Many lakes around the world have faced eutrophication and frequent cyanobacterial harmful algal blooms (cyanoHABs), where chlorophyll-a concentration (Chla) is often used to determine the trophic status of the lake and evaluate the effectiveness of nutrient reduction strategies. However, limitedin situsampling and the absence of robust remotely sensed data product often hinder the analysis of long-term Chla changes in such lakes. Here, using Lake Okeechobee (LO) as an example, we addressed this challenge by developing a novel method to harnesses the power of both multi-sensor remote sensing and machine learning. The method first usedin situChla to estimate Chla from OLCI imagery (ChlaOLCI), which then served as a surrogate forin situdata toward obtaining the large matchup datasets for MODIS model training. Subsequently, two deep neural network models, DNNSand DNNNS, were trained for the MODIS-saturated pixels with high algal biomass waters and the non-saturated pixels with lower Chla but more turbid waters, respectively. The daily and monthly Chla derived from MODIS (ChlaMODIS) were validated usingin situChla (R2~0.77, MAPD ~25%) and monthly ChlaOLCI(R2~0.90, MAPD ~11%), respectively. Monthly ChlaMODIS derived from Terra in 2000 — 2023 was used to analyze the spatiotemporal changes in LO. We observed a significant long-term increasing trend of Chla (~2 mg m-3 decade-1) and a ~30% increase relative to the multiple-year average. To explore the potential factors driving long-term Chla changes, water temperature, precipitation, inflow, stage, and the multivariate ENSO index were tested through multi-variative analysis. Although further test is required, the approach of using one sensor as a “bridge” to bring limitedin situdata and another sensor together to overcome the data scarcity and lack of spectral bands is believed to be extendable to other eutrophic lakes around the world.

Satellite ocean color remote sensing provides an effective tool for monitoring and tracking harmful algal blooms (HABs) in oceanic, coastal, and inland waters. However, traditional medium-resolution (300 m – 1 km) satellite data do not have the required spatial resolution to characterize nearshore HABs, whose harmful impacts are directly relevant to local communities. Here, we show that commercial imagery provided by the >100 SuperDove sensors in PlanetScope constellation can provide ~4 m resolution, 8-band spectral data every 1-2 days (resampled to 10 m in this study for consistency with Sentinel-2) to meet the critical need for detecting blooms of the toxic dinoflagellateKarenia brevis(often called red tides) in nearshore waters off the central west coast of Florida (USA). Despite the relatively low signal-to-noise ratios (SNRs), SuperDove Enhanced Red-Green-Blue (ERGB) images clearly depict the detailed distributions and morphological shapes ofK.brevisblooms in both nearshore coastal waters and small estuaries. These features, challenging or impossible to observe from traditional satellite sensors, are well captured and quantified with an optimized deep learning model that leverages both spectral features and spatial context. Comparison of model results with medium-resolution satellite data indicates a detection limit of 0.055 mW cm-2μm-1sr-1in the normalized fluorescence line height (nFLH) data product, corresponding to around 2-3 mg m-3surface chlorophyll-aconcentration or about 300,000 cells L-1ofK. brevis. Although this detection limit is higher than those offered by traditional satellite sensors with higher SNRs, because of its frequent coverage of global nearshore waters at much higher spatial resolution, the SuperDove constellation is expected to play an essential role in monitoring HABs in nearshore environments to help mitigate the negative impacts of these blooms.

