The biodiverse and rapidly changing Greater Cape Floristic Region (GCFR) of southern Africa is outlined by coastal bays that receive dissolved organic matter (DOM) from rivers draining complex catchments composed of natural, agricultural, and urban land classes. As part of NASA's BioSCape field campaign (October-November 2023), we characterized the optical properties of three GCFR coastal bays (St. Helena, Walker, and Algoa) and four inland systems (Rietvlei wetland, Zeekoevlei lake, Theewaterskloof dam, and the Klein River estuary) in relation to DOM biogeochemistry and carbon cycling. Measurements of the optical properties of colored DOM (CDOM), including absorption at 300 nm (a(g)300), spectral slope (S275-295), and fluorescence, highlighted the bio-optical complexity associated with terrestrial influences, urban disturbances, intense biological activity, and rapid transformations along this dynamic coastline. CDOM in the coastal bays was characterized by an order of magnitude lower a(g)300 (0.5-2.8 m(-1)) and considerably higher S275-295 (0.020-0.031 nm(-1)) compared to upstream waters (25-71 m(-1) and 0.012-0.019 nm(-1), respectively), suggesting intense biological production and/or photochemical degradation. Coastal DOM was mostly (>75%) composed of protein-like compounds indicative of primary production, whereas inland DOM was mostly composed of terrigenous humic materials and had higher overall fluorescence signal. Satellite retrievals, using Sentinel-3 OLCI and Sentinel-2 MSI imagery, captured the relative influence of different rivers on coastal DOM dynamics and revealed that episodic events-extended periods of drought punctuated by heavy precipitation-are the primary drivers of biogeochemical variability along this globally significant, coastal biodiversity hotspot.

The Ocean Color Instrument (OCI) onboard PACE satellite, launched in February 2024, provides continuous hyperspectral measurements from 340-890 nm at similar to 2.5 nm resolution and nine shortwave infrared bands, with similar to 1 km spatial resolution. While OCI products have been evaluated over ocean and coastal waters, their performance in inland lakes remains unknown. Here, we assessed the capability of state-of-the-art algorithms to retrieve remote sensing reflectance (R-rs), phytoplankton absorption [a(ph)(lambda)], and pigment concentrations from OCI hyperspectral imagery. Overall, OCI-derived R-rs retrievals have perform unsatisfactory accuracy for these optically complex inland lakes, with ACOLITE-DSF (mean absolute percentage error [MAPE] > 25%) outperforming the NASA's MBAC algorithm (MAPE > 40%). Errors were lower in the red and green regions and increased toward the blue and near-infrared. Likewise, a(ph)(lambda) retrievals exhibit low accuracy (error > 35%) across all spectral wavelengths, with the Mixtured Density Network (MDN) model (MAPE > 35%) outperforming the semi-analytical Quasi Analytical Algorithm (QAA-750E) (MAPE > 70%). Errors for a(ph) (lambda) were generally lower in the red bands than in the blue and green bands. In addition, MDN-estimated chlorophyll-a and phycocyanin show MAPE values of 68.73% and 88.73%, respectively, which is superior to existing empirical algorithms on this dataset. Overall, while OCI's hyperspectral measurements show potential for resolving phytoplankton optical features in inland lakes, this capability is limited by the performance of current atmospheric correction (AC) and a(ph)-retrieval algorithms. Continued development of inland water-specific AC schemes and robust hyperspectral inversion algorithms is therefore essential.

