ABSTRACT. The occurrence of oil seeps on the sea surface from active petroleum systems has been extensively documented by radar imaging using SAR (Synthetic Aperture Radar). Polarimetry is consolidating in the oil industry as a sophisticated technique for the study of marine seeps. The Cantarell Complex, located in the Gulf of Mexico, is currently the most prolific natural seep site in the world. This paper aims to add to the discussion on the physical properties of seepage slicks from full polarimetric data (quad-pol) from RADARSAT-2 satellite. The fact that the research was based on two images (ID#02 and ID#04) acquired at the same place and with the same mode of operation but with a time lapse between acquisitions and different incidence angles allowed to evaluate the influence of the imaging geometry on target signatures. To this end, samples of four classes (sea, offshore platform, oil and false targets) were collected. The data of the samples were afterwards shown in graphs and plotted in the classification plan of Cloude & Pottier – entropy (H) vs. alpha angle (¯α) – for backscattering mechanism analysis. Such elements made it possible to certify that (i) the sea is a Bragg-type surface regardless the incidence angle, (ii) platforms have double-bounce scattering, but small incidence angles are inadequate for their characterization, (iii) false targets (associated in ID#02 to regions of little wind) are moderately random/quasi deterministic surfaces, and (iv) oil behaves either as a Bragg (19.0◦–22.7◦ incidence angle range) or as a random/anisotropic surface (33.7◦–36.7◦ incidence angle range). Keywords: H-¯α diagram, incidence angle, radar imaging, marine seeps.RESUMO. A ocorrência de exsudações de óleo na superfície do mar a partir de sistemas petrolíferos ativos vem sendo extensivamente documentada por imageamento utilizando sistemas SAR (Synthetic Aperture Radar). A polarimetria está se consolidando na indústria do petróleo como uma técnica sofisticada para o estudo de seeps marinhos. O Complexo de Cantarell, situado no Golfo do México, é atualmente o local de exsudação natural de óleo mais prolífico do globo terrestre. Este trabalho visou contribuir com a discussão acerca das propriedades físicas de escapes de óleo a partir dos dados polarimétricos completos (quad-pol) do satélite RADARSAT-2. O fato da pesquisa valer-se de duas imagens (ID#02 e ID#04) adquiridas no mesmo local e modo de operação, mas com intervalo entre as aquisições e diferentes ângulos de incidência, possibilitou avaliar a influência da geometria de imageamento na assinatura dos alvos. Com esse propósito, foram geradas amostras de quatro classes (mar, plataforma, óleo e falso alvo), que tiveram os dados posteriormente relacionados em gráficos e no plano de classificação de Cloude & Pottier – entropia (H) vs. ângulo alfa (¯α) – para análise do mecanismo de espalhamento. Tais elementos permitiram atestar que (i) o mar é uma superfície do tipo Bragg independente do ângulo de incidência, (ii) as plataformas têm espalhamento double-bounce, mas baixos ângulos são inadequados para sua caracterização, (iii) falsos alvos (associados em ID#02 a regiões de baixo vento) são superfícies moderadamente aleatórias/quase determinísticas, e (iv) o óleo comporta-se ou como Bragg (19,0◦–22,7◦), ou como uma superfície aleatória/anisotrópica (33,7◦–36,7◦).Palavras-chave: Diagrama H-¯α, ângulo de incidência, imagem de radar, seeps marinhos.
The development and application of predictive models to distinguish seepage slicks from oil spills are challenging, since Synthetic Aperture Radars (SAR) detect these events as dark spots on the sea surface. Traditional Machine Learning (ML) has been used to discriminate the Oil Slick Source (OSS) as natural or anthropic assuming that the samples employed to train and test the models in the source domain (DS) follow the same statistical distribution of unknown samples to be predicted in the target domain (DT). When such assumptions are not held, Transfer Learning (TL) allows the extraction of knowledge from validated models and the prediction of new samples, thus improving performances even in scenarios never seen before. A database with 26 geometric features extracted from 6279 validated oil slicks was used to develop predictive models in the Gulf of Mexico (GoM) and its Mexican portion (GMex). Innovatively, these well-trained models were applied to predict the OSS of unknown events in the GoM, the American (GAm) portion of the GoM, and in the Brazilian continental margin (BR). When the DS and DT domains are similar, the TL and generalization are null, being equivalent to the usual ML. However, when domains are different but statically related, TL outdoes ML (58.91%), attaining 87% of global accuracy when using compatible SAR sensors in the DS and DT domains. Conversely, incompatible SAR sensors produce domains statistically divergent, causing negative transfers and generalizations. From an operational standpoint, the evidenced generalization capacity of these models to recognize geometric patterns across different geographic regions using TL may allow saving time and budget, avoiding the collection of validated and annotated new training samples, as well as the models re-training from scratch. When looking for new exploratory frontiers, automatic prediction is a value-added product that strengthens the knowledge-driven classifications and the decision-making processes. Moreover, the prompt identification of an oil spill can speed up the response actions to clean up and protect sensitive areas against oil pollution.
The amount of data from geochemical analysis using samples collected in oil wells grows simultaneously to the investment in the exploration and production sector. On the other hand, the treatment and interpretation of these results are still very dependent on experts and demand time. With the generation of extensive databases, data mining presents itself as a good alternative to explore them through statistical methods and computational algorithms, providing technological differential and agility to the system. In an experimental way, with data from 200 oils from the Potiguar Basin, these tools were implemented, with the consequent suggestion of a workflow that would, in the end, return a reasonable accuracy in predicting their genetic classification. Using multidimensional scaling (MDS) and clustering (dendrogram and k-means types) from 60 initial attributes, the optimal set was reduced to 26. Applying Machine Learning, 92.50% of median accuracy were obtained in the Decision Tree algorithm, 95.00% in Random Forest and 87.50% in Artificial Neural Network. Comparing to an analysis previously presented at the pertinent literature, the benefits in terms of efficiency can be realized with the adoption of the methodology herein proposed. Keywords: Organic geochemistry; Data Mining; Multivariate Statistics; Workflow.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
customersupport@researchsolutions.com
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.