Bias Adjustments Of Arctic Precipitation

1
Bias Adjustments Of Arctic Precipitation
Tianna A Bogart

• Department of Geography
• MS Thesis Presentation
• April 4, 2007

This research was funded by the National Science
Foundation under grant 0230083
2
Why are we so concerned with precipitation in
the Arctic?

• Model simulations show Climate Change will first
  be most noticeable in polar regions
• More accurate data are needed to monitor and
  analyze the hydroclimatology of the Arctic
• Ground truth for satellite data ?especially
  important since we simply cant get to and
  maintain a robust station network in the Arctic

3
What is wrong with the gage- recorded
precipitation?

• 2 types of Biases
• -Unsystematic discontinuities in station
  records, can be abrupt or gradual changes
• -Systematic mostly due to the physics associated
  with the can-type precipitation gages.
• This research focuses on Systematic Biases

4
Wind Bias

• Wind field deformation due to the physics of the
  gage.
• Most precipitation gages are located some
  distance above the ground, wind speeds tend to be
  higher at the gage than at ground level.

• Wind can cause a decrease in the gage-measured
  annual average of the worlds precipitation by
  about 81.
• Results from one study found a 30 to 120
  increase solely from adjusting for the wind bias
  for some Alaskan stations.2

1Legates, 1987 , 2Yang et al, 1998
5
Wetting Bias

• When precipitation collects on the inside walls
  of a gage, it is subject to evaporation or
  sublimation.
• Magnitude of wetting loss dependent on
• type of precipitation,
• number of times the gage is emptied,
• geometry of the gage, and
• materials used to construct the gage1
• A global estimate for wetting loss is about 2 of
  the gage-measured annual average1
• Average wetting losses have been measured to be
  as high as 14 of gage-measured yearly totals at
  some Alaskan staion.2

(former Soviet Union countries)
1Legates, 1987 , 2Yang et al, 1998
6
Evaporation Bias

• Underestimation of precipitation due to lag
  between the precipitation event and its
  measurement
• Magnitude of evaporation loss is dependent on
• the gage type,
• time of year,
• weather conditions,
• and observation methods1
• Studies have estimated the evaporative losses to
  be an average of 1 or less of the annual
  precipitation2.

1Sevruk, 1982, 2Sevruk, 1984
7
Trace Events (exclusion)

• amount of fallen precip is less than the finest
  resolution of the recording instrument
• no contribution to observed precip totals
• Individual trace event is about 0.05 to 0.15 mm
  of precip.

• Trace can account for 50 to 70 of all precip
  days in
• northern Alaska1
• Accounting for trace added 3 to 12 to the
  annual precip
• total for some Alaskan stations1.

1Yang et al 1998
8
(No Transcript)
9
Study Intro, Legates et al 2005

• Data from 2791 stations north of 50N from the
  Global Summary of the Day
• From1994 to 2002
• Bias adjustments applied were gage specific and
  dependent on site specific variables
• Adjustments applied to daily and monthly average
  data.

10
Legates et al 2005 Adjustments
General adjustment equation
Daily Adjustments
Monthly Adjustments
Only one trace event per day accounted for
Trace precipitation not included in adjustment
Type of precipitation for the whole day
determined from average temperature.
Fraction of precipitation in solid form (R)
calculated from average monthly temperature
Precip considered to be mixed if average temp
was btw 0 and 2C
Precipitation categorized as rain or snow only
11
WMO Solid Precipitation Measurement
Intercomparison Project

• In 1998 the World Meteorological Organization
  (WMO) released the Solid Precipitation
  Intercomparison Project1
• In this report a reference gage, the Double Fence
  Intercomparison Reference (DFIR), was used as the
  best estimate for ground truth precipitation.

1Goodison et al., 1998
12
WMO Solid Precipitation Measurement
Intercomparison Project
Snow Adjustment Coefficient
Rain Adjustment Coefficient
13
Current Research Goals

• Daily Gage-recorded vs. Daily-Adjusted
  sPrecipitation
• Daily-Adjusted vs. Monthly-Adjusted
  sPrecipitation
• Wind sensitivity, better estimate of R

14
1. Daily Gage-recorded vs. Daily-Adjusted
Precipitation
15
1. Daily Gage-recorded vs. Daily-Adjusted
Precipitation Methods

• Monthly totals computed from daily gage-recorded
  and daily-adjusted precipitation
• From these totals, monthly averages are computed
  over the study period (1994-2002)

• Criteria for data inclusion in this analysis
• - Month must have less than 3 days of missing
  data, otherwise that whole month was
  considered to be missing
• - For the monthly average, at least 5 of the 9
  years for a station must be present,
  otherwise that stations monthly average is
  not included in the analysis.

