Reproducibility of Research Algorithms in GOES-R Operational Software

The research to operations transition for satellite observations is an area of active interest as identified by The National Research Council Committee on NASA-NOAA Transition from Research to Operations. Their report recommends improved transitional processes for bridging technology from research to operations. Assuring the accuracy of operational algorithm results as compared to research baselines, called reproducibility in this paper, is a critical step in the GOES-R transition process. This paper defines reproducibility methods and measurements for verifying that operationally implemented algorithms conform to research baselines, demonstrated with examples from GOES-R software development. The approach defines reproducibility for implemented algorithms that produce continuous data in terms of a traditional goodness-of-fit measure (i.e., correlation coefficient), while the reproducibility for discrete categorical data is measured using a classification matrix. These reproducibility metrics have been incorporated in a set of Test Tools developed for GOES-R and the software processes have been developed to include these metrics to validate both the scientific and numerical implementation of the GOES-R algorithms. In this work, we outline the test and validation processes and summarize the current results for GOES-R Level 2+ algorithms.