Planet’s SuperDove (SD) sensors offer eight bands (seven visible, one near infrared (NIR)) at 3 m spatial and near-daily temporal resolution. The yellow (610 nm) and red-edge (705 nm) bands are valuable for retrieving water quality (WQ) parameters, supporting applications such as harmful algal bloom (HAB) and post-disaster monitoring. To enable scientific use, we assess signal-to-noise ratios (SNRs), with the highest (248:1) at 443 nm and the lowest (8:1) at 865 nm, and other visible bands ranging from 26:1–98:1. We cross-calibrated SD with Sentinel-2 Multi-Spectral Imager (MSI) using near-simultaneous observations over aquatic environments by comparing top-of-atmosphere (TOA; ρt ) reflectance across five shared visible bands (443, 490, 565, 665, and 705 nm), and derived calibration coefficients through linear regression. Before calibration, SD-MSI median ρt differences ranged from ∼0.7–13%, with highest differences at 705 nm. After applying the calibration, these differences reduced to −0.07% to −2.2%, including improvements at 665 nm (from ∼8% to −2.2%) and 705 nm (from ∼13% to −0.1%). Differences in atmospherically corrected remote sensing reflectance (Rrs ) also decreased from 16%–95% to 8%–72% post-calibration, with 565 nm showing the lowest (∼8%) and 705 nm the highest (∼72%) residual difference. Remaining Rrs discrepancies are attributed in part to SD’s inter-sensor differences and uncertainties in atmospheric correction. We qualitatively compared chlorophyll-a (Chla) and Secchi-disk depth (Zsd ) WQ products from SD and MSI, including a time-series analysis focused on the Dixie Fire and subsequent algal bloom in Lake Almanor (Sept–December 2021). The products captured expected trends, highlighting SD’s potential for WQ monitoring, while elevated uncertainties in ρt and Rrs suggest the need for improved calibration stability and atmospheric correction.

Ocean color remote sensing tracks water quality globally, but multispectral ocean color sensors often struggle with complex coastal and inland waters. Traditional models have difficulty capturing detailed relationships between remote sensing reflectance (Rrs), biogeochemical properties (BPs), and inherent optical properties (IOPs) in these complex water bodies. We developed a robust Mixture Density Network (MDN) model to retrieve 10 relevant biogeochemical and optical variables from heritage multispectral ocean color missions. These variables include chlorophyll-a (Chla) and total suspended solids (TSS), as well as the absorbing components of IOPs at their reference wavelengths. The heritage missions include the Moderate Resolution Imaging Spectroradiometer (MODIS) aboard Aqua and Terra, the Environmental Satellite (Envisat) Medium Resolution Imaging Spectrometer (MERIS), and the Visible Infrared Imaging Radiometer Suite (VIIRS) onboard the Suomi National Polar-orbiting Partnership (Suomi NPP). Our model is trained and tested on all available in situ spectra from an augmented version of the GLObal Reflectance community dataset for Imaging and optical sensing of Aquatic environments (GLORIA) (N = 9,956) after having added globally distributed in situ IOP measurements. Our model is validated on satellite match-ups corresponding to the SeaWiFS Bio-optical Archive and Storage System (SeaBASS) database. For both training and validation, the hyperspectral in situ radiometric and absorption datasets were resampled via the relative spectral response functions of MODIS, MERIS, and VIIRS to simulate the response of each multispectral ocean color mission. Using hold-out (80–20 split) and leave-one-out testing methods, the retrieved parameters exhibited variable uncertainty represented by the Median Symmetric Residual (MdSR) for each parameter and sensor combination. The median MdSR over all 10 variables for the hold-out testing method was 25.9%, 24.5%, and 28.9% for MODIS, MERIS, and VIIRS, respectively. TSS was the parameter with the highest MdSR for all three sensors (MODIS, VIIRS, and MERIS). The developed MDN was applied to satellite-derived Rrs products to practically validate their quality via the SeaBASS dataset. The median MdSR from all estimated variables for each sensor from the matchup analysis is 63.21% for MODIS/A, 63.15% for MODIS/T, 60.45% for MERIS, and 75.19% for VIIRS. We found that the MDN model is sensitive to the instrument noise and uncertainties from atmospheric correction present in multispectral satellite-derived Rrs. The overall performance of the MDN model presented here was also analyzed qualitatively for near-simultaneous images of MODIS/A and VIIRS as well as MODIS/T and MERIS to understand and demonstrate the product resemblance and discrepancies in retrieved variables. The developed MDN is shown to be capable of robustly retrieving 10 water quality variables for monitoring coastal and inland waters from multiple multispectral satellite sensors (MODIS, MERIS, and VIIRS).

