Resource Modeling Solutions Platform

Resource Modeling Solutions Platform (RMSP) is the python package for modern geostatistical modeling. Contact us for more information.

What is RMSP and how does it work?

The Resource Modeling Solutions Platform (RMSP) is a library for solving resource modeling challenges. RMSP implements core geostatistical modeling algorithms for the modeling and management of uncertainty in the earth sciences. RMSP is used by leading mining, petroleum, and consulting companies to quantify uncertainty, rapidly generate resource estimates, and apply modern geostatistical algorithms to their projects. RMSP enables you to use the right tools for your geostatistical modeling challenge, whether it is determining how many drill holes you need for your model to be within 15% on an annual basis, or generating short term models on the fly as new production data lands in your database.

RMSP provides more than 100 geostatistical modeling functions, from generating block models around your data, to variogram calculation and modeling, estimation with kriging, uniform conditioning, indicators, to advanced simulation technology with hierarchical truncated plurigaussian simulation and projection pursuit multivariate transform simulations. A tiny example using a set of copper blast holes:

RMSP works within the Python ecosystem so you can integrate directly with other data analysis and machine learning libraries. Interaction happens with Python, but underneath it is a fully parallel high performance library enabling you to model hundreds of millions of blocks for dozens of variables simultaneously. RMSP can be used on your laptop or on powerful cloud virtual machines and distributed environments. Set up a simulation locally then push it up to your corporate cloud to crunch hundreds of realizations and quantify uncertainty in your deposit on a Windows or Linux environment. RMSP is backed by our team of consultants who can assist with implementing and auditing workflows.

RMSP is fully documented with hundreds of small examples and more than 60 fully worked examples covering resource modeling challenges ranging from calibrating an ordinary kriging estimate for resources, to applying localized conditional simulation, to generating fully locally varying variograms and generate short term models for mine planning. When RMSP is combined with scripting interfaces such as Jupyter Notebooks, RMSP provides transparent and auditable workflows that are automated to the correct extent for rapid model deployment. Please refer to Highlights for a few more examples of the RMSP python wrapper, or get in touch to try RMSP for yourself on a cloud environment.

What can RMSP do?

RMSP is compliant with best practice estimation using legacy techniques such as variograms and kriging techniques, though it is aimed at modern probabilistic resource modeling. It seamlessly handles:

  • Multiple data types, vintage, data with error and missing data
  • Realistic decisions of stationarity, including local anisotropy
  • Hierarchical structural, large scale and stochastic categorical modeling
  • Multivariate and multiscale modeling with metallurgical and mechanical rock properties
  • Uncertainty in all modeling parameters
  • Uncertainty in all modeling outputs
  • Optimization with appropriate utility
  • Version and change control for data, parameters and models
  • Disclosure of probabilistic resources

Specific RMSP features include:

  • Compositing
  • Desurveying (minimum curvature and tangential)
  • Duplicate removal
  • Reading and writing RMSP files including simulation caches
  • Regular and subblock grid creation and slicing
  • Mesh flagging and calculations
  • Spatial rotations and translations

Exploratory Data Analysis tools:

  • Despiking (random, multivariate, spatial)
  • Declustering (cell or any arbitrary estimation weight)
  • Gaussian mixture models
  • Univariate and multivariate statistics including outlier analysis
  • Normal score transformation
  • Point signed distance calculation
  • Variograms
  • Data spacing calculation
  • Change of support (affine, corrected indirect lognormal, discrete Gaussian model with diffusion or mosaic assumption)
  • Boundary analysis (contact and swath calculations and plots)
  • Categorical proportions and transition probabilities
  • Paired bias analysis

Estimation functions:

  • Kriging (simple, ordinary, intrinsic collocated, local anisotropy)
  • Multiple indicator kriging, localized indicator kriging, and indicator post-processing
  • Cross validation with leave-one-out, leave-one-hole-out, etc
  • Nearest neighbour
  • Inverse distance
  • Spatial search queries
  • Cokriging (full)
  • Multipass and high-yield estimation
  • Trend estimation with kernels
  • Moving averages and other spatial filters
  • Grid averaging (regularization)
  • Multigaussian kriging post-processing
  • Localized uniform conditioning

Simulation and additional advanced functionality:

  • Linear decorrelation transforms including principal component analysis (PCA), sphereing, and min/max autocorrelation factors (MAF)
  • Logratio transformations
  • Non-linear decorrelations including the stepwise conditional transformation with Gaussian mixture models (GMM), and projection pursuit multivariate transform (PPMT)
  • Unfolding using lower and upper triangulations
  • Bootstrap and multivariate spatial bootstrap
  • Trend transformation
  • Continuous realization filtering and snapping transformations
  • Constrained imputation of continuous features
  • Maximum a-posteriori transformation (MAPS)
  • Multiple imputation for spatial data with complex multivariate relationships
  • Hierarchical truncated plurigaussian transforms
  • Domain boundary simulation
  • Model sampling (redrilling for drill hole spacing studies)
  • Tabular vein simulation
  • Turning bands simulation
  • Turning bands simulation (simple, intrinsic collocated)
  • Ore waste proxy filters (for drill hole spacing studies) and spatial mixers (for block caving or similar)
  • Simulation histogram correction filters
  • Simulation post-processing for realization summaries
  • Multivariate conditional simulation localizations
  • Simulation validation (checking, accuracy plots, multivariate relationships, histograms, variograms)

Composites with shells

Interested in applying RMSP to your resource modeling project? Our team would love to hear from you.