240 likes | 348 Views
CIMMSE Improving Inland Wind Forecasts October 2011 Project Update. Project Focuses This Period. HWind vs. ASOS Station Verification NDFD Verification: Irene (2011) ASOS wind data outages?. HWind/Observation Station Comparison.
E N D
CIMMSE Improving Inland Wind ForecastsOctober 2011 Project Update
Project Focuses This Period • HWind vs. ASOS Station Verification • NDFD Verification: Irene (2011) • ASOS wind data outages?
HWind/Observation Station Comparison • Motivation: last conference call, it was suggested to use HWind analyses for NDFD forecast verification • Advantages of HWind analyses: • Data assimilation from multiple sources (METAR, satellite, buoys, etc.) • Gridded data available (much easier to work with) • Disadvantages of HWind analyses: • Gridded data only available since 2000 • Different data sources for each storm, based on availability (inter-storm comparison may have bias) • Analysis stops shortly after landfall for most storms • Grid moves with storm (area of interest sometimes outside of HWind domain • Three hour resolution
HWind/Observation Station Comparison • NDFD wind data available since 2005 • Storms affecting region since 2005: - Ernesto (2006) -Gabrielle (2007) • Hanna (2008) -Irene (2011) • Cristobal (2008) -Earl (2010) • Can interpolate NDFD/HWind analysis to a common grid and compare to station data
HWind Analysis: Irene (2011) 8/27/2011 0130 UTC: Wind Speeds (m/s)
HWind - NDFD • Interpolated HWind and NDFD data at each time to common grid • Used latest forecast cycle when making comparisons (eliminate as much as possible track/NHC guidance bias)
HWind – NDFD: Irene (2011) • Boundaries of WFOs appear in analysis • Raleigh WFO appears to have smallest different between forecast and HWind analysis, at least when comparing to most recent forecast cycles • Strongest overprediction of wind speeds present in coastal regions • Suggests the 33% reduction used by Raleigh forecasters worked well
HWind – NDFD: Various Forecast Cycles • Several collaborators have suggested verifying different forecast cycles (inter-forecaster bias) • Examined forecasts issued: • 8/25 at 22 UTC (Thursday evening) • 8/26 at 10 UTC (Friday morning) • 8/26 at 22 UTC (Friday evening)
HWind – NDFD: Various Forecast Cycles Forecasted Issued 8/25 at 22 UTC
HWind – NDFD: Various Forecast CyclesForecasted Issued 8/26 at 10 UTC
HWind – NDFD: Various Forecast Cycles Forecasted Issued 8/26 at 22 UTC
Different Forecast Cycles • Some offices differ rather significantly between forecast cycles versus other offices • Provides further evidence for the lack of scientific processes going into the forecast • Future work will better quantify these differences, relative to the differences in track/intensity forecasts
ASOS: Missing Data Effect? • Several collaborators have suggested observation stations can go down during storms • We need to incorporate this influence in developing the climatology • Compared wind data while storm was influencing the region to four days prior
ASOS: Missing Data Effect? • Up to 45% less data when storm is present • Most reduction for: Able (1952), Barbara (1953), Connie (1955), Hazel (1954), and Ione (1955) • Temporal element appears to be present • Other noteworthy reductions: Fran (1996): 11% Floyd (1999): 13% Bertha (1996): 20% Isabel (2003): 23% Hanna (2008): 20%
Goal for Final Product • It has become clear that a more systematic approach to the land reduction factor and gust factor be developed • Adjustment factor to account for TCM wind inflation (gross) • Distance from storm center • Local topography • Strength of system, speed of propagation • Mesoscale environmental adjustment (thermodynamic factors, etc.)
Upcoming goals • Gust factors derived for select stations (similar to analysis of Larry Brown) • Quantitative results for comparing HWind and surface wind obs. • Begin modeling studies (with assistance of Dr. Sukanta Basu) • Other NDFD verification: compare 4 quadrant data to NDFD vs. HWind (added value from forecasters) • Final climatology report (gust factors, Weibull distributions, NDFD verification, Obs. vs. HWind analysis)—to be posted on blog for comments
Key areas for discussion • Suggestions for improvement in the TCM tool for GFE developers (led by JB?) • Other ways to verify NDFD forecasts • Irene (2011) forecaster observations/notes • Other suggestions for final product?