Cyanobacterial harmful algal blooms are an increasing threat to coastal and inland waters. These blooms can be detected using optical radiometers due to the presence of phycocyanin (PC) pigments. The spectral resolution of best-available multispectral sensors limits their ability to diagnostically detect PC in the presence of other photosynthetic pigments. To assess the role of spectral resolution in the determination of PC, a large ( $N =905$ ) database of colocated in situ radiometric spectra and PC are employed. We first examine the performance of selected widely used machine-learning (ML) models against that of benchmark algorithms for hyperspectral remote sensing reflectance ( $R_{\mathrm {rs}}$ ) spectra resampled to the spectral configuration of the Hyperspectral Imager for the Coastal Ocean (HICO) with a full-width at half-maximum (FWHM) of < 6 nm. Results show that the multilayer perceptron (MLP) neural network applied to HICO spectral configurations (median errors < 65%) outperforms other ML models. This model is subsequently applied to $R_{\mathrm {rs}}$ spectra resampled to the band configuration of existing satellite instruments and of the one proposed for the next Landsat sensor. These results confirm that employing MLP models to estimate PC from hyperspectral data delivers tangible improvements compared with retrievals from multispectral data and benchmark algorithms (with median errors between $\sim 73$ % and 126%) and shows promise for developing a globally applicable cyanobacteria measurement approach.

Linear Mixed Effect (LME) models are applied to the CoastColour atmospherically-corrected Medium Resolution Imaging Spectrometer (MERIS) reflectance, L2R full resolution product, to derive chlorophyll-a (chl-a) concentration and Secchi disk depth (SDD) in Lake Erie, which is considered as a Case II water (i.e., turbid and productive). A LME model considers the correlation that exists in the field measurements which have been performed repeatedly in space and time. In this study, models are developed based on the relation between the logarithmic scale of the water quality parameters and band ratios: B07:665 nm to B09:708.75 nm for log₁₀chl-a and B06:620 nm to B04:510 nm for log₁₀SDD. Cross validation is performed on the models. The results show good performance of the models, with Root Mean Square Errors (RMSE) and Mean Bias Errors (MBE) of 0.31 and 0.018 for log₁₀chl-a, and 0.19 and 0.006 for log₁₀SDD, respectively. The models are then applied to a time series of MERIS images acquired over Lake Erie from 2004–2012 to investigate the spatial and temporal variations of the water quality parameters. Produced maps reveal distinct monthly patterns for different regions of Lake Erie that are in agreement with known biogeochemical properties of the lake. The Detroit River and Maumee River carry sediments and nutrients to the shallow western basin. Hence, the shallow western basin of Lake Erie experiences the most intense algal blooms and the highest turbidity compared to the other sections of the lake. Maumee Bay, Sandusky Bay, Rondeau Bay and Long Point Bay are estimated to have prolonged intense algal bloom.

A global constant value of the extinction coefficient (K_d) is usually specified in lake models to parameterize water clarity. This study aimed to improve the performance of the 1-D freshwater lake (FLake) model using satellite-derived K_d for Lake Erie. The CoastColour algorithm was applied to MERIS satellite imagery to estimate K_d. The constant (0.2 m⁻¹) and satellite-derived K_d values as well as radiation fluxes and meteorological station observations were then used to run FLake for a meteorological station on Lake Erie. Results improved compared to using the constant K_d value (0.2 m⁻¹). No significant improvement was found in FLake-simulated lake surface water temperature (LSWT) when K_d variations in time were considered using a monthly average. Therefore, results suggest that a time-independent, lake-specific, and constant satellite-derived K_d value can reproduce LSWT with sufficient accuracy for the Lake Erie station. A sensitivity analysis was also performed to assess the impact of various K_d values on the simulation outputs. Results show that FLake is sensitive to variations in K_d to estimate the thermal structure of Lake Erie. Dark waters result in warmer spring and colder fall temperatures compared to clear waters. Dark waters always produce colder mean water column temperature (MWCT) and lake bottom water temperature (LBWT), shallower mixed layer depth (MLD), longer ice cover duration, and thicker ice. The sensitivity of FLake to K_d variations was more pronounced in the simulation of MWCT, LBWT, and MLD. The model was particularly sensitive to K_d values below 0.5 m⁻¹. This is the first study to assess the value of integrating K_d from the satellite-based CoastColour algorithm into the FLake model. Satellite-derived K_d is found to be a useful input parameter for simulations with FLake and possibly other lake models, and it has potential for applicability to other lakes where K_d is not commonly measured.

