The course presents what is at stake in resources classification and assessment of the estimates uncertainty. The kriging variance is a convenient tool for comparing different sampling strategies, but it is not an appropriate tool for the derivation of a true confidence interval, because the kriging variance is not conditioned to the data. A better solution consists in computing many conditional simulations, but this is time consuming. To obtain a confidence interval, simple models, like the discrete gaussian model may be sufficient and give an acceptable answer for the purpose of resources classification.
- Reminders on resources and reserves notions.
- Concepts of variances calculated in geostatistics (dispersion variance, estimation variance).
- The kriging variance and its properties.
- Data distribution and kriging.
- Gaussian based methods: normal score transformation using the anamorphosis function.
- Change of support within the discrete gaussian model.
- Calculation of the confidence intervals using the discrete gaussian model.
- Geostatistical simulations for achieving the same tasks and comparison with the previous results.
- Available tools for analysing the results.
- Discussion on the limitations of the method and solutions.
Half of the course is dedicated to practical computer exercises, using Isatis, that reinforce the previously presented theoretical notions.