Sri Lanka Bird Diversity — Spatiotemporal Data Pipeline

01Overview

Citizen-science platforms have produced an enormous record of where and when birds are seen in Sri Lanka. On its own that record answers very little: observation effort is wildly uneven across space and time, and the environmental context that would explain a pattern lives in entirely separate satellite products.

This project built the pipeline that makes the two speak to each other, and the analysis that followed became a co-authored paper.

02Building the dataset

The integration layer links 1.5M+ observations — 429 species, 25 districts, 2014 to 2024 — to satellite-derived environmental features: NDVI, land cover, climate variables, aerosols, elevation, and Artificial Light At Night as an urbanization proxy.

Because raw counts mostly measure who was holding binoculars, analysis runs on multiple spatial grids (2 km, 5 km, 10 km) with spatial thinning to reduce sampling bias, and on effort-corrected temporal metrics such as rarefied richness and occupancy rather than naive totals.

03What the analysis found

Environmental drivers were examined with multivariate models including Poisson GLMs alongside correlation analysis, plus community structure, dominance, and beta diversity across regions and years.

Land-cover type turned out to be a stronger predictor of bird diversity than any single continuous variable such as NDVI or temperature on its own. Urbanization measured through ALAN showed scale-dependent effects: it supports high abundance of a few generalist species while reducing overall richness.

04Publication

The work is written up as "How Environment and Urbanization Shape Bird Diversity in Sri Lanka," accepted to MERCon 2026 in the Data Science and Artificial Intelligence track and available as arXiv:2607.00582.