Seismic full-waveform inversion (FWI) produces high resolution images of the subsurface by exploiting information in full acoustic, seismic or electromagnetic waveforms, and has been applied at global, regional and industrial spatial scales. FWI inverse problems are traditionally solved by using optimization, in which one seeks a best model by minimizing the misfit between observed waveforms and model-predicted waveforms. Due to the nonlinearity of the physical relationship between model parameters and waveforms, a good starting model is often required to produce a reasonable result. In addition, the optimization methods cannot produce accurate uncertainty estimates, which are required to better interpret final model estimates.

In principle, nonlinear Bayesian methods can be deployed to solve both issues. Monte Carlo sampling is one such class of algorithms which are computationally expensive, and all Markov chain Monte Carlo-based methods are difficult to parallelise fully. Variational inference provides a fully parallelisable alternative methodology. This is a class of methods that optimize an approximation to a probability distribution describing post-inversion parameter uncertainties. Both Monte Carlo and variational full waveform inversion have been applied previously to solve 2D Bayesian FWI problems, but neither of them have been applied in 3D.

In this study we apply three variational methods to a 3D FWI problem and analyse their performance. Specifically we apply automatic differential variational inference (ADVI), Stein variational gradient descent (SVGD) and stochastic SVGD (sSVGD), and compare their results and computational costs. These tests show that ADVI is the least computationally demanding method, but its results are systematically biased as uncertainty is underestimated. The method might therefore be used to provide relatively rapid but approximate insights into the Bayesian solution. SVGD demands the highest computational cost, yet produces equally biased results. Adding a randomized term in the SVGD dynamics produces sSVGD, a Markov chain Monte Carlo method based on variational principles. This provides the most accurate results, at intermediate computational cost. We conclude that 3D variational full-waveform inversion is practically applicable, at least in small problems, and can be used to image the Earth’s interior and to provide reasonable uncertainty estimates on those images.

How to cite: Zhang, X., Lomas, A., Zhou, M., Zheng, Y., and Curtis, A.: 3D Bayesian Variational Full-Waveform Inversion, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-3788, https://doi.org/10.5194/egusphere-egu23-3788, 2023.

Oil and gas industry is particularly interested in subsalt imaging because salt bodies often serve as seals for hydrocarbon reservoirs especially on the flanks and underneath the salt body. Classical imaging or full-wave inversion have encountered many difficulties in imaging such subsalt regions. The high velocity contrast between the salt body and its surroundings, the shape of the salt body, and the strong attenuation of waves propagating in the salt make accurate imaging difficult.

Imaging from surface seismic data has already been addressed. We propose to apply the full-wave inversion approach to borehole seismic data.  These data are supposed to provide more informative seismic signals (one-way, transmitted waves, scattered field near receivers, …) and should better constrain the salt flanks and bottom imaging. Unfortunately, this inverse problem is severely ill-posed due to the lack of data redundancy.  Introducing geological prior information through the parameterization of the salt geometry using the level set method partly overcomes this problem. This approach implicitly defines the salt/sediments interface through a smooth function. This hybrid  full-wave inverse problem combines the classical field gradient and the level set geometric gradient, allowing us to retrieve respectively the physical parameters and the salt body geometry.

The forward wave equation for a heterogeneous elastic domain is solved using the spectral element method. This method allows us to have a better representation of the salt/sediments interface, thus a more accurate modeling of the scattered wavefield due to interactions at the interface. Regarding the level set inversion, we define the implicit function in a parametric framework. We use compactly supported B-spline basis functions for their ability to represent a wide range of geometries compared to radial basis functions, for instance.

During the level set inversion process, when the salt/sediments interface evolves, updating the physical model parameters near the interface becomes an issue. We propose to perform a continuous deformation of the medium while conserving the mesh topology. The interface nodes are moved consistently with the implicit level set function. Furthermore, this approach  preserves a meshed representation of the interface through the inversion process allowing a more accurate seismic modeling. The mesh deformation is obtained by geostatistical kriging of node locations, constrained by the updated interface position.

