The reviewed record of science sign in
Pith

arxiv: 2606.02592 · v1 · pith:7YIJNLU2 · submitted 2026-05-21 · stat.AP · cs.AI

Tracking Urban Atmospheric Pollutants using Sentinel-5P Satellite Data

Alice Gomez-Cantos , Henry O. Velesaca This is my paper

Reviewed by Pith2026-06-30 15:54 UTCgrok-4.3pith:7YIJNLU2open to challenge →

classification stat.AP cs.AI
keywords NO2Sentinel-5PTROPOMIK-means clusteringurban air qualitysatellite observationsEcuador
0
0 comments X

The pith

Clustering of satellite NO2 percentiles distinguishes urban pollution patterns without ground measurements.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper develops a satellite-based method to track urban nitrogen dioxide pollution using only Sentinel-5P/TROPOMI tropospheric column data over Guayas Province, Ecuador. It summarizes multi-year observations into median and upper-tail percentiles at the canton scale, then applies K-means clustering to identify characteristic pollution regimes. The approach deliberately skips surface concentration conversion and ground-truth validation. A sympathetic reader would care because it supplies an air-quality assessment option for regions that lack dense local monitoring networks. Results indicate that highly urbanized cantons show elevated extreme values and greater variability while less urbanized areas remain lower and more uniform.

Core claim

Unsupervised K-means clustering applied to the median and upper-tail percentiles of annual Sentinel-5P/TROPOMI NO2 observations identifies distinct pollution regimes at the canton scale using satellite data alone.

What carries the argument

K-means clustering performed on the median, P90, P95, and P99 of aggregated tropospheric NO2 column values at canton resolution.

Load-bearing premise

Tropospheric column observations summarized by median and upper percentiles and grouped by clustering can reliably separate local pollution regimes without surface conversion or ground validation.

What would settle it

Ground-based surface NO2 measurements collected in the same cantons that show no systematic differences between the resulting clusters.

Figures

Figures reproduced from arXiv: 2606.02592 by Alice Gomez-Cantos, Henry O. Velesaca.

Figure 1
Figure 1. Figure 1: Map of Guayas Province in Ecuador, illustrating the spatial distribution of cantons across the province. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Spatial distribution of annual NO2 median (x˜), P90, P95, and P99 tropospheric column values over Guayas Province, illustrating background levels and localized pollution hotspots. and P99 maps highlight compact, spatially coher￾ent hotspots associated with dense traffic corri￾dors, industrial zones, and port-related activities. These results confirm that extreme percentiles are effective for isolating pers… view at source ↗
Figure 3
Figure 3. Figure 3: Interannual evolution (2020–2025) of provincial-scale [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Silhouette-based automatic K selection for K-means clustering using [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: K-means clustering map of Guayas Province based on [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
read the original abstract

Urban nitrogen dioxide ($NO_2$) is a key indicator of combustion-related air pollution and exhibits strong spatial and temporal variability in cities. This study presents a satellite-based framework for tracking urban $NO_2$ pollution using tropospheric column observations from Sentinel-5P/TROPOMI over Guayas Province, Ecuador. Rather than estimating surface concentrations, the methodology emphasizes robust distributional metrics, including the median and upper-tail percentiles ($P_{90}$, $P_{95}$, and $P_{99}$), to characterize background conditions and localized pollution extremes at the canton scale. Multi-year satellite observations are aggregated annually and analyzed using unsupervised K-means clustering to identify characteristic pollution regimes without predefined thresholds. Results show that highly urbanized cantons consistently exhibit elevated extreme $NO_2$ values and greater variability, while less urbanized areas display lower and more homogeneous patterns. The proposed approach provides an interpretable and scalable tool for urban air-quality assessment in data-scarce regions using satellite observations alone. The implementation is publicly available on GitHub https://hvelesaca.github.io/sentinel-5P-clustering/.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The manuscript proposes a satellite-based framework for tracking urban NO2 pollution in Guayas Province, Ecuador, using Sentinel-5P/TROPOMI tropospheric column observations. It aggregates multi-year data to canton-scale median and upper-tail percentiles (P90, P95, P99), applies unsupervised K-means clustering to identify characteristic pollution regimes without predefined thresholds or surface conversion, and concludes that the method offers an interpretable, scalable tool for air-quality assessment in data-scarce regions using satellite data alone.