Freshwater cyanobacterial blooms pose a major threat to local ecosystems, economies, and public health. Monitoring these occurrences is essential for water resource managers worldwide. Satellite remote sensing techniques can detect and quantify blooms in large inland and estuarine water bodies but monitoring blooms in small water bodies (300 m) or low re-visit frequency (>10 days) of most operational satellite sensors. The ephemeral nature of cyanobacterial blooms and form dense surface mats (or ‘scums’) that aggregate nearshore further highlight the need for sensors with higher spatial and temporal resolutions. In this study, a deep learning model based on Convolutional Neural Network U-net was developed to detect cyanobacterial blooms (i.e., scums) in the highly modified and managed Caloosahatchee River (i.e., C-43 canal) and Caloosahatchee River Estuary (CRE) (Florida, USA) using Dove imagery (3-m resolution) obtained near-daily from the PlanetScope satellite constellation. The approach consisted of three steps: 1) training and validating the U-net model with “ground truth” images; 2) classifying bloom pixels; and 3) quantifying bloom area using linear unmixing. Validation results indicate an overall F1 score of 89.6% when assessing bloom area. Application of the model revealed the westward expansion of a cyanobacteria bloom from C-43 to CRE in summer 2018, indicating the physical transport of the bloom originating upstream in Lake Okeechobee to the estuary. This approach was tested on other inland water bodies, indicating potential for monitoring cyanobacterial blooms on a global scale.

Sargassumbeaching events have been reported in recent years around the Caribbean Sea and FL, USA, causing numerous environmental and economic problems. Satellite remote sensing has been widely used to monitorSargassumblooms in open waters, yet due to either coarse spatial resolution or low-revisit frequency, it is difficult to provide timely information onSargassuminundation from traditional satellite instruments. In this study, we demonstrate the capacity of 3-m resolution daily Dove imagery in monitoringSargassumbeaching events on Miami beach (FL, USA) and Cancun beach (Mexico). A U-net deep learning (DL) computer model is developed to extractSargassumfeatures from Dove imagery over beaches and nearshore waters. Application of the model to Dove image sequences between May and August 2019 shows two major inundation events on both Miami beach and Cancun beach, consistent with local reports. With the availability of 3-m resolution PlanetScope/Dove and PlanetScope/SuperDove data around the globe, the findings suggest that it is possible to monitor dynamic inundation events of not onlySargassumbut also other macroalgae in many other regions.

Numerous assumptions and approximations are employed when translating satellite-derived radiance to surface remote sensing reflectance (RRS) for ocean color applications. Among these is the vicarious calibration coefficient (g) of the ‘long’ near infrared band (NIRL) used for atmospheric correction. For this band, the pre-launch calibration has always been deemed sufficient [thus g(NIRL) = 1.00] as long as other bands are vicariously calibrated. Recent research, however, suggests that MODIS/AquaRRStime series are quite sensitive to g(NIRL) (and associated vicarious gains in other bands). In this work, we assessed the sensitivity of VIIRS/SNPPRRSto NIRLcalibration, and compared our results to previous MODIS/Aqua and SeaWiFS/OrbView2 analysis. In doing so, we noteg(NIRL) sensitivities of mission-averagedRRStimeseries are lower for VIIRS and SeaWiFS, relative to MODIS. At the scale of monthly climatologies, however, all sensors show prominentg(NIRL) sensitivity, with that of SeaWiFS being the most substantial. These findings informed simulation analyses, whereby we identified signal-to-noise ratio (SNR) and radiant path geometry, as well as their interaction, as having notable impacts ong(NIRL) sensitivity. As such,g(NIRL) sensitivity is a necessary consideration for reflectance uncertainty budgets, especially for sensors with higher NIR SNR or particular prevailing radiant path geometries. Given the geometry components embedded within g(NIRL) sensitivity, such studies should be coupled with cross-sensor intercalibrations (e.g., using simultaneous same view measurements) toward minimizing NIRLerrors between satellite instruments, but such efforts will not completely remediate remaining cross-sensor biases inRRS.

