Air pollution measurement errors: Is your data fit for purpose?

This repository contains Python and R code for reproducing the diagnostic plots shown in Air pollution measurement errors: Is your data fit for purpose? (Diez et al, 2022) published in Atmospheric Measurement Techniques (https://doi.org/10.5194/amt-15-4091-2022).

Python package

To install the Python package, run pip install git+https://github.com/wacl-york/quant-measurement-errors-tools#subdirectory=quantpy.

The quantpy.plots module contains functions to reproduce all 4 of the main plots in the functions: time_series, scatter, bland_altman, reu_plot. These can be used as follows:

import quantpy.plots as plots
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt

# Creating a simple dataframe with random reference data
times = pd.date_range('2021-10-01', periods = 1000, freq ='60min')
df = pd.DataFrame(np.random.lognormal(mean = 3, sigma = 0.4, size = 1000),
                  columns = ['NO2'], index = times)
# Simulate a LCS with noise and bias
df['LCS1'] = (df['NO2'] + np.random.normal(0,3,len(df.index)).tolist())*1.2

plots.bland_altman(df, 'NO2', 'LCS1')

plots.scatter(df, 'NO2', 'LCS1')

plots.reu_plot(df, "NO2", "LCS1", DQO = 25, ylim=[0, 200])

plots.time_series(df, 'NO2', 'LCS1')

See the accompanying documentation for further support (i.e. help(plots.bland_altman)). Furthermore, the quantpy.reu module contains the function reu to calculate the REU.

R package

To install the R code, run devtools::install_github("wacl-york/quant-measurement-errors-tools/quantr") to install the quantr package. This package contains functions to reproduce all 4 of the main plots in the functions: plot_time_series, plot_scatter, plot_bland_altman, plot_reu.

These can be used as follows:

library(quantr)

# Create dummy data frame
n_vals <- 100
df <- data.frame(timestamp=seq.POSIXt(from=as.POSIXct("2020-09-24 09:00:00"),
                                      to=as.POSIXct("2020-09-30 09:00:00"),
                                      length.out=n_vals),
                 ref=rnorm(n_vals, 30, 5),
                 lowcost=rnorm(n_vals, 32, 8))

plot_reu(df, lcs_column="lowcost", reference_column="ref", y_limits=NULL)

plot_scatter(df, lcs_column="lowcost", reference_column="ref")

plot_bland_altman(df, lcs_column="lowcost", reference_column="ref")

plot_time_series(df, lcs_column="lowcost", reference_column="ref")

See the accompanying help pages (e.g. ?plot_reu) for further details.

Real world data

The real world air quality measurements that were used to generate some of the figures in the paper are available in the data sub-folder. See the README in that directory for details on the data structure.