Air pollution from particulate matter with size below 10 µm (PM10) has been shown to be a major issue, being responsible of adverse effect on human health, climate change and environmental deterioration. Especially over developing countries, where air quality is often poor and the few available ground-based monitoring networks are generally sparse and mainly located over urban areas, there is a need for an advanced approach giving a continuous output in time and space. Satellite-borne instrumentation and data retrieval protocols now allow detailed monitoring of atmospheric properties and thus of air quality conditions with almost global, homogeneous, and near-real-time spatial coverage. The present study is aimed to provide a statistical framework to quantify ground-level PM10 from NASA satellite-retrieved quantities. The focus is on Equatorial Asia, where the availability of long-term PM10 observations at high resolution on Malaysia peninsula and Borneo allowed the development of an effective statistical approach to quantify outdoor air pollution and evaluate its spatio-temporal patterns and effects. Satellite-retrieved AOD550 (Aerosol Optical Depth at 550 nm), water column, key chemical species amounts and land use properties were analyzed in terms of their predictive skills on PM10, in reference to different spatial and temporal scales. The effect of cyclic seasonal phenomena (e.g. monsoons and wildfires), typical of the study region, was also included in the analysis. Focusing on the period 2005-2015, the performances of univariate/multivariate linear models and Artificial Neural Networks (ANN) in predicting PM10 were compared. Referring to the same period, ANN were also applied to develop PM10 annual maps over a broad region comprising Malaysia and Indonesia. The predicted annual PM10 concentration patterns were finally used to apply an epidemiological model aimed to estimate the yearly amount and spatial distribution of premature deaths connected to chronic exposure to fine particulate matter in outdoor air.
L'inquinamento atmosferico da particolato solido di dimensione inferiore a 10 µm (PM10) rappresenta un problema di grande rilevanza scientifica e sociale, dati i significativi impatti sulla salute umana, sui cambiamenti climatici e più in generale sul deterioramento ambientale. Specialmente nei paesi in via di sviluppo, dove la qualità dell'aria è spesso scarsa e le poche reti di monitoraggio disponibili hanno generalmente una copertura lacunosa, è necessario un approccio avanzato che dia un output continuo nel tempo e nello spazio. Le attuali strumentazioni satellitari e i relativi algoritmi di acquisizione dei dati consentono ad oggi il monitoraggio ad alta risoluzione di fondamentali proprietà atmosferiche, e quindi delle condizioni di qualità dell'aria, con una copertura quasi globale, omogenea ed in tempo reale. Il presente studio ha lo scopo di sviluppare un approccio statistico atto alla quantificazione del PM10 a livello del suolo, partendo da dati acquisiti da satelliti NASA. L’area d’interesse è l’Asia equatoriale, dove la disponibilità di osservazioni di PM10 a lungo termine e ad alta risoluzione sulla penisola della Malesia e nel Borneo ha permesso lo sviluppo di alcuni modelli statistici predittivi dell'inquinamento dell'aria esterna e della sua distribuzione spazio-temporale. Valori satellitari di AOD550 (Aerosol Optical Depth a 550 nm), umidità, quantità di specie chimiche chiave e proprietà d'uso del suolo sono state analizzate in termini di potere predittivo sul PM10, in riferimento a diverse scale spaziali e temporali. Nell'analisi è stato incluso anche l'effetto di fenomeni ciclici stagionali (ad es. Monsoni e incendi), tipici dell’area di studio. Relativamente al periodo 2005-2015, sono state confrontate le prestazioni dei modelli lineari univariati e multivariati, e di reti neurali artificiali (ANN) nella previsione del PM10 al livello del suolo. In riferimento allo stesso periodo, reti neurali sono state in seguito applicate per la ricostruzione di mappe annuali di PM10, su una vasta regione comprendente Malesia e Indonesia. Le suddette mappe hanno infine rappresentato l’input essenziale per l’applicazione di un modello epidemiologico finalizzato alla stima del numero e della distribuzione spaziale annuale dei decessi prematuri dovuti all'esposizione cronica al particolato fine.
A statistical approach to monitor particulate matter concentration from space
BRUNI ZANI, NICOLA
2018/2019
Abstract
Air pollution from particulate matter with size below 10 µm (PM10) has been shown to be a major issue, being responsible of adverse effect on human health, climate change and environmental deterioration. Especially over developing countries, where air quality is often poor and the few available ground-based monitoring networks are generally sparse and mainly located over urban areas, there is a need for an advanced approach giving a continuous output in time and space. Satellite-borne instrumentation and data retrieval protocols now allow detailed monitoring of atmospheric properties and thus of air quality conditions with almost global, homogeneous, and near-real-time spatial coverage. The present study is aimed to provide a statistical framework to quantify ground-level PM10 from NASA satellite-retrieved quantities. The focus is on Equatorial Asia, where the availability of long-term PM10 observations at high resolution on Malaysia peninsula and Borneo allowed the development of an effective statistical approach to quantify outdoor air pollution and evaluate its spatio-temporal patterns and effects. Satellite-retrieved AOD550 (Aerosol Optical Depth at 550 nm), water column, key chemical species amounts and land use properties were analyzed in terms of their predictive skills on PM10, in reference to different spatial and temporal scales. The effect of cyclic seasonal phenomena (e.g. monsoons and wildfires), typical of the study region, was also included in the analysis. Focusing on the period 2005-2015, the performances of univariate/multivariate linear models and Artificial Neural Networks (ANN) in predicting PM10 were compared. Referring to the same period, ANN were also applied to develop PM10 annual maps over a broad region comprising Malaysia and Indonesia. The predicted annual PM10 concentration patterns were finally used to apply an epidemiological model aimed to estimate the yearly amount and spatial distribution of premature deaths connected to chronic exposure to fine particulate matter in outdoor air.File | Dimensione | Formato | |
---|---|---|---|
TESI.pdf
non accessibile
Descrizione: Testo della tesi
Dimensione
19.91 MB
Formato
Adobe PDF
|
19.91 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/154124