Hydrological Utilization of Radar Data'

1
Hydrological Utilization of Radar Data.
2
Each source of error in precipitation estimates
by radar is evaluated separately
Wet-radome Attenuation
Beam broadening Height increase Bright Band
Contamination How to get Z at ground !
Strong attenuation by precipitation at C-band
Z ? R
3

• C0 Removal of Ground Clutter. No VPR correction
• C1 C0 Correction of 1-hr accums on the basis
  of the observed 1-hr VPR
• Accumulation are derived from 1.5 km CAPPIs
  (1-km res., 120km range)
• The correction is applied to the 1-hr accums
• C2 C1 Simulation of the observed 1-hr VPR to
  ranges up to 200 km
• Correction of the 1-hr accums on the basis
  of the observed VPR within
• 110 km and of the simulated VPR at farther
  ranges. (2 km resolution)
• C3 The 1-hr accumulations are generated from the
  VPR-corrected surface precipitation maps. The
  lowest reflectivity data above the ground echo
  and not affected by the bright band are used. A
  VPR correction is applied when forced to take
  data at higher heights (1-km res., 120 km range)
• A more recent version of C3 attempts to correct
  data inside the bright band (rather than avoiding
  it) and extends the correction to the full range
  of 240 km by simulating the observed VPR at a
  closer range.
• In C1 and C2 the VPR correction is applied once
  an hour
• In C3, the correction is applied to the 12 maps
  included in that hour

4
One-hour accumulations (C1, 1-km res. 120 km
range)uncorrected and corrected )
5
24-hr accumulations, 2-km res. 240 km
range 0200 GMT 16-Jun-2002 (Event E later)
6
Progress to date

• Archived hourly precipitation data from the
  McGill and Franktown Radars since Dec. 5. 2001
  (Up to 7 data streams).
• Set up the WATFLOOD model for 73 gauged
  watersheds within the McGill radar coverage in
  Ontario, Quebec and the US.
• Archived streamflow and temperature data within
  the study area (Mostly from web sites).
• Compared the various radar products on the South
  Nation and Raisin Rivers in Ontario.

7
Water System Modeling
Evapotranspiration
Interception
Depression Storage
Wetting Front
Unsaturated Zone
API
Saturated Zone
Channel Flow
8

• Parameters are for land cover classes A, B, C D
• Parameters do not change with percentage of each
  land cover
• Each grid is represented by a watershed with its
  own cover allocation.
• There are NO watershed parameters

9
Chateauguay
DEM of Study Area (Red spots are streamflow
stations)
10
Study area
11
Castor R.
S. Nation R.
Raisin R.
Urban Bare Low crops Woodland??? Wetland Dec
. Forest Conf. Forest Water
Land Cover Map
12
Ottawa
Cornwall
Total Precipitation for 2002 in mm
Brockville
13
error in annual runoff Computed flow using
RADAR precip. Data are somewhat range dependent
Why do we have these differences ??
McGill RADAR
Focus on S. Nation River Near Ottawa
14
Fig. 6 Cumulative precipitation in mm. for the
period Dec. 5 2001 to June 30, 2002
15
875 750 625 500 375 250 125
F
Snow
Cumulative precipitation for four locations
(shown in previous slide) on the South Nation
and Raisin River watersheds. Need case by case
analysis to determine deficiencies.
16
cms
SNOW

• South Nation River near Plantagenet (02LB005)
  3810 km2
• C2 best estimate of precipitation Closest to
  observed hydrographs
• Continuous Simulation

17
cms
SNOW

• Raisin River near Williamstown (02MC001) 404 km2
  (Model 345 km2)
• C2 best estimate of precipitation Closest to
  observed hydrographs

18
Success Story

• Snow on the ground is accumulated using the C3
  RADAR algorithm (Clutter Mask Anomalous
  Propagation removed VPR correction out to 110
  km Simulated VPR out to 200 km) but was
  multiplied by 2 by the hydrologists!
• The melt hydrographs are modeled using WATFLOOD

19
2002 snow accumulation and ablation
Blackobserved Red computed
20
2002 snow accumulation and ablation
Blackobserved Red computed
21
Study area and percent error in runoff for the
South Nation and Raisin Rivers for the period
Dec. 5, 2001 to June 30, 2002 (Winter and Spring
Seasons) Most of the error resulted from Event A
in late March when serious bright band problems
often occur.
22
Inspect Event E

• At Spencerville No observed no computed
  hydrograph. (GOOD)
• At Delisle, slightly more precipitation (i.e.
  lots) and good agreement between computed
  observed hydrograph. (GOOD)
• For the Raisin River at Williamstown, low
  observed precipitation leading to no computed
  hydrograph but there is a large observed
  hydrograph. (BAD)

23
CMS
Event E
mm
CMS
CMS
24
Inspect Event F

• Highest precipitation at Spencerville Computed
  hydrograph compares reasonable well with the
  observed.(GOOD)
• At Delisle, less precipitation and decent
  agreement between second peaks of computed
  observed hydrograph. (GOOD)
• For the Raisin River at Williamstown, lowest
  observed precipitation again leading to no
  computed second hydrograph but there is a large
  observed hydrograph. (BAD)

25
Event F
26
Summary

• Snow accumulation and ablation was modeled very
  well.
• For the two particular back-to-back events in
  June, precipitation was underestimated in one
  area and not the others. (Thresholds for runoff
  not reached).
• Delisle (nearest the radar) gave the best
  results.
• The rainfall events were relatively small. An
  extrapolation to large events can not be made.
  (But small events are important in setting up
  antecedent conditions for events that follow).
• Clutter Mask Removal of Anomalous Propagation
  VPR correction out to 110 km Simulated VPR out
  to 200 km provide the best results.

27
Conclusions

• Radar measurements of snowfall appear to very
  useful. Errors are compensated over the time that
  snow is accumulated.
• Many rainfall events still need to be analyzed to
  look for recurring problems.
• Suspect that even for same event in close
  proximity variations in the Z-R relationship need
  to be identified.
• Hopefully the more recent version of C3 which
  attempts to correct data inside the bright band
  (strongest signal) will be more successful.