Monitoring the health of essential freshwater and coastal ecosystems like lakes and estuaries requires estimating water quality indicators (WQIs) like chlorophyll-a from remote sensing data. Machine learning (ML)-based regression tools are a popular choice for this estimation task. Training such models requires a labeled dataset comprising of co-located measurements of remote sensing variables and WQIs. To address this need, practitioners primarily use co-located in situ measurements of remote sensing reflectance and WQIs to train ML models. While these ML models outperform other analytical/empirical methods for the chlorophyll-a estimation tasks on labeled in situ datasets, their performance deteriorates quite significantly when applied to satellite datasets. Using satellite-based training datasets is infeasible given the paucity of co-located measurements of satellite-derived variables and WQIs, especially for newer sensors. To address the discrepancy between the training (in situ) datasets and application (satellite) datasets, we will leverage domain adaptation (DA), i.e., an ML trick to transfer knowledge from a source (in situ) domain with abundant labeled examples to a related target domain with little or no labeled data (satellite). To demonstrate the effectiveness of this approach, we implement Domain Adaptation for REgression by aligning inverse GRAM matrices (DARE-GRAM) for the chlorophyll-a estimation task. This regression-specific domain adaptation technique uses the labeled source (in situ) data and unlabeled satellite data (i.e., only the remote sensing reflectance) to build a model with improved chlorophyll-a estimation for satellite datasets. We test this approach using data from two popular multispectral sensors, namely, Sentinel-2 MultiSpectral Instrument (MSI) and Sentinel3 Ocean and Land Colour Imager (OLCI). The DARE-GRAM shows improvements of between 20-50% across a wide range of regression metrics, indicating a comprehensive improvement in the estimation performance relative to classical ML methods. The DARE-GRAM method also shows significantly improved agreement for chlorophyll-a estimates from near-simultaneous acquisitions from the different sensors. This improved harmony is expected to be an important step in creating consistent crosssensor products.

Constructing a robust ocean color (OC) record(e.g., water transparency, phytoplankton absorption) for long-term assessments of coastal and inland water ecosystems from past, present, and future missions requires high-quality spectral remote sensing reflectance (R-rs) products. Using the GLORIA dataset (Lehmann et al., 2023), we evaluated the quality of R-rs products from the moderate resolution imaging spectroradiometer (MODIS on Terra and Aqua), medium resolution imaging spectrometer (MERIS), and visible infrared imaging radiometer suite (VIIRS) processed via the two-band heritage atmospheric correction method (a combination of near-infrared and shortwave infrared bands) available in the Sea WiFS Analysis Data Analysis System (SeaDAS). Overall, retrieval residuals are consistent within a few percentages among the four missions. Median residuals ranged from similar to 20% in the similar to 550-nm band to>60% in the similar to 412-nm bands. Spectrally averaged root mean squared differences for all the missions were similar to 0.0024 sr(-1)with one standard deviation of similar to 0.001 sr(-1). The corresponding(median) biases in the visible bands varied from-60% to-3%,with the largest biases identified in MERIS and VIIRS products. Despite the lower sensitivity of band-ratio algorithms to residuals in specific spectral regions [e.g., OC3 chlorophyll-a algorithm is less prone to residuals in R-rs(lambda>600 nm)], other algorithms or downstream products that leverage all the visible bands are highly compromised. We underscore the need to improve the quality of Rrsproducts, thereby enabling the reconstruction of baseline OC products of high caliber in global coastal and inland waters that are often near human activity.

