The ocean is a strategic resource for human survival and sustainable development. However, the rapid development of industries such as maritime transport and oil extraction has dramatically increased the risk of pollution from oil spills. In the past 30 years, there have been over 60 major oil spill accidents worldwide, creating more than 45×10⁶ m² of polluted area (Leifer et al., 2012; Fingas and Brown, 2014). These oil spill accidents have seriously damaged the ecological environment of the ocean and caused significant economic losses. Moreover, ocean oil spill pollution often affects a large area, is long lasting, and causes great damage to marine wildlife. The surveillance and monitoring of oil spills have thus attracted considerable attention from marine management departments in many countries (Cai et al., 2016; Zou et al., 2016).

Remote sensing technology is an important method used to monitor and track oil spills. Microwave remote sensing, especially synthetic aperture radar (SAR), can collect data during the day and night and in all weather conditions. SAR is not affected by either clouds or light and produces high-resolution two-dimensional images (Brekke and Solberg, 2005; Fingas and Brown, 1997; Girard-Ardhuin et al., 2005). This technique has been widely applied to ocean oil spill monitoring. SAR can detect oil spills on the sea surface because the Marangoni effect produced by oil films on the sea surface inhibits the formation of short-scale waves or capillary waves in the ocean (Alpers and Hühnerfuss, 1988; Skrunes et al., 2016). The resulting weaker radar echo intensity creates dark areas in radar images that indicate the sea areas covered by the oil film (Fingas and Brown, 1997 ; Zheng et al., 2017). To date, scholars in many countries have carried out extensive research on SAR oil spill detection. These studies have focused on the oil spills detection and discrimination between oil spills and lookalikes. Oil film detection has largely been based on detecting dark spots in images. Gemme and Dellepiane (2018) proposed an automatic, unsupervised method to detect sea surface oil spills and extract spill information. Topouzelis et al. (2007, 2008) and Del Frate et al. ( 2000) applied neural networks to high-resolution SAR satellite images to detect oil spills. Solberg et al. ( 2003) designed a classifier for oil spill detection, defined 10 characteristics, and applied Bayesian theory to discriminate oil spills from lookalikes. Chehresa et al.(2016) proposed an algorithm for selecting the optimal features from SAR images to distinguish oil spills from lookalikes . Nunziata and Migliaccio successfully detected ocean oil spills and distinguished oil spills from biological oil films using co-polarized phase differences (Migliaccio et al., 2009).

The increasing demand for oil spill monitoring has made it necessary to not only accurately detect and locate oil spills but also obtain key information about the oil spill, such as the type, thickness, and state, which are of considerable significance for the response and subsequent cleanup of oil spill accidents. Thus, it is essential to study the properties of oil spills, including dielectric properties and microwave scattering characteristics. It is not sufficient to analyze only the grayscale and texture characteristics of oil spill areas in radar images. In recent years, scholars in many countries have used microwave equipment to carry out various types of oil spill experiments, which are mainly divided into indoor and outdoor experiments (Guo et al., 2016). Outdoor experiments refer to the intentional pouring of oil into natural marine environments to simulate the occurrence of an oil spill and the monitoring of the spill through satellite-borne (Skrunes et al., 2014, 2016), airborne (Angelliaume et al., 2018; Ulaby et al., 1986), land-based, and ship-based equipment (Wismann et al., 1998). This method offers the advantage of simulating the real environment but is difficult to implement, is costly, and causes environmental pollution. Therefore, scholars are currently more inclined to carry out indoor experiments by building wind-wave tanks. Regarding the experimental aspect of distinguishing types of oil spill, in 1992, Alpers et al. (1988) built a 26.0 m×1.0 m×0.5 m indoor wind-wave tank, using two types of oil, namely, light fuel oil and heavy fuel oil, and carried out the experiments using scatterometers with L-, S-, C-, and X-bands (Gade et al., 1998). Yang et al. (1993) studied oil film thickness inversion in 1987 by using a microwave radiometer to analyze the relationship between the thickness and brightness temperature of an oil film. Pang and Sun (2003) used a microwave radiometer to analyze the relationship between microwave radiation and the film thicknesses of petroleum, fuel oil, and edible oil. Loor et al. (1978) studied multimethod oil spill detection in 1971 by launching a remote sensing ocean monitoring plan and building 100.0 m×8.0 m×0.5 m indoor wind-wave tanks to simulate a marine environment. The microwave scattering characteristics of ocean wind waves and oil spills were analyzed by using airborne side vision radar and infrared sensor detection equipment.

Numerous indoor and outdoor spill experiments have been carried out, but the effect of changes in the oil film properties on the film surface scattering characteristics, especially the effect of the emulsification of oil spills on microwave scattering, have not been considered. However, the emulsification caused by wave action and weathering will not only increases the viscosity, surface tension, and volume of oil film (Fingas, 1995; Khan et al., 2011) but also cause the oil to mix with seawater and form a water-in-oil emulsion (Thingstad and Pengerud, 1983). The considerable modification of the nature of the oil, and the emulsification state and properties of the oil film change continuously over a period of hours to days (Minchew et al., 2012). The resulting water-in-oil emulsion exhibits characteristics between those of a liquid and a solid with a loose and porous surface (Fingas and Fieldhouse, 2003, 2004; Fingas, 1995).

Therefore, it is not known whether the emulsification state change affects the microwave scattering characteristics or the detection accuracy for oil spill. In this study, oil spill experiments were performed in an outdoor water tank. Different types of oil were used with a C-band fully-polarimetric microwave scatterometer (HH, VV, HV, and VH) and a vector network analyzer to analyze the changes in the microwave scattering characteristics in the emulsification of oil spills. The differences in the emulsification-induced microwave scattering among different types of oil were also investigated.

