You can view this vignette any time by running:

spanishoddata::spod_codebook(ver = 2)

The mobility data v2 (2022 onwards) was originally released by the Ministerio de Transportes, Movilidad y Agenda Urbana (MITMA) , now Ministerio de Transportes y Movilidad Sostenible (MITMS) (Secretaría de Estado de Transportes y Movilidad Sostenible 2024).

The dataset is produced by Nommon using the raw data from Orange España. Even though the raw data is only from one mobile phone operator, the resulting flows and other counts of number of individuals in the data set are already resampled to be representative of the total population of Spain (see details in the official methodology).

The tables in the data set provide hourly flows between zones across Spain for every day of the observation period (2022-01-01 onwards), the number of individuals making trips for each zone, the number of individuals spending the nights in one location while regularly residing in a different location, and many more advanced datasets. This document will introduce you to the available data and provide brief code snippets on how to access it using the {spanishoddata} R package.

Compared to the v1 data (2020-2021) this dataset has many additional variables, such as age, sex, and income, has better spatial resolution (the zones are more spatially granular) and covers a continuous period¹ (2022-01-01 onward), rather than only a limited period (in v1 - 2020-02-14 to 2021-05-09).

In v2 data, the datasets comparable to the v1 (2020-2021) data (hourly daily origin destination matrices and number of trips at each location) are called “basic studies” (“estudios basicos”), however there are even more advanced datasets that are already available or will be made available soon: “complete studies” (“estudios completos”) and “road routes” (“rutas de carretera”). At the moment, the {spanishoddata} package only provides the interface to the “basic studies” datasets, but support for “complete studies” and “road routes” will be added in the future.

Key sources for this codebook/cookbook include:

• original data collection methodology and codebook in Spanish + automatically translated English version of methodology and codebook

• original data download page

• original data license

• homepage of the v2 study and open mobility data project of the Ministry of Transport and Sustainable Mobility of Spain

To access the data we reference in this codebook, please follow these steps:

Install the package

The package is not yet available on CRAN.

You can install the latest version of the package from rOpenSpain R universe:

install.packages("spanishoddata",
  repos = c("https://ropenspain.r-universe.dev",
    "https://cloud.r-project.org"))

if (!require("remotes")) install.packages("remotes")

remotes::install_github("rOpenSpain/spanishoddata",
  force = TRUE, dependencies = TRUE)

gh repo clone rOpenSpain/spanishoddata
code spanishoddata
# with rstudio:
rstudio spanishoddata/spanishoddata.Rproj

devtools::load_all()

Load it as follows:

library(spanishoddata)

Using the instructions below, set the data folder for the package to download the files into. You may need up to 400 GB to download all data and another 400 GB if you would like to convert the downloaded data into analysis ready format (a DuckDB database file, or a folder of parquet files). You can find more info on this conversion in the Download and convert OD datasets vignette.

Set the data directory

Choose where {spanishoddata} should download (and convert) the data by setting the data directory following command:

spod_set_data_dir(data_dir = "~/spanish_od_data")

The function above will also ensure that the directory is created and that you have sufficient permissions to write to it.

Sys.setenv(SPANISH_OD_DATA_DIR = "~/spanish_od_data")

usethis::edit_r_environ()
# Then set the data directory globally, by typing this line in the file:

SPANISH_OD_DATA_DIR = "~/spanish_od_data"

file.edit(".Renviron")

Overall approach to accessing the data

If you only want to analyse the data for a few days, you can use the spod_get() function. It will download the raw data in CSV format and let you analyse it in-memory. This is what we cover in the steps on this page.

If you need longer periods (several months or years), you should use the spod_convert() and spod_connect() functions, which will convert the data into special format which is much faster for analysis, for this see the Download and convert OD datasets vignette. spod_get_zones() will give you spatial data with zones that can be matched with the origin-destination flows from the functions above using zones ’id’s. Please see a simple example below, and also consult the vignettes with detailed data description and instructions in the package vignettes with spod_codebook(ver = 1) and spod_codebook(ver = 2), or simply visit the package website at https://ropenspain.github.io/spanishoddata/. The Figure 1 presents the overall approach to accessing the data in the spanishoddata package.

1. Spatial data with zoning boundaries

The boundary data is provided at three geographic levels: Distrtics, Municipalities, and Large Urban Areas. It’s important to note that these do not always align with the official Spanish census districts and municipalities. To comply with data protection regulations, certain aggregations had to be made to districts and municipalities”.

1.1 Districts

Districts correspond to official census districts in cities; however, in those with lower population density, they are grouped together. In rural areas, one district is often equal to a municipality, but municipalities with low population are combined into larger units to preserve privacy of individuals in the dataset. Therefore, there are 3792 ‘districts’ compared to the 10494 official census districts on which they are based. There are also NUTS3 statistical regions covering France (94 units) and Portugal (23 units). Therefore there is a total of 3909 zones in the Districts dataset.

districts_v2 <- spod_get_zones("dist", ver = 2)

The districts_v2 object is of class sf consisting of polygons.

Data structure:

1.2 Municipalities

Municipalities are made up of official municipalities in those of a certain size; however, they have also been aggregated in cases of lower population density. As a result, there are 2618 municipalities compared to the 8,125 official municipalities on which they are based. There are also NUTS3 statistical regions covering France (94 units) and Portugal (23 units). Therefore there is a total of 2735 zones in the Districts dataset.

municipalities_v2 <- spod_get_zones("muni", ver = 2)

The resulting municipalities_v2 object is type sf consisting of polygons.

