soon for RStudio desktop.

To get a summary of the data held within the shapefile data (attribute table) enter the following:

```{r, cache=FALSE}
summary(shape)
```

To have a quick look what the shapefile looks like enter the following:

```{r, cache=FALSE}
plot(shape)
```

That plots everything in the shapefile (all the attributes) if you just wanted the geometry (outline of the shape) you could use...

```{r, message=FALSE}
library(sf)
shape %>% 
  st_geometry() %>%
  plot()
```

From what we did in QGIS this should look familiar.

We now need to load our `.csv` file:

```{r message=TRUE, warning=FALSE, cache=FALSE}
library(tidyverse)
#this needs to be your file path again
mycsv <-  read_csv("C:/Users/Andy/OneDrive - University College London/Teaching/CASA0005/2020_2021/CASA0005repo/prac1_data/fly_tipping_borough_edit.csv")  
```

When you load `tidyverse` you might get a list of conflicts, this basically means that several packages have functions named the same thing. For example there is a function named `filter()` in the packages `dplyr` and `stats`. If you wanted to use the `dplyr` version you would use `dplyr::filter()` or for the `stats` version it would be `stats::filter()`...

::: {.infobox .warning data-latex="{note}"}
**Note** Your `.csv` might have a extra header present (a row before Row Labels). You can either: (a) stop and go and delete this in the `.csv` now and reload it or (b) change the code to... `mycsv <- read_csv("prac1_data/fly_tipping_borough_edit.csv",      skip = 1)`
:::

To view the data just input:

```{r cache=FALSE, warnings=FALSE, messages=FALSE}
mycsv 
```

#### Join data

In R we've given our London boroughs shapefile the name `shape` and our flytipping `.csv` the name `mycsv`. If you look in the Environment quadrant you should see them both listed.

Join the `.csv` to the shapefile. Here, replace Row Labels with whatever your GSS_CODE is called in the `.csv`:

```{r, cache=FALSE}
shape <- shape%>%
  merge(.,
        mycsv,
        by.x="GSS_CODE", 
        by.y="Row Labels")

```

Let's break this down a bit. We just created a tibble of `mycsv` (this is a new form of a dataframe) where each column has a variable and each row contains a set of values --- so basically a normal table. We did this simply by using `read_csv()` to read the data in. We then matched our `csv` to our `shape` based on the `GSS_CODE` values in both.

Check the merge was successful, this is just going to show the top 10 rows:

```{r cache=FALSE}
shape%>%
  head(., n=10)
```

Now, let's make a quick thematic map (or a qtm) using the package `tmap`. I've made mine for flytipping between 2011 and 2012 (column 2011_12). But check what your column name is from the above code `head(shape, n=10)` it might be slightly different like 2011-2012 or x2011_2012....

```{r echo=FALSE, out.width = "600pt", fig.align='center', cache=FALSE}
knitr::include_graphics('prac1_images/colname.png') 
```

```{r cache=FALSE, message=FALSE}
library(tmap)
tmap_mode("plot")
# change the fill to your column name if different
shape %>%
  qtm(.,fill = "2011_12")
```

### Export data

Finally write shape to a new GeoPackage (`.gpkg`) giving it the layer name of your choice:

```{r cache=FALSE, message=FALSE}
shape %>%
  st_write(.,"C:/Users/Andy/OneDrive - University College London/Teaching/CASA0005/2020_2021/CASA0005repo/prac1_data/Rwk1.gpkg",
           "london_boroughs_fly_tipping",
           delete_layer=TRUE)
```

So here, we are saying the shape is the object we want to save, then to the GeoPackage file path, with the layer name of `london_boroughs_fly_tipping`. I've set `delete_layer` to true so I could overwrite mine when I developed this practical. Changing it to false would generate an error message if you ever tried to re-run the code.

Let's also add the `.csv` as we did in QGIS. This is a bit more complicated as we have to use the SQLite database package. Firstly, connect to the `.gpkg` we just made:

```{r cahce=TRUE}
library(readr)
library(RSQLite)

con <- dbConnect(RSQLite::SQLite(),dbname="C:/Users/Andy/OneDrive - University College London/Teaching/CASA0005/2020_2021/CASA0005repo/prac1_data/Rwk1.gpkg")
```

Now examine what is in the `.gpkg`...you can see that i've already got my `original_csv` stored within the `.gpkg` as when i developed this practical i made sure it was working!

```{r cache=FALSE}
con %>%
  dbListTables()
```

Add your `.csv` and disconnect from the `.gpkg`:

```{r cache=FALSE}
con %>%
  dbWriteTable(.,
               "original_csv",
               mycsv,
               overwrite=TRUE)

con %>% 
  dbDisconnect()
```

Here `overwrite` let's you well..overwrite the exisiting file...if this wasn't specified as true you would get an error saying the file already existed if you tried to run this code again.

