Search for: All records

Abstract. In this work, we integrated the WAVEWATCH III model into the regional coupled model SKRIPS (Scripps–KAUST Regional Integrated Prediction System). The WAVEWATCH III model is implemented with flexibility, meaning the coupled system can run with or without the wave component. In our implementations, we considered the effect of Stokes drift, Langmuir turbulence, sea surface roughness, and wave-induced momentum fluxes. To demonstrate the impact of coupling we performed a case study using a series of coupled and uncoupled simulations of Tropical Cyclone Mekunu, which occurred in the Arabian Sea in May 2018. We examined the model skill in these simulations and further investigated the impact of Langmuir turbulence in the coupled system. Because of the chaotic nature of the atmosphere, we ran an ensemble of 20 members for each coupled and uncoupled experiment. We found that the characteristics of the tropical cyclone are not significantly different due to the effect of surface waves when using different parameterizations, but the coupled models better capture the minimum pressure and maximum wind speed compared with the benchmark stand-alone Weather Research and Forecasting (WRF) model. Moreover, in the region of the cold wake, when Langmuir turbulence is considered in the coupled system, the sea surface temperature is about 0.5 ∘C colder, and the mixed layer is about 20 m deeper. This indicates the ocean model is sensitive to the parameterization of Langmuir turbulence in the coupled simulations. 

Atmospheric rivers (ARs) and Santa Ana winds (SAWs) are impactful weather events for California communities. Emergency planning efforts and resource management would benefit from extending lead times of skillful prediction for these and other types of extreme weather patterns. Here we describe a methodology for subseasonal prediction of impactful winter weather in California, including ARs, SAWs and heat extremes. The hybrid approach combines dynamical model and historical information to forecast probabilities of impactful weather outcomes at weeks 1–4 lead. This methodology uses dynamical model information considered most reliable, that is, planetary/synoptic‐scale atmospheric circulation, filters for dynamical model error/uncertainty at longer lead times and increases the sample of likely outcomes by utilizing the full historical record instead of a more limited suite of dynamical forecast model ensemble members. We demonstrate skill above climatology at subseasonal timescales, highlighting potential for use in water, health, land, and fire management decision support.

 

Abstract Using a high-resolution atmospheric general circulation model simulation of unprecedented ensemble size, we examine potential predictability of monthly anomalies under El Niño Southern Oscillation (ENSO) forcing and back-ground internal variability. This study reveals the pronounced month-to-month evolution of both the ENSO forcing signal and internal variability. Internal variance in upper-level geopotential height decreases (∼ 10%) over the North Pacific during El Niño as the westerly jet extends eastward, allowing forced signals to account for a greater fraction of the total variability, and leading to increased potential predictability. We identify February and March of El Niño years as the most predictable months using a signal-to-noise analysis. In contrast, December, a month typically included in teleconnection studies, shows little-to-no potential predictability. We show that the seasonal evolution of SST forcing and variability leads to significant signal-to-noise relationships that can be directly linked to both upper-level and surface variable predictability for a given month. The stark changes in forced response, internal variability, and thus signal-to-noise across an ENSO season indicate that subseasonal fields should be used to diagnose potential predictability over North America associated with ENSO teleconnections. Using surface air temperature and precipitation as examples, this study provides motivation to pursue ‘windows of forecast opportunity’, in which statistical skill can be developed, tested, and leveraged to determine times and regions in which this skill may be elevated. 

Despite the well‐recognized initial value nature of the subseasonal forecasts, the role of subsurface ocean initialization in subseasonal forecasts remains underexplored. Using observing system experiments, this study investigates the impact of ocean in situ data assimilation on the propagation of Madden–Julian Oscillation (MJO) events across the Maritime Continent in the European Centre for Medium‐Range Weather Forecasts (ECMWF) subseasonal forecast system. Two sets of twin experiments are analyzed, which only differ on the use or not of in situ ocean observations in the initial conditions. Besides using the Real‐time Multivariate MJO Index (RMMI) to evaluate the forecast performance, we also develop a new MJO tracking method based on outgoing longwave radiation anomalies (OLRa) for forecast evaluation. We find that the ocean initialization with in situ data assimilation, though having an impact on the forecasted ocean mean state, does not improve the relatively low MJO forecast skill across the Maritime Continent. Moist static energy budget analysis further suggests that a significant underestimation in the meridional moisture advection in the model forecast may hinder the potential role played by the ocean state differences associated with data assimilation. Bias of the intraseasonal meridional winds in the model is a more important factor for such underestimation than the mean state moisture biases. This finding suggests that atmospheric model biases dominate the forecast error growth, and the atmospheric circulation bias is one of the major sources of the MJO prediction error and should be a target for improving the ECMWF subseasonal forecast model.

 

Stochastic parameterizations account for uncertainty in the representation of unresolved subgrid processes by sampling from the distribution of possible subgrid forcings. Some existing stochastic parameterizations utilize data‐driven approaches to characterize uncertainty, but these approaches require significant structural assumptions that can limit their scalability. Machine learning models, including neural networks, are able to represent a wide range of distributions and build optimized mappings between a large number of inputs and subgrid forcings. Recent research on machine learning parameterizations has focused only on deterministic parameterizations. In this study, we develop a stochastic parameterization using the generative adversarial network (GAN) machine learning framework. The GAN stochastic parameterization is trained and evaluated on output from the Lorenz '96 model, which is a common baseline model for evaluating both parameterization and data assimilation techniques. We evaluate different ways of characterizing the input noise for the model and perform model runs with the GAN parameterization at weather and climate time scales. Some of the GAN configurations perform better than a baseline bespoke parameterization at both time scales, and the networks closely reproduce the spatiotemporal correlations and regimes of the Lorenz '96 system. We also find that, in general, those models which produce skillful forecasts are also associated with the best climate simulations.