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.
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.
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.
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.
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.
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.
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.
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.