This paper is organized as follows. In Section 1, background information and the goal of the study are presented. The experimental setup, data collection, and meteorological conditions during the experiment are presented in Section 2. The experimental results and analysis are given in Section 3, and conclusions are drawn in Section 4.

The experimental site was located at the Muping Coastal Environment Research Station of the Chinese Academy of Science. The experiment consisted of two stages. The first stage lasted for 5 d from August 23 to August 27, 2017, and the second stage lasted for 4 d from September 15 to September 18, 2018. The experiments were performed in the same fiberglass tank (length×width×depth, 6.0 m×2.2 m×0.7 m). Three types of crude oil were tested: crude oil with an asphalt content below 3% that was prone to emulsification (type A), fresh crude oil extracted from an oilfield (type B), and industrial crude oil that was dehydrated and purified (type C). The parameters of the oils are given in Table 1. The main measuring equipment included a fully-polarimetric C-band microwave scatterometer and a vector network analyzer. Figure 1 shows the experimental setup. The microwave scatterometer was mounted on a platform at the narrow end of the tank. Before the experiments were performed, the tank was filled to a depth of 0.35 m with seawater with a density of 1.02 g/mL.

In the first stage of experiment, the emulsification of one type of oil and the impact of the amount of oil on the backscatter of the oil film was mainly investigated. Type A crude oil was used throughout this stage. The oil was added to the tank in eight increments to reach a cumulative amount of 6 996.7 g. To simulate the stirring effect of waves on an oil film in a natural marine environment, the oil film was stirred artificially after each increment and then measured with the scatterometer after the water surface calmed down. The stirring times and scatterometer measurements are shown in Fig. 2.

In the second stage of experiment, the effect of the emulsification of different types of oil on the backscatter was mainly studied, and the differences in the backscatter between oil types for different emulsification states were analyzed. Type B and C crude oils were used in this stage. The crude oil emulsification process was simulated using oil samples with moisture constants of 0%, 10%, 20%, and 30% and 0%, 20%, 40%, and 50% for type B and C oils, respectively. The moisture constant is an important indicator of the emulsification state of an oil spill and can be calculated as follows:

where $ {m}_{\rm{water}} $ is the mass of water, and $ {m}_{\rm{oil}} $ is the mass of oil. A water-in-oil emulsion was prepared as follows. A mixed solution of the desired ratio of crude oil to seawater was added to a JJ-2A mixer and mixed at a rate of 600 r/min. After stirring, the solution was allowed to rest for 24 h. If water accumulated at the bottom of the container, the solution was mixed again until all of the water was incorporated. After the scatterometer measurement for each oil sample was completed, the tank was treated to remove any oil residue. The detailed cleaning process and verification are given in Appendix A. The specific experimental procedure and amounts of oil added are given in Fig. 2. The entire experimental process was recorded by a high-resolution video camera to document the diffusion and distribution of the oil film.

Outdoor tank experiments were conducted to investigate whether an emulsification-induced state change affects the microwave scattering characteristics of an oil film or the detection accuracy of an oil spill. A C-band fully-polarimetric microwave scatterometer was used to observe the backscatter of the oil film surfaces of three different types of crude oil (A, B and C) during the emulsification process. Under the conditions of a low wind speed and no waves, the difference in the backscatter between the emulsified oil film and the surface of the calm water and the influence of the emulsification process on the backscatter were analyzed in detail. The differences in the backscatter between the different types of oil were also analyzed.

The experimental results showed that the water-in-oil emulsion of crude oil exhibited characteristics between those of a liquid and a solid (Fig. 8) and increased the surface roughness. Under the VV, HH and HV/VH polarization modes, the backscatter of the emulsified oil film was greater than that of calm water by 71.15%, 78.21%, and 64.1%, respectively (Fig. 7). The preliminary conclusion is that crude oil emulsification can change the surface roughness.

The changes in the backscatter $ {\sigma }^{0} $ and the dielectric constant $ {\varepsilon }_{r} $ were simulated using SEM. The results showed that there was a linear positive correlation between $ {\sigma }^{0} $ and $ {\varepsilon }_{r} $ (Fig. 11). The vector network analyzer data showed that the dielectric constant of the oil film was positively correlated with the moisture constant of the oil (Table 6). The experimental results were in agreement with the simulation results. The backscatter increased with the moisture constant of the oil film (Fig. 9). It was concluded that the emulsification modulated the radar backscatter through changes in the dielectric constant and the surface roughness. Comparing the experimental results of the type B and C oils showed that the backscatter was dominated by the surface roughness.

The differences in the backscatter between the type B and C crude oils were compared for the emulsified and non-emulsified states. The mean differences in the backscatter for the non-emulsified state were 2.19 dB, 2.63 dB, and 2.21 dB under the VV, HH, and HV/VH polarization modes, respectively (Fig. 13a). The corresponding differences in the emulsified state with a 20% moisture constant were much smaller: 0.98 dB, 1.49 dB, and 1.5 dB (Fig. 13b). The differences in the oil compositions and oil film properties may have affected the surface roughness of the oil films to some extent.

The experiments were carried out in an outdoor water tank under a low-wind and no-wave environment. The experimental results showed that the emulsification process had a significant impact on the microwave scattering characteristics of the oil films. This study serves as a reference for the detection of oil spills and the improvement in the oil spill detection accuracy of a C-band microwave sensor under a calm sea state. Future research should explore the effect of oil spill emulsification on the microwave scattering characteristics of an oil film surface in an actual marine environment. Different band sensors are also needed to evaluate the responses of different bands to the emulsification of an oil spill.

We thank the Muping Coastal Environment Research Station of the Chinese Academy of Science for allowing the experiment to be performed and providing ground truth data.