This PDF file contains the front matter associated with SPIE Proceedings Volume 11752, including the Title Page, Copyright information, and Table of Contents.

NOAA provides sea surface temperature (SST) products from multiple Earth observing satellites in low Earth orbits (LEO) using its Advanced Clear-Sky Processor for Ocean (ACSPO) system. Historically, ACSPO SST products from individual LEO platforms have been provided as 10-minute granules (144/day) in L2P (swath) and 0.02° L3U (gridded uncollated) formats. With the large, and increasing number of LEO sensors currently in orbit (two VIIRSs onboard NPP/N20, three AVHRRs onboard METOP-A/B/C and two MODISs onboard Aqua/Terra) and soon to be launched (N21/VIIRS and Metop-Second Generation METImage), the data volumes and number of files has grown dramatically and is now challenging to manage by an average user. Moreover, data from different sensors and overpasses may not be fully consistent. In response to multiple users’ requests, the NOAA SST team has developed the 0.02° gridded super-collated (L3S) line of LEO SST products, which collate L3U data from individual sensors into a multi-sensor products with higher information density, lower data volume consistent datasets. The L3S-LEO line comprises two products: from the afternoon (‘PM’) orbits (currently, two VIIRSs onboard NPP and N20) and from the mid-morning (‘AM’) orbits (currently, three AVHRR FRACs onboard Metop-A/B/C). Both products are reported twice daily, one nighttime and one daytime file, resulting in four files every 24 hours. The data are validated in the NOAA SST Quality Monitor (SQUAM) online system, and distributed to users via the CoastWatch service, in near real time. This work describes recent L3S-LEO algorithm developments, aimed at the reduced impact of cloud leakages from individual sensor L3U data, and improved SST imagery. We also present initial checks of the diurnal cycle in the L3S-LEO vs. GEO SST from the Advanced Baseline Imager (ABI) flown onboard GOES-16, and find the two datasets largely consistent.

In response to users’ needs, NOAA is developing a suite of multi-sensor gridded (0.02º resolution) super-collated (L3S) Advanced Clear Sky Processor for Ocean (ACSPO) SST products, which reduce number of files and data volume, while retaining high spatial and temporal resolution of the data. In 2020, two ACSPO L3S-LEO products were initiated from Low Earth Orbit satellites. Our next priority is the development of an hourly super-collated product generated from geostationary satellites (L3S-GEO). This paper describes the initial implementation of the L3S-GEO product, including end-to-end data production and initial evaluation, including consistency of satellite-specific GEO products where they overlap.

The first full-mission AVHRR FRAC SST dataset with nominal 1.1km resolution at nadir was created from three Metop First Generation (FG) satellites: A (2006-present), B (2012-present) and C (2018-present), using the NOAA Advanced Clear-Sky Processor for Ocean (ACSPO) enterprise system. Historical reprocessing (Reanalysis-1, RAN1) starts at the beginning of each mission and continues into near-real time (NRT) processing. ACSPO generates two SST products: Global Regression (GR) SST, highly sensitive to the skin SST, and Piecewise Regression (PWR) SST, a proxy for the depth SST. The effect of orbital and sensor drift on the stability of the SST time series is mitigated by retraining the regression coefficients daily against matchups with the drifting and tropical moored buoys. Those matchups are collected within moving windows: 91-day for GR and 361-day for PWR, with the offsets adjusted within a 31 day window. In RAN1, all training and offset correction windows are centered at the processed day. In NRT processing, the training and offset delayed windows of the same sizes and ending in 4-10 days prior to the processed day are used. This mitigates longterm calibration trends on scales from 1-2 months in both RAN1 and NRT. Short-term variations in SST biases in NRT are higher than in RAN1 but do not exceed ~0.05 K. Delayed-mode RAN processing follows the NRT with a lag of ~2 months, resulting in higher quality, more consistent Metop-FG SST record. The presentation evaluates the performance of the ACSPO AVHRR FRAC dataset and compares it with the EUMETSAT OSISAF Metop-A and -B FRAC SSTs available in PO.DAAC.

