An End-to-End Framework for Probabilistic Uncertainty Characterization of Climate Satellite Data and Products

1. Observational data sets have been utilized in various
capacities including surface wind stress fields as forcing to
ocean circulation models, sea surface temperature fields as
boundary conditions to atmospheric circulation models, and
high quality in-situ data as reference time series for validating
or calibrating either models or/and other observational data
sets including satellite data.
Although it has been widely-recognized that accurate, global
high-resolution observational data sets with uncertainty
estimates for each model grid point or cell are crucial in
improving numerical weather and climate predictions, it is a
challenge providing this uncertainty information systematically
due to the nature of the complexity, high dimensionality, and
huge volume of data that one has to deal with in characterizing
satellite data and products. In addition, the sources for the
uncertainty of any final product may be accumulative but the
impacts of all sources/processes are not equal. Therefore,
identifying possible sources of uncertainties during each stage
of creating satellite data and associated products will provide a
useful framework for systematically estimating overall
uncertainties and exploring their impact on climate predictions
and decision-making.
An end-to-end framework for probabilistically characterizing
uncertainty sources associated with climate satellite data and
products is proposed with uncertainty types identified from
the level 0 data to level 4 products. Ocean surface winds are
used to demonstrate the utility of this end-to-end system.
N O A A’s N a t i o n a l C l i m a t i c D a t a C e n t e r, A s h e v i l l e , N o r t h C a r o l i n a
AMS 94th Annual Meeting Paper Number - 332
10th Annual Symposium on New Generation Operational Environmental Satellite Systems
We thank scientists from or affiliated with the Product Branch of NOAA/NCDC/RSAD
for beneficial discussions and input, especially Richard Reynolds, Hilawe Semunegus, Lei Shi, Ken Knapp, and Huai-min Zhang.
CONTACT US
Ge.Peng@noaa.gov
DeWayne.Cecil@noaa.gov
Bryant.Gramer@noaa.gov
MEASUREMENTS PRODUCTS APPLICATIONS
Uncertainty Types
• Satellite deterioration (orbital drift
and decay)
• Instrument accuracy/precision
• Instrument degradation
• Instrumental calibration (pre-, on-
board, and post-operation)
• Mission ground system
• Geo- and time-referencing process
(RDRs, SDRs)
Uncertainty Types
• Retrieval algorithm (ECVs,
EDRs, FCDRs, TCDRs, …)
• Sampling (space/time)
• Remapping (space/time)
Uncertainty Types
• Product algorithm (CDRs,
CIRs, … )
• Remapping (space/time)
• Inter-calibration
• Source (in situ or model)
systematic error
• Model uncertainty
Ground-based measurements are often used in calibration/validation/evaluation
and creation of satellite-based products: valuable but also induce error and uncertainties
Uncertainty Types
• Downscaling
• Uncertainty propagation
• Model uncertainty
• Social uncertainty
Platforms
• DMSP
• TRMM
• EOS
• QuikSCAT
Sensors
• SSM/I
• TMI
• AMSR-E
• QuikSCAT
Input Data
• Satellite-based wind
speed
• Model wind direction
Use Cases
• Energy (wind turbine)
• Hydrological cycle
(evaporation)
• Marine monitoring &
transport (coral-reef, ship
routing, …)
• Weather & climate
prediction
Input Data
• Brightness temperature
• Sigma-0
Remote-Sensed Ocean Surface Wind Products
Level 0
unprocessed telemetry
data as received from the
observing platform
excluding communications
artifacts by the ground
system
Level 1a
telemetry data that have been
extracted but not decommutated
from Level 0 and formatted into
time-sequenced datasets for easier
processing. The Level 1a formats
are NOAA’s internal formats and
only exist for the purpose of
creating the Level 1b datasets
Level 1b
discrete, instrument-specific
datasets from level 1a containing
unprocessed data at full
resolution, time-referenced, and
annotated with ancillary
information including data quality
indicators, calibration coefficients
and georeferencing parameters
Level 2
derived geophysical
variables at the same
resolution and locations
as the level 1 source data
Level 3
variables mapped on
uniform space-time grid
scales, usually with some
completeness and
consistency
Level 4
variables derived from
multiple measurements
(e.g., model outputs or
results from analyses of
lower level data)
NOAA
Product
Levels
An End-to-End Framework for Probabilistic Uncertainty Characterization
of Climate Satellite Data and Products
Ge Peng1, L. DeWayne Cecil2, and Bryant Cramer2
1 Cooperative Institute for Climate and Satellite, North Caroline State University (CICS-NC) and
NOAA’s National Climatic Data Center (NCDC)/Remote Sensing Application Division (RSAD), Asheville, North Carolina, USA
2 Global Science & Technology, Inc. and NCDC’s Climate Data Record Program, Asheville, North Carolina, USA
BACKGROUND