Скачать презентацию Operational Flood Forecasting for Bangladesh Tom Hopson NCAR Скачать презентацию Operational Flood Forecasting for Bangladesh Tom Hopson NCAR

4ae716dbeb170e23a718666a0fec9186.ppt

  • Количество слайдов: 19

Operational Flood Forecasting for Bangladesh: Tom Hopson, NCAR Peter Webster, GT A. R. Subbiah Operational Flood Forecasting for Bangladesh: Tom Hopson, NCAR Peter Webster, GT A. R. Subbiah and R. Selvaraju, ADPC Climate Forecast Applications for Bangladesh (CFAB): USAID/CARE/ECMWF/NASA/NOAA Bangladesh Stakeholders: Bangladesh Meteorological Department, Flood Forecasting and Warning Center, Bangladesh Water Development Board, Department of Agriculture Extension, Disaster Management Bureau, Institute of Water Modeling, Center for Environmental and Geographic Information Services, CAREBangladesh Contact: [email protected] edu

Overview: Bangladesh flood forecasting I. CFAB forecasting context II. Forecasting techniques -- using Quantile Overview: Bangladesh flood forecasting I. CFAB forecasting context II. Forecasting techniques -- using Quantile Regression for: 1. precipitation forecast calibration 2. Post-processing to account for all errors III. 2007 Floods and Warning System Pilot Areas

River Flooding Damaging Floods: large peak or extended duration Affect agriculture: early floods in River Flooding Damaging Floods: large peak or extended duration Affect agriculture: early floods in May, late floods in September Recent severe flooding: 1974, 1987, 1988, 1997, 1998, 2000, 2004, and 2007 1998: 60% of country inundated for 3 months, 1000 killed, 40 million homeless, 10 -20% total food production 2004: Brahmaputra floods killed 500 people, displaced 30 million, 40% of capitol city Dhaka under water 2007: Brahmaputra floods displaced over 20 million (World Food Program)

CFAB Project: Improve flood warning lead time Problems: 1. Limited warning of upstream river CFAB Project: Improve flood warning lead time Problems: 1. Limited warning of upstream river discharges (Dhaka: ~24 -48 hr warning, only) 2. Precip forecasting in tropics difficult Skillful CFAB forecasts benefit from: 1. Large catchments => river discharge results from “integrated” inputs over large spatial and temporal scales 2. Skillful data inputs: ECMWF, TRMM, CMORPH, CPC-rain gauge 3. Partnership with Bangladesh’s Flood Forecasting Warning Centre (FFWC) => daily border river stage readings useful for data assimilation

Daily Operational Flood Forecasting Sequence Daily Operational Flood Forecasting Sequence

Example of Quantile Regression (QR) Our application Fitting precipitation quantiles using QR conditioned on: Example of Quantile Regression (QR) Our application Fitting precipitation quantiles using QR conditioned on: 1) Reforecast ens 2) ensemble mean 3) ensemble median 4) ensemble stdev 5) Persistence 7

Probability Calibration Procedure obs Forecast PDF For each quantile: 1) Perform a “climatological” fit Probability Calibration Procedure obs Forecast PDF For each quantile: 1) Perform a “climatological” fit to the data 2) Starting with full regressor set, iteratively select best subset using forward step-wise cross-validation Precipitation – – Fitting done using QR Selection done by: a) Minimizing QR cost function b) Satisfying the binomial distribution Precip 3) 2 nd pass: segregate forecasts into differing ranges of ensemble dispersion, and refit models. => have different calibration for different atmospheric stability regimes => ensures ensemble skill-spread has utility observed Forecasts Time Regressors for each quantile: 1) corresponding ensemble 2) ens mean 3) ens median 4) ens stdev 5) persistence 8

Significance of Weather Forecast Uncertainty on Discharge Forecasts Calibrated Precipitation Forecasts Discharge Forecasts 1 Significance of Weather Forecast Uncertainty on Discharge Forecasts Calibrated Precipitation Forecasts Discharge Forecasts 1 day 4 day 3 day 1 day 7 day 10 day 7 day 4 day 10 day

