Nature-Based Solutions · Vietnam Coast

Which mangroves stand between Vietnam's coast and the next storm?

758 historical cyclone tracks from NOAA IBTrACS (1950–2025), classified using Vietnam's official tropical cyclone scale, were cross-referenced with actual mangrove extent polygons from the Global Mangrove Watch. I first prototyped the analysis locally using GeoPandas, then turned it into a dbt pipeline running directly on BigQuery.

Storm tracks vs. mapped mangrove extent

1950–2025. Hover a province for its mangrove hectares, or a highlighted track for the storm's intensity.

Mangrove extent
none77k ha
Storm tracks
largest mangrove exposure (top 12)
full historical record
Context
Inland Vietnam · Cambodia, Laos, China

Largest mangrove exposure

Storms ranked by mangrove hectares within 50km of their track

Most frequently exposed

Provinces by count of historical storms within 50km

Do the numbers hold up?

How does the 2020 Global Mangrove Watch extent compare with the multi-year historical mangrove mask used in an earlier project? I compare the two datasets by province to see where the estimates broadly agree and where they diverge.

ProvinceGMW 2020 (ha)Historical max (ha)Ratio

GMW 2020 is consistently lower than the historical maximum. This is expected, since the historical figure is a union of a 2000 land-cover mask and GMW's 2007–2020 range (the largest footprint ever mapped), while GMW 2020 is a single-year snapshot taken after any loss in between.

From prototype to production

The same spatial logic, tested locally in GeoPandas and rebuilt in SQL for BigQuery.

Sources IBTrACS + GMW v3 netCDF + shapefile read locally bq load Local · validation geopandas prototype buffered-corridor spatial join Cloud · transform dbt: staging → marts ST_MAKELINE, ST_UNION_AGG, ST_DWITHIN dbt run Live in BigQuery fct_mangrove_storm_impact 860 rows · 8/8 tests passing
Both paths start with the same raw IBTrACS and Global Mangrove Watch data. The GeoPandas version was used to prototype and validate the spatial-join logic locally. That same logic was then re-implemented in SQL through a dbt pipeline running on BigQuery, where it now powers the live production table.
Research use only This analysis is a research prototype intended to explore spatial relationships between historical tropical cyclone tracks and mangrove extent. The results have not undergone sufficient validation to support operational or official decision-making. They should not be treated as a substitute for official forecasts, warnings, hazard assessments, or government reports.