Targeted process-oriented diagnostics provide an essential pathway toward better understanding model behavior and can yield insights into how models can be improved. Well-established summary statistics are frequently used to cast a “large net”, objectively synthesizing the level of agreement between simulations and observations in ways that often holistically encompass many processes and phenomena. In this presentation we discuss how these distinct approaches can be used in complementary ways, as both play an important role in characterizing model behavior. We compare the goals of several mature open source analysis capabilities and how they are being exercised, including the MDTF-diagnostic and PMP packages. Via several examples, we identify natural intersections between these capabilities and describe possible paths forward that can lead to better understanding and ultimately improvement of earth system models.
Acknowledgement: This work was supported by the Office of Science (BER), U.S. Department of Energy through Lawrence Livermore National Laboratory contract DE-AC52-07NA27344.