16
Monthly Average January Precipitation from
Daily Data
1994
Total for Jan, 1994
1995
Total for Jan, 1995
1996
Daily Data
Total for Jan, 1996
9 year monthly average From 31x9 279 values
. . . . . .
. . . . . .
2002
Total for Jan, 2002
17
1. Daily Gage-recorded vs. Daily-Adjusted
Precipitation Methods

• Thiessen Polygons created for each station and
  then shaded
• Thiessen Polygons
  
  -encompass the area closest to a station relative
  to any other station.
• -Shape and size of the polygons, therefore, are
  based on station distribution and proximity to
  neighboring stations.

Non-missing average January stations (n1031) and
their Thiessen Polygons.
18
January Monthly Average Precipitation
Percent Change
19
July Monthly Average Precipitation
Percent Change
20
Percent Change btw gage-recorded and
daily-adjusted Monthly Average Precip
21
1. Daily Gage-recorded vs. Daily-Adjusted
Precipitation
22
1. Daily Gage-recorded vs. Daily-Adjusted
Precipitation
23
2. Daily-Adjusted vs. Monthly-Adjusted
Precipitation
24
2. Daily-Adjusted vs. Monthly-Adjusted
Precipitation

• Why compare this?
• -Historical datasets often contain only monthly
  averaged data
• -Monthly total precipitation is available for
  more stations compared to daily precipitation
• -Since daily-adjusted precipitation is considered
  to be the true amount,
• ?lets see if adjusting on the monthly time
  scale tends to over- or under-estimate the
  true amount

25
Monthly Average January Precipitation from
Monthly Data
Jan, 1994
Remember, adjustments were done on the monthly
data, pretending the daily data dont exist
Jan, 1995
Monthly Data
9 year monthly average From 9 values
Jan, 1996
. . . . . .
Jan, 2002
26
Monthly Average January Precipitation from
Daily Data
1994
Total for Jan, 1994
1995
Total for Jan, 1995
1996
Daily Data
Total for Jan, 1996
9 year monthly average From 31x9 279 values
. . . . . .
. . . . . .
2002
Total for Jan, 2002
27
January Monthly Average Precipitation
Monthly-adj minus Daily-adj
28
July Monthly Average Precipitation
Monthly-adj minus Daily-adj
29
Monthly-adjusted minus Daily-adjusted Monthly
Average Precipitation
30
2. Daily-Adjusted vs. Monthly-Adjusted
Precipitation
31
Why are the daily-adjusted values and monthly
adjusted values so different? Some variables
needed for adjustment on the monthly time scale
have to be estimated
32
Variables needed to be estimated for Monthly
Adjustments

• Percent of Precipitation in solid form (R)
  calculated from monthly average temperature

where Ta avg temperature for the month
33
Variables needed to be estimated for Monthly
Adjustments

• Percent of Precipitation in solid form (R)
• Average wind speed during precipitation days
  (whp) is calculated from mean monthly wind speed

Logarithmic coef of the wind speed profile
Avg wind speed during precipitation events
Exposure coefficient
h2 precip gage height h1 anemometer height z0
roughness length
34
Variables needed to be estimated for Monthly
Adjustments

• Percent of Precipitation in solid form (R)
• Average wind speed during precipitation days
  (whp)
• Number of Precipitation Days (M) must be
  estimated from average daily precipitation (Pd),
  average temperature (Ta), and total number of
  days in the months (Dm).

35
How can we make the Monthly Adjustments better??

• Identify what adjustment makes the largest
  influence during the winter months
• See if we can come up with a better relationship
  between monthly averages and the variables needed
  for that adjustment.

