Loading...
AR2GAS
Geostats as a Service
AR2GAS (Geostats As a Service) is a newly-designed geostatistical library to support operations that are data-centric and rely on machine-learning. It can run both traditional and advanced geostatistical algorithms (kriging, cokiring, simulation and co-simulation) on the cloud and on the desktop. These capabilities can be applied on cartesian grids or specialized gridding systems such as stratigraphic or unstructured grids. AR2GAS can be used as a stand-alone python library or be integrated into a third party software.

FLEXIBILITY & INTEGRATION

AR2GAS is a series of geostatistical libraries with C++ and Python APIs. All the algorithms can be run from a script or a Jupyter notebook, locally or on high-performance servers. AR2GAS is ready to be integrated into machine learning workflows for state-of-the-art modelling.

DESIGNED FOR SPEED

AR2GAS is the result of a complete overhaul of our geostatistical libraries. Everything was designed from scratch to work efficiently both locally or on distributed computing systems. The same geostatistical workflow can now be executed on the desktop, the cloud or on private clusters. On the Google cloud, AR2GAS can simulate billion nodes properties in minutes.

CLOUD READY

Users are able to write and test python code on the desktop then run it on the cloud simply by adding the web address of additional cloud hosted workers. The complexity of the cloud is hidden from the users.

HIGHLY CUSTOMIZABLE

AR2GAS algorithms are designed to work across scales, gridding systems and with data that varies in density and quality. Because geomodeling situations and objectives vary, AR2GAS offers a Python and Jupyter interface to access its building blocks. The user has full freedom to create custom geomodeling workflows at any level of complexity.

Features

Key components are:

CONNECT TO DATA SOURCE AND DATABASE

AR2GAS is built to connect to an external data source and only extract data as needed for efficent workflows on very large models (billion of cells). The data stream is optimized for each algorithm providing performance and scalabiilty.

RUN ALGORITHMS AS PROCESSES

Easily build complex modelling algorithms in Python as processes with strong-typed parameters. This feature greatly decreases the chance of user errors when initializing algorithms and provides robustness during execution by isolating the execution of geostatistical algorithms from the main program. AR2GAS comes with processes for its geostatistical algorithms. Additionally, a user can build their own processes and inject them to remote (cloud) workers for execution.

STREAMING DATA TO ALGORITHMS

Whenever possible (e.g. kriging and turning bands) AR2GAS streams data to and from the data source to the algorithms. This allows 1) to avoid loading a large data set into memory 2) keep the memory requirement low, 3) with little or no compromise on performance on the desktop and 4) improve performance on the cloud.

UNSTRUCTURED GRIDS

With grid blocks of various geometric shapes and sizes, unstructured grids offer a better representation of geological complexities seen in deposits. The current standard of building block models with rectangular cuboids forces simplifying engineering concepts onto geological modeling processes. Unstructured grids separate the geologist's needs from the regularity of engineering grids though the use of flexible modelling frameworks.

GRIDLESS GEOSTATISTICS

AR2GAS provide an environment where a space is fully informed by geostatistical values. The user is then free to retrieve information on the support needed, be along a well, in a grid (cartesian, stratigraphic, or unstructured) or on any user defined volume. The geostatistical space ensures that all the values retrieved on any support are always consistent with one another and with all available data.

ACCESS TO GEOSTATISTICAL BUILDING BLOCKS

In addition to providing geostatistical algorithms, AR2GAS gives access to the building blocks of the algorithms. The Python interface provides access to low-level functionalities such as variograms, search neighborhood and kriging systems. Advanced users can build their own geostatistical algorithms or incorporate these low-level components into machine learning applications.

CREATION OF INTERACTIVE WEB-BASED DASHBOARD

AR2GAS creates dashboards to automatically analyse models and provide validation. These dashboards are designed to be actionable and provide real insight about the validity of a geostatistical models, all from an internet browser.

GEOSTATISTICS FOR EMBEDDED SYSTEMS (IOT)

The AR2GAS library can run on small device (such as Rasberry Pi) to provide advanced spatial algorithms to embedded systems. This is in active development.


About AR2Tech

Since 2010, AR2Tech offers geostatistics libraries and applications.

We provide cloud and server-based geostatistics for your geomodeling solutions.

Email Us!

Founder and CEO