Constraining storm-scale forecasts of deep convective initiation with surface weather observations
Madaus, Luke Edward
MetadataShow full item record
Successfully forecasting when and where individual convective storms will form remains an elusive goal for short-term numerical weather prediction. In this dissertation, the convective initiation (CI) challenge is considered as a problem of insufficiently resolved initial conditions and dense surface weather observations are explored as a possible solution. To better quantify convective-scale surface variability in numerical simulations of discrete convective initiation, idealized ensemble simulations of a variety of environments where CI occurs in response to boundary-layer processes are examined. Coherent features 1-2 hours prior to CI are found in all surface fields examined. While some features were broadly expected, such as positive temperature anomalies and convergent winds, negative temperature anomalies due to cloud shadowing are the largest surface anomaly seen prior to CI. Based on these simulations, several hypotheses about the required characteristics of a surface observing network to constrain CI forecasts are developed. Principally, these suggest that observation spacings of less than 4—5 km would be required, based on correlation length scales. Furthermore, it is anticipated that 2-m temperature and 10-m wind observations would likely be more relevant for effectively constraining variability than surface pressure or 2-m moisture observations based on the magnitudes of observed anomalies relative to observation error. These hypotheses are tested with a series of observing system simulation experiments (OSSEs) using a single CI-capable environment. The OSSE results largely confirm the hypotheses, and with 4-km and particularly 1-km surface observation spacing, skillful forecasts of CI are possible, but only within two hours of CI time. Several facets of convective-scale assimilation, including the need for properly-calibrated localization and problems from non-Gaussian ensemble estimates of the cloud field are discussed. Finally, the characteristics of one candidate dense surface observing network are examined: smartphone pressure observations. Available smartphone pressure observations (and 1-hr pressure tendency observations) are tested by assimilating them into convective-allowing ensemble forecasts for a three-day active convective period in the eastern United States. Although smartphone observations contain noise and internal disagreement, they are effective at reducing short-term forecast errors in surface pressure, wind and precipitation. The results suggest that smartphone pressure observations could become a viable mesoscale observation platform, but more work is needed to enhance their density and reduce error. This work concludes by reviewing and suggesting other novel candidate observation platforms with a potential to improve convective-scale forecasts of CI.
- Atmospheric sciences