A Scientific-Workflow-Based Execution Environment for Ensemble Prediction

Ensemble prediction is a great method for severe weather forecasting and a new direction for operational forecast. With the requirement and the demand of ensemble prediction, this paper presents a scientific- workflow-based solution for the forecasting process in grid environment. As a result, it can overcome the drawbacks brought by controlling manually and can automate the whole process. Moreover, this paper presents a two layered scheduling strategy which separates the job scheduling by workflow engine from the scheduling by the proxy in the ensemble forecasting service. Comparing with the traditional methods in this field, we conclude that the two layered scheduling strategy can satisfy the requirement of the ensemble prediction, reduce the time cost of the whole forecasting process and increase the utilization of the resource in the grid.