WP4 Numerical


stochastic modelling

training at

British Geological Survey







VENUE: BGS‑UKRI, Environmental Science Centre, Keyworth, Nottingham NG12 5GG, UK

Attendees: Staša Borović (HGI-CGS), Ozren Larva (HGI-CGS)

Lecturers: Marco Bianchi (BGS-UKRI), Dr Corinna Abesser (BGS-UKRI) and Dr Richard Haslam (BGS-UKRI)

Duration: May, 28 – June, 07 2019

The first week of the training, from 28th to 31st May 2019, was dedicated to geostatistical approach and methods (deterministic and stochastic) for modelling spatial distribution of hydrogeological parameters. The trainees were introduced to the basics of SGeMS (The Stanford Geostatistical Modelling Software), which is a user-friendly, state-of-the-art geostatistical software package used for general purposes. The training was performed using real-project data (MADE – Macro Dispersion Experiment) within SGEM software environment. After the trainees got familiar with workflow within SGeMS, they were introduced to Gaussian simulation methods and were exercising using MADE data for sequential Gaussian simulation within SGeMS. After covering sequential Gaussian simulation within SGeMS, introduction to methods for modelling categorical variables, such as two-point indicator based methods, and Markov chain / transition probability approach was presented. In the beginning participants learnt how to apply Indicator kriging and Sequential indicator simulation methods using hard and soft data within SGeMS. During the rest of the training week, participants were presented with detailed presentation of Markov chain / transition probability concept, within which, unlike the previously described methods that rely on the variogram as the model of spatial correlation, the spatial structure of categorical variable is represented by transition probabilities.

At the end of the first training week, trainees were introduced to the concept of Multiple Point Statistics (MPS), which aims at overcoming the difficulties of 2-point geostatistical methods to represent geological heterogeneity in the most accurate way.

During the second week of training, from 3rd to 7th June 2019, courses in fault and fracture network descriptions (held by Dr Richard Haslam) and numerical modelling (held by Dr Corinna Abesser) were organised.

Fault modelling and stress-field component description were conducted on the example of the karst island Vis in Croatia using (a) World Stress Map data and (b) existing structural measurements from Vis. The course was short, but informative, and their work has shown the congruence of stress definitions derived by both applied methodologies. In the next days the group was dedicated to numerical fluid flow and heat transport modelling using FEFLOW code. The trainees were presented with the functionalities of the software and learned how to make representation of 2D and 3D environments and conduct some simple numerical simulations. FEFLOW is a powerful tool which can be applied in the future for local- and regional-scale modelling of fluid flow and / or heat transport (e.g. geothermal aquifers, hydrothermal systems, heat pump systems).

On the last day the hosts also provided a guided tour through the BGS Core Store and the presentation of 3D visualisation room.

The attendees, Dr Staša Borović and Dr Ozren Larva from the Department of hydrogeology and Engineering Geology of the HGI-CGS, would like to thank the organisers for their cordial welcome and a pleasant stay.