Back to Digital Crust Home

One of the goals of the Digital Crust project is to link a dataspace containing observational thematic geoscience data with a flexible framework for representing locations within the Earth. This is essentially a cloud and web-based, geoscience-oriented geographic information system. The Digital Crust Data Framework page describes our approach to handling the heterogeneous realm of observation data that is the basic for characterizing the distribution of materials (Geologic Formations, aquifers, mineral deposits, etc.) within the Earth. As a starting point to demonstrate the framework for linking observation data with geometric models of the Earth's interior, we are working on a project to collect observation data related to the permeability or hydraulic conductivity of Earth materials, and constructing a framework to extrapolate from a selection of data in the dataspace to assign hydraulic conductivity values to grid cells to construct a continental-scale model of upper-crustal hydraulic conductivity.

For this demonstration, we will be using a very sparse data collection, and the resulting model will be very approximate. This initial model is not our goal -- the point of our Digital Crust building block is to build the scaffolding that would allow geometric models to be updated continuously as new data are made available in the dataspace. The Digital Crust system will be designed to allow continuing evolution, for example adding new geometric model types (3-D solid object models, irregular grid geometries, multiscale grids…), improving algorithms for automating the extrapolation from observation data to a 3-D model, or adding inverse capabilities to work from geophysical remote sensing data backwards to spatial distributions of materials. We will work with hydraulic conductivity, but the technique is intended to work with any physical property.

This section outlines the approach we are taking to this initial observation to model workflow.

 

Figure 1. Macrostrat spatial model.

Define polygons for areas that can be approximated by a single stratigraphic column.  Macrostrat was originally developed using time-stratigraphic columns, representing the sequence of sedimentary deposition within the polygon, irrespective of deformation, metamorphism or intrusion.   For the purposes of representing material properties, a columnar model representing the lithostratigraphy is necessary.  In regions with complex histories involving faulting, metamorphism, and intrusion, the columnar model of the material crust will be quite differernt from the time-stratigraphic column representing the depositional history. 
Thus, for the initial demonstration, we are restricting our model to parts of mid-continent region of United States (shown in Figure 1), where stratigraphy can be approximated with a layer cake.
The original Macrostrat polygons were defined based on a voronoi tesselation of the continent using point locations assigned to the time stratigraphic columns defined by COSUNA (1985).  Because we are interested in the material stratigraphy, we have updated the polygon boundaries to reflect geologic domains and boundaries more accurately, using the collection of digital versions of state geologic maps compiled by the USGS mineral resources group (see http://mrdata.usgs.gov/geology/state/). 

 

Figure 2. Some examples of tabular data containing hydraulic property information.

For the demonstration, we are interested in hydraulic conductivity. Hydraulic conductivity is functionally related to permeability at a given value for viscosity of water (a function of temperature and water chemistry). We have located a variety of datasets containing hydraulic conductivity or permeabilty value observations. The measurements are indexed to a particular location (e.g. depth interval in a well), formation, or rock type, and reported in tables that vary widely. Heterogeneity includes different field names, table layouts, observations keyed to location, lithology, or formation, and different but related properties measured (intrinsic permeability, local permeability, hydraulic conductivity…). 

These heterogenesous tabular data will be fed into the transformation pipeline (Figure 1 at http://earthcube.org/digitalCrustDataFramework) to generate AVRO formatted data that can be queried to access the observation values. The transformation process includes inspect the field names and values to propose mapping to known data types in a prototype data type registry (Figure 3). The mappings will be inspected by project team members to work with developers refining the algorithms used to suggest mappings. A query process will be designed to request data for each grid cell, first by location, then by formation name, then by lithology to obtain the best estimate.  Initially extrapolation will be done using a nearest neighbor assignment, but in the future, more sophisticated extrapolation algorithms will improve results.

Figure 3. Assignment of hydraulic conductivity values to grid cells.