36
Daily- Adjusted
Monthly-Adjusted
December
January
37
Daily- Adjusted
Monthly-Adjusted
February
March
38
3. Wind sensitivity, better estimate of R
39
Wind Adjustment Coefficients
Rain Adjustment Coefficient
Snow Adjustment Coefficient
40
Sensitivity Analysis of the Effect of Wind during
Snow Events Data

• 8 types of gage/shield combinations
• Stations with the least amount of missing data
  and most snow events are used for the wind
  sensitivity analysis.

41
Sensitivity Analysis of the Effect of Wind during
Snow Events Methods

• Daily recorded anemometer wind speeds are
  systematically perturbed
• The wind adjustment coefficient is then
  calculated from the daily perturbed wind speeds.
• Monthly averages of and whp computed. Only
  winter months (Dec-Mar) analyzed

42
Perturbed January Wind Speeds for ONE year
Wind Perturbed -90
Ave wind from -90 pert
Wind Perturbed -80
Ave wind from -80 pert
n 20 for one month from one year
1994
Daily Data
Wind Speeds ( of precip days)
. . .
. . .
n 20x9180 for one month from all years
Wind Perturbed 0
Original average
. . .
. . .
Wind Perturbed 100
Ave wind from 100 pert
43
010620, Norway
042020, Greenland
121950, Poland
222350, Russia
44
718250, Canada
116030, Czech Republic
504340, China
700260, Alaska
45
All Winter months, All years, All perturbations
max n4920720
010620, Norway
042020, Greenland
121950, Poland
222350, Russia
46
All Winter months, All years, All perturbations
max n4920720
116030, Czech Republic
718250, Canada
504340, China
700260, Alaska
47
Results from Wind Sensitivity Analysis
Wind sensitivity analysis may be the main
cause for under-estimation of monthly adjustments

What about the places where monthly adjustments
overestimated the precip?
?Need define the ratio of snow (R) to rain (1-R)
better.
Average January (1994-2002) monthly adjusted
minus daily adjusted precipitation.
48
Snow Fraction (R)
Arrrrrrrrrrrrrrrrrrrr!

• Can we define a better relationship between Snow
  ratio and avg monthly temperature?

• Use the actual ratio and avg temp (which we know
  from the daily data) to recalculate the
  coefficients

49
Snow Fraction (R)
a 3.29 b 1.695
50
Summary and Conclusions

• Snow adjustments much larger than rain
  adjustments
• Assuming the daily adjustments are a more
  accurate estimate of precipitation than monthly
  adjustments monthly adjustments both over- and
  under-estimate precipitation, with a net effect
  of overestimation
• The wind adjustment coefficient influences the
  monthly-adjustments underestimation of true
  precipitation
• The ratio of the type of precipitation (solid vs.
  liquid) influences the monthly overestimation.

51
Future Work

• Recalculate monthly adjustments using the new
  coefficients calculated from the wind sensitivity
  analysis.
• are the adjustments closer to those that were
  done on a daily basis?
• Sensitivity analysis on the assumption that every
  precipitation gage height matches that on the
  national standard.
• Use hourly data to estimate the bias associated
  with using daily data

52
Questions?
53
References
Goodison, B. E., P. Y. T. Louie, and D. Yang
(1998). WMO Solid Precipitation Measurement
Intercomparison, Final Report. Geneva, World
Meteorological Organization 212. Legates, D. R.
(1987). A climatology of global precipitation.
Publications in Climatology, 40 1-91. Legates,
D. R., D. Yang, S. Quiring, K. Felter, and T.
Bogart (2005). Bias adjustment to arctic
precipitation A comparison of daily versus
monthly bias adjustments. 8th Conference on
Polar Meteorology and Oceanography, San Diego,
CA, American Meteorological Society. Sevruk, B.
(1982). Method of correction for systematic
error in point precipitation measurement for
operational use. WMO-No. 589, pp. 91. Sevruk,
B. (1984). Comparison of evaporation losses from
standard precipitation gages. TECEMO/WMO, pp.
57-61. Yang, D., B. E. Goodison, S. Ishida, and
C.S. Benson (1998). "Adjustment of daily
precipitation data at 10 climate stations in
Alaska Application of World Meteorological
Organization intercomparison results." Water
Resources Research 34(2) 241-256.
For more information, please visit our
website http//www.deos.udel.edu/arctic.html