Significance. If the resulting clusters can be shown to align with actual surface-level pollution differences rather than retrieval artifacts, the approach would offer a practical, ground-data-independent method for regime identification in regions with limited monitoring infrastructure. The public GitHub implementation is a clear strength for reproducibility.

major comments (3)
  1. [Abstract] Abstract and Methods: The central claim that the distributional summaries and K-means clustering 'reliably distinguish characteristic pollution regimes' is unsupported because the pipeline performs no ground-truth comparison to surface measurements, no conversion from column to surface concentrations, and no sensitivity tests to confounders such as boundary-layer height variability or cloud/aerosol retrieval artifacts.
  2. [Methods] Methods: The number of clusters K is treated as a free parameter with no justification, elbow-plot analysis, or stability assessment across K values; this directly affects the robustness of the identified regimes and the scalability assertion.
  3. [Results] Results: The statement that 'highly urbanized cantons consistently exhibit elevated extreme NO2 values' is presented without quantitative cluster-separation metrics, statistical tests against urban-extent covariates, or comparison to independent pollution indicators, leaving open the possibility that clusters reflect data artifacts rather than pollution signals.
minor comments (2)
  1. [Abstract] The GitHub repository link is provided and the code is stated to be publicly available; this aids reproducibility and should be retained.
  2. [Methods] Notation for percentiles (P_{90}, etc.) is clear but the exact aggregation procedure (e.g., how daily pixels are combined per canton per year) could be stated more explicitly for readers unfamiliar with TROPOMI processing.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive comments, which highlight important aspects of validation and robustness. We address each major point below, with planned revisions to strengthen the manuscript while preserving its focus on satellite-only analysis for data-scarce regions.

read point-by-point responses

  1. Referee: [Abstract] Abstract and Methods: The central claim that the distributional summaries and K-means clustering 'reliably distinguish characteristic pollution regimes' is unsupported because the pipeline performs no ground-truth comparison to surface measurements, no conversion from column to surface concentrations, and no sensitivity tests to confounders such as boundary-layer height variability or cloud/aerosol retrieval artifacts.

    Authors: The study is explicitly framed for data-scarce regions where surface measurements are unavailable, so the method relies on satellite column distributions alone. We will revise the abstract and methods to clarify that the regimes characterize satellite-observed patterns and their spatial association with urbanization, without claiming direct surface-level validation. We will add a dedicated discussion subsection on potential confounders (boundary-layer height, clouds, aerosols) using TROPOMI quality flags and metadata, including qualitative sensitivity checks. revision: partial

  2. Referee: [Methods] Methods: The number of clusters K is treated as a free parameter with no justification, elbow-plot analysis, or stability assessment across K values; this directly affects the robustness of the identified regimes and the scalability assertion.

    Authors: We agree that K selection requires explicit justification. The revised methods will include an elbow plot of within-cluster sum of squares, silhouette scores across K=2 to 6, and stability assessment via multiple random initializations and bootstrap resampling of the canton-level feature vectors. revision: yes

  3. Referee: [Results] Results: The statement that 'highly urbanized cantons consistently exhibit elevated extreme NO2 values' is presented without quantitative cluster-separation metrics, statistical tests against urban-extent covariates, or comparison to independent pollution indicators, leaving open the possibility that clusters reflect data artifacts rather than pollution signals.

    Authors: We will augment the results with quantitative cluster-quality metrics (silhouette score and Davies-Bouldin index) and add a correlation analysis between cluster membership and independent canton-level urban extent derived from land-cover products, including Spearman coefficients and p-values. revision: yes

Circularity Check

0 steps flagged

No circularity; purely observational pipeline

full rationale

The paper describes a data-processing pipeline that ingests Sentinel-5P/TROPOMI column observations, computes canton-level median and upper-tail percentiles, and applies unsupervised K-means clustering. No equations are present that derive a quantity from itself, no parameters are fitted to a subset and then called a prediction, and no self-citations are invoked to justify uniqueness or an ansatz. The central output (cluster labels) is a direct algorithmic result of the chosen summaries and algorithm; it does not reduce to a redefinition or statistical tautology of the input data. The analysis is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The framework rests on the domain assumption that TROPOMI column densities can be used directly for regime identification without surface conversion or local calibration; K is an implicit free parameter in the clustering step.

free parameters (1)
  • number of clusters K
    Unsupervised K-means requires choosing K; value not stated in abstract and affects regime identification.
axioms (1)
  • domain assumption Sentinel-5P/TROPOMI tropospheric NO2 columns are sufficiently accurate and spatially resolved to characterize canton-scale pollution variability.
    Invoked when the paper states that column observations alone suffice for the distributional metrics and clustering.

pith-pipeline@v0.9.1-grok · 5722 in / 1149 out tokens · 34648 ms · 2026-06-30T15:54:46.750292+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

19 extracted references · 3 canonical work pages

  1. [1]

    Atmospheric Measurement Techniques Discussions , volume=

    Validation of methane and carbon monoxide from Sentinel-5 Precursor using TCCON and NDACC-IRWG stations , author=. Atmospheric Measurement Techniques Discussions , volume=. 2021 , publisher=

    2021

  2. [2]

    Atmospheric Measurement Techniques , volume=

    Methane retrieved from TROPOMI: improvement of the data product and validation of the first 2 years of measurements , author=. Atmospheric Measurement Techniques , volume=. 2021 , publisher=

    2021

  3. [3]

