1. Streamflow Predictability Tom Hopson

2. Conduct Idealized Predictability Experiments Document relative importance of uncertainties in basin initial conditions and weather and climate forecasts on streamflow Account for how uncertainties depend on – type of forcing (e.g. precip vs. T) – forecast lead-time – Regions, spatial-, and temporal-scales Potential implications for: – how to focus research efforts (e.g. improvements in hydrologic models vs data assimilation techniques) – observational network resources (e.g. SNOTEL vs raingauge) – anticipate future needs (e.g. changes in weather forecast skill, impacts of climate change)

3. Initial efforts Start with SAC lumped model and SNOW-17 – (ignoring spatial variability) Applied to different regions – Four basins currently Drive with errors in: – initial soil moisture states (multiplicative) – SWE (multiplicative) – Observations (ppt – multiplicative; T – additive) – Forecasts with parameterized error growth Place in context of climatological distributions of variables to try and generalize regional and seasonal implications (e.g. forecast error in T less important in August compared to April in snow-dominated basins)

4. Historical Simulation Q SWE SM Historical Data PastFuture SNOW-17 / SAC Sources of Predictability 1.Run hydrologic model up to the start of the forecast period to estimate basin initial conditions; Model solutions to the streamflow forecasting problem…

5. Historical Simulation Q SWE SM Historical DataForecasts PastFuture SNOW-17 / SAC 1.Run hydrologic model up to the start of the forecast period to estimate basin initial conditions; 2.Run hydrologic model into the future, using an ensemble of local-scale weather and climate forecasts. Sources of Predictability Model solutions to the streamflow forecasting problem…

6. Physically based conceptual model Two-layer model – Upper layer: surface and interception storages – Lower layer: deeper soil and ground water storages Routing: linear reservoir model Integrated with snow17 model Sacramento Soil Moisture Accounting (SAC-SMA) model Rainfall - Evapotranspiration - Changes in soil moisture storage = Runoff Model parameters: 16 calibrated parameters Input data: basin average precipitation (P) and Potential Evapotranspiration (PET) Output: Channel inflow (Q) Model parameters: 16 calibrated parameters Input data: basin average precipitation (P) and Potential Evapotranspiration (PET) Output: Channel inflow (Q)

7. Study site: Greens Bayou river basin in eastern Texas Drainage area: 178 km 2 Most of the basin is highly developed Humid subtropical climate 890-1300 mm annual rain Unit Hydrograph: Length 31hrs, time to conc 5hr Greens Bayou basin DJ Seo

8. Forcing and state errors Observed MAP – multiplicative – [0.5, 0.8, 1.0, 1.2, 1.5] Soil moisture states (up to forecast initialization time) – multiplicative – [0.5, 0.8, 1.0, 1.2, 1.5] Precipitation forecasts – error growth model

9. Forecast Error Growth models Lorenz, 1982 – Primarily IC error E small E large Another options: Displacement / model drift errors: E ~ sqrt(t) (Orrell et al 2001)

10. Error growth, but with relaxation to climatology Probability/m Precipitation [m] Error growth around climatological mean Probability/m Precipitation [m] Error growth of extremes Short-lead forecast Longer-lead forecast Climatological PDF => Use simple model Where: p f (t) = the forecast prec err static = fixed multiplicative error w(t) = error growth curve weight p o (t) = observed precip q c = some climatological quantile Err static = [0.5, 0.8, 1.0, 1.2, 1.5] q c = [.1,.25,.5,.75,.95] percentiles

11. Greens Bayou Precip forcing fields – Nov 17, 2003 tornado Perturbed obs ppt [mm/hr] Perturbed fcst ppt All perturbed (including soil moisture) Q response [mm/hr] Perturbed soil moisture (up to initialization) Note: high ppt with low sm (aqua) Low ppt with high sm (green)