Comparison of two methods for evaluating the performance of particle filtration systems in underground railway stations
Abstract
Filtration is one of the strategies that can be implemented to mitigate the public health challenge due to elevated concentrations of particles in underground railway stations (URS). During a particle (PM 2.5 and PM 10 ) monitoring campaign performed in a Paris URS over more than 5 years, two filtration systems were tested. One of these systems employs positive ionization (Experiment 1), the second one is based on water filtration (Experiment 2). The present study focuses on evaluating the performance of these experiments, using two distinct methods: the daily amplitude coefficient (DAC), and a generalized additive model (GAM). The DAC method requires only particulate matter measurements, while the GAM method necessitates additional measurements (indoor CO 2 concentration, temperature and humidity, outdoor pollutant concentration) as it analyses the nonlinear relationships between all these factors. The results show that both filtration technologies are more effective at reducing PM 2.5 concentrations than PM 10 , but Experiment 2 was less efficient than Experiment 1. Specifically, DAC analysis leads to a reduction of 17.8 ± 3.2 % for PM 10 and 25.9 ± 6.6 % for PM 2.5 for Experiment 1, compared to only 0.4 ± 0.1 % for PM 10 and 3.7 ± 1.0 % for PM 2.5 for Experiment 2. GAM analysis gives similar results, with a reduction of 13.1 ± 0.8 % for PM 10 and 26.0 ± 0.8 % for PM 2.5 for Experiment 1, and for Experiment 2 a reduction of 1.3 ± 2.0 % for PM 10 and 4.6 ± 1.3 % for PM 2.5 . Yet, because of differences in the filtration setups, such as the position of the modules and the maximal air flow capacity, and of unspecified internal characteristics of the systems, no conclusion can be drawn as to whether one of the filtration techniques is more efficient than the other. The two analysis methodologies lead to similar results and allow to fully and rigorously exploit long time series. The DAC method can be applied using only PM measurements, while the GAM method requires additional measurements, but at the same time provides valuable insights about the effects of each parameter considered. • PM was continuously monitored in an underground railway station in Paris over a 5-year period. • Two separate filtration experiments were conducted during the measurement period. • For each filtration system, similar filtration performances were assessed using both the DAC and GAM approaches. • DAC relies only on PM time series, while GAM uses additional parameters to identify and quantify key influencing factors.
Related
- Evaluating Underground CO2 Storage Potential across Sedimentary Facies of the Michigan Basin
- Development and Evaluation of Filtration Performance of ZnO-Doped PEI Composite Membranes
- Evaluating the performance of the surface water and ocean topography (SWOT) mission for monitoring reservoirs across the conterminous United States
- Evaluating the performance and reliability of CA-Markov and CA-ANN models for land use/land cover change simulation in Southwestern Nigeria
- Evaluating CMAQ Representation of Wildfire PM₂.₅ in California (2008–2018): Regime-Dependent Performance and Nonlinear Error Structure
BibTeX
@article{rakotonirinjanahary2026comparison,
title = {{Comparison of two methods for evaluating the performance of particle filtration systems in underground railway stations}},
author = {Valisoa Rakotonirinjanahary and Suzanne Crumeyrolle and Mateusz Bogdan and Benjamin Hanoune},
journal = {Indoor Environments},
year = {2026},
month = oct,
doi = {10.1016/j.indenv.2025.100144},
url = {https://hal.science/hal-05774948v1},
note = {CC BY},
}
RIS
TY - JOUR
TI - Comparison of two methods for evaluating the performance of particle filtration systems in underground railway stations
AU - Valisoa Rakotonirinjanahary
AU - Suzanne Crumeyrolle
AU - Mateusz Bogdan
AU - Benjamin Hanoune
PY - 2026
DA - 2026/10/02
JO - Indoor Environments
DO - 10.1016/j.indenv.2025.100144
UR - https://hal.science/hal-05774948v1
AB - Filtration is one of the strategies that can be implemented to mitigate the public health challenge due to elevated concentrations of particles in underground railway stations (URS). During a particle (PM 2.5 and PM 10 ) monitoring campaign performed in a Paris URS over more than 5 years, two filtration systems were tested. One of these systems employs positive ionization (Experiment 1), the second one is based on water filtration (Experiment 2). The present study focuses on evaluating the performance of these experiments, using two distinct methods: the daily amplitude coefficient (DAC), and a generalized additive model (GAM). The DAC method requires only particulate matter measurements, while the GAM method necessitates additional measurements (indoor CO 2 concentration, temperature and humidity, outdoor pollutant concentration) as it analyses the nonlinear relationships between all these factors. The results show that both filtration technologies are more effective at reducing PM 2.5 concentrations than PM 10 , but Experiment 2 was less efficient than Experiment 1. Specifically, DAC analysis leads to a reduction of 17.8 ± 3.2 % for PM 10 and 25.9 ± 6.6 % for PM 2.5 for Experiment 1, compared to only 0.4 ± 0.1 % for PM 10 and 3.7 ± 1.0 % for PM 2.5 for Experiment 2. GAM analysis gives similar results, with a reduction of 13.1 ± 0.8 % for PM 10 and 26.0 ± 0.8 % for PM 2.5 for Experiment 1, and for Experiment 2 a reduction of 1.3 ± 2.0 % for PM 10 and 4.6 ± 1.3 % for PM 2.5 . Yet, because of differences in the filtration setups, such as the position of the modules and the maximal air flow capacity, and of unspecified internal characteristics of the systems, no conclusion can be drawn as to whether one of the filtration techniques is more efficient than the other. The two analysis methodologies lead to similar results and allow to fully and rigorously exploit long time series. The DAC method can be applied using only PM measurements, while the GAM method requires additional measurements, but at the same time provides valuable insights about the effects of each parameter considered. • PM was continuously monitored in an underground railway station in Paris over a 5-year period. • Two separate filtration experiments were conducted during the measurement period. • For each filtration system, similar filtration performances were assessed using both the DAC and GAM approaches. • DAC relies only on PM time series, while GAM uses additional parameters to identify and quantify key influencing factors.
SN - 2950-3620
ER -