The GCOS Reference
Upper Air Network: Assuring
the 21st Century Climate
record?
Peter Thorne, CICS-NC
With thanks to GRUAN Lead
Centre (DWD) and Working
Group on Atmospheric
Reference Observations
Meteorological Observatory Lindenberg – Richard Assmann Observatory
What is GRUAN?
 GCOS Reference Upper Air Network
 Network for ground-based reference observations for climate in
the free atmosphere under the auspices of GCOS
 Initially 15 stations, envisaged to be a network of 30-40 sites
across the globe
See www.gruan.org for more detail
Meteorological Observatory Lindenberg – Richard Assmann Observatory
GRUAN tasks
 Provide long-term high-quality upper-air climate records
 Constrain and calibrate data from more spatiallycomprehensive global observing systems (including
satellites and current radiosonde networks)
 Fully characterize the properties of the atmospheric
column
Meteorological Observatory Lindenberg – Richard Assmann Observatory
GRUAN goals
 Maintain observations over several
decades for accurately estimating
climate variability and change
 Focus on characterizing
observational biases, including
complete estimates of measurement
uncertainty
 Ensure traceability of measurements
by comprehensive metadata
collection and documentation
 Ensure long-term stability by
managing instrumental changes
 Tie measurements to SI units or
internationally accepted standards
 Measure a large suite of co-related
climate variables with deliberate
measurement redundancy
Priority 1: Water vapor,
temperature, (pressure and wind)
Priority 2: Ozone, clouds, …
Meteorological Observatory Lindenberg – Richard Assmann Observatory
GRUAN structure
 GCOS/WCRP AOPC Working Group on Atmospheric Reference
Observations (WG-ARO)
 GRUAN Lead Centre at the Lindenberg Meteorological Observatory
(DWD)
 GRUAN sites world wide (currently 15 to be expanded to 30-40)
 GRUAN task teams for
 Radiosondes
 GNSS-Precipitable Water
 Measurement schedules and associated site requirements
 Ancillary measurements
 Site representation
 GRUAN Analysis Team for Network Design and Operations Research
(GATNDOR)
See www.gruan.org for more detail
Meteorological Observatory Lindenberg – Richard Assmann Observatory
Why is GRUAN required?
• Historical observations of the atmospheric column have
been made primarily for operational monitoring purposes
• Change has been ubiquitous, poorly managed, and rarely
adequately quantified
• Has led to substantial ambiguity in the rate and details of
climatic changes
• Significant impediment to understanding climate change
and its causes.
Meteorological Observatory Lindenberg – Richard Assmann Observatory
Ubiquitous change
Meteorological Observatory Lindenberg – Richard Assmann Observatory
Tropospheric temperature trend uncertainties
Meteorological Observatory Lindenberg – Richard Assmann Observatory
Implications
• Surface-troposphere trends issue has been ‘hot’ since 1990 paper in
Science by Spencer and Christy using terms such as ‘precise’ to
describe MSU.
• Since then 200+ papers and two dedicated reviews on the subject
(NRC, CCSP)
• Several congressional hearings
• BUT …
• No resolution to the issue – simply a better understanding of the true
degree of uncertainty
• Lesson 1: Never trust a single observational analysis. Structural
uncertainty is key.
• Lesson 2: It doesn’t have to be this way going forwards. We need
traceable measures in future to assure the record.
• Lesson 2 is where GRUAN comes in …
Meteorological Observatory Lindenberg – Richard Assmann Observatory
Focus on reference observations
A GRUAN reference observation:
 Is traceable to an SI unit or an accepted standard
 Provides a comprehensive uncertainty analysis
 Is documented in accessible literature
 Is validated (e.g. by intercomparison or redundant
observations)
 Includes complete meta data description
Meteorological Observatory Lindenberg – Richard Assmann Observatory
Establishing reference quality
Meteorological Observatory Lindenberg – Richard Assmann Observatory
Establishing Uncertainty
• Error is replaced by uncertainty
 Important to distinguish contributions from systematic
error and random error
• A measurement is described by a range of values
 generally expressed by m ± u
 m is corrected for systematic errors
 u is random uncertainty
Literature:

Guide to the expression of uncertainty in measurement (GUM, 1980)

Guide to Meteorological Instruments and Methods of Observation, WMO 2006, (CIMO Guide)

