Global_Environmental_Research_Vol.27No.1
61/80

The objective of our review was to provide a concise overview of the available resources used when exploring 3.5 Blended Models 4. Conclusions and Recommendations higher in vertical transmission and thus found it difficult to link to exposure-health response relationships alone (Burke et al., 2021). Conversely, Holloway et al. (2021) noted that newer satellite technology has been able to detect emissions from single power plants or industrial facilities and these will likely be further improved to be able to detect individual fires in the near future. Mueller, et al. (2021) utilized NASA’s Visible Infra-red Imaging Radiometer Suite sensor on the daily number of fires in Thailand, measuring both PM10 and O3, in addition to ground-monitoring data. They noted the inability to distinguish effects on the measured health outcome of low birth weight between exposure levels of either chemical. They are also limited by the long-term changes in satellite functionality, resulting in establishing connections between trends and ground monitoring stations (Holloway et al., 2021). Blended models use a combination of different methods to determine air pollution characteristics and data sources. Given the large uncertainties in air pollution data arising from the limitations of a single method, the application of a blended model may significantly enhance the accuracy and reliability of the data collected (Johnson et al., 2020; Pennington et al., 2019). One study (Jegasothy et al., 2023) for example used MODIS (satellite imaging) to estimate PM2.5 emitted by bushfires. They did a two-stage study where they first measured PM2.5 from monitoring sites to check for air quality, then visually confirmed any measurements exceeding the 95th percentile through MODIS if the source was indeed from bushfires, albeit with the potential for false negatives if the smoke produced was too thin. Magzamen et al. (2021) similarly used satellite imaging to determine the presence of smoke plumes while using surface monitors for actual PM2.5 measurements. Another emerging approach is the use of all the previous models in conjunction with machine learning for improved accuracy. Reid et al. (2015) employed a unique approach of combining CTMs with outputs from satellite imaging and meteorological data used in conjunction with gradient-boosted machine learning to provide accurate predictions of out-of-sample air pollution concentrations. O’Neill et al. (2021) and Zou et al. (2022) used a similar approach, using multiple data sources (CTM with satellite imaging) while three machine-learning approaches that similarly improved predictions when compared to surface monitors. The addition of machine learning however will require even more computational power than CTMs alone. applying in challenges satellite likely Review on Exposure Assessment of Biomass Burning the link between air pollution produced from biomass burning and its potential impact on public health. Our study revealed that air pollution caused by biomass burning can be observed and studied through a diverse array of data sources and modeling approaches. From chemical (atmospheric) transport models to satellite observations, many various techniques—usually in combination—to understand the complex dynamics of pollutants released during biomass burning events. Broadly categorized, the three main methods of quantifying and collecting air quality data are through direct site/station monitors, satellite data and air quality models (i.e., CTMs), each with its own strengths and weaknesses. In our findings, the use of combinations or blended approaches have been the prevailing method for collecting air pollution data. It is likely that the inherent limitations of each data collection method necessitate supplementary approaches, and also pave the way for the birth of novel techniques, as in the case of machine-learning algorithms that improve the accuracy of existing approaches. Both the traditional and the novel approaches would benefit from qualitative and ad hoc assessments to grasp their limitations and the complexity to make of generalizations that could potentially impact public health policy later on. their While certain data collection methods may be that of impractical to monitoring stations, which are resource-abundant free methods are available through the use of NASA's satellite imaging as well as through simulations, as seen in CTMs. These advancements will hopefully alleviate the need for such stations to provide quality evidence when linking air pollution to negative health impacts. While the emergence of machine to improving the accuracy of these methods when compared to direct local measurements of air quality seen in monitoring stations, the bulk of existing literature uses air pollution modeling methods that are calibrated for the global north, leaving behind the global south, which suffers from more frequent biomass burning and bears the brunt of air pollution health impacts due to climate change. Additionally, most modeling methods were developed for the primary purpose of surveillance and not necessarily directly considered for policy development or public health programming. These technological advancements can potentially be leveraged to create systems that can communicate risks in a timely manner. Recalibration or validation of these models is needed to increase the reliability of the results for the global south, as well as to explore the possibility of further developing these air pollution modeling initiatives to create early warning systems, especially in the global south, where monitoring stations are fewer and further in between. It is likely that our study was limited in terms of true researchers results, especially when used real-life for for many regions—particularly countries—alternative learning holds significant promise employ scenarios limited mostly and 55

元のページ  ../index.html#61

このブックを見る