A Data Driven Framework for Integrating Regional Climate Models

There are increasing needs for research addressing complex climate sensitive issues of concern to decision-makers and policy planners at a regional level. Decisions about allocating scarce water across competing municipal, agricultural, and ecosystem demands is just one of the challenges ahead, along with decisions regarding competing land use priorities such as biofuels, food, and species habitat. Being able to predict the extent of future climate change in the context of introducing alternative energy production strategies requires a new generation of modeling capabilities. We will also need more complete representations of human systems at regional scales, incorporating the influences of population centers, land use, agriculture and existing and planned electrical demand and generation infrastructure. At PNNL we are working towards creating a first-of-a-kind capability known as the Integrated Regional Earth System Model (iRESM). The fundamental goal of the iRESM initiative is the critical analyses of the tradeoffs and consequences of decision and policy making for integrated human and environmental systems. This necessarily combines different scientific processes, bridging different temporal and geographic scales and resolving the semantic differences between them. To achieve this goal, iRESM is developing a modeling framework and supporting infrastructure that enable the scientific team to evaluate different scenarios in light of specific stakeholder questions such as "How do regional changes in mean climate states and climate extremes affect water storage and energy consumption and how do such decisions influence possible mitigation and carbon management schemes?" The resulting capability will give analysts a toolset to gain insights into how regional economies can respond to climate change mitigation policies and accelerated deployment of alternative energy technologies. The iRESM framework consists of a collection of coupled models working with high resolution data that can represent the climate, geography, economy, energy supply, and demand of a region under study; an integrated data management framework that captures information about models, model couplings (workflows), observational and derived data sets, numerical experiments, and the provenance metadata connecting them; and a collaborative environment that enables scientific users to explore the datasets, register models and codes, launch workflows, retrieve provenance, and analyze results. In this presentation we address the challenges of coupling heterogeneous codes and handling large data sets. We describe our integration approach, which is based on a loosely coupled software architecture that supports experimentation and evolution of models on different datasets. We present our software prototype and show the scalability of our approach to handle a large number ( > 17,000) of model runs and a significant quantity of data in the order of terabytes. The resulting environment is now used by domain scientists and has proven useful to improve productivity in the evolving development of iRESM model coupling.