The variogram is where geostatistics is usually decided and usually rushed. It encodes how quickly a property loses correlation with distance, and in doing so it controls the connectivity, the flow behaviour and ultimately the recovery predicted by any model built with it. Experimental variograms from sparse well data are noisy and frequently uninterpretable in the horizontal direction, which is precisely the direction that matters most for flow. Ranges are therefore often taken from analogue outcrops, from seismic, from depositional reasoning or from convention, and the basis for that choice deserves to be documented because the model's dynamic behaviour depends on it.

Kriging and simulation answer different questions and are frequently confused. Kriging produces the best local estimate at each location, which means it smooths: it reproduces neither the histogram nor the variogram of the underlying property. That is appropriate for mapping a surface and inappropriate for populating a flow model, because a smoothed permeability field has different connectivity and different flow behaviour from a realistic one. Simulation sacrifices local accuracy to reproduce the correct statistics and spatial structure, and produces many equally probable realisations rather than one.

Multiple realisations are only useful if they span the uncertainty that matters. Generating fifty realisations by varying the random seed while holding the variogram, the facies model, the trend and the histogram fixed produces fifty models that differ in detail and agree in everything consequential. Genuine uncertainty spanning requires varying the interpretive choices, and that means treating the variogram model, the facies scheme, the trend and the conditioning data as uncertain rather than as fixed inputs.

Finally, geostatistical models must eventually explain production. A model that honours every well and reproduces every statistic but cannot match observed pressure communication, breakthrough timing or connected volume is wrong in a way that the static statistics cannot detect. Dynamic validation is part of geostatistical model building rather than a subsequent activity performed by someone else.

By the end of this training, participants will be able to:

• Apply univariate and bivariate statistical analysis to reservoir data including declustering and transformation
• Calculate, interpret and model experimental variograms including anisotropy and nested structures
• Justify variogram parameters using geological reasoning and analogue information
• Apply kriging and its variants and explain why kriged fields are unsuitable as flow model properties
• Select and apply pixel-based, object-based and multi-point simulation methods appropriate to the depositional setting
• Model facies and populate properties conditionally within them
• Condition models to seismic data using co-simulation and collocated methods
• Generate realisation sets that span the uncertainties that matter to the decision
• Rank and select realisations for dynamic modelling
• Validate geostatistical models against dynamic data and revise them when they fail

The training develops each method through its mathematics and then applies it to reservoir data sets, with participants calculating variograms, fitting models, and generating and comparing realisations. The effect of each modelling choice is demonstrated directly by rebuilding models with alternative variogram ranges, facies schemes and simulation methods and observing the change in connectivity and flow behaviour. Kriged and simulated fields are compared statistically and dynamically. Models are validated against production data from the fields they represent, including cases where a statistically sound model failed dynamically.

Organisations sending participants to this training will:

• Improve static model quality and the dynamic behaviour of models built from them
• Produce uncertainty ranges from realisation sets that genuinely span the relevant uncertainty
• Reduce time lost to dynamic models that cannot be history matched because the static model is wrong
• Improve integration between geoscience and reservoir engineering in model building
• Strengthen technical review of geostatistical modelling produced internally and by contractors
• Improve documentation and reproducibility of reservoir modelling work

Participants will:

• Calculate and model variograms with geological justification
• Select simulation methods appropriate to the depositional environment
• Understand why models behave dynamically the way they do
• Generate realisation sets that represent real uncertainty
• Validate static models against dynamic evidence
• Build a quantitative capability that bridges geoscience and reservoir engineering

• Reservoir engineers building or using static models
• Reservoir geologists and modellers
• Geophysicists working on seismic conditioning of property models
• Petrophysicists supplying property inputs to models
• Simulation engineers requiring understanding of the static model basis
• Technical staff reviewing reservoir modelling studies
• Graduate geoscientists and engineers entering modelling roles

Upon successful completion of this training course, delegates will be awarded an official Certificate of Completion issued by the Petroleum Engineers Association (PEA), an ISO 9001:2015 certified training organization. The certificate carries 10 Credits and formally records the total learning hours completed.

Each certificate is signed by the Course Facilitator and the CEO of the Petroleum Engineers Association, and serves as verifiable proof of professional training that delegates can present to employers and professional bodies worldwide.