Secchi Disk Depth (Z(sd)) is one of the most fundamental and widely used water-quality indicators quantifiable via optical remote sensing. Despite decades of research, development, and demonstrations, currently, there is no operational model that enables the retrieval of Z(sd) from the rich archive of Landsat, the long-standing civilian Earth-observation program (1972 - present). Devising a robust Z(sd) model requires a comprehensive in situ dataset for testing and validation, enabling consistent mapping across optically varying global aquatic ecosystems. This study utilizes Mixture Density Networks (MDNs) trained with a large in situ dataset (N = 5689) from 300+ water bodies to formulate and implement a global Z(sd) algorithm for Landsat sensors, including the Thematic Mapper (TM), Enhanced Thematic Mapper Plus (ETM+), and Operational Land Imager (OLI) aboard Landsat-5, -7, -8, and -9, respectively. Through an extensive Monte Carlo cross-validation with in situ data, we showed that MDNs improved Z(sd) retrieval when compared to other commonly used machine-learning (ML) models and recently developed semi-analytical algorithms, achieving a median symmetric accuracy (epsilon) of similar to 29% and median bias (beta) of similar to 3%). A fully trained MDN model was then applied to atmospherically corrected Landsat data (i.e., remote sensing reflectance; R-rs) to both further validate our MDN-estimated Z(sd) products using an independent global satellite-to-in situ matchup dataset (N = 3534) and to demonstrate their utility in time-series analyses (1984 - present) via selected lakes and coastal estuaries. The quality of R-rs products rigorously assessed for the Landsat sensors indicated sensor-/band-dependent epsilon ranging from 8% to 37%. For our Z(sd) products, we found epsilon similar to 39% and beta similar to 8% for the Landsat-8/OLI matchups. We observed higher errors and biases for TM and ETM+, which are explained by uncertainties in R-rs products induced by uncertainties in atmospheric correction and instrument calibration. Once these sources of uncertainty are, to the extent possible, characterized and accounted for, our developed model can then be employed to evaluate long-term trends in water transparency across unprecedented spatiotemporal scales, particularly in poorly studied regions of the world in a consistent manner.

With an identical design and build, the Operational Land Imager-2 (OLI2) aboard Landsat-9 (L9) complements OLI observations by reducing the global revisit rate of Landsat to 8 days. This study takes advantage of near-coincident OLI2 and OLI observations obtained on 11-17 November 2021 to assess the relative quality of the standard United States Geological Survey (USGS) top-of-atmosphere (TOA) reflectance (rho(t)) and atmospherically corrected reflectance (aquatic reflectance; rho(AR)(w)) products over bodies of water. The TOA assessment was carried out for all the visible bands, including the panchromatic band, as well as the near-infrared (NIR) and shortwave infrared (SWIR) bands, whereas rho(AR)(w) products were analyzed in the 443, 482, 561, and 655 nm bands. The overlapping areas of OLI-OLI2 rho(w) product pairs were further analyzed for the rigor in corrections for the viewing geometry implemented in selected atmospheric correction (AC) processors, including SeaDAS, ACOLITE, and POLYMER, with products denoted as rho(AR)(w), rho(ac)(w), and rho(pol)(w) , respectively. Overall, we found the OLI2-OLI rho(t) products to be consistent within 0.4% in the visible-near-infrared (VNIR) bands except in the green band (561 nm), where OLI2 records similar to 0.8% larger values than OLI. The two SWIR bands (1610 and 2200 nm) were also found to agree within similar to 2.2% and 2.1%, respectively, with OLI2 being lower in magnitude. The median differences in the standard rho(w) (rho(AR)(w)) were estimated to be similar to 2.4%, 1.5%, 2.3%, and 2.5% in the 443, 482, 561, and 655 nm bands, respectively, which are well within the accepted differences in cross-mission data merging schemes. Further, we show that OLI2's signal-to-noise ratio (SNR) is 7-30% higher than that of OLI in all the bands, likely due to its 14-bit digitization rate, as compared to OLI's 12-bit digitization rate. Our self-consistency assessment of the AC processors for handling the differences in the view zenith angles (Delta VZA) and relative azimuth angles (Delta RAA) of OLI2 and OLI observations suggests that, overall, the processors account for the Sun-sensor geometry differently at different spectral bands. These differences (inconsistencies) amount to average median absolute percentage differences (MAPD) in rho(w) from 0.3 to 5.3% (e.g.,Delta rho(AR)(w) (443 nm, Delta VZA, Delta RAA) = rho(AR,OLI)(w) (443, +6.3,10) - rho(AR,OLI2)(w) (443, +3.9,-13) < 0.5%).More specifically, we found that SeaDAS provides the most optimal corrections for the angular variability as a function of VZA, rho(ac)(w) are most consistent for different ranges of RAAs, and POLYMER performs best in the 443 nm band. We surmise that future (and other existing) AC methods should be tested with Landsat-8/-9 underfly imagery to quantify their performance for tackling the variability in imaging geometry and minimize associated uncertainties. With several missions planned for launch by the end of this decade, it is further emphasized that post-launch tandem maneuvers are essential to creating harmonized multimission data products and, therefore, similar operations should be considered and extended to ensure a broad range of environmental conditions are captured for comprehensive cross-mission analyses.