Instrumented buoys are used to monitor water quality, yet there remains a need to evaluate whether in vivo fluorometric measures of chlorophyll a (Chl a) produce accurate estimates of phytoplankton abundance. Here, 6 years (2014–2019) of in vitro measurements of Chl a by spectrophotometry were compared with coeval estimates from buoy‐based fluorescence measurements in eutrophic Buffalo Pound Lake, Saskatchewan, Canada. Analysis revealed that fluorometric and in vitro estimates of Chl a differed both in terms of absolute concentration and patterns of relative change through time. Three models were developed to improve agreement between metrics of Chl a concentration, including two based on Chl a and phycocyanin (PC) fluorescence and one based on multiple linear regressions with measured environmental conditions. All models were examined in terms of two performance metrics; accuracy (lowest error) and reliability (% fit within confidence intervals). The model based on PC fluorescence was most accurate (error = 35%), whereas that using environmental factors was most reliable (89% within 3σ of mean). Models were also evaluated on their ability to produce spatial maps of Chl a using remotely sensed imagery. Here, newly developed models significantly improved system performance with a 30% decrease in Chl a errors and a twofold increase in the range of reconstructed Chl a values. Superiority of the PC model likely reflected high cyanobacterial abundance, as well as the excitation–emission wavelength configuration of fluorometers. Our findings suggest that a PC fluorometer, used alone or in combination with environmental measurements, performs better than a single‐excitation‐band Chl a fluorometer in estimating Chl a content in highly eutrophic waters.

One essential optical parameter to specify in lake models is water clarity, which is parameterized based on the light extinction coefficient (K_d). A global constant value of K_d is usually specified in lake models. One-dimensional (1-D) lake models are most often used as lake parameterization schemes in numerical weather prediction and regional climate models. This study aimed to improve the performance of the 1-D Freshwater Lake (FLake) model using satellite-derived K_d for Lake Erie. The CoastColour algorithm is applied to MERIS satellite imagery to estimate K_d and evaluated against K_d derived from Secchi disk depth (SDD) field-based measurements collected during Lake Erie cruises. A good agreement is found between field and satellite-derived K_d (RMSE = 0.63 m^-1, MBE = −0.09 m^-1, I_a = 0.65) (in situ data was collected in 2004, 2005, 2008, 2011, 2012). The constant (0.2 m^-1) and satellite-derived K_d values as well as radiation fluxes and meteorological station observations are then used to run FLake at the location of a buoy where lake surface water temperature (LSWT) was measured in 2008. Results improved compared to using a constant K_d value (0.2 m^-1) (lake-specific yearly average K_d value: RMSE = 1.54 ºC, MBE = −0.08 ºC; constant K_d value: RMSE = 1.76 ºC, MBE = −1.26 ºC). No significant improvement is found in FLake simulated LSWT when K_d variations in time are considered using a monthly average. Therefore, results suggest that a time-independent, lake-specific, and constant satellite-derived K_d value can reproduce LSWT with sufficient accuracy. A sensitivity analysis is also performed to assess the impact of various K_d values on the simulation of mean water column temperature (MWCT), mixed layer depth (MLD), water temperature isotherms as well as ice dates and thickness. Results show that FLake is sensitive to variations in K_d to estimate the thermal structure of Lake Erie. Dark waters result in warmer spring and colder fall temperatures compare to clear waters. Dark waters always produce warmer MWCT, shallower MLD, longer ice cover duration, and thicker ice. The sensitivity of FLake to K_d variations is more pronounced in the simulation of MWCT and MLD. The model is particularly sensitive to K_d values below 0.5 m^-1. This is the first study to assess the value of integrating K_d from the satellite-based CoastColour algorithm into the FLake model. Satellite-derived K_d is found to be a useful input parameter for simulations with FLake and possibly other lake models, and with potential for applicability to other lakes where K_d is not commonly measured.