Fant 10178 publikasjoner. Viser side 408 av 408:
Evaluating the role of low-cost sensors in machine learning based European PM2.5 monitoring
We evaluate the added value of integrating validated Low-Cost Sensor (LCS) data into a Machine Learning (ML) framework for providing surface PM2.5 estimates over Central Europe at 1 km spatial resolution. The synergistic ML-based S-MESH (Satellite and ML-based Estimation of Surface air quality at High resolution) approach is extended, to incorporate LCS data through two strategies: using validated LCS data as a target variable (LCST) and as an input feature via an inverse distance weighted spatial convolution layer (LCSI). Both strategies are implemented within a stacked XGBoost model that ingests satellite-derived aerosol optical depth, meteorological variables, and CAMS (Copernicus Atmospheric Monitoring Service) regional forecasts. Model performance for 2021–2022 is evaluated against a baseline trained on air quality monitoring stations without any form of LCS integration. Our results indicate that the LCSI approach consistently outperforms both the baseline and LCST models, particularly in urban areas, with RMSE reductions of up to 15–20 %. It also exhibits higher accuracy than the CAMS regional interim reanalysis with a lower annual mean absolute error (MAE) of 2.68 μg/m3 compared to 3.32 μg/m3. SHapley Additive exPlanations based analysis indicates that LCSI information improves both spatial and temporal representativeness, with the LCSI strategy better capturing localized pollution dynamics. However, the LCSI's dependency on the spatial LCS layer limits its ability to capture inter-urban pollution transport in regions with sparse or no LCS data. These findings highlight the value of large-scale sensor networks in addressing spatial coverage gaps in official air quality monitoring stations and advancing high-resolution air quality modeling.
2026
New Approach Methodologies (NAMs) are gaining significant momentum globally to reduce animal testing and enhance the efficiency and human relevance of chemical safety assessment. Even with substantial EU commitment from regulatory agencies and the academic community, the full regulatory adoption of NAMs remains a distant prospect. This challenge is further complicated by the fact that the academic world, oriented toward NAMs development, and regulatory agencies, focused on practical application, frequently operate in separate spheres. Addressing this disconnect, the present paper, developed within the European Partnership for the Assessment of Risks from Chemicals (PARC), provides a clear overview of both the available non-animal tests and current evaluation practices for genotoxic and carcinogenic hazard assessment, while simultaneously highlighting existing regulatory needs, gaps, and challenges toward greater human health protection and the replacement of animal testing through NAMs adoption.
The analysis reveals a complex landscape: while the EU is deeply committed to developing and adopting NAMs, as outlined in its Chemical Strategy for Sustainability and supported by initiatives like PARC, prescriptive regulations such as Classification, Labelling and Packaging (CLP) and Registration, Evaluation, Authorisation and Restriction of Chemicals (REACH) still heavily mandate in vivo animal data for hazard classification, particularly for germ cell mutagenicity and carcinogenicity. This reliance creates a “too-short-blanket-problem,” where efforts to reduce animal testing may impact human health protection because of the current in vivo-based classification criteria. In contrast, sectors such as cosmetics and certain European Food Safety Authority (EFSA)-regulated products demonstrate greater flexibility toward progressive integration of NAMs. While the deep mechanistic understanding of genotoxicity and carcinogenicity has significantly advanced the integration of alternatives to animal tests into regulatory chemical hazard assessment, their broader and full implementation faces considerable challenges due to both scientific complexities (i.e., the development and validation of fit-for-purpose NAMs) and existing legislative provisions.
2026
Abstract Hierarchical agglomerative clustering is a useful analysis technique which allows for a level of stability, interpretability and flexibility not available in other similar techniques such as K‐means, density‐based clustering or positive matrix factorization. Previous studies using hierarchical clustering on atmospheric model output have been limited to small domain sizes (roughly 100 × 100 grid cells) by the computational expense and memory requirements of the algorithm. Here we present a scalable hierarchical clustering implementation that we apply to two year‐long, hourly atmospheric data sets: model concentration and deposition timeseries at 290,520 locations over Alberta and Saskatchewan (538 540 grid); and 366,427 multi‐pollutant observations from 51 national air pollution surveillance stations located across Canada. When combined with other information such as emissions source locations, orography, and prevailing meteorological conditions, the method yields coherent, interpretable structures. In the case of model time series, the clustering provides regions of similar air quality (airsheds) which can be used to inform air quality monitoring network placement, or regions of similar deposition which can inform critical load assessment as well as monitoring site locations. In the case of the multi‐pollutant observations, we show that a single low‐primary pollutant cluster appears the most frequently at all but one of 51 stations across Canada, accounting for 62% of all station‐hours, while elevated SO 2 appears in factor profiles at certain monitoring locations near industrial and shipping activity. Together, these results demonstrate that hierarchical clustering can efficiently summarize patterns relevant to airshed mapping and source apportionment at previously unreachable scales.
2026