In short, we propose a hybrid full-wave inversion method to estimate both material parameter fields and geometric parameters of the salt/sediments interface. The method is validated on several elastic models using synthetic seismic well data in a subsalt imaging context.

How to cite: Khazraj, K., Barnes, C., and Maillot, B.: Hybrid  full-wave inversion based on a B-spline level set mesh deformation method of borehole seismic data for a subsalt imaging context, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-5219, https://doi.org/10.5194/egusphere-egu23-5219, 2023.

The Earth’s crust hosts most of the geo-resources of societal interests (e.g. minerals, geothermal energy etc.). Integrative approaches combining geophysical and petrological observations to study the mantle assuming thermodynamic equilibrium are relatively common nowadays. However, in contrast to the mantle, where thermodynamic equilibrium is prevalent, vast portions of the crust are thermodynamically metastable. This is because equilibration processes are essentially temperature activated and the temperature in the crust is usually too low to trigger them. Consequently, the mineralogical assemblage of crustal rocks is mostly decoupled from the in situ pressure and temperature conditions, reflecting instead the conditions present at the moment of rock formation. Here we present a new methodology for integrated geophysical-lithological multi-data modelling of the crust. Our primary constraining data are fundamental mode Rayleigh wave surface wave dispersion curves determined by interstation cross-correlation measurements and teleseisms, as well as surface elevation (isostasy) and heat flow. Additional prior information is provided by P-wave velocities coming from controlled source and body wave tomography data. The inversion is framed within an integrated geophysical-petrological setting where mantle seismic velocities and densities are computed thermodynamically as a function of the in situ temperature and compositional conditions. In the crust we invert for a three-layered crust defined by Vs, density and Vp/Vs ratios (or Poisson coefficients) linked according to statistical correlations from global petrophysical data sets. The new methodology is applied to the Iberian Peninsula and adjacent margins where we jointly invert for both the crustal and lithospheric mantle structure. Our results show that the Iberian upper-middle crust is characterized by a clear dichotomy between the high Vs and felsic lithologies (Vp/Vs<1.76) in the Iberian Massif, and the low Vs and mafic lithologies (Vp/Vs>1.81) in the Betic, Pyrenees and Cantabrian Alpine mountain chains. The pattern changes in the lower crust where we obtain felsic lithologies  in  the Central system, NE Betics and N Mediterranean margin, and mafic lithologies in the Ossa-Morena, South-Portuguese, Galician, and Asturian-Leones terranes in the Variscan Iberian Massif. Overall we find a good correlation with previous geophysical studies (receiver functions, controlled source seismics) and the petrology of the main magmatic episodes since the Neoproterozoic (575 Ma).

How to cite: Clemente-Gómez, C., Fullea, J., and Arnaiz-Rodríguez, M. S.: Integrated geophysical-petrological inversion of surface wave and other data for the lithological structure of the Iberian crust, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-14389, https://doi.org/10.5194/egusphere-egu23-14389, 2023.

High-quality maps of the geothermal gradient and temperature are essential when assessing the geothermal potential of a region. However, determining geothermal potential is a challenge as direct measurements of in situ temperature are sparse and individual geophysical methods are sensitive to a range of parameters, not solely temperature. Here, we develop a novel approach to determine the geothermal gradient using a joint geophysical-petrological inversion which requires thermal property data, seismic and additional geophysical datasets. The seismic data provide new constraints on lithospheric boundaries which influence crustal geotherms. We utilise large seismic datasets and extract Rayleigh- and Love-wave phase velocity dispersion curves, measured for pairs of stations. The measurements were performed using two methods with complementary period ranges; cross-correlation of teleseismic earthquakes and waveform inversion, yielding measurements in a broad period range (4-500 s).

The joint analysis of Rayleigh and Love measurements constrains the isotropic-average shear-wave velocity, relatable to temperature and composition providing essential constraints on the thermal structure of a region's lithosphere. We demonstrate this by inverting the data using an integrated joint geophysical-petrological thermodynamically self-consistent approach (Fullea et al., GJI 2021), where seismic velocities, electrical conductivity, and density are dependent on mineralogy, temperature, composition, water content, and the presence of melt. The multi-parameter models produced by the integrated inversions fit the surface-wave and other data and reveal the temperatures and geothermal gradients within the crust and mantle which will be used for future geothermal exploration and utilisation.