Global long-term SST record is being created under the AVHRR GAC Reanalysis (RANs) project, by historical reprocessing 4 km data of the AVHRR/2 and /3 instruments flown onboard multiple NOAA satellites from 1981 – present. The AVHRR data are reprocessed with the NOAA Advanced Clear-Sky Processor for Oceans (ACSPO) system. During RANs, the ACSPO algorithms are being adjusted to mitigate various issues intrinsic to AVHRR sensors, especially AVHRR/s on the earlier NOAA missions in 1980s and 1990s. The focus of the latest interim release of RAN2, Beta 02, is the contamination of retrieved SSTs with massive cold biases, originating from two sources. First, in all AVHRR missions, periodic cold SST outliers occur at night due to solar impingement on the black body calibration target, when the satellite orbit approaches the terminator. Second, multiple cold outliers appear in the NOAA-7, -11 and -12 SSTs following three major volcanic eruptions of Mt. El Chichon (1982), Mt. Pinatubo (1991) and Mt. Hudson (1991). The current mitigation algorithm exploits the fact that in both cases, the spatial densities of the cold outliers exhibit well-expressed latitudinal dependencies. The algorithm identifies 5° latitudinal bands with abnormally high density of outliers and makes the cloud mask more conservative within those bands. This improves filtering cold SST outliers in the contaminated areas without increasing the false cloud detection rate in the unaffected parts of the ocean. We also discuss the ongoing development to mitigate cold SST biases by correcting AVHRR L1b calibration (rather than eliminating the affected SST data).

The potential for hydrodynamic drag reduction using small-scale actuation to influence turbulent boundary layer flow is investigated. When coupled with lab data, numerical simulations allow for the most complete understanding of the effect of various parameters on boundary layer turbulence. Two computational fluid dynamics (CFD) model sets are created in COMSOL, a multiphysics software package, to compare to experimental data and inform future research in the flowSiTTE laminar-to-turbulent flow tank. Effects on boundary layer dynamics and turbulence are studied through the implementation of an actuated boundary along the tank lid, using a standing wave to counter shear stress streaks characteristic of fully developed boundary layer turbulence. Numerical velocity data from the tank lid model are consistent with laboratory data collected with Particle Image Velocimetry (PIV). Additionally, boundary layer actuation of transitional turbulence is studied by analyzing the formation and dampening of Tollmien-Schlichting (TS) waves on a NACA 0012 airfoil. Standing wave actuation with amplitudes of 10-100μm along the airfoil is shown to reduce hydrodynamic drag by up to 15% at a wide range of frequencies. Boundary actuation delays the formation of the separation layer along the airfoil’s trailing edge – the region of flow responsible for much of the airfoil’s drag – keeping the flow attached for nearly the entire chord length and significantly reducing pressure drag.

Recent advances in technology, in particular soft robotics and micro-electronics, have renewed the interested in the impact of viscoelastic boundaries and active boundary modulation on hydrodynamic drag and boundary layer turbulence. Viscoelastic boundary materials, such as those found in dolphin skin, are known to have the potential to reduce boundary drag, by delaying the transition from laminar to turbulent flow in the boundary layer around the body and minimizing boundary layer turbulence. The possible mechanisms to reduce boundary layer turbulence include counteracting boundary layer coherent structures or impacting momentum transfer near the boundary. Actuating a deformable membrane in a channel flow allows the investigation of the impact of boundary actuation on boundary layer turbulence for a range of actuation parameters and flow channel speeds. We developed a deformable boundary and tested the system in channel flow, in direct contact with the water, actuating at various wave patterns and frequencies. The impact on boundary layer velocity was investigated with Particle Image Velocimetry, as well as numerical simulations (see companion paper). Boundary actuation is shown to impact the boundary layer velocity profile and near boundary momentum transfer. We characterize the parameter space most likely to reduce boundary layer turbulence in a natural environment, which could lead to more energy-efficient platforms and underwater vehicles.

