b'AEM 2023Short abstractsAEM 2023: Short abstracts *Automated data processing of a large-scale airbornereceiver is achieved by GPS, within less than 50 nanoseconds. time-domain electromagnetic survey by a deepThe drone is equipped with two lasers for determination of the learning expert system attitude and real-time image processing has been developed to measure and control the movement of the receiver coil Muhammad Rizwan Asif1, M. Andy Kass1, Anders V. Christiansen1,with the drone in the airspace. Data from all sensors are Zara Rawlinson2 and Rogier Westerhoff2 continuously streamed to the ground station. The system can be used for mapping of deep targets. However, as the current 1. Department of Geoscience, Aarhus University, Aarhus C,waveform and system bandwidth is well defined also more MIDTJYLLAND, Denmark shallow layered targets can be mapped. The latter makes 2. GNS Science, Taupo, New Zealand it possible to use the system to map shallow ground water The new generation of airborne electromagnetic (AEM) surveysaquifers in terrains inaccessible for traditional ground-based yield large data sets of thousands of line km. Parts of theseTEM systems.data are often contaminated by noise from various sources, e.g., fences, power lines, which corrupts the data to a degreeFinding geology structures in depth sections from that it can no longer be used. The problem intensifies inairborne geophysics: Automatic workflowsurban areas where the risk of data corruption is highest due to dense infrastructure. The inversion of corrupted data risksSimge Ayfer and Desmond FitzGeraldinterpreting spurious subsurface features and flawed geologicalIntrepid Geophysics, Melbourne, VIC, Australiainterpretations. Therefore, in many cases, the corrupted data is identified and culled prior to inversion. This process of cullingThe explosion in new airborne electro-magnetic surveys is corrupted data is generally a manual task requiring specialists tocreating the need for less cutting of corners, better honouring examine the data in detail, which is an extremely complex andof the known physics in the algorithms, and proper use of time-consuming process. all the system monitors. The importance of a good starting model in a deterministic, iterative, non-linear inversion, such as Recently, we proposed a deep learning expert system tothat provided by the 2.5D Moksha code, has been recognised automate the complex AEM data processing workflows. Thefor many years. This study touch bases on two project scale proposed method uses a deep convolutional auto-encoder toexamples that were collected by the same aircraft. Clearly in identify corrupted data and was trained such that it generalisesthe context of an emerging continent wide AEM campaign to diverse geological conditions and various survey areas.to acquire prospective surveys, the implications for these In this study, we investigate the generalisation capabilitiesdevelopments are critical, in that these tools can also manage of our deep learning method on a large AEM survey area incomplete surveys, no matter what line length are involved. Northland, New Zealand. Our approach takes ~ 600 s to processThis concentration of predicting geology structures in depth 3984 line-km of data and displays strong spatial correlationsections has demonstrated the ability to identify possible for the data identified as corrupted. The inversion resultsexploration targets and map steeply dipping and folded show very few potential anomalies in the model space whichgeology in a deformed terrain. Equally important, is the creation are being inspected by a manual operator. In general, theof workflows and visualisation toolkits to help interpreters, no proposed approach is generalisable and displays high-qualitymatter what scale, or which aspect of geology or rock properties data processing within short amounts of time, which requiresthey wish to interrogate. The laissez faire situation of accepting minimal further quality inspection sub-optimal methods for estimating potential field gradients has plagued, and held back, the successful use of potential field An early time semi-airborne loop source TEM system geophysics for too many years now. Almost all interpretation methods are based upon estimating these gradients.Esben Auken1, Pradip K Maurya1, Anders Christiansen2, LichaoLiu2, Jacob Naundrup3, Anders la Cour-Harbo3 and Michael J Nielsen3 Beyond conductive targets: Characterising lithium-prospective lacustrine evaporite mineral 1. Aarhus GeoInstruments, Aarhus, Denmark systems of North Americas Basin and Range 2. Department of Earth Sciences, HydroGeophysics Group, Aarhus3. Aalborg Univeristy, Aalborg, Denmark Provincewith regional-scale AEMWe present a new semi-airborne transient electromagneticLyndsay Ball, Paul Bedrosian and Chloe Gustafson(TEM) system, dTEM, for subsurface imaging. The dTEM systemU.S. Geological Survey, Denver, Colorado, United Statesis designed for both imaging of groundwater and mineral resources. The system uses a large ground loop for transmittingThe Basin and Range province of North America hosts energy into the ground. Its a dual moment system with peaksubstantial lacustrine evaporite mineral systems prospective current up to 30 A for high moment and 1-2 A for low moment.for lithium, a critical mineral currently listed for mineral The fast LM turn off time is around 8 s from the beginningresource assessment by the U.S. Geological Survey. Airborne of the turn-off ramp. The receiver coil is a high frequency, lowelectromagnetic (AEM) surveys are being conducted to support noise, open air coil carried by the drone as a slung load. Thethese assessments by identifying shallow clays and brines, as high accuracy synchronisation between the transmitter and thewell as through improving the shallow subsurface geologic *Presentations are listed alphabetically by first author.AUGUST 2023 PREVIEW 52'