Seasonal Forecasting Albrecht Weerts `
Societal themes Mission: developing and applying top level expertise in the area of water, subsurface and infrastructure for people, planet and prosperity. Foundation under Dutch law 800-900 people 33 nationalities Two main locations: Delft and Utrecht albrecht.weerts@deltares.nl Turnover: 110 million euros/y
Delft-FEWS worldwide (see also http://oss.deltares.nl) albrecht.weerts@deltares.nl
Some projects involving seasonal forecasting CHPS used by NOAA NWS, BPA, TVA etc ESP snow updating + ESP pre/postprocessing (BPA) piloting state updating (TVA: using OpenDA together with NCAR/RTI) EU FP7 GLOWASIS paper HESS :Skill of a global seasonal streamflow forecasting system, relative roles of initial conditions and meteorological forcing Candogan et al. 2013 paper J Hydromet. (forth coming) : Added skill of NWP for global seasonal hydrologic forecasting Candogan et al. 2014 EU FP7 DEWFORA paper HESSD (in review): Hydrological drought forecasting and skill assessment for the Limpopo river basin, Southern Africa, Trambauer et al. 2014 paper HESS: The potential value of seasonal forecasts in a changing climate Winsemius et al., 2014 paper?: Drought vulnerability assessment for prioritising drought warning implementation: Pan African level and Kenya Naumann et al. 2014
What provides prediction skill? Initial hydrological state: may provide memory (e.g. snow built up, soil moisture, surface water, taking some time to flow downstream) Meteorological forcing, beyond memory: Climatology: only if the climate is not so exciting, but every year more or less the same (e.g. dry seasons) (bias corrected) seasonal probabilistic weather forecasts Quality of model => If you have skill how to use the information provided by seasonal forecasts
GLOWASIS experiments Ensemble Streamflow Prediction (ESP): t 0 Reverse ESP: t 0 ECMWF forecast: (with/without bias correction) t 0 Hydr. state based on observed weather Hydr. states based on climatology observed weather forcing Weather forcing, drawn from climatology ECMWF seasonal forecast
Relative importance memory and forcing Lead time, at which skill of reverse ESP becomes higher than skill of ESP February
Relative importance memory and forcing Lead time, at which skill of reverse ESP becomes higher than skill of ESP May
Relative importance memory and forcing Memory of the groundwater level 250x 250m model Colloquium Thomas Berends 29th of October, 2012 9/36
Relative importance memory and forcing Memory of the soil moisture content Colloquium Thomas Berends 29th of October, 2012 10/36
Conditional weather resampling for ensemble streamflow forecasting Joost Beckers, Albrecht Weerts, Erik Tijdeman, Edwin Welles ` *submitted to Water Resoru. Res.
PNW Columbia River BPA Many dams Hydropower From http://www.usbr.gov/pn/fcrps/hydro/index.html
ESP Seasonal streamflow forecasting by Bonneville Power Administration (BPA) : Classical ESP Meteo from 55 historical years to represent climate Run a hydrologic model starting from the current initial state Initial state: Soil moisture Snow pack Reservoir levels precip/ temp time SAC-SMA, SNOW17, RES Q time Day, G.N., 1985. Journal of Water Resources Planning and Management 111, 157 170.
Jay Day, 1985
El Niño effects on local weather in PNW El Niño Warm and dry La Niña Cold and wet
Climate Mode Indices MEI QBO.esrl QBO.org QBO.anom QBO.std SOI.anom SOI.std NINO TNI NAO EA WP PNA EA/WR EP/NP SCA TNH POL PT AMO AO PDO MJO.phase MJO.ampl Historical and current phases available online: http://www.cpc.ncep.noaa.gov/ http://www.esrl.noaa.gov/psd/ http://www.cawr.gov.au/ (MJO) http://jisao.washington.edu/ (PDO)
Challenge Use climate mode information to improve the skill of the ESP Add climate mode info Initial state: Soil moisture Snow pack Reservoir levels Streamflow precip/ temp Hydrologic model Q time time
Approach: conditioning of the ESP by subsampling Select the years with most similar climate indices (at time of forecast) Select ESP members with similar phases Time of the forecast Dismiss ESP members with dissimilar phases Hamlet, A., Lettenmaier, D., (1999) Journal of Water Resources Planning and Management 125, 333 341.