Reference Quality Upper-Air Measurements: Guidance for developing GRUAN data products,
Immler et al. (2010), Atmos. Meas. Techn.
Meteorological Observatory Lindenberg – Richard Assmann Observatory
Uncertainty, Redundancy and Consistency
 GRUAN stations should provide redundant measurements
 Redundant measurements should be consistent:
m1  m2  k u12  u22

No meaningful consistency analysis possible without uncertainties

if m2 has no uncertainties use u2 = 0 (“agreement within errorbars”)
Meteorological Observatory Lindenberg – Richard Assmann Observatory
Uncertainty, Redundancy and Consistency
Understand the uncertainties:
 Analyze sources:
Identify, which sources of measurement uncertainty are
systematic (calibration, radiation errors, …), and which are
random (noise, production variability …). Document this.
Synthesize best uncertainty estimate:
 Uncertainties for every data point, i.e. vertically resolved
Use redundant observations:
 to manage change
 to maintain homogeneity of observations across network
 to continuously identify deficiencies
Meteorological Observatory Lindenberg – Richard Assmann Observatory
Consistency in a finite atmospheric region
Co-location / co-incidence:
 Determine the variability () of a variable (m) in time and space
from measurement or model
 Two observations on different platforms are consistent if
m1  m 2  k  2  u12  u 22
 This test is only meaningful, i.e. observations are co-located
or co-incident if:
  u12  u22
Meteorological Observatory Lindenberg – Richard Assmann Observatory
Uncertainty example:
Daytime temperature Vaisala RS92
Sources of measurement uncertainty
(in order of importance):
 Sensor orientation
 Radiative heating of sensor
 Unknown radiation field
 Ventilation
 Ground check
 Calibration
 Time lag
Meteorological Observatory Lindenberg – Richard Assmann Observatory
Uncertainty example:
Comparison Vaisala RS92 with Multithermistor
 Minor systematic difference
at night
 Significant systematic
difference during the day
 But observations are
consistent with the
understanding of the
uncertainties in the Vaisala
temperature measurements
 Lack of uncertainties in
Multithermistor
measurements precludes
further conclusions
Meteorological Observatory Lindenberg – Richard Assmann Observatory
Principles of GRUAN data management
 Archiving of raw data is mandatory
 All relevant meta-data is collected and stored in a meta-data
base (at the lead centre)
 For each measuring system just one data processing center
 Version control of data products. Algorithms need to be
traceable and well documented.
 Data levels for archiving:
 level 0: raw data
 level 1: raw data in unified data format (pref. NetCDF)
 level 2: processed data product → dissemination (NCDC)
• Data streams reprocessed as necessary as new knowledge
accrues
Meteorological Observatory Lindenberg – Richard Assmann Observatory
GRUAN data flow
Distributed data processing
DATA dissemination (at NCDC)
documentation
GRUAN
Meta-database
(at GRUAN
lead center)
data
Data
processing
center
raw
data
archive
GRUAN sites
Meteorological Observatory Lindenberg – Richard Assmann Observatory
Future steps
•
•
•
Bring in additional data streams
•
Frostpoint hygrometer sondes (WV in UTLS)
•
GNSS-PW
•
Lidar, FTIR, MWR etc.
Additional sites
•
Workshop to be held summer 2012 (let me know if interested)
•
Need to ascertain optimal mix of sites to meet the varied demands
Building user base
•
GRUAN will only be successful if the data are used on a regular
basis.
Meteorological Observatory Lindenberg – Richard Assmann Observatory
Next challenge: How to use these measures to calibrate more
globally complete networks
• Statistical and physical problem
• Geographical and temporal coincidence will be important.
• For satellite calibration use RTMs to convert the
geophysical observations to radiance equivalents?
• Does sustained cal/val require launch coincident
measurements? What is the cost/benefit? Who pays?
• Use of sites as opportunities to perform regular
instrumentation suite intercomparisons?
• Could help in calibrating ground based remote sensing
and in-situ sounding capabilities.
Meteorological Observatory Lindenberg – Richard Assmann Observatory
Summary of GRUAN
 GRUAN is a new approach to long term observations of upper air
essential climate variables
 Focus on priority 1 variables to start: Water vapor and temperature
 Focus on reference observation:
 quantified uncertainties
 traceable
 well documented
 Understand the uncertainties:
 analyze sources
 synthesize best estimate
 verify in redundant observations
 GRUAN requires a new data processing and data storage approach
Meteorological Observatory Lindenberg – Richard Assmann Observatory