Exploratory analysis of crowdsourced acoustic open data

This short analysis explore the data collected by the Noisecapture Android application between 2017 and 2020.

Several exploratory analysis has been done, focusing on the tracks recorded in France.

These preliminary works are part of the research carried out within the framework of the LASSO project led by the UMRAE laboratory (Univ. Gustave Eiffel/CEREMA)

A presentation of the analysis at FOSDEM 2023 is available here (slides).

Data source

The raw data are available here :

https://research-data.ifsttar.fr/dataset.xhtml?persistentId=doi:10.25578/J5DG3W

How to reproduce

Build the database

Database configuration

• Ubuntu 18.04 or higher
• PostgreSQL 10.15 or higher (14.0 is recommended)
• Postgis 2.5 or higher

Steps

• Create an empty database named noisecapture with the PostGIS extension
• Copy in your home folder the SQL script 01_drop_foreign_keys.sql if your available storage is less than 200 Gb
• Execute the script 00_prepare_database.sh, comment the second line if you want to keep foreign keys
• Execute the SQL script 02_load_country_data.sql to load additional data from NaturalEarth used by the analysis
• Execute the SQL script 03_create_views.sql to compute the views that prepare the data used in the analysis.

Get the source code

As the analysis part of project as been treated as R package, there is several ways to get the code source:

• using git

git clone https://github.com/ifsttar/lasso-data-analysis

• using R and the remotes package:

# You can clone 

# We suggest to use the remotes packages to install required packages
# install.packages("remotes")
remotes::install_github("ifsttar/lasso-data-analysis")

• download as a zip archive

Setting up R

This analysis use several packages that you'll need to install beforehand.

# Package list
pkgs <- c("RPostgreSQL",
          "DBI",
          "sf",
          "dplyr",
          "purrr",
          "ggplot2",
          "scales",
          "lubridate",
          "hydroTSM",
          "suncalc",
          "xfun")

# Packages installation from CRAN
# Already installed packages won't be reinstalled
remotes::install_cran(pkgs)

Set connection parameters to the database

Please be sure to adapt the connection parameters to your database. Those parameters are presented as an example, the database is not available online.

drv <- DBI::dbDriver("PostgreSQL")

con <- DBI::dbConnect(
drv,
dbname ="noisecapture",
host = "noisecaptureDB", #server IP or hostname
port = 5432, #Port on which we ran the proxy
user="noisecapture",
password=Sys.getenv('noisecapture_password') # password stored in .Renviron. Use this to edit it : usethis::edit_r_environ()
)

Render analysis

In order to facilitate reproductibility, the analysis have been set in several vignettes that are stored in the vignettes folder. Each document is autonomous and can be executed independently (except for the Main_doc.Rmd document which executes others).

The crowdsourced_acoustic_data_analysis_with_foss4g_2022.Rmd vignette is the source code to the published article.

The temporal_exploratory_analysis.Rmd vignette corresponds to the first raw analysis.

The Main_doc.Rmd vignettes calls and execute every [Computing] and [Analysis].

Those whose name begins with [Analysis] contain the analytical part. They are based on pre-processed data that can be either downloaded from Zenodo or generated using the documents whose name starts with [Computing]. The [Computing] documents must have a functional connection to the database containing the noisecapture data and the corresponding views.