This effort designed and tested new algorithms and deployable scintillometer hardware for ocean optical turbulence characterization. Novel features include a hand-deployable design, a non-laser optical source, a rapidly adjustable propagation length, and a collocated multi-instrument environmental sensor package. Undersea testing was contingent on several accomplishments, including developing robust algorithms and data logging methods, integrating compact optics and electronics, and engineering handheld-sized pressure vessels suitable for field experimentation. The test assembly was deployed in 428-m Pacific Ocean water from a small boat. Direct measurements revealed the ocean’s refractive-index structure parameter (C_n² from 1.9×10⁻¹¹ m^−2/3 to 2.3×10⁻¹⁰ m^−2/3) and the inner scale of optical turbulence (l₀ from 0.5 mm to 1.5 mm). Onboard temperature, depth, beam attenuation, and backscattering sensors corroborated key regions of interest, namely the thermocline. By integrating turbulence, temperature, depth, attenuation, and backscattering measurements within a single hand-portable assembly, we increased our understanding of ocean optical dynamics while demonstrating the practicality of a low size, weight, and power scintillometer.

Current Ocean Color (OC) algorithms for the estimation of chlorophyll-a concentration (Chla) from satellite imagery, which are based primarily on 440, 490 and 550 nm bands, work well in the open ocean areas and often break in coastal waters because of failure of the atmospheric correction in blue bands and water complexity. A recently developed neural network (NN) algorithm for VIIRS avoids blue bands utilizing 490, 550 and 671 nm bands. This algorithm was proven to work well to detect Karenia brevis algal blooms along the West Florida Shelf even near the coast and to determine Chla in many other areas but is limited to Chla below 10-15 mg/m³. Algorithms that are based on the red-NIR bands, work well at high Chla for the near surface measurements. However, they require a band near 709 nm which is available only on the Sentinel-3 OLCI satellite sensor, and the current OLCI atmospheric correction is not reliable enough at this band. To detect algal blooms in very complex areas like the Chesapeake Bay, a new approach is presented, which avoids blue bands but includes data from the VIIRS I1 imaging band 600 – 680 nm with a center at 638 nm. Remote sensing reflectance for this I1 band is now routinely included in the standard set of processed VIIRS bands at NOAA CoastWatch with 750 m spatial resolution. This band integrates part of the remote sensing reflectance spectrum with variable phytoplankton absorption together with a complex combination of spectra from other water components and is helpful in the estimation of Chla in a broad range. Results are compared with the performance of the OLCI red-NIR algorithm with an alternative atmospheric correction, showing good qualitative agreements in the Chesapeake Bay area in conditions from clear waters to algal blooms.

Traditional seawater chemical oxygen demand (COD) monitoring methods are based on manual operations, which require various operating procedures and long duration of measurement, are prone to secondary pollution and hence unsuitable for in-situ monitoring. In this paper, we developed a prototype of in-situ seawater COD monitoring sensor based on UV-Vis absorption spectroscopy, and integrating it to a buoy for coastal trials. During the trials, several measures were applied to reduce the influence of biofouling, including coating sensor housing with an environment-friendly anti-fouling paint, and designing a motorized underwater wiper for optical window cleaning. The in-situ COD sensor had been continuously working underwater for more than 6 months, obtaining 1536 sets of seawater UV-Vis absorption sectrum.

We developed two underwater fluorimeters (VIS&UV) for in-situ assessments of aquatic fluorescence constituents. Two prototypes had been developed to assess chlorophyll a and BOD₅, respectively, and were deployed under a buoy platform for long-term field tests. Design considerations include exciting light use efficiency, weak fluorescence signal detection, ambient light suppression, corrosion resistance and anti-biofouling. The prototypes demonstrated excellent linearity in response to fluorescence emissions in laboratory calibrations and good environment suitability during the field tests. We had obtained a large amount of observational data and maintenance experience.

