River Discharge from Space: Key Measurement Strategies

1
River Discharge from Space Key Measurement
Strategies
____________________

• University of New Hampshire/USGS Study Team
• C.J. Vörösmarty, D.M. Bjerklie, S.L. Dingman, B.
  Fekete, C.H. Bolster, R.G. Congalton, W. Kirby

____________________
See Bjerklie et al. 2003. J. Hydrology (in press)
____________________
River Wetland Processes NASA Working Group
Meeting Chicago, 4/5 November 2002
2
Goals for Discussion

• Candidate Space-Based Variables
• Review of General Approaches
• Error Analysis

3
Overall Objective of Study

• Knowledge of hydraulic relationships key to
  assessing potential for satellite-derived
  estimates of flow
• To evaluate univariate/multivariate river
  discharge estimation equations that use hydraulic
  variables potentially observed from space

4
Basic Relations

• Long history (e.g. Leopold et al.)
• Q VA VWY
• with mean velocity(V), cross-sectional
  area(A), width(W), and mean depth(Y)
• Relation exploited at site-specific gaging
  stations to measure Q
• Rating curve based on site-specific Qa(Z-e)m
  for operational application
• Typical accuracies -5

5
Basic Relations (cont.)

• Satellite-derived stage (i.e. altimetry) could
  supplant operational stage monitoring w/ periodic
  ground-truthing
• Major advantage would be to develop and apply a
  universal rating to poorly-monitored areas
• Predictable geometry relations link mean V, mean
  Y, W, S (slope) with Q

6
Four Models Tested

• 1) Width Depth Slope Model
• Based on Dingman and Sharma (1997) (mean
  accuracy gt80 n128)
• Q c1WaYbSd
• Q discharge, W width, Y mean depth
  (estimated from stage),
• S slope (geomorphic, constant channel
  slope from topo maps)
• Width Velocity Model
• Q c2WeVf
• V mean velocity (estimated from surface
  velocity)
• 3) Depth Slope Model
• Assumes parabolic channel
• Q c3WmgYmhSiYj
• Wm bankfull width, Ym bankfull mean
  depth
• Slope - Width Model
• Assumes parabolic channel
• Q c4WmkYmlSmWn

7
Hydraulic Observables from Space
8
Error Assessment

• Based on n1012 Q measurements
• (US, New Zealand, Amazon split 5050 into
  cal/val min Q0.01m3/sec max216,000m3/sec)
• - Many reach-integrated
• Regression analysis
• Regression analysis w/ random error applied to
  observation variables

9
Regression Model Comparisons
Model 4 omitted -- poor performance slope factor
not significant
10
Regression Model Comparisons
11
Model 1 WYS Model 2 WV
Model 3 WmYmSY Model 4 Dingman Sharma
12
Model 1 WYS Model 2 WV
Model 3 WmYmSY Model 4 Dingman Sharma
13
Regression with Uncertainty

• Maximum Minimum
• (95 range)
• Width (W) 10 m 1 m
• Depth (Y) 0.5 m 0.1 m
• Velocity (V) 0.5 m/s 0.1 m/s
• Unquantified additional error
• Uncertainty of slope measured from topography
• Uncertainty of estimating mean depth from stage
• 3) Uncertainty of estimating mean velocity from
  surface velocity

14
Regression with Uncertainty
Model 1 WYS Model 2 WV
Model 3 WmYmSY Q VWY
15
Regression with Uncertainty
Model 1 WYS Model 2 WV
Q VWY
16
Conclusions

• General and reliable empirical hydraulic
  relations can be constructed
• These relations can be applied in a space-borne
  context
• Regression and error analysis indicates that
• - No silver bullet -- alternatives can be invoked
  when all variables not available
• W,S,Y,V-dependent relations accurate generally to
  20
• Errors in retrievals add numerical dispersion
• Not a replacement of operational networks
• - Reliability decreases rapidly below 100 m3/sec
• - WV relations require additional calibration
• Additional errors (and new research on)
• - Converting observed V on surface to mean V
  (surface wind distortion, in-water V
  distribution)
• Converting observed stage to mean Y
• Explore reach-integrated variables (Wave, Area)
• - Determining bankfull Wm,Ym,
• Deriving slope S from topo maps or satellite
  ranging
• - Overbank conditions yet to be articulated
• - Flow downstream of hydraulic modifications