#### Making a script

To convert this bit of analysis into a script that we could save and run again in future, I would write the following in the script quadrant:

```{r, eval=FALSE, cache=FALSE}
library(sf)
library(tmap) 
library(tmaptools)
library(RSQLite)
library(tidyverse)
#read in the shapefile

shape <- st_read(
  "C:/Users/Andy/OneDrive - University College London/Teaching/CASA0005/2020_2021/CASA0005repo/prac1_data/statistical-gis-boundaries-london/ESRI/London_Borough_Excluding_MHW.shp")
# read in the csv
mycsv <- read_csv("C:/Users/Andy/OneDrive - University College London/Teaching/CASA0005/2020_2021/CASA0005repo/prac1_data/fly_tipping_borough_edit.csv")  
# merge csv and shapefile
shape <- shape%>%
  merge(.,
        mycsv,
        by.x="GSS_CODE", 
        by.y="Row Labels")
# set tmap to plot
tmap_mode("plot")
# have a look at the map
qtm(shape, fill = "2011_12")
# write to a .gpkg
shape %>%
  st_write(.,"C:/Users/Andy/OneDrive - University College London/Teaching/CASA0005/2020_2021/CASA0005repo/prac1_data/Rwk1.gpkg",
           "london_boroughs_fly_tipping",
           delete_layer=TRUE)
# connect to the .gpkg
con <- dbConnect(SQLite(),dbname="C:/Users/Andy/OneDrive - University College London/Teaching/CASA0005/2020_2021/CASA0005repo/prac1_data/Rwk1.gpkg")
# list what is in it
con %>%
  dbListTables()
# add the original .csv
con %>%
  dbWriteTable(.,
               "original_csv",
               mycsv,
               overwrite=TRUE)
# disconnect from it
con %>% 
  dbDisconnect()
```

You can then save your script through File \> Save As.


### What will I use

Well... it depends. If you wanted to quickly open a dataset to explore its contents then I'd use QGIS. However, if you had 100 raster images that you wanted to clip to your study area, I'd automate it in R. There are also specific packages developed for each type of software that might dictate what you use, for example I recently made use of the Urban Multi-scale Environmental Predictor (UMEP) plugin in QGIS. That said, as I needed to match different hourly meteorological variables over a three year period I automated the first part of the analysis in R and loaded a `.csv` into QGIS. Recent advancments have also closed the gap somewhat between Graphic User Interface (GUI) GIS software (such as QGIS) and programming languages (such as R) with 'bridges' that allow you to control GUI software through code. To learn more about these bridges read [Lovelace et al. (2019) chapter 9](https://geocompr.robinlovelace.net/gis.html#introduction-6).

## Data sources and task

Below I've listed a few good data sources. Explore these and any others you can find and get an interesting dataset (e.g. in this practical our flytipping `.csv`) that you could join to some spatial data (e.g. in this practical the London boroughs `.shp`). This could be for any location in the world.

### UK Data Service

The UK Data Service geography service (<https://census.edina.ac.uk/>) has a library of hundreds of current and former boundary datasets for which attribute data are produced in the UK.

### ONS

The Office for National Statistics (ONS) are the national statistical agency for England and Wales and have recently started to provide access to boundary data for the statistics they produce for various geographic areas.

Many of the boundaries on the [ONS Geoportal](http://geoportal.statistics.gov.uk/) are also available from the Edina Census Geography website in a more flexible fashion, however the ONS website provides very quick access to bulk-downloads --- something which can be very useful when reading data directly from the web using computer software. From the ONS website you can also extract the URL at which the data is stored to use directly within your future code...

### nomis

Nomis is provided by the ONS giving free access to UK labour market statistics. You can also [bulk download census data](https://www.nomisweb.co.uk/census/2011/bulk/r2_2) from the site too!

### OS

The Ordnance Survey (OS) are the national mapping agency for the UK. A few years ago, they opened up a number of their data products for public use including greenspace, OS Open Map and OS Terrain.

For the full range see: [https://www.ordnancesurvey.co.uk/business-and-government/products/finder.html?Licensed%20for=OpenData%20(Free)&withdrawn=on](https://www.ordnancesurvey.co.uk/business-and-government/products/finder.html?Licensed%20for=OpenData%20(Free)&withdrawn=on){.uri}

### Edina Digimap

Before the Ordnance Survey opened up much of its data for public use, academics and students in the UK could access OS data using the Edina Digimap Service ---- this service is still available today and provides access to a number of products in addition to those available from OS Open Data.

Perhaps the most exciting of the additional OS data products available from Digimap is OS MasterMap. MasterMap is a framework for all OS data and contains layers of data that include details of real world objects such as buildings, roads, paths, rivers, physical structures and land parcels, as well as the complete UK transport network. Whilst we still are required to go through Edina OS have recently announced plans to make this dataset free in the near future under the new Geospatial Commission.

```{r echo=FALSE, out.width = "500pt", fig.align='center'}
knitr::include_graphics('prac1_images/mastermap.png') 
```

### OSM

Open Street Map (OSM) is a fantastic resource ---- as the name suggests, all data contained in Open Street Map are open and free for anyone to use. Much like Wikipedia, anyone can contribute content to OSM and this brings with it its own benefits (frequent updates, very large user-base) and problems (data quality and patch coverage). OSM is a very good example of Volunteered Geographic Information (VGI).

It's possible to download OSM data straight from the website, although the interface can be a little unreliable (it works better for small areas). There are, however, a number of websites that allow OSM data to be downloaded more easily and are directly linked to from the 'Export' option in OSM. Geofabrik (<https://www.geofabrik.de/data/download.html>) allows you to download frequently updated Shapefiles for various global subdivisions.

### DEFRA

The Department for Environment and Rural Affairs (DEFRA) have recently created the Data Services Platform to openly distribute environmental data. See: <https://environment.data.gov.uk/>

### Data lists

Another good place to start searching for data are data lists. They simply provide a comprehensive overview of all available data conveniently categorised by discipline and country.

I normally use this one: <https://freegisdata.rtwilson.com/>

## Summary

Within this practical we have explored the different types, formats and software used to store, analyse and manipulate spatial data. In reflecting upon this practical you should consider the (dis)advantages of each, where and when they might be appropriate and the overall practicality. Next week we will delve further into R and RStudio.

## Feedback

Was anything that we explained unclear this week or was something really clear...let us know using the [feedback form](https://forms.gle/w2GUDYc7tSavGy7r6). It's anonymous and we'll use the responses to clear any issues up in the future / adapt the material.