Download presentation
Presentation is loading. Please wait.
Published byRoderick Cannon Modified over 8 years ago
1
Term Project: Correlation of AO Indices vs. Corresponding Tornadic Centroid Location and Total Tornado Count April Haneklau EAS 4480 – Environmental Data Analysis April 23, 2013
2
Introduction Idea: Positive AO = Tornadic Activity Reason: Warmer temperatures Sharper temperature gradients Cold air Warm air
3
Overview Analyzing correlation between AO indices and tornadic activity of April of each year. Monthly = 1990-2012. Daily = 2000-2012 Data collected from NCDC Storm Events Database and CPC Teleconnection Archives. April 2011!
4
Primary Calculations Weighted Average: From daily storm reports, a weighted average (by EF/F rating) latitude and longitude value was calculated for each year's April. Purpose: to show location of centroid of tornadic activity Correlation Coefficient & Least-Squares Regression Calculated the correlation and least- squares regression between AO and monthly weighted latitudes & longitudes, monthly tornado count and daily tornado count. Covariance & CPSD Calculated cross power spectral density to identify possible shared cycles observed in periodogram. Calculated lag time from covariance of daily data to identify lag between observed AO & tornadic activity. Periodogram Calculated and plotted periodogram of monthly weighted latitudes & longitudes, monthly tornado count, daily tornado count, monthly AO indices and daily AO indices. Purpose: to identify possible cycles.
5
Weighted Average Calculation April 2011!
6
Histogram: Positive vs. Negative AO
7
Correlation Coefficient & Lag Monthly VariablesDaily Variables R 2 Valuep value Tornado Count 0.56470.005 Weighted Latitude 0.00390.9860 Weighted Longitude 0.000980.9965 Significant! R 2 Valuep value Tornado Count (total) 0.04660.3597 Tornado Count (tornadic days) 0.08180.2047 Lag time-29 days(?)
8
Least-Squares Regression East West North South April 2011!
9
Periodogram
10
CPSD Estimate
11
Houston, we have a problem... Tornado rating scale changed from F to EF in 2007. Thus, weights could potentially be slightly skewed. Data not well obtained/archived before 2000; 107 days in 1990's had incomplete data and were excluded from calculations. Only analyzed 23 years; likely would see significant or real peaks with larger dataset.
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.