Dense Air Pollution Estimation from Sparse in-situ Measurements and Satellite Data

Ruben Gonzalez Avilés, Linus Scheibenreif, Damian Borth · 2025

Read Dense Air Pollution Estimation from Sparse in-situ Measurements and Satellite Data by Ruben Gonzalez Avilés, Linus Scheibenreif, Damian Borth (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-17039.txt.

Introduction

Ruben Gonzalez Avilés, Linus Scheibenreif, and Damian Borth present a 2025 research paper on mapping air pollution over wide areas. Ground stations measure nitrogen dioxide at only a few points, while satellites cover much more ground. Older methods link the two sources point by point, which takes heavy computing power when a full map is needed.

The authors describe a new method that spreads each ground reading across a larger patch and then estimates a whole grid in one step. Tests show lower error than older point methods and far less computing cost. Readers get a clear plan for fast, wide maps that can support health and policy work.

Catalog record

CREATORRuben Gonzalez Avilés, Linus Scheibenreif, Damian Borth
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE34 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-17039

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Dense Air Pollution Estimation from Sparse in-situ Measurements and Satellite Data Ruben Gonzalez Avilés, Linus Scheibenreif, Damian Borth Abstract This paper addresses the critical environmental challenge of estimating ambient Nitrogen Dioxide (NO 2 ) concentrations, a key issue in public health and environmental policy. Existing methods for satellite-based air pollution estimation model the relationship between satellite and in-situ measurements at select point locations. While these approaches have advanced our ability to provide air quality estimations on a global scale, they come with inherent limitations. The most notable limitation is the computational intensity required for generating comprehensive estimates over extensive areas. Motivated by these limitations, this study introduces a novel dense estimation technique. Our approach seeks to balance the accuracy of high-resolution estimates with the practicality of computational constraints, thereby enabling efficient and scalable global environmental assessment. By utilizing a uniformly random offset sampling strategy, our method disperses the ground truth data pixel location evenly across a larger patch. At inference, the dense estimation method can then generate a grid of estimates in a single step, significantly reducing the computational resources required to provide estimates for larger areas. Notably, our approach also surpasses the results of existing point-wise methods by a significant margin of 9.45%, achieving a Mean Absolute Error (MAE) of 4.98 μ ​ g / m 3 \mu g/m^{3} . This demonstrates both high accuracy and computational efficiency, highlighting the applicability of our method for global environmental assessment. Furthermore, we showcase the method’s adaptability and robustness, by applying it to diverse geographic regions. Our method offers a viable solution to the computational challenges of large-scale environmental monitoring. Figure 1 : Comparison of point-wise and dense estimation approaches for NO 2 concentration estimation from satellite imagery. The point-wise approach (left) generates predictions at a single location, denoted by the red dot. The dense estimation approach (right) estimates NO 2 concentrations over a designated area, indicated by the red square, enhancing computational efficiency for regional-scale estimations. Figure 2 : High-level overview showcasing the process of estimating NO 2 concentrations using a deep learning model. The model inputs are randomly sampled windows consisting of satellite data from Sentinel-2 and Sentinel-5P, which then pass through a convolutional neural network to extract relevant features. The NO 2 Regression Head produces a concentration estimate, which is subsequently aligned with ground truth data for validation. Concurrently, the Land Cover Classification Head processes the same features to classify land cover, aiding in the interpretation of NO 2 distribution. Losses from both heads are combined to optimize the model, with the ultimate goal of providing accurate, high-resolution estimations of ground-level NO 2 . 