Daily Operational Flood Forecasting Sequence Daily Operational Flood Forecasting Sequence

Step 1: generate discharge ensembles from precipitation forecast ensembles (Qp): Probability Producing a Reliable Step 1: generate discharge ensembles from precipitation forecast ensembles (Qp): Probability Producing a Reliable Probabilistic Discharge Forecast 1 PDF 1/51 Qp [m 3/s] Step 2: a) generate multi-model hindcast error time-series using precip estimates; b) conditionally sample and weight to produce empirical forecasted error PDF: forecast a) 1000 b) 1 Residuals PDF horizon [m 3/s] time => -1000 Residual [m 3/s] 1000 -1000 Probability Step 3: combine both uncertainty PDF’s to generate a “new-and-improved” more complete PDF forecasting (Qf): 1 Qf [m 3/s]

2004 Brahmaputra Ensemble Forecasts and Danger Level Probabilities 7 -10 day Ensemble Forecasts 7 2004 Brahmaputra Ensemble Forecasts and Danger Level Probabilities 7 -10 day Ensemble Forecasts 7 day 3 day 5 9 day 7 -10 day Danger Levels 8 day 4 day 3 day 10 day 5 day 7 day 4 day 9 day 8 day 10 day

Brahmaputra Discharge Forecast Verification Rank Histograms Brier Skill Scores CRPS Scores Brahmaputra Discharge Forecast Verification Rank Histograms Brier Skill Scores CRPS Scores

Daily Operational Flood Forecasting Sequence Daily Operational Flood Forecasting Sequence

Five Pilot Sites chosen in 2006 consultation workshops based on biophysical, social criteria: Rajpur Five Pilot Sites chosen in 2006 consultation workshops based on biophysical, social criteria: Rajpur Union -- 16 sq km -- 16, 000 pop. Uria Union -- 23 sq km -- 14, 000 pop. Kaijuri Union -- 45 sq km -- 53, 000 pop. Bhekra Union -- 11 sq km -- 9, 000 pop. Gazirtek Union -- 32 sq km -- 23, 000 pop.

2007 Brahmaputra Ensemble Forecasts and Danger Level Probabilities 7 -10 day Ensemble Forecasts 7 2007 Brahmaputra Ensemble Forecasts and Danger Level Probabilities 7 -10 day Ensemble Forecasts 7 day 7 -10 day Danger Levels 8 day 7 day 9 day 8 day 9 day 10 day

Community level responses to 2007 flood forecasts • Planned evacuations to identified high grounds Community level responses to 2007 flood forecasts • Planned evacuations to identified high grounds with adequate communication and sanitation facilities Economically, they were also able to: • Move livestock to high lands with additional dry fodder. • Early harvesting of rice and jute anticipating floods. • Protected fisheries by putting nets in advance Selvaraju (ADPC-UNFAO)

Conclusions § 2003: Daily operational probabilistic discharge forecasts “experimentally” disseminated based on lumped model Conclusions § 2003: Daily operational probabilistic discharge forecasts “experimentally” disseminated based on lumped model and 51 -member ECMWF ensemble § 2004: -- Multi-model and post-processing approach operational -- initializing watersheds using TRMM / CMORPH -- Forecasts automated -- CFAB became Bangladesh federal government entity -- forecast the severe Brahmaputra floods § 2005: CFAB became HEPEX test bed § 2006: -- Forecasts incorporated into national flood warning program and hydraulic model -- 5 vulnerable pilot areas designated and trained on using 1 -10 day probabilistic forecasts. § 2007: 5 pilot areas warned many days in-advance during two severe flooding events § 2008 -2009: Ongoing expansion of the warning system thoughout Bangladesh Further technological improvements through HEPEX test bed collaborations

Thank You! hopson@ucar. edu Thank You! [email protected] edu