Optical interpretation of oil emulsions in the ocean – Part I: Laboratory measurements and proof-of-concept with AVIRIS observations

Highlights

• Oil emulsions are found to form two types: oil in water (OW) and water in oil (WO).

• Spectral reflectance and absorption of oil emulsions are determined from experiments.

• OW and WO emulsions with various concentrations have different spectral features.

• For fixed oil type, oil thickness of up to 0.4 mm can be estimated from reflectance.

• The proof-of-concept is demonstrated with AVIRIS data over the DWH oil spill.

Abstract

Optical identification and quantification of various marine-spilled oils play an important role in oil spill monitoring, assessment, and response. Through weathering processes, oil may become emulsified in two forms of oil-water mixture: water in oil (WO) and oil in water (OW). These two forms of oil emulsion are significantly different in their volume concentration (oil/water ratio), physical properties (viscosity, density, thickness), and optical properties (spectral reflectance (R_u(λ), sr⁻¹), and spectral absorption (a(λ), m⁻¹)). In this study, the optical properties of both types of oil emulsion, with different volumetric concentrations, are determined from carefully prepared oil emulsion samples, with the aim of helping to interpret optical remote sensing imagery. The concentrations of stable WO and OW emulsions range from 45% to 95% and from 0.025% to 3%, respectively. They exhibit different R_u spectral shapes in the near-infrared and shortwave-infrared wavelengths, with five “-CH” molecular bonds evident in the WO emulsion spectra. R_u (600–1400 nm) of OW emulsions increases with volume concentrations from 0% to 3.0%, but R_u (600–2400 nm) of the WO emulsions decreases with volume concentrations from 45% to 100%. On the other hand, for a fixed concentration (80%), R_u (600–2400 nm) of WO emulsions increases monotonically with thicknesses of up to ~0.4 mm, beyond which R_u (600–2400 nm) no longer increases with oil thickness. The difference between the R_u spectral shapes of OW and WO emulsions, as well as the statistical relationships between volume concentrations and R_u (NIR-SWIR) and between oil thickness and R_u (NIR-SWIR), provide the basis for developing optical models to classify oil emulsion types and for quantifying oil volume from remote sensing imagery. The potential of such an application is demonstrated using hyperspectral AVIRIS imagery collected over the Deepwater Horizon (DWH) oil spill in the Gulf of Mexico (GoM).

Introduction

Spilled oil can harm marine and coastal environments, and major spills can have a negative impact on the environment for a long time, such as in the case of the Deepwater Horizon (DWH) oil spill in the Gulf of Mexico (GoM) between late April and mid-July 2010 (Kessler et al., 2011; Mariano et al., 2011; Leifer et al., 2012; Murawski et al., 2018). During the weathering process, oil undergoes physical and chemical changes such as spreading, drifting, mixing, evaporation, sedimentation, dissolution, emulsification, photo-oxidation, and biodegradation, thus forming different pollution types (Zhong and You, 2011; Lu et al., 2013). Many treatment methods use booms, skimmers, chemical dispersants, and in situ burning to mitigate the effects of these different oil types (Zhong and You, 2011). Timely detection, classification, and quantification of the different weathered oils is a key research direction in marine oil spill monitoring and the resulting response (Lu et al., 2013).

Many remote sensing techniques have been used for this purpose, including optical remote sensing (Hu et al., 2009, Hu et al., 2018; Chust and Sagarminage, 2007; Kuhn et al., 2004; Bulgarelli and Djavidnia, 2012; Lu et al., 2013; Sun et al., 2016; Sun and Hu, 2016; Lu et al., 2016a; Leifer et al., 2012; Kokaly et al., 2013; Carolis et al., 2014; Svejkovsky et al., 2016; Teodosio et al., 2017), synthetic aperture radar (Zheng et al., 2001; Brekke and Solberg, 2005; Zhang et al., 2011; Garcia-Pineda et al., 2013), laser fluorescence (Brekke and Solberg, 2005; Brown et al., 1996), and thermal remote sensing (Asanuma et al., 1986; Cross, 1992; Salisbury et al., 1993; Tseng and Chiu, 1994; Lu et al., 2016b; Niclòs et al., 2014). Among these techniques, although SAR is perhaps the most frequently used technique because of its all-weather capability (i.e., it does not suffer from cloud cover), SAR is generally regarded as being only able to detect surface oil presence or absence while discriminating oil emulsions from the detected surface oil footprint is difficult. In contrast, although passive optical sensors that measure the reflected sunlight suffer from clouds, the many spectral bands of these sensors carry more information (than SAR signals) on the spilled oil, therefore having been increasingly used on oil spill response and assessment in recent years (Clark et al., 2010; Leifer et al., 2012; Svejkovsky et al., 2016; Hu et al., 2018). However, how to interpret the information from the multiple bands is not straightforward, and understanding oil-water spectral contrasts still requires further research (Fingas and Brown, 2018).