We use Ireland as a case study (part of the De-risking Ireland's Geothermal Potential project - DIG) and find that our new methodology produces results comparable to past temperature and geophysical measures, and enhances resolution. Lithospheric and crustal thickness play a key control on the temperature gradient with areas of thinner lithosphere resulting in elevated geotherms. In some locations we observe geotherms elevated beyond expectations which result from high radiogenic heat production from granitic rocks. This new methodology provides a robust workflow for determining the geothermal potential in areas with limited direct measurements.

The DIG project is funded by the Sustainable Energy Authority of Ireland under the SEAI Research, Development & Demonstration Funding Programme 2019 (grant number 19/RDD/522) and by the Geological Survey of Ireland.

How to cite: Chambers, E., Xu, Y., Bonadio, R., Fullea, J., Lebedev, S., Kiyan, D., O'Reilly, B., Meere, P., Rezaeifar, M., Ye, T., Scully, A., and Tomar, G. and the DIG Team: Developing Joint Geophysical and Petrological Inversion to Determine Temperature and Image the Lithosphere and Asthenosphere: De-risking Ireland's Geothermal Potential (DIG), EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-6864, https://doi.org/10.5194/egusphere-egu23-6864, 2023.

Spectacular Paleogene uplift and volcanism in Ireland and Britain are thought to be associated with the Iceland Plume but their mechanisms are still unclear, considering, in particular, that the Iceland Hotspot was many hundred kilometres away from the volcanism. We obtain new insights into the mechanisms by combining new evidence from seismic tomography, petrological inversion of seismic data, and the geological data on uplift and volcanism. Optimal resolution tomography is a new approach, developed for surface wave tomography, that allows us to find the optimal resolving length at every point of a tomographic model grid. With this approach we evaluate the posterior model error at a point of the model grid empirically, estimating it by isolating the roughness of the phase-velocity curve that cannot be explained by any Earth structure. We apply this method to the region of Ireland and Britain, using more than 11,000 interstation phase-velocity curves measured at station pairs recording simultaneously, to image the lithosphere and underlying mantle beneath the area. The use of cross-correlation of teleseismic earthquakes and waveform inversion produces measurements in a very broad period range, leads to an unprecedented data coverage of the region, and allows us to unveil exciting new insights into the structure and evolution of the area, from the crust to the deep asthenosphere at a unprecedented level of detail.

The composite, optimal resolution phase-velocity maps are inverted for a 3-D V_S model, which reveals pronounced, previously unknown variations in the lithospheric thickness beneath the area. The model shows evidence of a robust, low-velocity anomaly beneath the Irish Sea and its surroundings that persists in the models from ~60 to at least 140 km depth, indicating an anomalously thin lithosphere and demonstrating that the assumption of a nearly constant lithospheric thickness across the area, previously adopted, is not valid. Phase velocity data at key locations are inverted using integrated geophysical-petrological inversion, to estimate the thermal structure of the lithosphere-asthenosphere system consistent with the seismic data, surface elevation, and heat-flow. The circum-Irish Sea area reveals a pronounced lithospheric thinning and matches the region of the Paleogene uplift previously suggested to be caused by a lateral branch of the Iceland mantle plume, which may have flowed into thin lithosphere areas surrounded by continental lithosphere during the evolution of the North Atlantic Ocean over the past 60 M.y. Our results show a striking correlation between lithospheric thickness and exhumation thermochronological measurements (as well as proposed underplating thickness, denudation, and the locations of the intraplate volcanism of the enigmatic North Atlantic Igneous Province) suggesting a significant lithospheric control on the volcanism of the area.

How to cite: Bonadio, R., Lebedev, S., and Chew, D.: Lithospheric control on the Paleogene uplift and volcanism in Ireland and Britain, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-7579, https://doi.org/10.5194/egusphere-egu23-7579, 2023.

