Toward self-describing and workflow integrated Earth system models: A coupled atmosphere-ocean modeling system application

Abstract

The complexity of Earth system models and their applications is increasing as a consequence of scientific advances, user demand, and the ongoing development of computing platforms, storage systems and distributed high-resolution observation networks. Multi-component Earth system models need to be redesigned to make interactions among model components and other applications external to the modeling system easier. To that end, the common component interfaces of Earth system models can be redesigned to increase interoperability between models and other applications such as various web services, data portals and science gateways. The models can be made self-describing so that the many configuration, build options and inputs of a simulation can be recorded. In this paper, we present a coupled modeling system that includes the proposed methodology to create self-describing models with common model component interfaces. The designed coupled atmosphere-ocean modeling system is also integrated into a scientific workflow system to simplify routine modeling tasks and relationships between these tasks and to demonstrate the enhanced interoperability between different technologies and components. Later on, the work environment is tested using a realistic Earth system modeling application. As can be seen through this example, a layered design for collecting provenance and metadata has the added benefit of documenting a run in far greater detail than before. In this way, it facilitates exploration and understanding of simulations and leads to possible reproducibility. In addition to designing self-describing Earth system models, the regular modeling tasks are also simplified and automated by using a scientific workflow which provides meaningful abstractions for the model, computing environment and provenance/metadata collection mechanisms. Our aim here is to solve a specific instance of a complex model integration problem by using a framework and scientific workflow approach together. The reader may also note that the methods presented in this paper might be also generalized to other types of Earth system models, leading to improved ease of use and flexibility. The initial results also show that the coupled atmosphere-ocean model, which is controlled by the designed workflow environment, is able to reproduce the Mediterranean Sea surface temperature when it is compared with the used CCSM3 initial and boundary conditions.

Introduction

With the continual advances in high-performance computing systems and observation networks over the last few decades, scientific tasks related to modeling the Earth system have become increasingly complex and computationally demanding. Examples of the complexities in modeling systems include components with non-standard programming interfaces, preparation or processing input data with various file formats and grid types, running a modeling system on a variety of computational resources (computational grids, cluster or shared memory systems), keeping track of the model parameter changes, and analyzing results with ad hoc metadata structures. To reduce these complexities and promote understanding, programming interfaces among different Earth system models must be standardized and self-describing modeling systems must be designed to achieve increased interoperability among models and external applications and technologies such as scientific workflow systems, metadata/data portals, web services, scientific gateways, etc.

In this context, ‘self-describing’ means that metadata and provenance information about the model components, inputs, and parameters used in a simulation are produced along with data output. A similar concept is widely used in data formats such as NetCDF¹ and HDF² that are specialized for Earth system science and remote sensing. The convention for ‘self-describing’ models basically provides a definitive description of model and its parameters, metadata about transferred data among different model components, input and output fields, build system and runtime environment. Using common conventions for designing ‘self-describing’ models promotes easy to use and efficient Earth system models for users and developers. Because the implementation of this concept in Earth system modeling is a relatively new approach, the development of tools and methodologies that can enable the design of self-describing Earth system models is still an open research area. Fortunately, many efforts are currently underway to create easy to use multi-component Earth system models with common component interfaces and their applications. Basically, these efforts can be categorized in two main groups: modeling frameworks and scientific workflow systems.

A modeling framework is an environment for coupling model components and couplers of different kinds of Earth subsystem models through a common calling interface. The main advantage of the modeling framework approach is that it reduces the complexity of the regular tasks (i.e. interpolation of different grids, transferring data among model components) to design coupled modeling systems and helps to increase the efficiency and interoperability of the different model components. It also simplifies the synchronization of the execution of individual model components and the exchange of data/metadata among them. The Earth System Modeling Framework (ESMF) is one of the most popular examples for this approach. The ESMF consists of a superstructure for coupling components of Earth system applications in a standardized way and an infrastructure of robust, high-performance utilities and data structures that ensure consistent component behavior (Hill et al., 2004, Hill et al., 2006; Collins et al., 2005). In this context, standardization means that the multi-agency consortium responsible for developing ESMF agreed to conform to a set of specific interfaces in order to improve the interoperability of their models. Unlike many other examples of the framework approach such as Model Coupling Toolkit (MCT; Jacob et al., 2005; Larson et al., 2005), Model Coupling Environmental Library (MCEL; Bettencourt, 2002) and OASIS (Redler et al., 2010), ESMF also provides the ability to store and export component (physical model and coupler), grid and field-level (physical variable such as temperature) metadata as XML and other documents. The latest public release of ESMF (5.2.0r) also allows use of different metadata conventions such as Climate and Forecast³ (CF), ISO standards and METAFOR⁴ Common Information Model (CIM) to store and write the gathered metadata. The main goal of the European Union (EU) founded project METAFOR is to describe climate simulations. The 5th Coupled Model Intercomparison Project⁵ (CMIP5) is the latest example to show the importance of the usage of such conventions (i.e. CIM) in Earth system science applications. It is clear that this feature can be used to create preliminary examples of self-describing Earth system models that conform to popular standards and conventions, which are also used in portals. In this way, it also achieves interoperability between designed self-describing Earth system model and data/metadata portals (i.e. Earth System Grid⁶ – ESG).

