Bolton 2

1
Bolton 2

• Link Adjustment of Radar Data

3 October 2006
2
Aims of the project

• To investigate how effectively we can adjust
  radar rainfall estimates using information from a
  microwave link
• To investigate the usefulness of differential
  phase

3
Data
Radar - New scan every 5 minutes - Gives
estimate of start and end time of
events Link - 1 minute averages Gauge -
Tipping bucket (0.2mm)
4
Path-averaged Rain-rate

• Cartesian Radar Weighted average of the squares
  through which the link passes
• Gauge Weighted average of the gauges closest to
  link path
• Which gives most accurate estimate of true
  path-average?

5
(No Transcript)
6
Hourly accumulations of rainfall
Only consider hours where

• Link, radar and gauges all give path-averaged
  accumulations greater than 0.2mm.
• The temperature does not fall below 4C
• We have 12 radar scans in the hour

7
RMS RMSF Bias
Definitions
8
(No Transcript)
9
Outlier 2
Outlier 1
10
Outlier 1 2nd November 2005
11
Outlier 1 2nd November 2005, 1500-1600
12
Tip-plot 2 Nov 2005 0000-2400
13
Tip-plot 2 Nov 2005 1500-1600
14
(No Transcript)
15
Outlier 2 20th June 2006
16
Outlier 2 20th June 2006, 2100-2200
17
Tip-plot 20 Jun 2006 0000-2400
18
Tip-plot 20 Jun 2006 2100-2200
19
(No Transcript)
20
Hameldon Hill Link Attenuation 17.6GHz
H Attenuation 12.8GHz H Attenuation 12.8GHz
V Differential Phase 12.8GHz Path-Length 23.3km
21
Event definition (new method)

• We aim to have a method that does not need gauge
  information.
• It is raining on the link if
• at least one relevant radar pixel shows rainfall
  of at least 1 mm/hr, and
• the radar path average is at least 0.25 mm/hr

22
Link estimates (new method informed by Radar)
versusLink estimates(old method informed by
Gauges and Correlation) shows link can be used
without the gauges
23
(No Transcript)
24
Dual Frequency Estimates
25
(No Transcript)
26
(No Transcript)
27
(No Transcript)
28
(No Transcript)
29
17.6GHz Single Frequency Estimates
30
(No Transcript)
31
(No Transcript)
32
(No Transcript)
33
(No Transcript)
34
(No Transcript)
35
Radar Adjustment
36
Gauge Adjustment

• The Met Office currently compare hourly
  accumulations of gauge and radar data in order to
  calculate their adjustment factors
• The adjustment factor is then applied to the
  whole radar domain (Mean Field Bias Adjustment)

37
The gauges used for adjustment of the Hameldon
Hill radar
Radar pixels have a side of 2 km --- (All gauges
are within 100 km)
38
Review of Adjustment Process (I)

• Where
• w is a function of n(i), the number of
  gauge-radar pairs available in hour i,
• rj and gj are the radar and gauge accumulations
  in hour i at location j

39
Review of Adjustment Process (II)

• The weight used in hour k is given by
• w(k)n(k)/ ?(k)
• where n(k) is the number of non-zero
  gauge-radar comparisons available in hour k,
• ?(k)exp(-1/?) ?(k-1)n(k)
• and ? is one of

1, 10, 20, 50, 100, 200, 500, 1000, 2000

• The Met Office keeps running calculations for
  each ? value
• The adjustment factor chosen for hour k comes
  from the scheme with lowest ? that has ?(k)
  greater than 50

40
(No Transcript)
41
Link Adjustment

• Similar procedure but comparing path-averaged
  hourly accumulations instead of point
  accumulations
• Weights are proportional to the length of link
  that experiences rainfall gt1mm/hr in hour
  (determined using radar data)

42
(No Transcript)
43
Verifying Link Adjustment

• Compare hourly accumulations of radar and
  gauges at test gauges
• 1. Using Met Office adjusted radar data
• 2. Using link (17.6GHz) adjusted radar data

44
(No Transcript)
45
RMS RMSF Bias
Definitions
46
Gauge and Radar both greater or equal than 0.2mm
47
Monthly Bias
48
Monthly RMSE
49
Monthly RMSF
50
Gauge and Radar both greater or equal than 0.2mm
51
Gauge and Radar both greater or equal than 1mm
52
Gauge and Radar both greater or equal than 2mm
53
Gauge and Radar both greater or equal than 3mm
54
Gauge and Radar both greater or equal than 4mm
55
Gauge and Radar both greater or equal than 5mm
56
Does the effectiveness of the link adjustment
depend upon range?
57
Gauge and Radar both greater or equal than 0.2mm,
Gauges lt 40km from Radar
58
Gauge and Radar both greater or equal than 0.2mm
, Gauges gt 40km from Radar
59
Future work on adjustment

• Obtain adjustment factors using the Winter Hill
  non-radial link (see next slide)
• Experiment with integration times other than 1
  hour, (minimum of 5 minutes)

60
Winter Hill
Only High Frequency (22.9GHz) data available
61
(No Transcript)
62
(No Transcript)
63
(No Transcript)
64
Comparison of Horizontally and Vertically
Polarised Estimates at 12.8GHz
65

• Differences in attenuation at horizontal (H) and
  vertical (V) polarisation occur solely due to
  drop shape
• In rain expect to see H experience greater
  attenuation than V

66
A Raindrop
67
(No Transcript)
68
(No Transcript)
69
Hours containing sleet
70
Problems at High Rain-rates

• Hameldon 17.6GHz link is saturated at
  path-averaged rain-rates greater than 22mm/hr
• The 12.8GHz can reach 32mm/hr
• Variations in time spent off-line after
  saturation occurs

71
1230pm-130pm 8November 2005
72
1pm-4pm 2July 2006
73
Tip-plot 2 July 2006 1300-1600
74
Tip-plot 2 July 2006 0000-2400
75
(No Transcript)
76

• Differential Phase

77
Differential Phase - Rainrate relationship

• Has a greater dependence on drop size
  distribution, and
• Is less linear
• than Attenuation-Rainrate relationships.

78
(No Transcript)
79

• Scatter diagrams
• Same conditions as before with additional
  requirement that the phase accumulation is
  greater than 0.2mm

80
(No Transcript)
81
(No Transcript)
82
(No Transcript)
83
END
84
Tip-plot 2 Nov 2005 1500-1600
85
Tip-plot 20 Jun 2006 0000-2400
86
Tip-plot 2 Nov 2005 0000-2400
87
Tip-plot 20 Jun 2006 2100-2200