About me

Rocky Dunlap is a research scientist in Georgia Tech’s College of Computing. His research interests including software engineering for scientific applications, software and data modeling, software architecture, generative programming, frameworks for building scientific applications, and cloud computing for e-science.

Klaus Advanced Computing Building, Room 3319

email address


I have wide-reaching research interests at the intersection of scientific computing and software engineering.  The overall thrust of my research is to help computational scientists do their jobs better.  This involves understanding how computational scientists work and how software engineering principles can be applied.  So far, my primary working domain has been numerical climate modeling.

Specific interests include:

  • Cyberinfrastructure for e-science
  • Software frameworks for writing Earth System Models
  • Couplers for numerical models
  • Software architecture of climate models
  • Configuration of climate models
  • Automatic code generation of scientific infrastructure
  • Metadata for climate models and datasets
  • Scientific computing on the Cloud

Software Infrastructure for Earth System Models

Not everything inside a scientific model is science.  Much of the code serves as infrastructure–that is, software capabilities that are in service of the scientific requirements.  Examples of infrastructure include software that standardizes internal data structures,  manages data transfer between two or more coupled components, and performs grid interpolation functions.  The thrust of this research is to characterize the relationship between the infrastructure and non-infrastructure parts of Earth System Models and to determine how to enable scientists to shift focus from infrastructure development to doing science.

Couplers are well-known abstractions in the geophysical and other scientific communities.  Because coupling numerical modeling components is a common need, a number of infrastructure technologies have emerged in the form of reusable software assets to facilitate building coupled scientific applications.  To understand better what services these coupling technologies provide, we have done a feature analysis of those coupling technologies currently used for building global climate simulations.  (more info)

Configuration of Climate Models

Numerical climate simulations are designed to be used by many scientists and in multiple contexts.  As such, the models are designed to be configured to meet the full range of requirements of the entire user base.  Therefore, a large portion of time spent working with models involves configuration–for example, which components to couple together, which initial and boundary conditions to supply, which grids to use, how often to communicate coupling fields, etc.  Configuration can be viewed as more than a mere technical task–it is the process of defining an experiment.  This works aims to get a handle on how configuration of climate models is currently done and how the configuration process can be improved.

Climate Models on the Cloud

This research seeks to answer the question of to what degree emerging Cloud platforms and application environments can be leveraged by the numerical climate modeling community.  The hypothesis is that by abstracting away details of the underlying computational environment (i.e., hardware, operating system, compiler, etc.) it becomes much easier for scientists to share simulations and leverage the work of other scientists.

We are in the process of building a prototype Cloud configurator for deploying the Community Earth System Model using Amazon EC2 cloud services.  (more info)

Metadata for Climate Models:  Earth System Curator

The Earth System Curator, sponsored by NSF and NASA, is developing metadata-rich software infrastructure in support of end-to-end modeling in the Earth sciences. The primary motivating observation of the project is that a model’s source code plus the configuration parameters required for a model run are a compact representation of the datasets generated when the model is executed. The end goal of the project is a convergence of models and data where both resources are accessed uniformly.


Rocky Dunlap, Leo Mark, Spencer Rugaber, V. Balaji, Julien Chastang, Luca Cinquini, Cecelia DeLuca, Don Middleton, and Sylvia Murphy. Earth System Curator:  Metadata Infrastructure for Climate Modeling.  Earth Science Informatics 1(3-4): 131-149, 2008.  DOI 10.1007/s12145-008-0016-1.  (Available via SpringerLink).