Intercomparison of CHARM Data

1
Intercomparison of CHARM Data and WSR-88D Storm
Integrated Rainfall Gary Jedlovec (NASA), Paul
Meyer (NASA), Anthony Guillory (NASA), Ashutosh
Limaye (USRA), Global Hydrology and Climate
Center,Huntsville, Alabama and Keith Stellman
(NOAA/NWS), LMRFC, Slidell, Louisiana

• Goal Produce a fine scale hourly rainfall
  product by combining radar and special rain gauge
  measurements.
• Outline
• CHARM rain gauge network
• NWS WSR-88D HDP products
• Case study intercomparison
• Data combination
• Discussion

2
COOPERATIVE HUNTSVILLE-AREA RAINFALL MEASUREMENT
NETWORK (CHARM)

• Local precipitation network ( est. 1/2001)
• 110 sites in Huntsville Madison County, AL
• NASA, Army, USGS, and NWS sites and weather
  enthusiasts
• Daily rainfall totals (1200UTC reports)
• 3600 km2 coverage (1 gauge per 6x6 km)
• Primarily 4 manual gauges (70) with remaining
  (30) manual or automated tipping bucket (6 and
  8)
• Plans to expand to 200 stations by 2003
• 4x4 km average spacing
• twice daily manual observations
• 1 minute data from 40 automated sites

http//wwwghcc.msfc.nasa.gov/charm
big picture

• Supports local weather and climate research at
  the GHCC
• validate weather radar and lightning data from
  satellites
• monitor spatial distributions of precipitation
  for modeling activities
• various satellite remote sensing studies of
  soil moisture and energy fluxes

3
TYPES OF RAIN GAUGES USED IN CHARM
4

• Details
• Strong north-south gradient on
• either side of the storm track.
• Width of heavy rainfall area is
• about 10 km
• Maximum of 2.96 on east of
• region, secondary max (2.73) in
• the western half of network

big picture
5
Hourly Digital Precipitation Product

• NWS offices produce their own precipitation
  estimates from their local WSR-88D radar. The
  Nashville and Hytop (northeast of Huntsville)
  radars captured this storm event.
• Figures present storm totals (sum of hourly
  product over storm lifetime 4-6 hours).
• Same general structure
• Nashville radar depicts more intense and
• widespread rainfall
• Radar calibration, elevation, scan patterns and
• distance from storm may all contribute to the
  
• difference.

4-5
3
6
NWS HDP Products
Example of stage III radar product for June 4,
2001 over Northern Alabama. Hourly stage III
product summed from 2100 0300 UTC.

• Stage I - integrated precipitation from local
  radar using standard or tropical Z-R
  relationships on
• 4 km grid dependent on Z-R relationship used
• Stage II - adjusted stage I product for bias
  using local (hourly) rain gauges
• Stage III - combination of stage II HDP from
  individual radars to produce a regional
  precipitation estimate
• minimal bias due to mis-calibration and
  different Z-R, local and seasonal adjustments
• regional continuity and consistency

Rainfall (cm)
Note the dual rainfall maximum in the stage III
radar product corresponds nicely both in position
and magnitude to the CHARM data.
Stellman et al. 2001 (Wea. Fore.)
CHARM
7
Stage III versus CHARM

• Resolution issues will affect comparison
• Radar
• radar volume varies with distance
• 4km grid cells arbitrary selection
• Rain gauge
• point measurement microscale variability of
  rainfall greatest in convective situations
• multiple rain gauges for each grid cell (a
  single best comparison used in analysis below)

ALL POINTS MD 0.19 SD 0.50
Comparison shows little or no bias (0.19)
between CHARM measurements and radar estimates.
Scatter is considerable especially for amounts
1.00.
1 MD 0.21 0.17 SD
0.15 0.60
8
Combined Stage III and CHARM Data
Can time-continuous stage III data products be
used to improve on 24h precipitation estimates
from spatially dense CHARM rain gauge network?

• Adjust hourly stage III data using CHARM 24h data
• Assume rain gauge (24h total) is correct and
  assume (radar rain gauge) bias is constant with
  time
• Calculate bias between stage III data and CHARM
  24h amount on a gauge-to-radar (4km data cell)
  basis
• Use point-to-point bias to scale (or adjust)
  stage III data

Actual gauge values

• For NSSTC site, bias factor was 1.36 (radar
  under-estimate).
• Rainfall duration is spread over wider span
• Adjustment produces radar under-estimate

9
Combined Radar Rain Gauge Product - Summary

• Individual radar estimates of precipitation vary
  greatly from radar-to-radar without some bias
  adjustment.
• The NWS stage III HDP product which uses (a few)
  regional rain gauges to adjust radar estimates of
  precipitation for calibration and Z-R
  inaccuracies shows considerable improvement in
  precipitation estimation.
• Hourly radar rainfall estimates can be used to
  increase the utility of daily precipitation
  measurements (24h totals) from gauges by applying
  local bias corrections and radar time estimates.
  However, as a result
• the duration of the rain event is often over
  estimated by up to 2dt
• (2 hours for this study),
• 2) the duration over estimate will lead to an
  storm intensity underestimate
• Despite these minor shortcomings, the combined
  radar and gauge rainfall product has a variety of
  uses for meso/microscale studies.

10
GAUGE OWNERSHIP MAP
80km
back
JANUARY 2002
11
back