How will SWOT observations inform hydrology models? Eric Martin and Kostas Andreadis SWOT Science Definition Team Meeting 17-20 June 2014, Toulouse
Background Models developed with motivation of understanding the water cycle Budget closure Reproducing variability of processes Impact and sensitivity studies Applications of water resources models can range Long-term re-analyses Prediction and forecasting at various scales (seasonal to decadal) Earth system modeling components for climate change simulations Some disparity among models that depends on specific application Number of deficiencies that complicates reconciling models with observations Trade-offs between resolving processes and model area size
Global & regional water balance Ability to close the water balance as a metric of model improvement How well do models perform over regional and global scales? What can we currently observe from remote sensing?
Observing each water balance term TRMM, GPM MODIS GRACE SWOT JASON, SENTINEL-3 SARAL, MODIS SMOS, SMAP Uncertainties exist in each satellite observation retrieval algorithms, representativeness of observations Observations of individual components do not close the water budget
Can models close the water budget? Models don’t agree with each other even when forced by identical data Example comparison of runoff from PILPS-2
Role of river discharge Discharge acts as a basin-wide “integrator” of water fluxes Discharge dynamics vary both spatially and temporally Extensively used to calibrate & validate hydrology models Level 2.5 data product from SWOT Indirectly estimated from SWOT observables Number of candidate “direct estimation” algorithms Instantaneous estimates at a reach-averaged scale Through data assimilation into river hydraulic models Potentially continuous estimates across river network and between observation times
State-of-the-art hydrology models Most large-scale hydrology models are essentially column models Flow routing schemes are rather simplified
Hydrodynamic models Water levels are usually not represented in hydrology models Need hydrodynamic models to accurately simulate processes in the river and floodplain Few models can be used to simulate large areas Downscaling formulations could allow transforming 1-km to <100-m scales Example of downscaling
What can SWOT improve? SWOT will provide observations of Water storage changes in lakes and reservoirs Water inundation and surface elevation River discharge Overview of improvements envisioned by SWOT Model calibration and validation Modeling and delineating lakes and wetlands Deriving information on reservoir operations Indirect estimation of water budget components (e.g. precipitation) Forecasting using either hydrologic or hydrodynamic models As with every novel type of observation, there will probably be improvements in models that are not included here GRACE is a favorite example, especially in hydrology
At what scales can SWOT be valuable? Spatial resolution of discharge observations will vary but on the order of few kilometers Should be adequate for current state-of-the-art hydrologic models Hydrodynamic models usually have finer spatial resolution SWOT will only observe rivers wider than 50-100 m SWOT will provide observations at varying temporal frequencies (~7-10 days) Should be adequate for model calibration It will be difficult to observe faster processes (e.g. flood wave propagation for most rivers) Higher latitudes will be better described
Spatial and temporal sampling Example of the Garonne River over an orbit cycle
Model calibration and validation Primarily hydrodynamic models Water levels are usually used for model calibration and validation Many examples of using in-situ or altimeter measurements for calibration Use of SWOT water level observations would be seamless Spatially-distributed measurements should provide order-of-magnitude improvement Discharge observations from SWOT can be directly used to calibrate hydrologic models Calibration for each sub-basin would lead to distributed parameters -> increased realism
Model calibration and validation (cont’d) Despite coarser resolution of discharge relative to hydraulic models SWOT can provide boundary inflows Provide a reach-averaged river channel bathymetry and roughness Adjust floodplain topography based on SWOT observations Example from the Amazon main stem Model agreement with Observations Model over- & under-prediction Water inundation observations have been used to calibrate against specific flood events
Reducing errors in water budget terms Given the role of discharge, SWOT observations can be assimilated to correct water budget imbalances Unconstrained Constrained Example of Mississippi Use of assimilation to constrain water budget Q from gauges – SWOT should further improve technique Water budget closure error
Lake water storage from SWOT Observations of water storage change Fine-scale determination of size and location of lakes Especially important for Arctic lakes WSE and delineation of lakes is an indirect combination of the water budget and dynamics Ability to extend information on storage by developing area-storage relationships Interpretation of the measurements requires models that incorporate lake dynamics No generic parameterization for lakes exists
SWOT observations of reservoir storage Some large-scale models incorporate reservoirs in their flow routing Simple linear schemes Optimizing releases according to reservoir type Coupling of hydrology models with dedicated water resources management models Observed WSE could be ingested directly or model parameters can be calibrated Issues of trans-boundary rivers persist for models of these systems Forecasting and assessment hydropower production
Mapping changes in wetlands Diversity and complexity of wetlands makes their modeling difficult Affect local water and energy exchanges due to relatively high ET Distributed versus areal modeling of wetlands Probably need to explicitly include groundwater in hydrologic models Changes in water storage and inundation extent can be used to calibrate model parameters Inferred ET can be assimilated directly Implications for eco-hydrologic models
Can SWOT help with forecasting? When there is a SWOT overpass initial conditions for a forecast can be estimated Estimation can be direct or indirect Direct example: flood forecasting Indirect example: hydrologic forecasting by estimating soil moisture that produced observed runoff Improvement in forecast skill by model calibration or identification of model biases Example of forecast error reduction when assimilating satellite WSE over the Ohio River
Challenges: discrepancy between observations and models What SWOT observes does not necessarily match the model state variable Transforming WSE to water depth depends on accuracy (or assumption) of topography Reach-averaged properties are not directly represented in the models Need to validate assumptions of aggregation with finer scale measurements and models Models lack or have simplistic representation of lakes and wetlands Do we need to modify existing model structures? What is the best approach for resolving these discrepancies?
Challenges: hyper-resolution modeling New satellite missions including SWOT are starting to provide information at high spatial resolutions Grand challenge of developing models at 1-km scale globally Not as simple as changing the grid cell size… Data assimilation must play key role Perhaps models need to be restructured with satellite observations in mind
Challenges: how to move forward Coupling of hydrology and hydrodynamic models Need to represent surface water inundation Leverage existing work or develop new schemes for Consumptive use Reservoir operations Lake and wetland dynamics Perform model validation experiments Continue work on data assimilation of SWOT and AirSWOT observations into both hydraulic and hydrodynamic models State estimation Model calibration Assess the feasibility of closing the water budget, in combination with other satellite observations Demonstrate value of approach in applications
Questions?