Satellite ocean color missions require accurate system vicarious calibrations (SVC) to retrieve the relatively small remote-sensing reflectance (Rrs, sr-1) from the at-sensor radiance. However, the current atmospheric correction and SVC procedures do not include calibration of the “long” near infrared band (NIRL-869 nm for MODIS), partially because earlier studies, based primarily on simulations, indicate that accuracy in the retrieved Rrsis insensitive to moderate changes in the NIRL vicarious gain (g). However, the sensitivity of ocean color data products to g(NIRL) has not been thoroughly examined. Here, we first derive 10 SVC “gain configurations” (vicarious gains for all visible and NIR bands) for MODIS/Aqua using current operational NASA protocols, each time assuming a different g(869). From these, we derive a suite of ~1.4E6 unique gain configurations with g(869) ranging from 0.85 to 1.2. All MODIS/A data for 25 locations within each of five ocean gyres were then processed using each of these gain configurations. Resultant time series show substantial variability in dominant Rrs(547) patterns in response to changes in g(869) (and associated gain configurations). Overall, mean Rrs(547) values generally decrease with increasing g(869), while the standard deviations around those means show gyre-specific minima for 0.97 <; g(869) <; 1.02. Following these sensitivity analyses, we assess the potential to resolve g(869) using such time series, finding g(869) = 1.025 most closely comports with expectations. This approach is broadly applicable to other ocean color sensors, and highlights the importance of rigorous cross-sensor calibration of the NIRL bands, with implications on consistency of merged-sensor data sets.

Nearly all calibration/validation activities for the satellite ocean color missions have focused on data quality to produce data products of the highest quality (i.e., science quality) for climate-related research. Little attention, however, has been paid to data quantity, particularly on how data quality control during data processing impacts downstream data quality and data quantity. In this letter, we attempt to fill this knowledge gap using measurements from the Visible Infrared Imaging Radiometer Suite (VIIRS) onboard the Suomi National Polar-orbiting Partnership (SNPP). For this sensor, the same level-1B data are processed independently using different quality control methods by NASA and NOAA, respectively, allowing for an in-depth evaluation of the interplay between data quantity and quality. The results indicate that the methods to identify stray light and sun glint are the two primary quality control procedures affecting data quantity, where the criteria for flagging pixels “contaminated” by stray light and sun glint may be relaxed in the NASA ocean color data processing to increase data quantity without compromising data quality.

The atmospheric correction approach currently being used operationally by NASA [termed as NASA standard atmospheric correction (NSAC) approach] to process ocean color data relies on traditional “black pixel” approach, with additional modifications to account for nonnegligible water-leaving radiance in the near-infrared (NIR) bands. The NSAC approach underestimates remote-sensing reflectance (Rrs, sr-1) in blue wavelengths in the presence of absorbing aerosols. Addressing this issue requires realistic absorbing-aerosol model and knowledge of the vertical distribution of aerosols, which are currently difficult to achieve. An alternative atmospheric correction approach has been evaluated in this paper for Moderate Resolution Imaging Spectroradiometer (MODIS) data. The approach is based on a previously developed spectra-matching optimization [POLYnomial-based approach established for the atmospheric correction of MERIS data (POLYMER)], where polynomial functions are used to express atmospheric contribution to the measured radiance and where a bio-optical model is used to estimate the water contribution. Evaluation against in situ data measured over the regions frequently affected by absorbing aerosols indicates that, compared with the NSAC approach, the POLYMER approach improves the Rrsretrievals in blue wavelengths while having a slightly worse performance in other wavelengths. Evaluation using NSAC-retrieved Rrs in adjacent days free of absorbing aerosols suggests that the POLYMER approach could improve the spectral shape and increase valid spatial coverage. When applied to time-series MODIS data, the POLYMER approach could generate more temporary coherent daily and monthly Rrspatterns than the NSAC approach. These results suggest that the POLYMER approach could be an alternative approach to partly correct for absorbing aerosols in the absence of explicit information on the aerosol type and the vertical distribution.

The Great Lakes contain 84% of the fresh water in North America and provide many critical and valuable ecosystem services to lakeside communities. In the past decade, a rapid ecosystem regime shift has occurred where changing precipitation and runoff patterns, along with invasive zebra and quagga mussels, have contributed to dramatic lake-specific responses in water clarity, reductions in phytoplankton, increases in toxic algal blooms, and disruptions in the food chain. Focusing on Lake Michigan, this study applies a newly improved remote sensing algorithm for water clarity and integrates meteorological data and in-situ sampling of mussels to understand the spatial and temporal linkages between historical weather patterns, water clarity and mussel abundance. Results from a decade of satellite-derived ocean color images reveal the complex spatio-temporal patterns of `bio-clarification' occurring in Lake Michigan. Mussel biomass was negatively correlated with phytoplankton concentrations and turbidity, but association was likely weakened by unmeasured variables such as nearshore runoff, patterns of plankton and nutrient dynamics, thermal structure and mixing.

