02319naa a2200193 a 450000100080000000500110000800800410001910000170006024501520007726000090022952016890023865000170192765000210194465000150196565000190198070000210199970000170202077300880203711354682025-01-22       2025    bl uuuu     u00u1 u    #d1 aMONTAGNA, T.  aAn improved algorithm for estimating chlorophyll-a in coastal waters of southern Brazil from multispectral satellite images.h[electronic resource]  c2025  aRemote  sensing  chlorophyll-A  (CLA)  estimates  from  global  models  have  been  used  to  support  decision making  in southern Brazil, the most important bivalve mollusks production region (~9 thousand tons in 2022) in the country, and a recent study indicated that these estimates poorly represent the actual levels of CLA. The aim of the study was to develop an improved algorithm for estimating CLA in these coastal waters from multispectral images. A CLA database generated in situ between 2007 and 2009 was used to calibrate  and validate  algorithms based on spectral  data  from the  Medium Resolution Imaging Spectrometer (MERIS) (ENVISAT satellite) (300m spatial resolution), including algorithms based on red and near-infrared bands: two bands (2B and M2B), three bands (3B) and the Normalized Difference Chlorophyll Index (NDCI and MNDCI). Outputs from the global modelsOC4ME and Neural Network were also evaluated. NIR-red algorithms  outputs  correlated  significantly  with  the  measured  CLA,  except  for  MNDCI.  The  best  performing  models during the calibration, those based on 2B and NDCI (R2 = 0.37, residual standard error = 2.57 mg.m-3), were validated and fitted better the measured data (R2 &gt= 0.22) and showed lower RMSE values (around 2.5 mg.m-3) than the global models? outputs, which did not even correlate significantly (p&gt0.05) with in situ CLA measurements. The developedmodels  performed  better  than  the  global  models  evaluated  nevertheless  they  have  a  limited  prediction  power  when compared to regional algorithms developed elsewhere and this is probably linked to the low range of CLA measurements used to train the models.  aAlgae blooms  aCross validation  aMonitoring  aRemote sensing1 aVIBRANS??, A. C.1 aSOUZA, R. V.  tRevista Brasileira de Geografia F??sica, Recife, PEgv. 18, n. 1, p. 633-645, 2025.