Three icddr,b surveillance sites

Coastal → riverine → urban inland

Same country, same pathogen, same 26-year window — three radically different water regimes. Distance to the Bay of Bengal is the design variable, and it is what makes the gradient testable.

Rural COASTAL

Chakaria HDSS

21.78940 N, 92.05500 E · 22.9 x 29.4 km · 1 polygon · 10 km to the sea

Water bodies: Matamuhuri river and estuary · Moheshkhali channel · Chakaria Sundarbans - former mangrove now converted to extensive SHRIMP AQUACULTURE PONDS · salt pans · Bay of Bengal within ~10 km of t…
Trophic state: Estuarine-brackish, strongly eutrophic in the shrimp-pond belt. Salinity gradient from fresh (east, hilly) to fully marine (west).
Ocean colour applies: YES — direct exposure
Datasets: 16 apply 8 partial 2 not applicable
Variables: BLOOM_AREACHL_OCXFAIMCIMNDWINDCIPHENOLOGYPHYCOCYANINSWMTURBIDITY

Rural riverine

Matlab HDSS

23.39130 N, 90.70120 E · 20.4 x 22.1 km · 251 polygons · 250 km to the sea

Water bodies: Meghna river (2-6 km wide), Dhonagoda river, Gumti, canals, beels, thousands of household ponds
Trophic state: Meso- to eutrophic. Agricultural nutrient runoff · high suspended sediment.
Ocean colour applies: REGIONAL FORCING ONLY
Datasets: 11 apply 11 partial 4 not applicable
Variables: BLOOM_AREAFAIMCIMNDWINDCIPHENOLOGYPHYCOCYANINTURBIDITY

Urban inland

Dhaka City

23.78660 N, 90.42350 E · 17.6 x 26.0 km · 1 polygon · 300 km to the sea

Water bodies: Hatirjheel, Gulshan/Banani/Dhanmondi/Ramna/Crescent lakes · Buriganga, Turag, Balu, Shitalakshya rivers · Tongi Khal · thousands of ponds
Trophic state: Hyper-eutrophic. Heavy untreated sewage and industrial load. Dense, persistent cyanobacterial blooms.
Ocean colour applies: REGIONAL FORCING ONLY
Datasets: 9 apply 6 partial 11 not applicable
Variables: BLOOM_AREAFAIMNDWINDCIPHENOLOGYPHYCOCYANINTURBIDITY

Why the gradient is the point

The three sites form a COASTAL -> RIVERINE -> URBAN INLAND gradient. This is the study design's greatest strength: the classic coastal-chlorophyll/plankton-intrusion hypothesis (Lobitz 2000, Jutla 2013, Constantin de Magny 2008) should be STRONGEST at Chakaria, INTERMEDIATE at Matlab, and WEAKEST at Dhaka, where local eutrophic pond and lake blooms should dominate instead. That gradient is directly testable and has never been published.

The testable prediction. Adding algae variables to a strong hydro-climate model should produce a large gain at Chakaria, a moderate one at Matlab and a small one at Dhaka — and any Dhaka gain should come from the local variables (NDCI, phycocyanin, bloom fraction), not the coastal ones (SWM, Bay of Bengal chlorophyll). A flat gradient refutes the intrusion hypothesis as a general explanation. Either result is publishable.

Known problems in the site geometry

MATLAB — 43 POLYGONS ARE NOT IN MATLAB

The shapefile carries 251 polygons across three upazilas: Matlab Uttar (112), Matlab Dakkhin (96) and Daudkandi (43) — which is in Cumilla district, outside the HDSS. 42 of those 43 sit in Block E. Filter them out before extraction (208 remain) or Block E is 56% contaminated with no matching population denominator.

Separately, the village ID (vid) is populated in 8 of 251 records, and 251 polygons carry only 245 distinct names — so a name-based join to the HDSS outcome data is not safe either.

CHAKARIA — NO INTERNAL BOUNDARIES

The boundary file is a single polygon with only OBJECTID, Shape_Leng and Shape_Area. The 11 unions described in the config are not in it. The east side is hilly and the west is coastal lowland with the shrimp-pond belt — completely different hydrology and salinity. Either request the union-level shapefile, or derive the stratum from SRTM elevation ≤ 5 m combined with distance to coast.

Pooling the two halves averages away exactly the exposure contrast the study is trying to detect.