In recent decades, the popularity of community-based modeling systems has grown so that multiple groups can address difficult modeling problems together. A community-based modeling system is defined as an open-source suite of modeling components coupled in a framework (Voinov et al., 2010). The technologies used to support these systems are improving, but many still face challenges with respect to interoperability and metadata. For example, the Community Modeling and Analysis System⁷ (CMAS) include many different components specific to air-quality modeling applications such as atmosphere and air-quality models, emission-processing tools etc. To couple the Community Multiscale Air Quality (CMAQ) model with the MM5 (Grell et al., 1995) atmosphere model, CMAS uses the MM5 Meteorology Coupler. The main disadvantage of this approach is that the coupler is designed to work with a specific atmosphere model and the user needs to develop new couplers to work with other models such as WRF (Michalakes et al., 2001; Janjic et al., 2001). It is clear that using generic coupler libraries that provide common component interfaces can help to solve this interoperability problem. The Community Surface Dynamics Modeling System⁸ (CSDMS) is another example of a community modeling system. The CSDMS system is based on the Common Component Architecture (CCA; Armstrong et al., 1999). The CCA can create generic calling interfaces for components based on wrappers, and can work with a variety of different tools underneath, including ESMF and MCT. The CSDMS also includes a graphical user interface to link different modeling system components. This system displays many advantages of the framework approach. However, metadata must still be captured manually and is not integrated into the modeling system.

In contrast to a modeling framework approach, scientific workflow systems create generic interfaces to a variety of technologies such as job schedulers, authentication mechanisms, data transfer protocols etc. and automate the execution and monitoring of a heterogeneous workflow (Altintas et al., 2006). A scientific workflow system is defined as a problem-solving environment that simplifies tasks by creating meaningful, easily understandable sub-tasks/modules and combining them to form executable data management and analysis pipelines (Bowers and Ludäscher, 2005; Ludäscher et al., 2006). It acts as an abstraction layer to keep the details of the computing environment, modeling system and external applications (i.e. pre-/post-processing and visualization application) away from the user. Through their use, interoperability of different technologies can be achieved to create an easy to use work environment for Earth system modelers.

There are several scientific workflow applications in existence such as Kepler (Ludäscher et al., 2006), Taverna (Oinn et al., 2004), Triana (Majithia et al., 2004), Trident (Barga et al., 2008) and VisTrails (Bavoil et al., 2005). In addition to standalone scientific workflow systems, the Linked Environments for Atmospheric Discovery (LEAD; Plale et al., 2006) project demonstrates how workflows can be used to solve problems specific to Earth system science by integrating together various technologies such as web and grid services, metadata repositories, and workflow systems. The LEAD portal integrates a sophisticated set of tools to enable users to access, analyze, run, and visualize meteorological data and forecast models, facilitating an interactive study of weather. In the LEAD portal, the XBaya⁹ workflow composer is used to generate BPEL (Business Process Execution Language) workflows and compose web services. After the success of the NSF funded LEAD project, LEAD II is designed to be a follow-on project. As such, the LEAD project is an important example that illustrates how scientific workflows can be used to simplify interfaces to Earth science applications and how these different technologies can be used to address a specific problem that includes many different tasks. While workflows have shown great value, a common disadvantage of the approach is that the steps for creating connections between models and the scientific workflow application are not straightforward and easy if the modeling applications used do not have common interfaces. As described in the beginning of this section, the framework approach might help to design Earth system models that have common interfaces that can be used to create standardized connections among models and scientific workflow applications.

The scientific workflow system also creates a work environment in which one can collect and archive provenance information about specific model runs. Provenance is defined as structured information that keeps track of the origin and derivation of the modeling applications (Bowers et al., 2006; Klasky et al., 2008). The collected information can be used to compare, reproduce, tune, debug or validate a specific set of simulation runs. The combination of provenance information and metadata collected from the modeling system itself can be also used to create self-describing Earth system models. The stored provenance and metadata information can be parsed and queried to reproduce the work environment (operating system, compilers, model and its configurations) that produced a particular result.

Provenance information has been categorized as system, workflow, data and process provenance (Bowers et al., 2006; Klasky et al., 2008). Briefly, system provenance is defined as data about the computing environment or host system in which the job executes. It includes information about the operating system, environment variables and libraries that are used and/or defined by the application, and system software and model versions. With this information, the user can reproduce the work environment later on. In addition to system provenance information, workflow provenance is the data about the evolution and structure of the workflow itself, i.e. different versions, designs, and internal structures of the workflow. The collected workflow provenance information can be stored as an incremental change or complete new version but the main structure of the data and the storage type (i.e. relational database, ASCII, XML) can vary based on the provenance collection tool and scientific workflow application itself. As mentioned above, Kepler and VisTrails workflow applications are able to capture the workflow provenance at different levels of detail but the proposed coupled modeling system needs additional tools to collect system and process provenance information from the model itself. In this study, we are only interested in recording system and workflow provenance information related to model simulation, not data and process provenance in the conventional sense.

The purpose of this paper is to demonstrate the advantages of using a modeling framework and scientific workflow approach together to create a self-describing Earth system modeling workflow, which provides capabilities that existing modeling workflow systems do not. The proposed modeling system and work environment aim at increasing the interoperability of different technologies. The automatic collection of metadata and provenance information also simplifies the interaction between the user and the modeling system. The provenance information enables users to track back from the results to debug and reproduce the simulation. Our contribution is developing a methodology for integrating all the necessary pieces to create this new type of workflow. To test the capabilities of this modeling environment, we use a particular instance of a complex model integration problem that includes three main components: an Earth system model, a scientific workflow application and a coupler library to create a coupled modeling system and collect metadata information from the models that were used.

The rest of this paper is organized as follows: in the next section, we describe the proposed methodology used to create a self-describing coupled modeling system and embed it in a scientific workflow application. The next section also introduces information about the workflow application, outlining the components and describing a prototype system that was constructed using the proposed methodology. In Section 3, we present detailed information about use case application and its components. In Sections 4 Results, 5 Conclusion and future work, we conclude and discuss possible future directions.