Addressing critical earth science questions often requires time scales beyond the life of any single satellite sensor. Overlap between satellite-based datasets allows for the quantification of continuity (and discrepancies) between sensors. Toward that end, collocated matchups between Sea-viewing Wide Field-of-View Sensor (SeaWiFS), Moderate Resolution Imaging Spectroradiometer (MODIS), and Visible Infrared Imager Radiometer Suite (VIIRS) water clarity data from the Gulf of Mexico were analyzed at simultaneous, daily, and monthly time scales. Simultaneous data indicated strong agreement between sensors, with unbiased percent difference (UPD) generally less than 10% for both SeaWiFS/MODIS and VIIRS/MODIS matchups, with no apparent temporal trends. Spatially, UPD was highest near frontal boundaries and at high sensor zenith angles, while bias showed nearshore/offshore trends. UPD and bias statistics did not diminish for daily matchups; however, large degradation was seen for comparisons of monthly means between sensors, particularly SeaWiFS/MODIS matchups. Data coverage represented an important factor contributing to uncertainties in monthly mean data, as higher UPD was observed when fewer valid satellite measurements were recorded. Requiring a minimum of 15 samples per pixel per month minimizes the uncertainties in monthly mean products, with UPD between satellites roughly equivalent to that for simultaneous matchups. Overall, these findings demonstrate high consistency between three satellite instruments for most locations, while several “hot spots” of inconsistency are also revealed, which should be avoided in time-series studies. The findings also highlight the need to quantify uncertainties in often-used satellite products (particularly monthly mean composites) as well as the need to have a sufficient number of observations to assure the fidelity of monthly means.

Cloud contamination can lead to significant biases in sea surface temperature (SST) as estimated from satellite measurements. The effectiveness of four cloud detection algorithms for the Moderate Resolution Imaging Spectroradiometer (MODIS) in retaining valid SST data and masking cloud-contaminated data was assessed for all 2125 daytime and nighttime images during 2010 over the eastern Gulf of Mexico and including the east coast of Florida. None of the cloud detection algorithms was found to be sufficient to reliably differentiate clouds from valid SST, particularly during anomalously cold events. The strengths and weaknesses of each algorithm were identified, and a new hybrid cloud detection algorithm was developed to maximize valid data retention while excluding cloud-contaminated pixels. The hybrid algorithm was based on a decision tree, which includes a set of rules to use existing algorithms in different ways according to time and location. Comparing with >10000 concurrent in situ SST measurements from buoys, images processed with the hybrid algorithm showed increases in data capture and improved accuracy statistics over most existing algorithms. In particular, while keeping the same accuracy, the hybrid algorithm resulted in nearly 20% more SST retrievals than the most accurate algorithm (Quality SST) currently being used for operational processing. The increases in both data coverage and SST range should improve MODIS data products for more reliable SST retrievals in near real time, thus enhancing the ocean observing capacity to detect anomaly events and study short- and long-term SST changes in coastal environments.

Cloud filters developed for high-resolution (1-km) Advanced Very High Resolution Radiometer (AVHRR) satellite-derived sea surface temperature (SST) observations are generally inadequate to capture extreme cold events. Such events impacted shallow waters in Florida Bay and other coastal regions in January 2010 with fatal consequences for large numbers of corals and associated organisms. Raw AVHRR images were reprocessed to understand whether historical knowledge of daily and interannual SST variations could be used to derive a practical cloud-filtering technique. This approach, however, misidentified valid water temperature pixels in nearly 20% of 2703 images collected during the month of January for each year between 1995 and 2010. To create an improved SST climatology, this cloud-filtering method was combined with manually delineated overrides of falsely masked regions. During the January 2010 cold event, this climatology indicated negative SST anomalies of up to 11.6°C in the Big Bend region and 14°C in Florida Bay, with high spatial heterogeneity throughout. Our findings highlight the need for improved autonomous cloud-masking techniques to detect cold events in near real time.