Employing Machine Learning for Climate Change Resilience Decision Models
Our research aims to understand how communities can take better climate change resilience decisions by employing data analytics, machine learning (ML), and artificial intelligence (AI) techniques. Climate resilience planning, supported by ML and AI, can enhance inclusive decision making by incorporating multitude of inputs from communities. More specifically, we aim to identify and assess coastal resilience with a focused case study of the greater Boston area with an emphasis on vulnerable communities. We develop an urban climate resilience decision model and process that employs machine learning to iteratively analyze climate resilience data sets, including social data from communities. The enhanced model can accommodate various sets of climate criteria as well as social attributes to improve data analyses. For instance, the real-time data of natural hazards, such as flooding conditions, can be analyzed with ML to understand the clusters of hazards. Decision makers considering social and regulatory factors can influence the ranking of climate vulnerabilities assessments in urban areas. In doing so, we also provide further understanding to the use and application of AI in IS literature (Collins, et al., 2021). AI, ML, and sustainability literature in information systems (IS) can aid in facilitating environmental governance and deliver solutions to improve organizational processes and individual practices (Nishant, Kennedy, and Corbett, 2020). Most of the previous research on climate change and coastal resilience is focused on scientific observations. For example, classification methods were applied to historical flood behaviors to understand flood resilience (Saraviet al., 2019). The impact of community and climate factors on agricultural vulnerabilities in the coastal regions of Bangladesh were measured using ML techniques to help improve resource planning and management (Jakariya et al., 2020). The analysis can help us understand climate risks on infrastructure needs and vulnerable communities. Boston initiatives, such as Climate Ready Boston and the Natural Hazard Mitigation Plan has identified some of these risks based on scientific data. Boston neighborhoods can be assessed in terms of climate resilience and climate equity based on their potential differences in environmental attributes (i.e., coastal flooding risk) by integrating it with social data. Our research will particularly focus on historically vulnerable communities, such as Black, Indigenous and People of Color (BIPOC) communities, that are historically both economically and racially discriminated as well as are likely to see disproportionate impact of climate change. We have conducted exploratory data collection and analysis from Twitter, Facebook, and climate-focused blogs to understand the climate criteria influencing the livelihood of BIPOC community members, such as climate change impact on infrastructure needs. For example, flooding during a 2018 storm required rescuing of trapped residents in a coastal Boston neighborhood. Our model aims to analyze data from these communities for coastal resilience planning and decision making. The theoretical foundations of the study will be adapted from IS literature, including AI, big data analytics, decision models in climate change, and empirical evaluation methods. Although data analytics techniques have been adapted for DSS, incorporating big data, such as images and videos, to address theory and practice is a challenge (Bharati, 2017, pp. 273-276). Social factors play an important role in decision making on sustainability (Carberry et al., 2019) and the role of information systems is integral to solving climate change challenges (Seidel et al., 2017). While the modeling of applications with a spatial element, such as flexible geographic information system (GIS) software, can be enhanced with big data and machine learning, the facets of AI (autonomy, learning, and inscrutability) can challenge the norms, processes and outcomes in computing, business, and society (Berente, Gu, Recker, and Santhanam, 2021). We plan to address these challenges by extending IS theories to further our understanding of climate change impact. We have collected data related to key coastal measurements, social media data on coastal resilience discourse from BIPOC Boston communities, and data on coastal resilience planning and strategies. The data sets on coastal measurements are collected from public datasets, such as from the United States Geological Survey (USGS) and the National Oceanic and Atmospheric Administration (NOAA). We plan to further examine the climate datasets and social data to develop a decision model that leverages ML techniques. The outcomes of this research can help community members and policy makers make more robust sustainability and climate focused decisions.
Lee, Carol and Bharati, Pratyush, "Employing Machine Learning for Climate Change Resilience Decision Models" (2021). NEAIS 2021 Proceedings. 10.