Global-scale seismic velocity models of the Northern Canadian Cordillera show high velocities to the east of the Cordilleran deformation front and low velocities to the west.    This velocity contrast is consistent with other geophysical observables, such as regional seismological studies, that indicate a weak and thin lithosphere to the west that transitions quickly to a strong and thick craton-like lithosphere at the deformation front.    We present new results using data collected by the Mackenzie Mountains EarthScope Project, which included an ~875 km-long line of 40 broadband seismographs across the Cordillera and into the craton extending from roughly Skagway, Alaska to Great Bear Lake, Northwest Territories.    The 3-year overlap of this deployment with other broadband seismic stations in the region, most notably the EarthScope Transportable Array and the Yukon Northwest Seismic Network, allows for detailed 3-D Rayleigh wave ambient noise imaging of the upper lithosphere.    Results show large velocity variations west of the deformation front.   Notably, we image a 5% Vs low that extends from the upper crust to the asthenospheric mantle.   This plume-like structure, and associated weakening, may be a primary cause for the ongoing uplift of the Mackenzie Mountains at their unusually eastward location.   We also image a low velocity feature in the lower crust extending to the west of the deformation front, which may facilitate eastward crustal translation along a large-scale (~800 km) decollement system driven by the Yakutat indentor consistent with the orogenic float hypothesis of Mazzotti and Hyndman (2002).    We also note strong lithosphere-scale lateral heterogeneity suggesting that 3-D effects are important in focusing deformation in the Mackenzie Mountain area.

How to cite: Schutt, D., Porritt, R., Estève, C., Audet, P., Gosselin, J., Schaeffer, A., Aster, R., Freymueller, J., and Cubley, J.: Large Lithospheric Seismic Velocity Variations Across the Northern Canadian Cordillera Imaged by Ambient Noise Tomography, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-1214, https://doi.org/10.5194/egusphere-egu23-1214, 2023.

The most common way to estimate the porosity of the mature oceanic crust, and hence its contribution to geochemical exchange, is from the inversion of active source seismic data. Previous studies have suggested that hydrothermal activity effectively ceases in crust older than 10 Ma. This is based on observations that the seismic velocity of the uppermost crust increases rapidly within 10 Ma but changes little beyond that. The velocity increase is widely explained by reduced porosity and permeability due to hydrothermal mineral precipitation and fracture closure due to sediment blanketing. However, a potential problem with conventional wide-angle Ocean Bottom Seismometer (OBS) modelling over mature oceanic crust is the imaging geometry, where the water wave obscures the onset of the crustal refractions for tomographic inversion needed to resolve the velocity of the upper hundred meters of the igneous crust.

To investigate the issue of imaging geometry and accurately extract the physical properties of the upper crust, we applied downward continuation on conventional OBS records across 65 Ma Atlantic Ocean crust. The method eliminates the effect of the thick water column by locating shots on a datum close to the seabed. This enables refractions from the uppermost 100-200 m of igneous crust to be viewed as first arrivals, and hence significantly improves the accuracy of the velocity inversion of the upper layers. Using travel time picks from downward continued and original OBS records, we applied tomographic inversion with three starting models taken from the latest compilation of crustal velocity-depth (VZ) models. The three models have a velocity variation of ±10% for the uppermost crust, representing a low, mean, and high bound for crustal VZ relations. By comparing the results, we show that with downward continued data the inverted velocity of the uppermost crust is less dependent on the starting model and converges to the same trend closer to the low bound of previous VZ relations. The average velocity of the uppermost crust inverted with downward continued data is ~0.3 km/s lower than that inverted with original data. These results would translate into a porosity of 14%, compared to 10% for the non-downward continued analysis. We also resolve stronger along-strike variation in the inverted velocity of the uppermost crust (4.2 km/s to 4.8 km/s) using downward continued data compared to original data, which may correspond to porosity as large as 18%, much higher than previously suggested porosity of 8 % for mature oceanic crust. This implies that open cracks may be still present and thick sediments may seal hydrologically but not close fractures that affect the seismic refraction data. Moreover, our results reconcile with the recent work in the South Atlantic performed with full-waveform inversion (FWI) on streamer data. Therefore, downward continuation, also FWI may need to be incorporated into the workflow of conventional wide-angle OBS modelling, especially for mature oceanic crust. This will improve the resolution and accuracy of the physical properties of the uppermost crust and may shed new light on its evolution and role in off-axis hydrothermal circulation and seawater chemistry.

