The 31st Conference on Hydrology is hosting a session joint with the Probability and Statistics committee on diagnostic model evaluation and improvement of land models. Advancements in earth system modeling require coupling of the atmosphere, hydrologic, land surface, ocean and cryosphere systems. Correspondingly, there are significant challenges in the systematic evaluation of each of the system components and their interactions. This session seeks to move beyond traditional verification studies that document accuracy of land models, and instead ask questions if models are adequately using the information available to them, what are specific weaknesses in land models, and how can models be improved. As such, the particular interest of this session is to explore the model evaluation challenges in hydrology, land processes and the corresponding impact on coupled land-atmosphere processes and hydrometeorological prediction. With the increase in the variety of model developers and users, more integrated approaches that encapsulate the key water cycle component are needed to improve individual model components and coupled earth system component interactions. In addition, the development of a common, systematic set of measures will improve the “observability” of various model outputs from these systems. This session solicits contributions on integrated diagnostic evaluation and benchmarking techniques and metrics that promote systematic error and uncertainty quantification across complex modeling components. The session also solicits contributions that pinpoint model weaknesses and introduce innovative modeling approaches to address long-standing challenges simulating heterogeneity, emergent behavior, and process coupling across a range of different space and time scales. A key focus of the session, supporting the conference theme of “observations lead the way” is contributions that use multivariate and multiscale observations to diagnose model weaknesses and evaluate the fidelity of competing modeling approaches. This session will also address the significant challenges associated with assessing the quality and informativeness of both models and data products that are largely related to scale, heterogeneity, complexity, and representativeness. These challenges compound when assessing spatially and temporally distributed products and the fusion of models and data via approaches such as parameter estimation and data assimilation. This session therefore also solicits contributions related to innovative methods for assessing quality of model-data fusion and assessing the fidelity of models of complex terrestrial hydrologic systems in both offline and coupled modes.