Modeling Urban Dynamics for Climate Resilience: Understanding Threats and Vulnerabilities

Global urban expansion is rapidly increasing, with over 50% currently living in cities and an expected rise to 66% by 2050. This growth poses challenges for peri-urban areas, including agricultural land loss, biodiversity destruction, and climate change impacts. Comprehensive research, integrating urban and rural perspectives, is vital, emphasizing the need for quantitative methods like the model analysis approach. The complexity of urban systems, inspired by ecological theories, requires geosimulation for improved decision-making. Geosimulation enhances decision-making by considering a wide array of variables in complex real-world dynamics. Focused on modeling and analyzing sophisticated systems, it enriches geography with diverse methods. Uncertainties in geosimulation often revolve around computational challenges in simulating intricate systems like urban landscapes. Agent-based modeling (ABM) offers a bottom-up approach for simulating complex dynamics and has become a well-established tool for studying land use changes. GeoAI extends artificial intelligence (AI) to support geospatial problem-solving, leveraging advancements in machine learning and computing power. As geography transforms into a domain of big data science, integrating AI and machine learning becomes crucial for unlocking the full scientific potential of geospatial data. The main focus of our research developments is to integrate advanced ABM simulations with GeoAI methodologies to analyze threats and vulnerabilities in urban and peri-urban landscapes. These analyses aim to guide decision-making towards the sustainable management of climate-resilient cities. 

We present here two study cases, the first studies the environmental factors impacting urban beehive health. Honey bees (Apis mellifera) are essential for ecosystems and agriculture, but global colony mortality rates have risen. Our study in Greater Montreal, Canada, with data from 1000 hives, used machine learning to pinpoint factors affecting bee colony health. Higher hive density correlated with lower health, and specific months, particularly September and October, showed a significant reliance on nearby vegetation for hive development. The machine learning model results were used to map predicted honey bee habitat quality, offering insights for conservation and management strategies. The second study located in the capital city of Bogota, Colombia, aims to assess water provision and habitat quality while analyzing their response to land cover changes through the application of the Integrated Valuation of Ecosystem Services and Tradeoffs (InVEST) model. A multitemporal analysis utilized three analogous images to capture land cover changes. Artificial Neural Networks (ANN) were employed to establish relationships between land uses and associated drivers, estimating the probability of occurrence for each land use. Subsequently, the InVEST model was utilized to map the Ecosystem Services (ES) of wetlands, and spatial analyses were conducted to investigate potential correlations between the city's population socioeconomic profile and access to the evaluated ES. This project aimed to position and establish Geosimulation as an important tool for analyzing the crucial implications of policy-making.