How to cite: Li, L., Collier, J., Henstock, T., and Goes, S.: Evidence for a higher porosity upper crust in the North Atlantic Ocean  , EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-4317, https://doi.org/10.5194/egusphere-egu23-4317, 2023.

The Indian Ocean Geoid Low (IOGL) is a significant negative geoid anomaly (-106 m), located south of the Indian subcontinent. Several studies have been carried out to investigate the causes responsible for IOGL, results indicate that its origin could be low velocity/density anomalies in the depth range of mid-to-upper mantle and/or high velocity/density anomalies in the depth range of lower mantle. However, a concrete model to explain the origin of IOGL and especially the effect of the shallow structure on IOGL is still enigmatic. In the present study, we investigate the high-resolution 3D shear velocity structure beneath the Indian Ocean region down to a depth of 300 km using surface wave tomography. For this analysis, collated extensive data from more than 700 broadband seismological stations of various data centres (IRIS: Incorporated Research Institutions for Seismology, FDSN: International Federation of Digital Seismograph Networks, IN: Indian Stations) to obtain a good azimuthal and spatial coverage. This dataset has been pre-processed and utilized to measure the fundamental mode of Rayleigh wave group velocities sampling the IOGL region. Further, visually checked the quality of the dispersion curves and considered only good-quality ones, which resulted in ~19,300 good-quality dispersion curves in a frequency range of 10-120 s and then applied the regionalization to extract the geographical distribution of local group velocities in different periods. Later, inverted the regionalized dispersion data using the trans-dimensional inversion approach. Regionalized shear wave velocity maps are in good agreement with surface tectonics such as low-velocity anomalies along the large-scale ridges and previous tomography studies. The western Indian Ocean shows very fast velocity anomalies at short-period (~20 s) and low-velocity anomalies in long periods (>100 s) which could be attributed to the magmatic underplating originating from the various hotspots and the presence of channelled flows asthenosphere towards eastward to spreading ridges. Further, the results rules-out that the contribution of the lithosphere could be negligible in explaining the IOGL as there are no low-velocity anomalies beneath the IOGL region. In addition, the obtained high-resolution regional surface wave tomography from this study enables the research community to obtain/measure the precise IOGL anomaly and understand the detailed tectonics and crustal/lithospheric deformation beneath the study region.

How to cite: Sundaran, S., Rao, B. P., and Maurya, S.: Crustal and Uppermost Mantle Structure beneath the Indian Ocean Geoid Low using Surface Wave Tomography, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-11063, https://doi.org/10.5194/egusphere-egu23-11063, 2023.

In recent years, the Marchenko method has proven to be a viable tool to create virtual seismic sources and receivers in the subsurface from reflection measurements at the surface. Applications range from suppressing internal multiples in seismic imaging to forecasting responses to induced seismic sources. One of the attractive aspects of the Marchenko method is that no detailed subsurface model is needed; a smooth background model suffices. All information needed to treat the internal multiples correctly comes from the reflection measurements at the surface. 

One of the underlying assumptions of the Marchenko method is that the seismic wave field can be decomposed into downgoing and upgoing waves at any position in the subsurface where one wants to create a virtual source or receiver. Although in many situations this implies no significant restrictions, it may hamper the imaging of steeply dipping flanks and it prevents the treatment of refracted and evanescent waves.

It can be shown that the Marchenko focusing function (the nucleus of the Marchenko method) can be expressed in terms of the so-called propagator matrix. The propagator matrix, which was introduced in geophysics in the nineteen-sixties for 1D systems and developed further in the nineteen-seventies for laterally varying 3D media, ‘propagates’ a wave field from one depth level to another. It does not rely on up-down decomposition and it accounts for propagating waves at all angles and for evanescent waves. By exploiting the link between the Marchenko focusing function and the propagator matrix, the applicability of the Marchenko method can be expanded. In the presentation we will review the underlying theory and discuss the potential application of the Marchenko method for refracted waves.