Lake Erie, the shallowest of the five North American Laurentian Great Lakes, exhibits degraded water quality associated with recurrent phytoplankton blooms. Optical remote sensing of these optically complex inland waters is challenging due to the uncertainties stemming from atmospheric correction (AC) procedures. In this study, the accuracy of remote sensing reflectance (Rrs) derived from three different AC algorithms applied to Ocean and Land Colour Instrument (OLCI) observations of western Lake Erie (WLE) is evaluated through comparison to a regional radiometric dataset. The effects of uncertainties in Rrs products on the retrieval of near-surface concentration of pigments, including chlorophyll-a (Chla) and phycocyanin (PC), from Mixture Density Networks (MDNs) are subsequently investigated. Results show that iCOR contained the fewest number of processed (unflagged) days per pixel, compared to ACOLITE and POLYMER, for parts of the lake. Limiting results to the matchup dataset in common between the three AC algorithms shows that iCOR and ACOLITE performed closely at 665 nm, while outperforming POLYMER, with the Median Symmetric Accuracy (MdSA) of -30 %, 28 %, and 53 %, respectively. MDN applied to iCOR- and ACOLITE-corrected data (MdSA < 37 %) outperformed MDN applied to POLYMER-corrected data in estimating Chla. Large uncertainties in satellite-derived Rrs propagated to uncertainties -100 % in PC estimates, although the model was able to recover concentrations along the 1:1 line. Despite the need for improvements in its cloud-masking scheme, we conclude that iCOR combined with MDNs produces adequate OLCI pigment products for studying and monitoring Chla across WLE.(c) 2022 The Authors. Published by Elsevier B.V. on behalf of International Association for Great Lakes Research. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/ licenses/by-nc-nd/4.0/).

Cyanobacteria harmful algal blooms (cyanoHABs) present a critical public health challenge for aquatic resource and public health managers. Satellite remote sensing is well-positioned to aid in the identification and mapping of cyanoHABs and their dynamics, giving freshwater resource managers a tool for both rapid and long-term protection of public health. Monitoring cyanoHABs in lakes and reservoirs with remote sensing requires robust processing techniques for generating accurate and consistent products across local and global scales at high revisit rates. We leveraged the high spatial and temporal resolution chlorophyll-a (Chla) and phycocyanin (PC) maps from two multispectral satellite sensors, the Sentinel-2 (S2) MultiSpectral Instrument (MSI) and the Sentinel-3 (S3) Ocean Land Colour Instrument (OLCI) respectively, to study bloom dynamics in Utah Lake, United States, for 2018. We used established Mixture Density Networks (MDNs) to map Chla from MSI and train new MDNs for PC retrieval from OLCI, using the same architecture and training dataset previously proven for PC retrieval from hyperspectral imagery. Our assessment suggests lower median uncertainties and biases (i.e., 42% and -4%, respectively) than that of existing top-performing PC algorithms. Additionally, we compared bloom trends in MDN-based PC and Chla products to those from a satellite-derived cyanobacteria cell density estimator, the cyanobacteria index (CI-cyano), to evaluate their utility in the context of public health risk management. Our comprehensive analyses indicate increased spatiotemporal coherence of bloom magnitude, frequency, occurrence, and extent of MDN-based maps compared to CI-cyano and potential for use in cyanoHAB monitoring for public health and aquatic resource managers.