Brier skill score Problem Smaller ensemble leads to more sampling uncertainty, less accurate quantile estimates and less forecast skill 0.2 0.1 0-0.1-0.2-0.3-0.4-0.5 0 10 20 30 40 50 Ensemble size analytical result Dworshak Hungry Horse Libby Dam C.A.T. Ferro: Weather and Forecasting 22, pp 1076-1088 (2007).
Solution: generate more ensemble members Instead of full historical years: MAP MAT MAP MAT MAP MAT MAP MAT MAP MAT MAP MAT MAP MAT MAP MAT past future Use parts of historical years: MAP MAT MAP MAT MAP MAT MAP MAT MAP MAT MAP MAT MAP MAT MAP MAT past future 27 permutations for 3 years 150,000 permutations for 55 years
Solution Our implementation: Monthly resampling period (1 seam per 30 days) Assemble historical MAP and MAT into forecast time series Condition on climate mode indices
Conditional sampling Historical time series of monthly climate index JUN JUL AUG SEP 1995 0.5 0.2 -.2 -.5 1996 0.0 -.2 -.4 -.5 1997 2.3 2.7 3.0 3.0 1998 1.1 0.2 -.4 -.7 1999 -.4 -.5 -.8-1 2000 -.2 -.2 -.1 -.2 2001 -.1 0.2 0.4 -.1 2002 0.9 0.6 0.9 0.8 2003 0.0 0.1 0.2 0.4 2004 0.2 0.4 0.7 0.5 2005 0.5 0.5 0.3 0.3 2006 0.5 0.6 0.7 0.8 2007 -.4 0.3 -.5-1.2 2008 0.1 0.0 -.3-0.7 2009 0.9 0.9 0.9 0.8 2010 -.5-1.2-1.8-2.0 2011 -.2 -.1 -.5-0.8 Simulated time series ENSO MAP/MAT 2014-JUN 0.9 2014-JUL 0.2 1998-JUL 2014-AUG 0.4 2001-AUG 2014-SEP 0.3 2005-SEP
Results Ensembles of synthetic ENSO index time series 1973 1978 1997 La Niña year Average year El Niño year Negative MEI Positive MEI
Results Ensembles of monthly precipitation 1973 1978 1997 La Niña year Average year El Niño year Wet winter Dry winter
Results Ensembles of monthly averaged temperature 1973 1978 1997 La Niña year Average year El Niño year Cold winter Warm winter
Results Ensembles of monthly averaged streamflow 1973 1978 1997 La Niña year Average year El Niño year High volume Low volume
Brier skill score Brier skill score relative to ESP Ensemble size effect is canceled out Improvement of skill is found for two out of three test catchments 0.2 0.1 0 Improved skill -0.1-0.2-0.3 More ESP traces replaced by resampled traces -0.4-0.5 analytical result Dworshak Hungry Horse Libby Dam 0 10 20 30 40 50 Nr historical years in ensemble
CRPSS CRPS skill relative to ESP Mix of 10 full historical years and 40 resampled traces seems optimal 0.1 0.05 0-0.05-0.1 Dworshak Hungry Horse Libby Dam 0 10 20 30 40 50 Nr historical years in ensemble
improvement in skill Forecast lead time (Dworshak) Skill as a function of lead time (10 historical years and 40 resampled): 15% 10% 5% 0% -5% -10% -15% -20% CRPS Brier RMSE 0 2 4 6 8 10 12 forecast lead time [months] Improvement of forecast skill for lead times of 3 months and more Improvement in the order of 5%, depending on the subbasin
Implemented in CHPS
Hybrid method ReduceESP selects historical traces from ESP StochResampler generates additional traces
Conclusion Running / Setting up hydrological seasonal forecasting system is very easy (extension of flow forecasting system for short/medium range); Data assimilation is important for most places and can improve skill up to one month (in some places even more); Pre- and postprocessing techniques of seasonal forecasts are essential; Usage depends very much on the setting;