| Summary: | Skillful subseasonal-to-seasonal (hereafter S2S; 10 days - 12 weeks) prediction can greatly benefit decision-making on resource management and agricultural planning that falls into the weekly to seasonal time ranges. The S2S prediction, bridging the traditional weather forecasting and climate prediction, remains a challenge for global numerical models. This is mostly because the sources of predictability on S2S timescales are not fully understood and/or not well represented in global models. My Ph.D. thesis research seeks to improve the model performance and the S2S prediction by 1) developing a suite of "physics-oriented" model evaluation metrics that can not only assess how model performs but also help to reveal possible error sources, and 2) investigating sources of predictability of high-impact weather phenomena with a special focus on tropical cyclones (TCs). The analysis is expected to provide useful guidance on model development and improvement. My thesis first evaluated the Madden-Julian oscillation (MJO) - the dominant mode of tropical subseasonal variability and an important source of predictability on S2S timescales. Both the Navy Operational Global Atmospheric Prediction System (NOGAPS) and Global Forecasting System (GFS) exhibit relatively low predictive skill when the MJO initiates over the Indian Ocean and when the active convection of the MJO is over the Maritime Continent. Further analyses indicated a dry bias within the marine boundary layer and a misrepresented shallow heating mode in the NOGAPS, suggesting a model deficiency in cumulus parameterization. The diabatic heating biases are associated with weaker trade winds, weaker Hadley and Walker circulations over the Pacific, and weaker cross-equatorial flow over the Indian Ocean. The TC prediction was evaluated in the National Centers for Environmental Prediction (NCEP) Global Ensemble Forecasting System (GEFS) with forecast lead time up to 2 weeks. It shows that the GEFS has large errors of TCs on the regional scale. The negative genesis biases over the western North Pacific are associated with a weaker-than-observed monsoon trough, the erroneous genesis pattern over the eastern North Pacific is related to a southward displacement of the ITCZ, and the positive genesis biases near the Cape Verde islands and negative biases farther downstream over the Atlantic can be attributed to the hyperactive African easterly waves and stronger deep convective heating. The biases are associated with the deficiencies in the cumulus schemes of the GEFS. The precipitation initiates too early with respect to the column water vapor. And there is a dry bias in the column water vapor, which increases with the forecast lead times. The GEFS also underpredicts moderate-to-heavy precipitation. The analyses suggest that improvement in the cumulus parameterization may reduce the model mean state errors and enhance the TC prediction on the regional scale. The predictability of TCs was investigated on interannual, subseasonal, and synoptic timescales using the GEFS reforecasts. It shows that the model skillfully captures the interannual variability of TC activity over the North Pacific and the North Atlantic, which can be attributed to the modulation of TCs by the El Nino-Southern Oscillation (ENSO) and the Atlantic meridional mode. The GEFS has promising skill in predicting the active and inactive periods of TC activity over the Atlantic. The skill, however, has large year-to-year fluctuations. The analyses suggest possible impacts of ENSO, the MJO, and the anticyclonic Rossby wave breaking on the TC subseasonal predictability. Lastly, the predictability associated with different synoptic flow regimes was evaluated using the TC development pathways. It shows that the extratropical influenced TCs have lower predictability than the ones dominantly modulated by tropical atmosphere. Such extratropical influenced storms, when developing near the coast, will pose a challenge for operational forecasts and emergency management.
|
|---|