SWOT DA estimate of river depth and slope

1
SWOT DA estimate of river depth and slope

• Why
• Highly uncertain for many rivers
• Temporally persistent errors lead to significant
  uncertainty in LISFLOOD discharge uncertainty
• Approach
• Retrospective ensemble data assimilation analysis
  using LISFLOOD similar to EnKF or EnKS
• Advantage
• Retrospective approach allows for rapid
  evaluation of various measurement characteristics
  (uncertainty, overpass frequency, pixel size,
  etc.)

2
Methods

• Observation vector hobs is timeseries of SWOT
  elevations
• ? is time invariant model input either river
  slope or river depth

• Generate ensemble of n20 LISFLOOD runs from
  ensemble of 20 values of ?
• Treat each of the ensemble members as truth in
  turn, estimate it using ensemble of n-119

3
Ensemble of LISFLOOD depth and slope inputs
4
LISFLOOD slope sensitivity Experiment results

• Model is highly sensitive to river slope
• Relationship is linear for slopes gt 2 cm km-1

5
River slope estimation

• Estimated river slope using timeseries of basin
  mean inundated area
• Each estimate is a mean across an ensemble
• SWOT slope RMSE estimated at 0.36 cm km-1

6
Estimate river depth from floodplain observations

• Define measurement window of pixels upstream from
  elevation node within spatial correlation length
• Use timeseries of SWOT water levels in the
  floodplain within the measurement window

7
River depth estimation
8
Future work

• Analysis
• Tradeoffs on slope/depth and discharge errors
  with overpass sampling frequency
• Tradeoffs on errors with pixel size
• Comparison of profiling and swath altimeters?
• Tradeoffs with and without a radiometer
• Timeline
• ASLO meeting (March 1)
• GRL paper (submit by March 15)
• Next steps
• Simultaneous river depth floodplain flow
  estimation for the Amazon
• Assimilate JERS measurements to estimate river
  slope/depth