In this work we investigate the use of pattern classification algorithms to enhance detection performance of the underwater radar-encoded laser system. A challenge encountered with this system is the automatic detection of the return from an underwater object in highly-scattering and/or low signal-to-noise ratio (SNR) conditions. Previous efforts were largely based on threshold detection and result in detection errors in such challenging conditions. Other efforts attempt to use signal processing to remove scatter returns, but this does not address low SNR cases. We take a different approach here, investigating the use of machine learning to develop classifiers which combine various shape and statistical features to discriminate between object and non-object returns. Such pattern classifiers are commonly used in a variety of applications; the novelty in this work is applying such techniques to the problem of automatic object detection in a degraded visual environment, namely turbid water. We describe our framework and features, then demonstrate the performance of three pattern classification detectors using a series of test data collected in a variety of water conditions in a laboratory test tank. All three pattern classification detectors outperform a standard detection method. There are subtle performance differences between the classifiers that may result in application-specific tradeoff considerations.

Environmental conservation is an area where AI can provide significant help for many types of tasks. Oil, plastic, anthropogenic noise, overfishing and global warming are known to affect marine ecosystems (flora, fauna) inducing a drastic decrease of marine biodiversity and ecosystem services. The assessment of marine animals’ distribution could benefit from automatic recognition of the presence of a species in a specific location. For this purpose, the passive acoustics monitoring can use underwater audio recordings and try to recognize the sound produced by the species. This work compares the performance of classical computer vision algorithms and modern deep learning methods for the task of identifying if a spectrogram contains the characteristic sound produced by the brown meagre. An accuracy of 95% was achieved using a deep convolutional neural network based on a recent architecture and partially pretrained, outperforming classical computer vision algorithms.

Coastline segmentation is the process of separating the coastal and backshore zones on aerial images. With a large world population living close to the coast, monitoring coastline changes is critical. Classical computer vision techniques were used to segment the coastline in high quality grayscale images where the difference between the zones was easy to distinguish. However, these techniques are limited to low resolution images and in areas with similar colors or textures. In this work we propose deep convolutional architectures for coastline segmentation using aerial images. An F1 score above 96% was obtained by the best performing model. The obtained results show that our deep models are capable of automatically and accurately detecting coastlines which will help in speeding-up the coastline localization process in large aerial images and improve the efficiency of monitoring coastal areas.

The performance of lidar systems used for high resolution ranging and imaging in coastal water environments is primarily limited by optical scattering. Understanding how scattering effects the optical phasefront of laser light has the potential to improve the performance of these systems. In this paper, light is transmitted through a scattering underwater environment, and the transmitted light is then encoded with optical phase. This encoding allows us to understand the optical phase distribution of the transmitted light. Specifically, we demonstrate through a combination of theory, simulation, and experiment that we can determine the statistics of the optical phase distribution of light by measuring the spatial intensity distribution of the encoded optical return. These results advance our understanding of the relationship between optical phase and scattering, as well as inform the performance enhancements and limitations associated with this spatial discrimination, optical signal processing approach.

This paper investigates a total variation (TV) regularization image processing algorithm to restore underwater range images taken with a modified commercial time-of-flight (ToF) camera. The ToF camera illuminator was modified to support 532 nm flood illumination for underwater operation. This approach can produce highresolution amplitude and range images while rejecting a significant amount of ambient light. However, scattering due to the water turbidity adversely impacts image quality by introducing high amounts of image noise and image blurring that affect both the amplitude and range images. The TV regularization algorithm is applied to experimental images taken in a small test tank in the presence of a scattering agent to simulate a range of practical turbidities. Algorithm details are provided, and baseline and processed images are presented. The processed images demonstrate image restoration that retains the downrange edge features of the object being imaged is possible for a range of practical turbidities.