新資料時代下以風險評估為觀點的環境資料分析應用

Applications of Environmental Data Analysis From the Perspective on Risk Assessment in the New Data Era

李杰翰(Chieh-Han Lee)

指導教授 ： 余化龍

國立臺灣大學/生物資源暨農學院/生物環境系統工程學研究所/博士(2020年)

https://doi.org/10.6342/NTU202000011

摘要

近年來，在電腦科學與資訊科技領域，無論是在硬體或演算法上取得的巨幅進展，大數據與機器學習皆成為了現今最熱門的兩個詞彙，也使得其他不同領域皆想藉由這股新的科技力量在應用上有所突破。也因為這股熱潮，資料成為了最重要的資產之一；資料分析方法成為了不可或缺的技術；資料科學家成為了人力市場中的熱門職缺。相同地，環境科學領域也極力嘗試結合此新型科技，來找出嶄新的應用方式。然而，由於環境資訊與民眾風險感知是緊緊相扣的，因此，在現今的這股資料浪潮中，從風險分析的觀點，在環境資訊的應用上有許多重要且需被關注的議題。

在本論文中，嘗試由風險評估的觀點出發，探討在現今的的新資料時代下，環境資料分析在應用上其可能性及衍生的重要課題。新資料時代下的多樣應用，加速了政府在開放資料上的進展，然而，環境擁有屬於公共財的特性，環境資料的蒐集與揭露主要掌握在政府部門手中。民眾對於環境資訊知的權利，往往與政府部門形成了對立關係。另外，在政府與民眾對於新科技在環境領域應用上的不熟悉，進而產生環境風險認知上的歧見。在這其中，環境資料科學家藉由其專業的科學知識與能力，在政府部門和民眾之間，形成一交互三角關係。此三角關係中，為了因應新資料時代的發展，每一個角色對於其餘兩個角色皆為不同的利益關係者。

本論文利用三個實際應用案例，作為闡述本論文所提出在新資料時代下，臺灣環境資料分析可能的未來發展方向以及問題所在。首先，本研究在開放政府與開放資料架構下，建立一南臺灣登革熱預警系統，經由過往難以取得的登革熱發病資料結合氣象因子，提供政府部門在登革熱防治上的預先部署依據，以及民眾對於自身所處環境的登革熱風險認知。第二，本研究利用建立特定商用物聯網空氣感測器的校正模型，經由比較不同可信度監測資料，了解環境數據除數字本身之外，數據的不確定性與民眾風險感知之間的關係，需要謹慎的對待。最後，本研究利用發展具備高效能的資料融合架構，整合確定性與不確定性資料，凸顯在大量含有不確定性的環境資料之下，如何以資料融合方式，達到正確的風險溝通結果。

本論文以風險評估的觀點，檢視現今在這個以資料引領的時代中，環境科學結合資料分析方法在政府、社會與科學三方中所扮演的角色，以及對於環境保護助益的可能性。希望此論文能夠給予未來環境資料分析在風險管理中的一個初步方向。

關鍵字

環境資料分析 ； 風險分析 ； 預警系統 ； 資料校正 ； 資料融合

並列摘要

In recent years, the world has made tremendous progress in computer science and information technology. Either computer hardware development or algorithms evolution lead Big Data and Machine Learning become two most popular words nowadays. Other applied fields also have seen great opportunities on using these emerging technologies to make a breakthrough. Because of this global trend, data has become one of the most valuable asset; data analysis methods have become the essential techniques; data scientists have become the most favored job in human resources market. Likewise, environmental science attempts to apply the new technology and finds innovations. However, environmental information is strongly associated with public risk perception. Hence, there are many important issues from the perspective on risk assessment need to be concerned while surfing on this new data wave.

The dissertation aims to explore application potentials of environmental data analysis and its related issues from the aspect of risk assessment today. The new data era has accelerated the progress of open governmental data. Environmental information is considered as public asset. However, government agencies mostly have authorization of environmental information in collection and reveal. Public's environmental information right-to-know often stands on the opposite side of government agencies. In addition, the reason for the controversy between government agencies and public is unfamiliar with the new technology. Besides, environmental scientists with professional knowledge and expertise forms the interaction triangle with the other characters that governments, public, and scientists are stakeholders to each other.

This dissertation illustrates the future possibility and problems for Taiwan's environmental data analysis in the new data era by three applications. First of all, under Open Data and Open Government framework, the study constructed an early warning system of dengue fever in southern Taiwan through combining incidences with meteorological factors. The results could provide the disease prevention and control for government agencies and provoke public risk awareness from the disease. Secondly, the study built a calibration model for particular commercial low-cost air quality sensors. By assessing the reliability of measurements, to have understanding that except for the numbers on devices, the relationship of measurement uncertainty and risk perception should be taken into consideration seriously. Lastly, the study developed a high performance data fusion framework that integrated certain and uncertain data to highlight the achievement for proper risk communication with large amount of uncertain environmental information.

The dissertation stands at the perspective of risk analysis to inspect what kind of role that environmental data sciences play in the relationship triangle. In conclusion, the dissertation seeks to open the way for environmental data analysis which is associated with risk management, in further, possible contributions to environmental protection.

並列關鍵字

Environmental data analysis ； Risk analysis ； Early warning system ； Data calibration ； Data fusion

參考文獻

[1] A. R. Al-Ali, I. Zualkernan, and F. Aloul, A mobile GPRS-sensors array for air pollution monitoring, IEEE Sens. J., 10 (2010), pp. 1666–1671.

[2] J. Angulo, H. L. Yu, A. Langousis, A. Kolovos, J. Wang, A. E. Madrid, and G. Christakos, Spatiotemporal Infectious Disease Modeling: A BME-SIR Approach, PLoS One, 8 (2013).

[3] J. M. Angulo, H.-L. Yu, A. Langousis, A. E. Madrid, and G. Christakos, Modeling of space–time infectious disease spread under conditions of uncertainty, 2012.

[4] Array of Things, Array of Things.

[5] N. S. Asmarian, A. Ruzitalab, K. Amir, S. Masoud, and B. Mahaki, Area-to-Area Poisson Kriging analysis of mapping of county- level esophageal cancer incidence rates in Iran., Asian Pac. J. Cancer Prev., 14 (2013), pp. 11–3.