Advancing landscape sustainability science using scalable spatio-temporal landscape analytics with big data and machine learning

Urban landscape designs typically attempt to balance human experience with ecosystem performance. Social and ecological data however, rarely align spatially or temporally; social-behavioral data can be highly localized, event-based, and rapidly changing, while ecological data can be coarser, slower to update, and collected at varying scales. The mismatch in data can hinder understanding the potential co-benefits and trade-offs across locations and timeframes. This dissertation operationalizes a Landscape Sustainability Science (LSS) lens to synchronize human perceptions and behaviors with ecological structure and function. I develop scalable spatio-temporal analytics that fuse “big” passively collected data with “small” curated datasets and machine-learning models to connect human well-being and cultural ecosystem services (CES) to carbon-related ecosystem functions. My thesis comprises four studies across multiple places and scales. Using geotagged microblogs and deep-learning sentiment models in Shanghai, Study 1 links park accessibility and biophysical qualities to expressed happiness, providing spatially explicit evidence for design and park planning. Study 2 examines the COVID-19 period and shows that shocks to mobility and risk perception reshaped park preferences and sentiment, with implications for equitable access and resilience. Study 3 estimates carbon-sequestration assets at a statewide scale in Illinois by integrating inventory, species/structure, and geostatistical modeling, demonstrating how higher-resolution characterization can support finer scaled climate planning. Study 4 integrates surveys, social-media text, and imagery across Shanghai, New York, and Stockholm to compare CES and perceptions, revealing both cross-cultural regularities and context-specific patterns. Together, these studies demonstrate the measurement and synchronization of data and scales to align heterogeneous social and ecological signals into shared space–time units. The approach improves reliability and representativeness through data fusion and yields decision-ready evidence. The contributions are conceptual (an operational LSS triangle for planning), methodological (reproducible, scalable analytics), and practical (tools for monitoring, scenario testing, and back-casting). The findings support landscape planning and governance aimed at healthier, more inclusive, and climate-aligned cities. The concluding chapter brings vertex- and edge-level evidence back into the LSS triangle and articulates a practical workflow that embeds machine learning with heterogeneous “big-and-small” data, geostatistics, and domain theory to strengthen inference and decision relevance. It demonstrates how analytics can be translated into design practice via transparent, auditable steps – e.g., multi-criteria weighting of carbon, well-being, cost, and equity; simple rulebooks that map model signals to design moves; and monitoring loops that feed outcomes back into models, showing evidence pathways from models to implemented park upgrades. It also assesses limitations (e.g., proxy measures, platform biases, cross-language model risks, and case-scope constraints) and proposes concrete remedies such as time-locking online/offline data and paired intercept surveys to improve validity and transportability across contexts.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Landscape Architecture

Grantor

University of Illinois Urbana-Champaign

Year dc:date

2025

Author dc:creator

• Lai, Siqi

Contributors dc:contributor

• Deal, Brian
• Sullivan, William
• Wang, Shaowen
• Cong, Cong

• Landscape sustainability science
• Cultural ecosystem services
• Urban parks and happiness
• Carbon sequestration
• Machine learning

Statement dc:rights

• Copyright 2025 Siqi Lai

Language dc:language

en

Handle dc:identifier

https://hdl.handle.net/2142/132502

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/132502