Urban reservoirs are important for drinking water services and urban living. However, potentially toxic cyanobacteria blooms are frequently present due to human pollution and might threaten the urban water supply. Conveniently, cyanobacteria can be monitored by remote sensing-based approaches based on the spectral features of C-Phycocyanin (PC). Furthermore, methods leveraging Machine Learning Algorithms (MLA) for PC estimation from hyperspectral data have highlighted the potential to estimate PC more accurately - even at low concentrations. Since relatively few methodologies for PC retrieval in tropical environments have been developed or validated, this research evaluated PRISMA hyperspectral data processed with three MLA (Random Forest, Extreme Gradient Boost, and Support Vector Machines) to estimate PC concentrations in the Billings reservoir, Brazil. The same MLA were used to generate PC models using Wordview-3 and Landsat-8/OLI simulated data to assess the potential gain of using hyperspectral over multispectral data. A PRISMA image was processed with three atmospheric correction methods and validated with co-located in-situ data, where the best atmospherically corrected product was used to generate synthetic Landsat-8/OLI and Worldview-3 images. The PC models were calibrated and validated through Monte Carlo simulation using field radiometric and biological data (Chlorophyll-a, PC, and phytoplankton taxonomy) collected in eight field campaigns (N = 115). The PRISMA and the synthetic multispectral images were used for a second round of models' validation using colocated PC measurements (match-up window +/- 4 h). The global PC Mixture Density Network was also applied to the PRISMA data, and the estimates were compared with the other MLA. The results showed that the standard PRISMA surface reflectance product provided the best atmospheric correction (MAE < 20% for the 500-700 nm bands), while ACOLITE and 6SV underperformed it from two to more than ten-fold. Cyanobacteria species were abundant in 96% of the taxonomical samples, even though relatively low PC concentrations were found (PC from 0 to 301.81 mu g/L and median PC = 2.9 mu g/L). The global Mixture Density Network sharply overestimated PC (MAE = 280% and Bias = 280%), potentially due to Billings reservoir's low PC:Chlorophyll-a ratio relative to the original training dataset. PRISMA/Random Forest (MAE = 45%) achieved the lowest error for orbital PC estimate, while Extreme Gradient Boost outperformed the other MLA using Worldview-3 (MAE = 49%) and Landsat-8 (MAE = 74%) synthetic imagery. Therefore, the results suggest hyperspectral and multispectral orbital data aligned with MLA are feasible for monitoring PC, even for waters containing low PC concentrations and reduced PC:Chlorophyll-a ratios.

