""At CGG, everyone around you is very friendly, supportive and knowledgeable. This creates an amazing working atmosphere, and on top of that, the work never gets boring.""
MSc in Natural Sciences (Physics), University of Cambridge.
Current project: Cornerstone Evolution 3D, a merge of CGG’s bandwidth-extended conventional and true broadband seismic surveys over the UK Central North Sea (CNS). The 35,000 km² project focuses on re-processing and imaging existing seismic data with the latest signal processing, model building and imaging technologies.
Current focus: Removing free-surface multiples through modelling and subtraction.
At CGG, everyone around you is very friendly, supportive and knowledgeable. This creates an amazing working atmosphere, and on top of that, the work never gets boring.
I studied physics at university, and was mostly interested in experimental physics. This lead me to work on several imaging projects during my studies, and naturally I was looking to do something related to data analysis after graduation.
Going into geophysical imaging was a bit of a leap of faith for me, since I hadn’t done any geology before that. However, it turned out great! With the training that CGG provides and with the assistance of my super-helpful colleagues, I learnt a great deal about geophysics in a very short time. I am still learning a lot every day, which is both challenging and exciting.
I joined CGG in 2018 as a graduate imaging geophysicist, and have really enjoyed my time here. The work is interesting, and the environment within the company is great. Crawley is a nice place to live as well, especially if you are a fan of travelling like me – being ten minutes away from Gatwick Airport and 30 minutes from London, you can get basically anywhere you want.
The project my team is working on at the moment is the reprocessing of several true broadband CGG multi-client seismic surveys from the Central North Sea as part of the larger 35,000 km² Cornerstone Evolution 3D project. With 6000 km² of high-resolution seismic data to process, there are lots of challenges which we have to deal with. Although my current focus is demultiple when I first joined I worked on removing noise from the data, which requires lots of different approaches since the types of noise vary greatly.
Noise falls into two broad and in practice rather fluid categories: random and coherent and is strongly influenced by the acquisition method and environment.
For marine seismic surveys where receivers are towed through the water column arrayed along a cable, random noise is most commonly generated by the ocean swell, something which obviously does not affect land surveys. Random noise can also be generated by instrumentation: for instance if water intermittently leaks into the cable (this is exceptionally rare) you can see cross-feed where for all receivers there is a temporary electrical spike. Amplitude spikes can also occur in the presence of strong currents where the recording cables have been tied together to avoid them drifting apart, these are obviously not random in positioning but can be random in amplitude and duration depending on the speed of the ship, variations in current strength and the direction & steering of the vessel.
What unites different sources of random noise is that by definition they are unpredictable in occurrence and are difficult to model, so industrial noise from ship engines from a nearby staging area for vessels about to enter port or from offshore wind turbines turning or a drilling rig in operation although not necessarily intrinsically random energy sources all qualify. Earthquakes, passing aircraft and marine wildlife (such as sharks biting the cables) though far less common also fall into the category of random noise.
Coherent noise on the other hand is by definition predictable and can usually be modelled. It is often related to instrumentation: on older marine seismic surveys which used kerosene to keep the receivers buoyant, energy from the source could travel within the cable itself exhibiting as a low velocity linear arrival; a more blurred example is if one of the receivers is faulty (a not uncommon occurrence) a single high or low amplitude spike usually occurs for all recorded times, which has the appearance of random noise being a spike but is not random in occurrence. The seismic source itself can generate noise in the form of bubble energy which is the subsequent oscillation of the air bubble created by the seismic pulse, this usually has a distinct amplitude behaviour with time so can be removed using predictive methods. The direct arrival of the seismic energy from the source to the receivers also falls into this category.
External coherent noise sources include seismic reflections from rig platforms and undersea infrastructure such as pipelines which are easy to model and remove particularly if you know their location.
The most pernicious source of noise is probably seismic interference, which are seismic arrivals generated by another survey vessel. Usually different seismic survey vessels even if they are competitors co-ordinate their acquisition to minimise this; however this does not always happen and is not always possible due to narrow time windows for acquisition because of weather, fishing or other industrial activities such as subsea infrastructure maintenance.
Seismic interference can be particularly difficult to model and predict especially if there is no co-ordination so overlaps both categories. In terms of the techniques used to remove it, modelling works but often looking at it in different domains can randomise the signal allowing its easier removal using the techniques for attenuating random noise.
Finally geology can also generate noise: rugose surfaces and broken up sediments can scatter energy resulting in a loss of signal and an increase in random background noise; on land or where receivers are placed at the seabed, unconsolidated sediments in the near surface can generate ground roll and other surface waves which are coherent in behaviour and although unwanted signal can sometimes be used to make inferences about the near surface velocity.
The image below shows a cross-section some 100s of kms long from the vintage Cornerstone MCNV dataset from 2014 showing the geology of the UK Central North Sea. The final seismic image is overlaid by the final velocity model used to image it. Our new re-processing of this data with the latest denoising, deghosting, demultiple, modelling and imaging techniques has already given a step-change in resolution and imaging of the data.