Weathered oil can form various targets with different visual characteristics, such as light sheen, silver sheen, rainbow sheen, dark patch, and brown-colored mousse. However, these visual characteristics are difficult to apply to optical remote sensing imagery for identification and classification of marine oil spill (Lu et al., 2013). This is because (1) the visual effects are smeared by the atmosphere; (2) oil slicks may occupy only a small portion of an image pixel (i.e., mixed pixel effect); (3) human eyes are only sensitive to the visible wavelengths but a lot of signals can be found in the near infrared and shortwave infrared (see below). Nevertheless, for oil spill optical remote sensing, these targets can be reclassified into three main types according to their optical properties: oil sheen (with different thicknesses), oil emulsions (with different concentrations), and floating crude oil (Lu et al., 2013). These targets have different interactions with light as they reflect, absorb, and scatter the incident sunlight, resulting in different optical contrasts and spectral features from the surrounding oil-free water (Carolis et al., 2014; Wettle et al., 2009; Clark et al., 2010). Although laboratory measurements (Kokaly et al., 2013; Cloutis, 1989; Clark et al., 2010; Kukhtarev et al., 2011; Lu et al., 2008; Wettle et al., 2009) and numerical models (Otremba, 2003, Otremba, 2005; Otremba and Piskozub, 2004) have been used to interpret oil-water spectral contrasts, carefully designed laboratory experiments for this purpose are rare, especially for oil emulsions.

Oil emulsions generally have two different forms: water-in-oil (WO) emulsions and oil-in-water (OW) emulsions. A WO emulsion is formed by small water droplets that are entrained in oil, while an OW emulsions consists of small oil droplets incorporated in water (Zhong and You, 2011). In addition to the emulsification state (WO and OW), the most important parameter of oil emulsions is the volumetric oil: water ratio, equivalent to the volume concentration of oil used in this study (i.e., oil volume over total volume, dimensionless). Fig. 1 illustrates several characteristics of OW and WO emulsions produced in the laboratory: with increasing concentration, the color of OW emulsions changes from light yellow to dark yellow (Fig. 1b). A droplet of the OW emulsion was placed on a piece of white paper, where a tiny oil drop surrounded by water is clearly visible (Fig. 1c). The WO emulsions with different oil concentrations exhibit different colors (Fig. 1d), and these WO emulsions can float on the water surface (Fig. 1e). A droplet of the WO emulsion on a glass slide is black-colored (like black crude oil), and several very small water drops can be seen when the WO droplet is enclosed by two glass slides (Fig. 1f). These oil emulsions with different concentrations can remain in unstable, semi-stable, and stable forms. The viscous WO emulsion can contain up to ~70% water and is very difficult to clean (Zhong and You, 2011). Light, refined oil is difficult to emulsify (in WO form) due to the lack of appropriate hydrocarbon components to stabilize the water droplets, but crude oil emulsifies readily (Zhong and You, 2011). Usually, chemical dispersants can be used to reduce the oil-water interfacial tension and disperse the WO emulsion (or oil slicks, floating black oil) into water to form an OW emulsion with varying low concentrations. When the spilled crude oil forms stable oil emulsions in seawater, the rate of other weathering processes may decrease.

To date, the only laboratory experiment to determine the reflectance characteristics of different emulsions has been conducted by Clark et al. (2010). In the experiment, reflectance was measured from oil emulsion samples placed in a hole in a lid, which was screwed onto an empty glass jar (painted black) above the background water in a black bucket. The measurements were used to improve the interpretation and quantification of oil emulsions from measurements made by the airborne visual infrared imaging spectrometer (AVIRIS) instrument onboard ER-2 aircraft (Clark et al., 2010; Leifer et al., 2012). However, there remain several issues that need to be addressed with more carefully designed experiments. They include (but are not limited to) the following: (1) the spectral reflectance of OW emulsions with various oil concentrations are difficult to measure with this experimental design; (2) the small thicknesses of the WO emulsion samples used in the experiment result in the samples not being completely opaque due to the interference of the background glass jar, leading to uncertainties in the measured sample reflectance.

The objectives of the present study are therefore to determine the optical properties (reflectance and absorption) of OW and WO emulsions with different concentrations, to complement the experimental results of Clark et al. (2010) and to help interpret remote sensing imagery. Specifically, in this study, laboratory experiments were carefully designed to measure spectral reflectance (R_u, sr⁻¹) of WO and OW emulsions with different concentrations, and to measure the R_u of WO emulsions with different thicknesses. In addition, spectral absorption coefficients of WO and OW emulsions were determined with an integration sphere. Based on the results, optical models were established to classify oil emulsion types and to quantify oil concentrations, and the proof-of-concept models were applied to airborne hyperspectral AVIRIS data collected over the DWH oil spill.