Data structure:

1.3 LUAs (Large Urban Areas)

Large Urban Areas (LUAs) has essentially the same spatial units as Municipalities, but are not aggregated. Therefore, there are 2086 locations in the LUAs dataset. There are also NUTS3 statistical regions covering France (94 units) and Portugal (23 units). Therefore there is a total of 2203 zones in the LUAs dataset.

luas_v2 <- spod_get_zones("lua", ver = 2)

The resulting luas_v2 object is type sf consisting of polygons.

Data structure:

2. Mobility data

All mobility data is referenced via id_origin, id_destination, or other location identifiers (mostly labelled as id) with the two sets of zones described above.

2.1. Origin-destination data

The origin-destination data contain the number of trips between districts, municipalities, or large urban areas (LUAs) in Spain for every hour of every day between 2022-02-01 and whichever currently available latest data (2024-06-30 at the time of writing). Each flow also has attributes such as the trip purpose (composed of the type of activity (home/work_or_study/frequent_activity/infrequent_activity) at both the origin and destination, but also age, sex, and income of each group of individuals traveling between the origin and destination), province of residence of individuals making this trip, distance covered while making the trip. See the detailed attributes below in a table.

Here are the variables you can find in the district, municipality and large urban area level data:

Getting the data

To access the data, use the spod_get() function. In this example we will use a short interval of dates:

dates <- c(start = "2022-01-01", end = "2022-01-04")
od_dist <- spod_get(type = "od", zones = "dist", dates = dates)
od_muni <- spod_get(type = "od", zones = "muni", dates = dates)

The data for the specified dates will be automatically downloaded and cached in the SPANISH_OD_DATA_DIR directory. Existing files will not be re-downloaded.

Working with the data

The resulting objects od_dist and od_muni are of class tbl_duckdb_connection⁵. Basically, you can treat these as regular data.frames or tibbles. One important difference is that the data is not actually loaded into memory, because if you requested more dates, e.g. a whole month or a year, all that data would most likely not fit into your computer’s memory. A tbl_duckdb_connection is mapped to the downloaded CSV files that are cached on disk and the data is only loaded in small chunks as needed at the time of computation. You can manipulate od_dist and od_muni using {dplyr} functions such as select(), filter(), mutate(), group_by(), summarise(), etc. In the end of any sequence of commands you will need to add collect() to execute the whole chain of data manipulations and load the results into memory in an R data.frame/tibble like so:

library(dplyr)
od_mean_trips_by_ses_over_the_4_days <- od_dist |>
  group_by(date, age, sex, income) |>
  summarise(
    n_trips = sum(n_trips, na.rm = TRUE),
    .groups = "drop") |> 
  group_by(age, sex, income) |>
  summarise(
    daily_mean_n_trips = mean(n_trips, na.rm = TRUE),
    .groups = "drop") |> 
  collect()
od_mean_trips_by_ses_over_the_4_days

# A tibble: 39 × 4
   age   sex    income daily_mean_n_trips
   <fct> <fct>  <fct>               <dbl>
 1 NA    NA     <10              7002485.
 2 NA    NA     10-15           16551405.
 3 NA    NA     >15              2651481.
 4 0-25  NA     <10               539060.
 5 0-25  NA     10-15            1950892.
 6 0-25  NA     >15               401557.
 7 0-25  female <10              1484989.
 8 0-25  female 10-15            5357785.
 9 0-25  female >15              1764454.
10 0-25  male   <10              1558461.
# ℹ 29 more rows
# ℹ Use `print(n = ...)` to see more rows

In this example above, becaus the data is with hourly intervals within each day, we first summed the number of trips for each day by age, sex, and income groups. We then grouped the data again dropping the day variable and calculated the mean number of trips per day by age, sex, and income groups. The full data for all 4 days was probably never loaded into memory all at once. Rather the available memory of the computer was used up to its maximum limit to make that calculation happen, without ever exceeding the available memory limit. If you were doing the same opearation on 100 or even more days, it would work in the same way and would be possible even with limited memory. This is done transparantly to the user with the help of DuckDB (specifically, with {duckdb} R package Mühleisen and Raasveldt (2024)).

The same summary operation as provided in the example above can be done with the entire dataset for multiple years worth of data on a regular laptop with 8-16 GB memory. It will take a bit of time to complete, but it will be done. To speed things up, please also see the vignette on converting the data into formats that will increase the analsysis performance.

2.2. Number of trips data

For each location, the “number of trips” data provides the number of individuals who spent the night there, with breakdown by the number of trips made, age, and sex.

Getting the data

To access it use spod_get() with type set to “number_of_trips”, or just “nt”.

dates <- c(start = "2022-01-01", end = "2022-01-04")
nt_dist <- spod_get(type = "number_of_trips", zones = "dist", dates = dates)

Because this data is small, we can actually load it completely into memory:

nt_dist_tbl <- nt_dist |> dplyr::collect()

2.3. Overnight stays

This dataset provides the number of people who spend the night in each location, also identifying their place of residence down to the census district level according to the INE encoding.

Here are the variables you can find in the district, municipality and large urban area level data:

Getting the data

To access it use spod_get() with type set to “number_of_trips”, or just “nt”.

dates <- c(start = "2022-01-01", end = "2022-01-04")
os_dist <- spod_get(type = "overnight_stays", zones = "dist", dates = dates)

Because this data is small, we can actually load it completely into memory:

os_dist_tbl <- os_dist |> dplyr::collect()