The simultaneous remote estimation of biogeochemical parameters (BPs) and inherent optical properties (IOPs) from hyperspectral satellite imagery of globally distributed optically distinct inland and coastal waters is a complex, unsolved, non-unique inverse problem. To tackle this problem, we leverage a machine-learning model termed Mixture Density Networks (MDNs). MDNs outperform operational algorithms by calculating the covariance between the simultaneously estimated products. We train the MDNs on a large (N = 8237) dataset of co-aligned, in situ measured, hyperspectral remote sensing reflectance (R-rs), BPs, and absorbing IOPs from globally representative optically distinct inland and coastal waters. The estimated IOPs include absorption due to phytoplankton (a(ph)), chromophoric dissolved organic matter (a(cdom)), and non-algal particles (a(nap)). The estimated BPs include chlorophyll-a, total suspended solids, and phycocyanin (PC). MDNs dramatically reduce uncertainty in the retrievals, relative to operational algorithms, when using a 50/50 dataset split, where the MDNs are trained on a randomly selected half of the in situ dataset and validated on the other half. Our model is shown to have higher, or equivalent, generalization performance than the calculated operational algorithms available for all BPs and IOPs (except PC) via a leave-one-out cross-validation assessment. The MDNs are sensitive to uncertainties in the hyperspectral satellite R-rs, resulting from instrument noise and atmospheric correction; there is a difference of similar to 37.4-62.8% (using median symmetric accuracy) between the MDNs' estimates derived from co-located satellite-derived R-rs and in situ R-rs. Of the IOPs, acdom and anap are less sensitive to uncertainties in hyperspectral satellite imagery relative to aph, with remote estimates of a(ph) exhibiting incorrect spectral shape and magnitude relative to in situ measured IOPs. Despite the uncertainties in satellite derived R-rs, the spatial distributions of BPs and IOPs in MDN-derived product maps of Lake Erie and the Curonian Lagoon, based on imagery taken with the Hyperspectral Imager for the Coastal Ocean (HICO) and PRecursore IperSpettrale della Missione Applicativa (PRISMA), are confirmed via co-aligned in situ measurements and agree with the literature's understanding of these well-studied regions. The consistency and accuracy of the model on HICO and PRISMA imagery, despite radiometric uncertainties, demonstrate its applicability to future hyperspectral missions, such as the Plankton, Aerosol, Cloud, ocean Ecosystem (PACE) mission, where the simultaneous estimation model will serve as a key part of phytoplankton community composition analysis.

The satellite ocean color remote sensing paradigm developed by government space agencies enables the assessment of ocean color products on global scales at kilometer resolutions.A similar paradigm has not yet been developed for regional scales at sub-meter resolutions, but it is essential for specific ocean color applications (e.g., mapping algal biomass in the marginal ice zone).While many aspects of the satellite ocean color remote sensing paradigm are applicable to sub-meter scales, steps within the paradigm must be adapted to the optical character of the ocean at these scales and the opto-electronics of the available sensing instruments.This dissertation adapts the three steps of the satellite ocean color remote sensing paradigm that benefit the most from reassessment at sub-meter scales, namely the correction for surface-reflected light, the design and selection of the opto-electronics, and the post-processing of over-sampled regions.First, I identify which surface-reflected light removal algorithm and view angle combination are optimal at sub-meter scales, using data collected during a field deployment to the Martha's Vineyard Coastal Observatory.I find that of the three most widely used glint correction algorithms, a spectral optimization based approach applied to measurements with a 40 ∘ view angle best recovers the remotesensing reflectance and chlorophyll concentration despite centimeter scale variability in the surface-reflected light.Second, I develop a simulation framework to assess the impact of higher optical and electronics noise on ocean color product retrieval from unique ocean color scenarios.I demonstrate the framework's power as a design tool by identifying hardware limitations, and developing potential solutions, for estimating algal biomass from high dynamic range sensing in the marginal ice zone.Third, I investigate a spectral super-resolution technique for application to spatially over-sampled oceanic regions.I determine that this technique more accurately represents spectral frequencies beyond the Nyquist and that it can be trained to be invariant to noise sources characteristic of ocean color remote sensing on images with similar statistics as the training dataset.Overall, the developed and critically assessed sub-meter ocean color remote sensing paradigm enables researchers to collect high fidelity sub-meter data from imaging spectrometers in unique ocean color scenarios.

