Gaussian process models embedded on an AUV enabled autonomous tracking and mapping of phytoplankton biomass in three dimensions.

An in situ particle imaging system for measurement of high concentrations of suspended particles ranging from 30μm to several mm in diameter, is presented. The system obtains quasi-silhouettes of particles suspended within an open-path sample volume of up to 5cm in length. Benchmarking against spherical standards and the LISST-100 show good agreement, providing confidence in measurements from the system when extending beyond the size, concentration and particle classification capabilities of the LISST-100. Particle-specific transmittance is used to classify particle type, independent of size and shape. This is applied to mixtures of oil droplets, gas bubbles and oil-coated gas bubbles, to provide independent measures of oil and gas size distributions, concentrations, and oil-gas ratios during simulated subsea releases. The system is also applied to in situ measurements of high concentrations of large mineral flocs surrounding a submarine mine tailings placement within a Norwegian Fjord.

Both oil droplets and gas bubbles have simultaneously been quantified in laboratory experiments that simulate deep-water subsea releases of both live oil (saturated with gas) and additional natural gas under high pressure. These data have been used to calculate particle size distributions (50-5000 μm) for both oil and gas. The experiments showed no significant difference in oil droplet sizes versus pressure (from 5 m to 1750 m) for experiments with live oil. For combined releases of live oil and natural gas, oil droplet sizes showed a clear reduction as a function of increased gas void fraction (increased release velocity) and a weak reduction with increased depth (increased gas density/momentum). Oil droplets were reduced by a factor of 3 to 4 during simulated subsea dispersant injection (SSDI) and no significant effect of pressure was observed. This indicates that SSDI effectiveness is not dependent on water depth or pressure.

Small-scale experiments performed at SINTEF, Norway in 2011-12 led to the development of a modified Weber scaling algorithm. The algorithm predicts initial oil droplet sizes (d₅₀) from a subsea oil and gas blowout. It was quickly implemented in a high number of operational oil spill models used to predict fate and effect of subsea oil releases both in academia and in the oil industry. This paper presents experimental data from large-scale experiments generating oil droplet data in a more realistic multi-millimeter size range for a subsea blow-out. This new data shows a very high correlation with predictions from the modified Weber scaling algorithm both for untreated oil and oil treated by dispersant injection. This finding is opposed to earlier studies predicting significantly smaller droplets, using a similar approach for estimating droplet sizes, but with calibration coefficients that we mean are not representative of the turbulence present in such releases.

Gas dissolution reduces the release of methane to the atmosphere from subsea sources. Being able to predict and assess the methane flux to the atmosphere requires knowledge on gas dissolution and mass transfer. This can be obtained by studying the size evolution of bubbles rising in water. New data of bubble size evolution have been obtained by releasing, tracking, and filming methane bubbles with an ROV in the Trondheimsfjord from depths varying between 100 and 300 m. Released bubbles had an initial diameter between 5 and 7 mm and were tracked until they reached a diameter of roughly 2 mm. The new data were compared against theory, applying established correlations for the mass transfer coefficient. There was an inconsistency between experiment and theory. Thus, new correlations for the mass transfer are proposed. The new correlations are consistent with both the new experiments and previously published experiments. They indicate that the conditions in the ocean can be labeled as partly contaminated with respect to mass transfer.

Measurements of morphometrical parameters on i.e., fish larvae are useful for assessing the quality and condition of the specimen in environmental research or optimal growth in the cultivation industry. Manually acquiring morphometrical parameters from microscopy images can be time consuming and tedious, this can be a limiting factor when acquiring samples for an experiment. Mask R-CNN, an instance segmentation neural network architecture, has been implemented for finding outlines on parts of interest on fish larvae (Atlantic cod, Gadus morhua). Using classical machine vision techniques on the outlines makes it is possible to acquire morphometrics such as area, diameter, length, and height. The combination of these techniques is providing accurate-, consistent-, and high-volume information on the morphometrics of small organisms, making it possible to sample more data for morphometric analysis. • Capabilities to automatically analyse a set of microscopy images in approximately 2-3 seconds per image, with results that have a high degree of accuracy when compared to morphometrics acquired manually by an expert. • Can be implemented on other species of ichthyoplankton or zooplankton and has successfully been tested on ballan wrasse, zebrafish, lumpsucker and calanoid copepods.