290 likes | 448 Views
Predicted Rainfall Estimation in the Huaihe River Basin Based on TIGGE. Fuyou Tian, Dan Qi, Jingyue Di, and Linna Zhao National Meteorological Center of China Meteorological Administration tianfy@cma.gov.cn 15 September 2009. Outline:. Data and Test Catchment
E N D
Predicted Rainfall Estimation in the Huaihe River Basin Based on TIGGE Fuyou Tian, Dan Qi, Jingyue Di, and Linna Zhao National Meteorological Center of China Meteorological Administration tianfy@cma.gov.cn 15 September 2009
Outline: • Data and Test Catchment • TIGGE 3 centers (CMA, ECMWF and NCEP) total precipitation data • Huaihe River Basin, and its sub-catchment • 19 Observations in Dapoling-Wangjiaba Reservoir 2.Method • Threat Score, Bias Score and Brier Score • Percentile • Results • TS, B and BS • Probabilistic forecast of Huaihe River Basin • Percentile-based precipitation probabilistic forecast • Summary and future works
Meteorological Centers of TIGGE Data • Total precipitation data from July 1 to August 6, 2008 • Accumulated rainfall from 00:00 to 00:00(GMT) of the next day
Huai-bin Wang-jia-ba Xi-xian The Test Catchment and Observation Stations 南四湖区 沂沭水系 涡河-蒙城以上 蚌埠-洪泽湖 淮河下游 颖河-阜阳 王家坝-蚌埠 Target basin 大别山库区 • Distribution of 19 observation stations in the Dapoling-Wangjiaba sub-region • Cumulative rainfall of every observation station from 00:00 to 00:00 of the next day • Using the bilinear interpolation method to obtain the grid value
The Variation of Daily Areal Rainfall (mm) Extreme event
Data and Test Catchment • TIGGE 3 centers (CMA, ECMWF and NCEP) total precipitation data • Huaihe River Basin, and its sub-catchment • 19 Observations in Dapoling-Wangjiaba Reservoir 2.Method • Threat Score, Bias Score and Brier Score • Percentile • Results • Probabilistic forecast of Huaihe River Basin • Percentile-based precipitation probabilistic forecast • Summary and future works
Criteria Adopted for Calculation of Rainfall Intensity • To calculate the threat score, bias score and brier score, observations and forecasts of daily rainfall are divided into four classes, but very heavy rainfall are not included. • The probabilistic and percentile rain not use this criteria, all rainfall intensities are take into consideration.
Threat Score (TS), Bias (B) and Brier Score (BS) • The Threat Score ( or CSI: Critical Success Index) and Bias (B) are given as (Wilks, D S, 1995) TS = a / (a + b + c) B = (a +b)/ (a + c) where a, b and c represents hits, false alarms , and misses, respectively. TS varies from 0.0 to 1.0, 1.0 indicates the perfect forecast. B=1.0 indicates that the event was forecast the same number of times it was observed. • The Brier Score is defined as BS = (fi - oi)2/N in which N is the sample size, the observations oi are all binary, 1.0 if the event occurs and 0 if it doesn’t. The BS ranges from 0 for a perfect forecast to 1.0 for the worst possible forecast.
Percentile • Firstly set the data in increasing order, a percentile is the value of a variable below which a certain percent of observations or values fall. • Values of the target percentiles are estimated using the experience-based equations (Hyndman R J, et al 1996) Qi(p) = (1 - r) X (j) + r X (j+1) in which j = integer (p*n + (1+p)/3) r = p*n + (1+p)/3) – j where Qi(p) presents the returned ith percentile, n is the sample number, X the ordered data.
Data and Test Catchment • TIGGE 3 centers (CMA, ECMWF and NCEP) total precipitation data • Huaihe River Basin, and its sub-catchment • 19 Observations in Dapoling-Wangjiaba Reservoir 2. Method • Threat Score, Bias Score and Brier Score • Percentile • Results • TS, B and BS • Probabilistic forecast of Huaihe River Basin • Percentile-based precipitation probabilistic forecast • Summary and future works
I II IV III Results: TS and B of Ensemble Mean over the sub-region
I II IV III Results: TS and B of Control Forecast Member over the sub-catchment
I II IV III Results: Brier Score over Dapoling-Wangjiaba
B C D A Observation
B C D A Observation
B C A 2008072300 observation Observation D The 95th percentile precipitation of the 19 observation stations in the Dapoling-Wangjiaba sub-region with a 1 day lead time of three EPSs and their grand ensemble.
B C A 2008072300 observation Observation D The 95th percentile precipitation of the 19 observation stations in the Dapoling-Wangjiaba sub-region with a 2 day lead time of three EPS and their grand ensemble.
A B D C A comparison of probabilistic forecast of daily areal rainfall of the three EPSs and their grand ensemble with a 1 day lead time. The 5th, 25th,50th, 75th, 95th and 99th percentile of daily rainfall are shown, black circles are observations.
Site Scale probability forecast A B Probabilistic forecast of Daily precipitation of Huaibin Station (115.41 32.45) with a 1 day lead time D C
Comparison of box and whisker plots for 22 July 2008 at Huaibin station. Black circles are observations (56.2mm).
Comparison of box and whisker plots for 16 July 2008 at Huaibin station. Black circles are observations (25.8mm).
Data and Test Catchment • TIGGE 3 centers (CMA, ECMWF and NCEP) total precipitation data • Huaihe River Basin, and its sub-catchment • 19 Observations in Dapoling-Wangjiaba Reservoir 2. Method • Threat Score, Bias Score and Brier Score • Percentile • Results • Probabilistic forecast of Huaihe River Basin • Percentile-based precipitation probabilistic forecast • Summary and future works
Summary • TS and B indicate that every EPS has its advantage, CMA is good at forecast little rain, EC is good at moderate rain • BS indicates that grand ensembletake all the probabilities into consideration, and improves the performance • Probability of daily rainfall exceeding 25mm/24hrs and 50mm/24hrs show that grand ensemble depicts the spatial distribution well
Summary • Variation of daily areal rainfall and site scale forecast indicate that grand ensemble has special advantage • For forecasters who know little about the performance of every EPS, grand ensemble would be a good choice • For hydrological users who pay special attention to key observation stations, grand ensemble based probabilistic forecast would be a good tool
Future works Probabilistic flood forecasts precipitation temperature WRF 3D-Var(15km×15km) VIC model How to show the probabilistic flood forecast? How to effectively use probabilistic forecast as input?
Thanks for your attention! Contact: tianfy@cma.gov.cn
Synoptic analyses of 00:00 22 July 2008 • A low pressure locates at the SW of Huaihe River Basin, wind shear shows a cyclonic vorticity clearly.
Synoptic analyses of 12:00 22 July 2008 • The low pressure and the cyclonic vorticity move slowly to the south-east, and produced very heavy localized rain .