How to cite: Wapenaar, K. and Brackenhoff, J.: Pushing the limits of the Marchenko method, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-12936, https://doi.org/10.5194/egusphere-egu23-12936, 2023.

We attempt the reconstruction of the solid earth’s interior three-dimensional structure using seismic wave observations. The interior structure of the mantle deviates moderately from spherically symmetrical reference models and therefore seismological observables also vary moderately from spherically symmetrical predictions. Hence we consider here the linearized inverse problem of seismic traveltime tomography.

Usually, the solution is approximated in a fixed basis system: either global (e.g. polynomials) or local (e.g. finite elements) basis functions. Here we use a dictionary-based approximation approach, called the Learning Regularized Functional Matching Pursuit (LRFMP). A dictionary is an intentionally redundant set of diverse trial functions from which iteratively an approximation in a best basis is built. The next best basis element is chosen such that the Tikhonov functional is minimized.

The methods have been used for a variety of spherical as well as tomographic tasks from the geosciences as well as medical imaging. Here we apply them to seismic traveltime tomography for the first time. We discuss relevant developments and challenges in the process of tailoring the methods to the problem and show first promising results.

How to cite: Schneider, N., Michel, V., Sigloch, K., and Totten, E.: A dictionary-based approximation approach for seismic traveltime tomography, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-6435, https://doi.org/10.5194/egusphere-egu23-6435, 2023.

Seismic methods are one of the most powerful existing geophysical tools to extract information on the structure of the Earth’s subsurface. These techniques continue to be widely used to obtain images of the sediments and crust and to map the variations in physical properties. Particularly the P-wave velocity (Vp) distribution is the most commonly modelled property. 

How to cite: Jimenez Tejero, C. E., R. Ranero, C., and Sallares, V.: Seismic tomography software tools at Barcelona Center for Subsurface Imaging, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-15581, https://doi.org/10.5194/egusphere-egu23-15581, 2023.

We have witnessed a transit of seismic tomography from ray-based traveltime inversion to wave equation-based waveform inversion during the past two decades. As is widely known, full waveform inversion outperforms traveltime tomography in resolving velocity variations with dimensions comparable and smaller than the dominant wavelength. However, this may not be always true in real practice, mainly due to the presence of unknown data noise, the lack of accurate initial models for the iterative inversion process (including material properties and source mechanisms), and the high demand for computational resources. Further efforts are required to develop techniques for using full waveform contents in seismic imaging studies, especially for the use of high-frequency waveform data (>1 Hz on regional scales). We have attempted to use common-source double-difference traveltime data in wave equation-based adjoint tomography studies of subsurface structures beneath Northeast Japan and Alaska. Because of the high level of waveform similarity, reliable common-source double-difference traveltime data at neighboring seismic stations are measured via cross-correlation approach. Insightful results are obtained. However, it is still computationally prohibitive to model high-frequency data by solving 3-D wave equations, limiting the resolution of seismic images. We admit that there is still a long way to go for the possible wide application of high-frequency full waveform inversion.

Traveltime is the most reliable information that can be extracted from raw seismological recordings.  We have developed a new modality of seismic imaging, called adjoint-state traveltime tomography, to unleash the full potential of traveltime in imaging subsurface structures. To avoid potential failure of ray tracing in 3D complex media, isotropic eikonal equation and anisotropic eikonal equations are used to model seismic traveltime field in heterogenous and anisotropic media, and the associated inverse problems are solved by the efficient adjoint state method. No ray tracing is required for the novel adjoint-state traveltime tomography method. Importantly, the sensitivities of traveltime-related objective functions to material parameters can be accurately measured even in complex media. We view adjoint-state traveltime tomography as a new framework for seismic tomography, as it naturally implements various body wave and surface wave tomographic inversions in a very similar way. Good performances of the adjoint-state traveltime tomography method will be reported via various case studies in regions with typically different tectonic settings. It is worth noting that an accompanying software package, TomoATT, is under development.

How to cite: Tong, P., Chen, J., Nagaso, M., and Hao, S.: Adjoint-state traveltime tomography: A new modality of seismic imaging, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-6305, https://doi.org/10.5194/egusphere-egu23-6305, 2023.