Definitions

The algae variables

What is computed, from which bands, with what expected lag to cholera — and the column name it lands under in the merged panel, so the number on this page and the number in your CSV are the same thing.

Applies at

Showing 10 of 10 variables.

Variable	Family	Formula	Sensors	Lag (wks)	Panel column	Applies	Evidence
MNDWI		`(B3 - B11) / (B3 + B11)`				Chakaria Dhaka Matlab
TURBIDITY		`B4 / B3`				Chakaria Dhaka Matlab
BLOOM_AREA	bloom extent	`fraction of the water mask exceeding a chlorophyll or FAI threshold`		2-8	`bloom_frac_ndci, bloom_frac_scum`	Chakaria Dhaka Matlab
MCI	chlorophyll	`L709 - L681 - (L754 - L681) * (709 - 681) / (754 - 681)`	Sentinel-3 OLCI / MERIS	4-8	`mci`	Chakaria Matlab	Anas et al. 2021 ↗
NDCI	chlorophyll	`(B5 - B4) / (B5 + B4)`	Sentinel-2 MSI / Sentinel-3 OLCI / MERIS	4-8	`ndci, ndci_max, ndci_wmax`	Chakaria Dhaka Matlab	Mishra & Mishra 2012 ↗ Anas et al. 2021 ↗ Sarker et al. 2019 (missing)
PHYCOCYANIN HIGHEST NOVELTY	cyanobacteria	`B3 / B5`	Sentinel-3 OLCI (best) / Sentinel-2 (proxy only) / …	2-8	`phyco_proxy, phyco_620, cyano_frac`	Chakaria Dhaka Matlab	Islam et al. 2007 (missing) Anas et al. 2021 ↗
PHENOLOGY	derived	`bloom initiation date, peak date, duration, amplitude, integrated bioma…`		seasonal	`derived downstream`	Chakaria Dhaka Matlab
CHL_OCX	ocean colour	`OC4 / OC3M maximum band ratio polynomial`	SeaWiFS / MODIS / VIIRS / OC-CCI / GlobColour	8-12	`chl_ocean, reg_chl_ocean`	Chakaria	Nusrat et al. 2024 ↗ Jutla, Akanda & Islam 2010 ↗ Jutla et al. 2013 ↗ Lobitz et al. 2000 Jutla et al. 2012 ↗
SWM	ocean colour	`Rrs(412) - Rrs(555)`	SeaWiFS / MODIS / VIIRS / OC-CCI	8	`swm, reg_swm`	Chakaria	Jutla et al. 2013 ↗
FAI	scum	`B8 - (B4 + (B11 - B4) * (842 - 665) / (1610 - 665))`	Sentinel-2 / Landsat / MODIS	2-6	`fai, fai_wmax`	Chakaria Dhaka Matlab	Hu 2009 ↗ Anas et al. 2021 ↗

Why each one is in the study

MNDWI lag wks

Water mask. Defines WHERE to measure algae each pass. Better than NDWI in built-up Dhaka.

(B3 - B11) / (B3 + B11)

TURBIDITY lag wks

Both a covariate and a QUALITY FLAG - extreme sediment loads corrupt chlorophyll retrievals.

B4 / B3

BLOOM_AREA bloom extent lag 2-8 wks

Bloom EXTENT and FREQUENCY are more epidemiologically meaningful than a mean concentration.

fraction of the water mask exceeding a chlorophyll or FAI threshold

MCI chlorophyll lag 4-8 wks

Maximum Chlorophyll Index. Slightly outperforms NDCI for very high biomass and surface blooms.

L709 - L681 - (L754 - L681) * (709 - 681) / (754 - 681)

Reference: Gower et al. 2005

NDCI chlorophyll lag 4-8 wks

The red-edge index. Works in TURBID productive water where blue-green ratios (OC4) fail - i.e. every water body at all three of your sites.

(B5 - B4) / (B5 + B4)

Reference: Mishra & Mishra 2012, Remote Sens. Environ. 117:394-406, doi:10.1016/j.rse.2011.10.016

PHYCOCYANIN cyanobacteria lag 2-8 wks

THE HIGHEST-NOVELTY VARIABLE IN THIS PROJECT. Phycocyanin is the CYANOBACTERIA-SPECIFIC pigment. Islam 2007 showed V. cholerae survives inside the mucilaginous sheath of Anabaena in Bangladeshi freshwater and is released when the bloom dies. Chlorophyll-a cannot distinguish a harmful cyanobacterial bloom from a benign green-algal one. Phycocyanin can.

B3 / B5

Reference: Islam et al. 2007 J. Appl. Sci. 7(13):1785-1789 for the cholera link · Simis et al. 2005 for the algorithm

PHENOLOGY derived lag seasonal wks

Racault argues bloom TIMING carries more cholera signal than bloom magnitude. High novelty, cheap to derive once you have the time series.

bloom initiation date, peak date, duration, amplitude, integrated biomass

Reference: Racault et al. 2019, Remote Sensing 11(23):2763, doi:10.3390/rs11232763

CHL_OCX ocean colour lag 8-12 wks

Standard open-ocean chlorophyll. KNOWN TO OVERESTIMATE in turbid case-2 water, so it is valid offshore of Chakaria but not inside the estuary.

OC4 / OC3M maximum band ratio polynomial

SWM ocean colour lag 8 wks

Satellite Water Marker. The spectral distance between clear blue and turbid green water. R2 = 0.78 for spring cholera vs 0.58 for chlorophyll-a alone.

Rrs(412) - Rrs(555)

Reference: Jutla et al. 2013, Remote Sens. Environ., PMID 23878762

FAI scum lag 2-6 wks

Floating Algae Index. Baseline subtraction in the NIR · robust to aerosol, thin cloud and sun glint. Detects SURFACE SCUM, which chlorophyll indices miss.

B8 - (B4 + (B11 - B4) * (842 - 665) / (1610 - 665))

Reference: Hu 2009, Remote Sens. Environ. 113(10):2118-2129, doi:10.1016/j.rse.2009.05.012

The gap that matters most. Phycocyanin is the highest-novelty variable here — it is the only one that can distinguish a harmful cyanobacterial bloom from a benign green-algal one, which chlorophyll-a cannot. Its key citation, Islam et al. 2007 on Vibrio cholerae surviving in the Anabaena mucilaginous sheath, is cited throughout the config but the PDF is not in the folder. And at Dhaka — the site where cyanobacteria dominate — only the Sentinel-2 B3/B5 proxy is available, because Sentinel-2 has no 620 nm band and Sentinel-3 at 300 m cannot see a Dhaka lake. That is the main methodological risk in the whole study. See Papers.