Publication Date

Fall 2024

Document Type

Student Project

First Advisor

Gensini, Vittorio

Department

Department of Geographic and Atmospheric Sciences

Abstract

Increasing the predictability of severe convective storms (SCSs) over extended time ranges (i. e. >8 days) is of utmost importance to forecasters, as modern numerical weather prediction (NWP) models exhibit little skill at this lead time. Analogs have been used to predict temperature and rainfall anomalies at this time range for decades, though modern developments in analog forecasting uses techniques such as analog ensembles to increase the skill of analog models. This study employs a simple analog ensemble (AnEn) method using the Climate Prediction Center (CPC) 6-10 Day 500hPa analogs and uses the Storm Prediction Center (SPC)’s report database to calculate report densities (Practically Perfect Hindcast, PPH) for each member of the AnEn. The climatology of PPH for the forecasted range is also created as a comparison with the analog model. These density estimations are then averaged and compared to the observed report densities within the 6-10 Day period, where non-probabilistic skill metrics are analyzed through a 2x2 contingency table. The Heidke Skill Score (HSS) is calculated by the metrics in the matrix, which is a scalar value that indicates the skill of the forecast model. Analyses of HSSs throughout the year 2017 showed significant daily and seasonal variability of both the analog and climatology model. In general, the analog model struggles to exceed the HSS of the climatology model. Lower thresholds exhibit high mean skill along with higher spatial coverage of grid point exceedances. Higher thresholds exhibit low mean skill, but lower spatial coverage of exceedances and better skill compared to the climatology model. Low thresholds favored all 10 analogs while high thresholds favored around 3-5 analogs.

Download

Share

COinS