Join us at the 2011 AGU Fall Meeting in San Francisco, CA for a demonstration of the MAD plug-in.
December 6, 2011 -1:40 PM to 6:00 PM
Session: H23D -- Computational Sciences and Water Applications in Hydrology and Groundwater Management
Posters: 1295 & 1297
Daniel P. Ames, Matthew W. Over, Carlos Osorio, & Yoram Rubin,
The Method of Anchored Distributions (MAD): Principles and Implementation as a Community Resource
We propose to develop a community-based, open-source computational platform for Bayesian inverse modeling and conditional simulations in hydrology. It is motivated by the need to provide easily-accessible tools for applications and a platform for broad-based,
long-term development, built to meet the challenges brought upon by the ever increasing complexity of modern computational tools, the diversity of data types and the breadth and depth of the subject matters needed for applications. The initiation of such an
effort could emulate the success experienced by other communities with a long tradition of community-based development efforts, and lead to a transformative change in hydrology. The absence of such an approach is hurting the science discovery process.
Our proposed project includes two elements. The first is a general statistical theory for inversion and uncertainty analysis. The second is an application and testing of the theory through a modular, open-source computational platform. Our approach is built
around two concepts that make it applicable to a wide range of data types and computational tools. The first concept is a classification of data into two broad categories, which allows us to formulate our theory around data sets and not around data types.
The second concept is a strategy for localization of data using anchors (hence Method of Anchored Distributions, or MAD) that captures the relevant information contained in a local or non-local process, respectively, and presents it in terms relevant for the
target variable(s).
We will construct a modular computational platform, with 3 distinct modules, including a pre-processing block, a likelihood (forward simulations) block and a post-processing/prediction block. The three modules combined comprise the MAD kernel. A user-interface
will allow users and developers to link the kernel with their models. Decisions over assumptions to be adopted or computational tools to be employed will be at the user’s discretion and will not be hard-wired into the kernel. The kernel will be built
using open-source architecture. Two versions of MAD will be available: a web-enabled version (for educational and training purposes) and a standalone (for large-scale applications at the user’s computers). The kernel will be interoperable with a wide
array of community-based resources (such as CUAHSI’s hydrologic information systems) and libraries (such as R). With these concepts and principles, various types of data can be assimilated, simplifying assumptions can be adopted at the user’s discretion,
and powerful numerical models and analytical tools, current or under development, could be used as a focal point for long-term community-based development.
We developed and secured a plan for the continuity of this effort after the development phase of this project is concluded. We will deliver the final product to CUAHSI. CUHASI agreed to act as the custodian of the MAD computational platform. This implies
securing the integrity of the kernel, updating it and making it available to users and developers. This effort is in line with CUAHSI overall strategy, and CUAHSI has the infrastructure to see it carried out.