Retrieval of the phycocyanin concentration (PC), a characteristic pigment of, and proxy for, cyanobacteria biomass, from hyperspectral satellite remote sensing measurements is challenging due to uncertainties in the remote sensing reflectance (∆Rrs) resulting from atmospheric correction and instrument radiometric noise. Although several individual algorithms have been proven to capture local variations in cyanobacteria biomass in specific regions, their performance has not been assessed on hyperspectral images from satellite sensors. Our work leverages a machine-learning model, Mixture Density Networks (MDNs), trained on a large (N = 939) dataset of collocated in situ chlorophyll-a concentrations (Chla), PCs, and remote sensing reflectance (Rrs) measurements to estimate PC from all relevant spectral bands. The performance of the developed model is demonstrated via PC maps produced from select images of the Hyperspectral Imager for the Coastal Ocean (HICO) and Italian Space Agency's PRecursore IperSpettrale della Missione Applicativa (PRISMA) using a matchup dataset. As input to the MDN, we incorporate a combination of widely used band ratios (BRs) and line heights (LHs) taken from existing multispectral algorithms, that have been proven for both Chla and PC estimation, as well as novel BRs and LHs to increase the overall cyanobacteria biomass estimation accuracy and reduce the sensitivity to ∆Rrs. When trained on a random half of the dataset, the MDN achieves uncertainties of 44.3%, which is less than half of the uncertainties of all viable optimized multispectral PC algorithms. The MDN is notably better than multispectral algorithms at preventing overestimation on low (<10 mg m−3) PC. Visibly, HICO and PRISMA PC maps show the wider dynamic range that can be represented by the MDN. The available in situ and satellite-derived Rrs matchups and measured in situ PC demonstrate the robustness of the MDN for estimating low (<10 mg m−3) PC and the reduced impact of ∆Rrs on medium-to-high in situ PC (>10 mg m−3). According to our extensive assessments, the developed model is anticipated to enable practical PC products from PRISMA and HICO, therefore the model is promising for planned hyperspectral missions, such as the Plankton Aerosol and Cloud Ecosystem (PACE). This advancement will enhance the complementary roles of hyperspectral radiometry from satellite and low-altitude platforms for quantifying and monitoring cyanobacteria harmful algal blooms at both large and local spatial scales.

Low-power, lightweight, off-the-shelf imaging spectrometers, deployed on above-water fixed platforms or on low-altitude aerial drones, have significant potential for enabling fine-scale assessment of radiometrically derived water quality properties (WQPs) in oceans, lakes, and reservoirs. In such applications, it is essential that the measured water-leaving spectral radiances be corrected for surface-reflected light, i.e., glint. However, noise and spectral characteristics of these imagers, and environmental sources of fine-scale radiometric variability such as capillary waves, complicate the glint correction problem. Despite having a low signal-to-noise ratio, a representative lightweight imaging spectrometer provided accurate radiometric estimates of chlorophyll concentration—an informative WQP—from glint-corrected hyperspectral radiances in a fixed-platform application in a coastal ocean region. Optimal glint correction was provided by a spectral optimization algorithm, which outperformed both a hardware solution utilizing a polarizer and a subtractive algorithm incorporating the reflectance measured in the near infrared. In the same coastal region, this spectral optimization approach also provided the best glint correction for radiometric estimates of backscatter at 650 nm, a WQP indicative of suspended particle load.

Optical remote sensing of aquatic environments using aerial drones is becoming more feasible as lightweight, low-power, spectral cameras increase in availability. Use of these cameras in such applications involves complex trade-offs in optical design and in deployment strategies, and simulations provide a means to examine this multidimensional design space to identify specific limitations on performance for a given measurement scenario. In this paper, such a simulation framework is developed, and its use in two realistic aquatic remote sensing scenarios is explored. Such a framework can provide insight into not only uses of existing camera systems, but also aspects of optical design or hardware that would lead to improved accuracy when using such cameras aerially over natural water bodies.

We assessed the signal-to-noise ratio (SNR) of a drone-deployable hyperspectral camera for water quality parameter estimation. We also developed a model to assess the SNR of comparable sensors in a range of sensing scenarios.