How do cities reshape the planet? My work answers that question through the lenses of land change, infrastructure, biodiversity, human–environment interactions, health, and the spatial models that hold them together.
Urban land is the single most concentrated form of human transformation of the planet, and yet long-term high-resolution maps of where it has actually grown remain scarce. My work uses the GLC_FCS30D 30-meter time series and SSP scenario projections to reconstruct multi-decadal urban footprints in Bangladesh, Indonesia, and 19 world regions, with attention to the regional differences SSP-driven projections often blur.
This thread sits underneath everything else on this page. You cannot meaningfully study road expansion or biodiversity loss without first knowing where cities have been, and where they are going.
My doctoral work produced the first global-scale road network growth model, calibrated against historical road expansion and projected forward under multiple development scenarios. AGU's Eos magazine featured the work shortly after publication.
The model addresses a simple question with hard implications. If we can forecast where new roads will appear, we can also forecast where forests will be opened, where protected areas will be punctured, and where new exposure to flooding and other hazards will accumulate. A manuscript extending the model is in revision at Nature Communications.
Two manuscripts in this thread are in revision or review. A PNAS-targeted paper with Burak Güneralp and Lee A. Fitzgerald examines the combined impacts of road development and urban expansion on biodiversity in Brazil, Indonesia, and Nigeria. A Nature Sustainability submission, also with Güneralp, quantifies how urban expansion has isolated the global protected-area network from 1992 to 2015 and proposes a quantitative typology of regional buffer-and-growth regimes.
Earlier work in this thread covers Bangladesh's protected areas and Key Biodiversity Areas, and threatened amphibian, reptile, and mammal ranges across Indonesia under SSP scenarios.
This is the umbrella that ties the previous three together. Cities, roads, and infrastructure are not external to ecosystems — they are coupled with them. The questions I keep returning to are about where that coupling produces conflict (habitat loss, hazard exposure, lost services) and where it produces opportunity (compact development, accessibility, conservation gains).
Methodologically the work draws on land systems science, landscape ecology, and the longer tradition of human–environment geography. Empirically it spans South and Southeast Asia, the Americas, and Africa, with an effort to keep regional differences visible rather than averaged away.
A parallel thread on spatial methods for public health. Earlier work mapped COVID-19 spread across multiple geographies, examined how geospatial analyses are used in research on homeless populations, and produced systematic reviews on the role of GIS in pandemic decision-making.
The thread is quieter now than it was during 2020–2022, but the methodological lessons — about exposure, accessibility, and how rapidly spatial data has to move during a crisis — carry into the climate and biodiversity work.
Across all of the themes above, the question of how we model spatial processes shows up again and again. My work in this thread brings together machine learning, statistical, and process-based approaches: hedonic and regression models for housing and accessibility, supervised and unsupervised ML for land-cover classification, statistical and AI-based flood susceptibility modeling, and large-scale process models for urban–transportation coevolution.
The common thread is a preference for models that take spatial structure seriously, communicate uncertainty honestly, and remain interpretable enough that someone other than the modeler can use them. Tools of choice: Python (scikit-learn, XGBoost, PyTorch), R, and a long-running soft spot for Bayesian thinking.