The development of the NMME seasonal climate forecast system is a significant success in demonstrating the potential for forging a collaboration of operational and research groups focusing on both the generation of forecasts and their analysis. Key issues that remain to be resolved are how to optimally combine the multiple model ensembles based on their hindcast skill or conditioned on the phase of teleconnection patterns for major climate variability patterns such as ENSO, the Pacific Decadal Oscillation, or the North Atlantic Oscillation. Promising new research is emerging in the area of hybrid dynami- cal/statistical approaches to climate prediction—for example, the finding that a skill-weighted statistical combination of dynamical model outputs improves upon simple ensemble averages (e.g., Wanders and Wood 2016).
Such advances have arisen since the operational community recognized the value of reforecasting (or "hindcasting"; Hamill et al. 2005), which enables forecast model post-processing and weighting, and multimodel efforts. New multimodel hindcast archives are becoming available and being tested for both medium range (1-15 days) and subseasonal to seasonal (S2S) predictions, complementing the NMME effort (e.g., as part of the SubX project). However, it is essential to improve the individual dynamical prediction model systems that contribute to systems such as the NMME via continued investments in a number of critical areas: model development; data assimilation; observational networks for model validation, assimilation, and initialization; and high performance computing infrastructure that can allow experimentation at high resolution; and the generation of more sophisticated models and larger ensembles. Ultimately, enhancing the understanding and modeling of predictable phenomena and processes is foundational to making improved predictions.
Important goals for enhancing US drought management capabilities include achieving seamless systems for monitoring and forecasting of drought, and advancing our ability to quantify uncertainties in monitoring and forecasting products. Toward the first objective, NOAA can build on success with the NLDAS and LSM-based hydrologic monitoring and prediction systems, in which the approaches for monitoring and predicting drought-related variables are consistent and integrated. To pursue the second goal, NOAA is working to link the four NLDAS LSMs and the NMME suite of seasonal climate forecasting models within an operational drought information system.
The overall system will provide broader estimates of uncertainty in both current land surface moisture states and future climate forcings, and enhance our probabilistic drought monitoring and prediction capabilities. Although the initial LSM ensemble in this system is small, limiting the depiction of land modeling uncertainty, the framework can be extended to leverage unified modeling concepts that can allow for more comprehensive and deliberate LSM uncertainty quantification (e.g., Clark et al. 2015).