1 Introduction Air quality degradation, particularly due to NO 2 , poses a significant risk to environmental sustainability and public health. This toxic and highly reactive gas, primarily emitted from the combustion of fossil fuels, is a major contributor to air pollution and precursor of greenhouse gases Chen et al. 2007 ; Lammel and Graßl 1995 . Moreover, NO 2 exposure is associated with a range of health issues. Various studies demonstrate the correlation between NO 2 levels and the prevalence of respiratory and cardiovascular diseases (see Latza et al. 2009 for a review). Therefore, the accurate assessment of NO 2 levels is crucial to safeguard public health and to support the development of effective policies aimed at mitigating its detrimental effects. For these reasons, national public health and environmental agencies monitor NO 2 concentrations through networks of localized ground-based air quality stations. While ground-based instruments provide accurate and reliable data, they offer limited spatial coverage and incur high operational costs. These factors result in various notable data gaps, particularly in remote or less developed areas where monitoring stations are sparse Lamsal et al. 2008 . To address these limitations, satellite-based remote sensing has emerged as a viable alternative for comprehensive environmental monitoring Scheibenreif et al. 2022a ; Cooper et al. 2020 . Satellites such as the European Space Agency’s Sentinel-5P provide extensive global coverage, enabling the monitoring of atmospheric constituents, including NO 2 , on a much broader scale Vîrghileanu et al. 2020 . However, a significant challenge persists in effectively converting satellite observations into precise ground-level pollutant concentrations. This difficulty is particularly pronounced when estimating over extensive areas, due to the inherent spatial sparsity of available ground truth data, which are often confined to localized measurements. This paper presents a novel method for ground-level NO 2 estimation, utilizing a dense estimation technique that leverages the capabilities of advanced deep learning models and the comprehensive data acquired from multi-spectral satellite imagery. We introduce a novel sampling technique to address the spatial sparsity of ground truth data, ensuring data continuity in areas without direct measurements. The dense estimation approach efficiently computes estimates for larger areas, significantly improving computational efficiency compared to individual point-wise estimations and enabling broader, more frequent assessment. This method marks a notable improvement in environmental monitoring, providing a scalable, efficient solution for global NO 2 estimation, which can in turn support enhanced environmental policy-making. 2 Related Work Local monitoring of NO 2 concentrations, primarily through ground-based stations, has been the traditional method for assessing air quality Guerreiro et al. 2014 . Despite their precision, these localized stations face significant limitations, including their inability to provide comprehensive spatial coverage. This results in notable data blind spots, especially in remote and underdeveloped regions where setting up and maintaining such stations is logistically challenging and costly Verghese and Nema 2022 . Complementing this, the utilization of satellite data for environmental monitoring introduces its own set of complexities. While satellite imagery offers expansive geographical coverage, transforming these high-altitude observations into accurate, ground-level pollutant concentration estimates poses a significant challenge Li et al. 2011 . In the face of these limitations, researchers have increasingly turned to deep learning as a potential solution. Convolutional neural networks (CNNs), in particular, have demonstrated their efficacy in remote sensing applications, from land cover classification to air quality estimation Zhu et al. 2017 . These advancements in deep learning have opened up new possibilities for interpreting and utilizing satellite data more effectively for environmental monitoring purposes Tuia et al. 2023 ; Liao et al. 2020 . Several recent methods have utilized deep learning approaches for air pollution estimation from satellite data at discrete geospatial locations, which we will henceforth refer to as the point-wise approach Rowley and Karakuş 2023 ; Scheibenreif et al. 2022a . This methodology, as highlighted in studies by Rowley and Karakuş 2023 , and Scheibenreif et al. 2022a excels in producing highly accurate pollution estimates at specific global locations. Despite its precision and significant improvement over traditional methods, the point-wise approach is constrained by its inherent focus on single spatial locations. When applied to expansive areas, the point-wise method necessitates repeated processing for each location, resulting in considerable computational demands for global-scale application. Addressing the limitations inherent in the point-wise approach, our research introduces a more efficient alternative: the dense estimation approach. This novel method aims to maintain the accuracy of point-wise estimations while significantly reducing computational requirements. Instead of producing individual estimates, the dense estimation approach simultaneously generates a grid of estimates for a designated area, effectively decreasing the total number of iterations required for larger-scale NO

Read the complete text