Seven Day Streamflow Forecasting - An ensemble streamflow forecasting system for Australia Dr. Sophie Zhang Water Forecasting Service Bureau of Meteorology 02 May 2018, Delft-FEWS Users Days and Workshop
Outline 1. Overview of water forecasting service 2. 7-day Forecast Service SDF 3. Ensemble 7-day streamflow forecasting 4. Future service development
Perspective Situational awareness Foresight PAST PRESENT FUTURE DECADES YEARS WEEKS DAYS DAYS WEEKS YEARS DECADES National water account Water balance reporting Water market website Flood & Long-term trends (HRS) Seasonal Forecasts (SSF) short-term forecasts (SDF)
7-day Forecast Service on website Public service website: http://www.bom.gov.au/water/7daystreamflow/ Video link: https://www.youtube.com/watch?time_continue=1&v=pougg0nkdoe Registered-user service website: http://reg.bom.gov.au/water/reg/stf/
7-Day Streamflow Forecast Service http://www.bom.gov.au/water/7daystreamflow/
Service Coverage Sites numbers after expansion Nov 2017: WA 25 28 NT 14 16 QLD 32 38 Public Regd. users TAS 22 24 WA 25 28 SA 3 4 NT 14 16 SA 3 4 VIC 34 43 NSW 38 56 NSW 38 56 QLD 32 38 VIC 34 43 Total 168 209 TAS 22 24 Total catchments: 100
7-Day Streamflow Forecasting Approach Inputs Outputs 7 days ahead hourly forecasts Rigorous forecast verification Past hourly rainfall 7 days ahead 3- hourly rainfall Past hourly streamflow Process Forecast post-processor
7 Day forecast operational system workflow HyFS DB ACCESS Real-time obs rain &WL Rainfall forecasts up to 10 days HyFS System (SWIFT/SDF models) Ingest data Update model states Process numerical weather prediction data Run Hydro model Export data Forecast flow & rainfall SDF Product Generator HTML images and forecast data Publish products in SDF website Operators HyFS: Hydrological Forecasting System the central national platform for modelling ACCESS: The Australian Community Climate and Earth-System Simulator weather models
Migrate Catchment Models to Operational Environment - HyFS
Ensemble 7-day streamflow forecasting
7-Day Ensemble Streamflow Forecasts Operational plans: June 2019 Rainfall uncertainty: Multi model ensemble mean (PME) ACCESS-GE ECMWF Rainfall post processing Hydrologic uncertainty Streamflow post processing
Hydrological model forecasting Forecast rainfall Hydrological model Forecast streamflow Forecast rainfall error Model structure error Forecast streamflow error Measurement error Model parameter error RPP ERRIS = Reliable estimate of streamflow forecast error
Improvements from rainfall post processing at catchment scale Characteristics OCF/PME ACCESS-GE ECMWF Observed rainfall 1 ensemble 24 ensembles 50 ensembles 50km, 3hourly, 228 leads 60km, 3hourly, 240 leads 20km, 3 hourly (144 steps), 6 hourly (steps 150 360) 00Z, 12Z 12Z 00Z and 12Z Day 1 Post- processed forecast Day 1 Raw forecast RPP
1 Modelling Residual Errors A single error model Complex Interactions between the hydrological model and the error model Difficult to calibrate the models jointly Hydrological model simulation Bias correction Error reduction and representation in stages (ERRIS) Deals with hydrological uncertainty only Runs a sequence of simple error models Each stage addresses different issues Autoregressive updating Residual distribution
Ensemble streamflow forecasting Service Development 7-Day ensemble streamflow forecasts Ensemble flood forecasts Research & Development Improved rainfall post processing techniques Advanced data QA/QC techniques Advanced rainfall interpolation techniques Further development of SWIFT model regulated catchments Implementation of a grid-based model for water forecasting
Future Service Development
Services Future Developments 30 day forecasts 3 9 months forecasts Ensemble 7-day forecasts Service expansion to more locations
Thank you Narendra Tuteja Mohammed Bari Prasantha Hapuarachchi Richard Laugesen Sophie Zhang Aynul Kabir Andrew MacDonald Kevin Plastow Daehyok Shin Jayaratne Liyanage Nilantha Gamage We honour and acknowledge the Ngunnawal people and Kulin nations as the traditional custodians of the land on which we live and work, and pay our respects to their Elders both past and present.