Probabilistic Forecast Verification. Yuejian Zhu EMC/NCEP/NOAA

Transcription

1 Probabilistic Forecast Verification Yuejian Zhu EMC/NCEP/NOAA

2 Review NAEFS Products (FY07) (December 4 th 2007) Bias corrected NCEP/GFS forecast 4 times daily, every 6 hours, out to 180 hours Bias corrected NCEP/GEFS forecast 4 times daily, 21 members, every 6 hours, out to 384 hours Bias corrected CMC/GEFS forecast 2 times daily, 21 members, every 6 hours, out to 384 hours Combine bias corrected GFS and GEFS ensemble forecast Dual resolution ensemble approach for short lead time Adjustable weight coefficient GFS has higher weights at short lead time NAEFS products at 1*1 degree globally Combine NCEP/GEFS (20m) and CMC/GEFS (20m) All bias corrected forecast Consider the difference between NCEP and CMC s analyses Statistical downscaling to 5km for CONUS, 6km for Alaska Use RTMA as reference Downscaled from 1degree forecast

3 Bias correction and downscaling Bias correction at 1*1 degree resolution (weight=0.02 for Kalman filter algorithm) Bias corrected NCEP/GEFS, GFS (out to 180 hours) and CMC/GEFS forecasts Consider the same bias for NCEP all ensemble members Consider the different bias for each model (member) Combine bias corrected high resolution GFS and low resolution ensembles Dual resolution ensemble approach for short lead time GFS has higher weights at short lead time NAEFS products based on all bias corrected forecasts Produce Ensemble mean, spread, mode, 10% 50%(median) and 90% probability forecast Climate anomaly (percentile) forecasts also generated for ensemble mean Statistical downscaling to NDGD grids (weight=0.2 for Kalman filter algorithm) Proxy for truth - RTMA at 5km/6km resolutions Variables (surface pressure, 2-m temp (and Max/min), and 10-m wind (and speed/direction) Downscaling vector Interpolate GDAS analysis to 5km/6km resolutions Compare difference between interpolated GDAS and RTMA Apply decaying weight to accumulate this difference downscaling vector Downscaled forecast Interpolate bias corrected 1*1 degree NAEFS to 5km/6km resolutions Add the downscaling vector to interpolated NAEFS forecast NAEFS products from downscaling CONUS NDGD grid/resolution (5km) 4 variables for Ensemble spread, mean, mode, 10%, 50%(median) and 90% forecasts Alaska NDGD grid/resolution (6km) 8 variables for Ensemble spread, mean, mode, 10%, 50%(median) and 90% forecasts

4 Bias Correction Method & Application Bias Correction Techniques array of methods Estimate/correct bias moment by moment Simple approach, implemented partially May be less applicable for extreme cases Moment-based method at NCEP: apply adaptive (Kalman Filter type) algorithm decaying averaging mean error = (1-w) * prior a.m.e + w * (f a) For separated cycles, each lead time and individual grid point, a.m.e = averaging mean error 6.6% 3.3% 1.6% Test different decaying weights. 0.25%, 0.5%, 1%, 2%, 5% and 10%, respectively Decide to use 2% (~ 50 days) decaying accumulation bias estimation Toth, Z., and Y. Zhu, 2001

5 NCEP/GEFS raw forecast 8+ days gain NAEFS final products From Bias correction (NCEP, CMC) Dual-resolution (NCEP only) Down-scaling (NCEP, CMC) Combination of NCEP and CMC

6 Application for probabilistic forecasts? How to verify 10% or 90% probabilistic forecast?

7

8 Reliability examine Method 1: (no climatology, no skill score, next 2 slides) Counts for each grid point, observation location Given 0 if obs > f(10%) or obs < f(90%) Given 1 if obs < f(10%) or obs > f(90%) Average for each grid point (weighted), or obs (weighted?) Reference: 10% or 90% (100% - 10%) of this count is for perfect. Method 2: (consider climatology, for discussion?) Using NCEP/NCAR 40y reanalysis data or CFSRR Consider each grid point, or observation Construct daily climatological mean and standard deviation (possible to consider non-gaussian distribution, too) Generate climatological numbers for 10% and 90% 10%, 90% probabilistic forecast at each gird point or observation location Best analysis or observation Using method one to count for analysis/observation and 10%, 90% climatology (sampling?) Using method one to count for analysis/observation and 10%, 90% forecast Possible to convert to skill score??? And how to???

9 2-meter temperature 10/90 probability forecast verification Northern Hemisphere, period of Dec Feb P10-expect P10-ncepraw P10-naefs P90-expect P90-ncepraw P90-naefs 100 Probabilities (percent) ~40% ~80% Lead time (hours)

10 2-meter temperature 10/90 probability forecast verification Northern Hemisphere, seasonal variation for NAEFS P10 P10-dec P10-feb P90 P90-dec P90-feb Probabilities (percent) NAEFS final 10% and 90% probability forecast for Dec and Feb Lead time (hours)