Research & systems

Forecasting systems built end-to-end

From raw NWP ensembles and reanalysis data to the thresholds, dashboards, and briefs decision-makers use. Status tags reflect where each system actually is — operational, pilot, or roadmap.

Forecast-X

A WMO-aligned, multi-hazard impact-based warning system covering 600 Bangladesh upazilas across 64 districts. Blends four NWP ensemble models — ICON-EPS, NCEP-GEFS, ECMWF-IFS, and UKMO — across 13 variables, applying a full multi-hazard threshold framework (11 hazard types with sub-types) against the WMO 4×4 risk matrix. Includes rolling multi-window rainfall computation, WBGT calculation, and a choropleth dashboard for at-a-glance risk visualisation.

Operational Multi-hazard Ensemble NWP Dashboard

Cascading Hazard Impact Framework

An impact-based forecasting framework for the upazila/district level, piloted for Chattogram District. Maps hydro-meteorological events through primary and secondary hazard chains to sector-level impacts and early actions. Delivered as a multi-sheet workbook: a master cascade matrix, event-specific sheets, and a risk-scoring guide.

Pilot Cascading risk Chattogram

ERA5-Land Climate Data Pipeline

A full climatology pipeline for 600 upazilas spanning 1990–2025: CDS API acquisition, derived-variable computation (relative humidity via the Magnus formula, feels-like temperature via NWS Rothfusz/wind chill, WBGT via ISO 7243/Stull 2011), and climatology generation — run in Google Colab with atomic write patterns to handle session disconnects.

Operational Reanalysis Heat stress

Automated BMD SYNOP Pipeline

A Google Apps Script pipeline fetching Bangladesh Meteorological Department SYNOP data hourly, handling zlib decompression via pako, and writing to structured Google Sheets — one tab per meteorological parameter, one row per station per date, across 48 weather stations.

Operational Data engineering

Multi-Model Ensemble Feels-Like Temperature

Ensemble forecast pipelines for 48 Bangladesh weather stations combining ECMWF IFS and GFS models, implementing comprehensive feels-like temperature logic (NWS wind chill + NOAA heat index + weighted blend) with null/NaN guards and daily min/max aggregation.

Operational Heat stress

Dengue Early Warning Trigger Framework

A scoping exercise for Mirpur Wards 3 and 5 (DNCC): critical assessment of an inception report and a revised three-layer trigger indicator matrix using proxy-based climate, remote sensing, and DGHS epidemiological data to substitute for unavailable ward-level entomological indices.

Scoping Health early warning
Long-term focus

Impact-based forecasting system design

Underpinning all of the above: a sustained focus on IBF system design spanning nine hazard modules, trigger and anticipatory-action frameworks, multi-level decision workflows, and visualisation dashboards — built with institutional precision around source attribution (BMD, FFWC, IWFM, IMD, GloFAS) and clear distinctions between what's operational today and what's on the roadmap.