    Remote Sensing , volume=

    Investigating the performance of carbon monoxide and methane observations from Sentinel-5 precursor in China , author=. Remote Sensing , volume=. 2022 , publisher=

    2022

  4. [4]

    Environmental Sciences Proceedings , volume=

    Sentinel-5P/TROPOspheric Monitoring Instrument CH4 and CO Total Column Validation over the Thessaloniki Collaborative Carbon Column Observing Network Site, Greece , author=. Environmental Sciences Proceedings , volume=. 2023 , publisher=

    2023

  5. [5]

    Atmospheric Measurement Techniques , volume=

    5 years of Sentinel-5P TROPOMI operational ozone profiling and geophysical validation using ozonesonde and lidar ground-based networks , author=. Atmospheric Measurement Techniques , volume=. 2024 , publisher=

    2024

  6. [6]

    Remote Sensing , volume=

    Evaluation of sentinel-5p Tropomi methane observations at northern high latitudes , author=. Remote Sensing , volume=. 2024 , publisher=

    2024

  7. [7]

    Atmospheric environment , volume=

    Urban air quality , author=. Atmospheric environment , volume=. 1999 , publisher=

    1999

  8. [8]

    Atmospheric Chemistry and Physics , volume=

    Emission of trace gases and aerosols from biomass burning--an updated assessment , author=. Atmospheric Chemistry and Physics , volume=. 2019 , publisher=

    2019

  9. [9]

    Atmosphere , volume=

    Atmospheric ozone and methane in a changing climate , author=. Atmosphere , volume=. 2014 , publisher=

    2014

  10. [10]

    Folkert and Compernolle, Steven and Pinardi, Gaia and Blechschmidt, Anne-Marlene and Peuch, Vincent-Henri and Colette, Augustin and Veefkind, J

    Douros, John and Eskes, Henk and van Geffen, Jos and Boersma, K. Folkert and Compernolle, Steven and Pinardi, Gaia and Blechschmidt, Anne-Marlene and Peuch, Vincent-Henri and Colette, Augustin and Veefkind, J. Pepijn , title =. Geoscientific Model Development , year =. doi:10.5194/gmd-16-509-2023 , url =

    work page doi:10.5194/gmd-16-509-2023 2023

  11. [11]

    and Al-Saadi, Jassim A

    Judd, Laura M. and Al-Saadi, Jassim A. and Szykman, James J. and Valin, Lukas C. and Janz, Scott J. and Kowalewski, Matthew G. and Eskes, Henk J. and Veefkind, J. Pepijn and Cede, Alexander and others , title =. Atmospheric Measurement Techniques , year =. doi:10.5194/amt-13-6113-2020 , url =

    work page doi:10.5194/amt-13-6113-2020 2020

  12. [12]

    and Eichmann, Kai-Uwe and Fj

    Verhoelst, Tijl and Compernolle, Steven and Pinardi, Gaia and Lambert, Jean-Christopher and Eskes, Henk J. and Eichmann, Kai-Uwe and Fj. Ground-based validation of the Copernicus Sentinel-5P TROPOMI NO _2 measurements with the NDACC ZSL-DOAS, MAX-DOAS and Pandonia global networks , journal =. 2021 , volume =. doi:10.5194/amt-14-481-2021 , url =

    work page doi:10.5194/amt-14-481-2021 2021

  13. [13]

    Remote Sensing Applications: Society and Environment , volume=

    Sentinel satellite data monitoring of air pollutants with interpolation methods in Guayaquil, Ecuador , author=. Remote Sensing Applications: Society and Environment , volume=. 2023 , publisher=

    2023

  14. [14]

    Air quality impacts of war detected from the Sentinel-5P satellite over Ukraine , author=. Conf. Series: Earth and Environmental Science , volume=. 2023 , organization=

    2023

  15. [15]

    SN Computer Science , volume=

    Air Quality Monitoring Using Sentinel-5p TROPOMI—A Case Study of Pune City , author=. SN Computer Science , volume=. 2024 , publisher=

    2024

  16. [16]

    Journal of the Air & Waste Management Association , volume=

    Air quality and climate connections , author=. Journal of the Air & Waste Management Association , volume=. 2015 , publisher=

    2015

  17. [17]

    Ortiz Morocho, Dayana Mishel and Montesdeoca Jara, Bryam Adri. Pron. 2023 , publisher=

    2023

  18. [18]

    Climate , volume=

    Satellite-Based Seasonal Fingerprinting of Methane Emissions from Canadian Dairy Farms Using Sentinel-5P , author=. Climate , volume=. 2025 , publisher=

    2025

  19. [19]

    International Conference on Smart Technologies, Systems and Applications , pages=

    Analysis of Hidden Patterns in Road Accident Dataset Using Clustering Techniques , author=. International Conference on Smart Technologies, Systems and Applications , pages=. 2024 , organization=

    2024