Intercomparison of Cloud Base Height at the ARM Southern Great Plains Site

1
Intercomparison of Cloud Base Height at the ARM
Southern Great Plains Site
Christina P. Kalb Oklahoma Weather Center REU,
Norman, Oklahoma The Ohio State University,
Columbus, Ohio kalb.29_at_osu.edu Mentors Andy
Dean, CIMMS Research Associate Randy Peppler,
CIMMS Associate Director Karen Sonntag, CIMMS
Research Associate
2
The ARM Program

• Climate models sometimes produce forecasts that
  are flawed
• Goals of ARM
• improve understanding of the formation,
  dissipation and radiative properties of clouds
• test and improve the accuracy of cloud
  parameterizations in models
• Ability to improve parameterizations is limited
  due to extent of knowledge on instrument
  performance

3
My Project

• Addresses differences in performance of
  instruments
• Specifically focuses on instruments that report
  cloud base height
• Importance many use data from instruments
  interchangeably
• Instruments perform unequally for different
  weather conditions and cloud types

4
Clouds or Not?
VCEIL first VCEIL second VCEIL third
MPL
Vaisala Ceilometer on 3/15/00
Micropulse Lidar on 3/15/00
5
Data

• Cloud base height measurements taken at Southern
  Great Plains Site
• Central Facility in Lamont, Oklahoma
• Blackwell Tonkawa Airport, temporary facility
• March 5 21, 2000, the Spring 2000 Cloud IOP
• Greater instrument examination and care
• Increased data

6
Instruments

• Central Facility in Lamont, Oklahoma
• Micropulse Lidar
• Belfort Laser Ceilometer
• Millimeter-wavelength Cloud Radar (MMCR),
  baseline instrument
• Blackwell Tonkawa Airport
• Micropulse Lidar
• Vaisala Ceilometer
• Adapted at Central Facility in Summer 2000
• Compared to Belfort Laser Ceilometer in 1997

7
Methodology

• Obtain data and set up plots
• Check instrument maintenance reports
• Qualitative analysis of Central Facility data
• Identify time periods where discrepancies occur
• Analyze weather data to determine probable causes

8
Preliminary Conclusions of Qualitative Analysis

• Micropulse Lidar
• Poor at depicting very low clouds
• Reports a large amount of scatter
• May pick up boundaries or Particle-laden regions
• Performs poorly during rain

9
Preliminary Conclusions

• Belfort Laser and Vaisala Ceilometers
• Report cirrus cloud bases too high
• Range of detection variable depending on ice
  content, temperature
• Report jagged clouds too high and with too little
  variation
• Perform poorly during rain
• All three instruments report low stratus and
  cumulus clouds similar to MMCR

10
Statistical Analysis

• First stage Statistical Comparison of mean cloud
  base height for all days and times
• Mean, standard deviation
• Hypothesis tests on the difference between means
  (zero or nonzero)
• Scatterplots
• Linear correlations, only where both instruments
  reporting
• Statistical results were inconclusive
• one false case
• Large standard deviations

11
Next Steps

• Screened data to throw out rain cases
• Break up remaining data to highlight specific
  cloud episodes
• Focus on Blackwell facility on March 15, 2000,
  low stratus
• Additional plots from
• March 13, cirrus and cumulus case
• March 19, low stratus case
• Further screened data to account for instrument
  limitations (most important and most difficult)

12
Measurement Ranges
VCEIL first VCEIL second VCEIL third
MPL
Vaisala Ceilometer 3/19/00
Micropulse Lidar 3/19/00
13
Scatterplot (VCEIL vs MPL)
Adjusted Range 0.5 to 7.5 km
Before Screening
14
Comparison Plots
MPL VCEIL first VCEIL second VCEIL third
MPL VCEIL
MPL VCEIL
Original Plot
Both Instruments within range
15
Second Stage

• Lag correlation
• 30 second data high autocorrelations
• Determine how far out we must go to get
  independent observation
• Used lag correlation of 0.1 to determine an
  effective n
• Recomputed standard deviation with effective n
• No change in Hypothesis Test results
• Reported cloud base heights still statistically
  equal
• Does not separate cases

16
Differing Cloud Base Heights
MPL VCEIL first VCEIL second VCEIL third
17
Quality Control Check
Scatterplot
Standardized difference Histogram
18
Outliers
MPL VCEIL Outliers
19
Results

• Micropulse Lidar reports low cloud bases slightly
  higher than Vaisala Ceilometer
• Most outliers are high MPL cloud base heights
• Histograms did not work for all cases
• When cloud base heights visibly unequal,
  distributions skewed
• Cirrus Cases
• Non-flat based Cumulus
• Adjustment of histograms needed to identify
  outliers

20
Overall Conclusions

• Ceilometers and Micropulse Lidar perform
  similarly for
• Stratus clouds above 0.5km
• Flatter based cumulus
• Micropulse Lidar bases slightly higher than
  Vaisala Ceilometer for above cases
• Outlier identification possible for these cases,
  within height range (0.5km to 7.5km)
• Discrepancies during cirrus and non-flat based
  cumulus make outlier identification difficult

21
Acknowledgements

• Mentors
• Andy Dean, CIMMS Research Associate
• Randy Peppler, CIMMS Associate Director
• Karen Sonntag, CIMMS Research Associate
• Statistics Aid
• Dr. Mike Richman, OU School of Meteorology
• Providing Data
• ARM Program

22
(No Transcript)
23
Plot Ranges
Millimeter Cloud Radar
24
MMCR
Millimeter Cloud Radar
25
MPL Higher
26
Cirrus Clouds