Using remotely sensed data for air pollution assessment

TL;DR

This work tackles the challenge of estimating surface air pollutant concentrations in locations lacking observations by leveraging remote-sensing data within a Random Forest framework. Using Sentinel-5P products, ERA5 meteorology, and Corine Land Cover for the Iberian Peninsula in 2019, the authors train pollutant-specific RF models and evaluate them with three cross-validation schemes to assess spatial and temporal generalization. NO$_2$ and $O_3$ concentrations are predicted with relatively strong performance, while $SO_2$, $PM_{10}$, and $PM_{2.5}$ exhibit weaker predictive power, highlighting data limitations and the need for richer features (e.g., tropospheric measurements, higher-resolution meteorology). The study demonstrates the potential to generate new, gridded air-quality data sets and outlines concrete directions for improving surface-concentration inference using remote sensing.

Abstract

Air pollution constitutes a global problem of paramount importance that affects not only human health, but also the environment. The existence of spatial and temporal data regarding the concentrations of pollutants is crucial for performing air pollution studies and monitor emissions. However, although observation data presents great temporal coverage, the number of stations is very limited and they are usually built in more populated areas. The main objective of this work is to create models capable of inferring pollutant concentrations in locations where no observation data exists. A machine learning model, more specifically the random forest model, was developed for predicting concentrations in the Iberian Peninsula in 2019 for five selected pollutants: $NO_2$, $O_3$ $SO_2$, $PM10$, and $PM2.5$. Model features include satellite measurements, meteorological variables, land use classification, temporal variables (month, day of year), and spatial variables (latitude, longitude, altitude). The models were evaluated using various methods, including station 10-fold cross-validation, in which in each fold observations from 10\% of the stations are used as testing data and the rest as training data. The $R^2$, RMSE and mean bias were determined for each model. The $NO_2$ and $O_3$ models presented good values of $R^2$, 0.5524 and 0.7462, respectively. However, the $SO_2$, $PM10$, and $PM2.5$ models performed very poorly in this regard, with $R^2$ values of -0.0231, 0.3722, and 0.3303, respectively. All models slightly overestimated the ground concentrations, except the $O_3$ model. All models presented acceptable cross-validation RMSE, except the $O_3$ and $PM10$ models where the mean value was a little higher (12.5934 $μg/m^3$ and 10.4737 $μg/m^3$, respectively).

Figures (6)

• Figure 1: Mean squared error graph for each "max_features" value and different number of estimators, for the nitrogen dioxide 2019 dataset.
• Figure 2: Execution time graph for each "max_features" value and different number of estimators, for the nitrogen dioxide 2019 dataset.
• Figure 3: Annual mean concentration of nitrogen dioxide (in $\mu g/m^3$) in 2019 in the Iberian Peninsula, calculated from model predictions.
• Figure 4: Box plot with temporal variation of nitrogen dioxide concentration (in $\mu g/m^3$) in 2019 in the Iberian Peninsula, obtained from model predictions. Each number corresponds to the respective month number in the year, that is January is 1 and December is 12.
• Figure 5: Annual mean concentration of ozone (in $\mu g/m^3$) in 2019 in the Iberian Peninsula, calculated from model predictions.