NPOESS Advisory Committee for Observing Systems Simulation Experiments
Report No. 1
March 1997
Introduction
The NPOESS Observing System Simulation Experiments (OSSE) Advisory Committee
was formed to provide technical oversight and scientific guidance to the investigators
involved in the NPOESS OSSE project. These investigators have requested the advice
of a committee representing the potential users of polar-orbiting satellite data that
would be provided by the sensor suite that eventually will be chosen for deployment on
the NPOESS satellites. The Advisory Committee has been asked to convene as
necessary to review progress on development and implementation of the OSSE system,
and report on the progress to the NPOESS Integrated Program Office. This offering
constitutes one of a series of such reports.
Progress Summary
The first committee meeting was held at the National Centers for Environmental
Prediction, Environmental Modeling Center's office in the World Weather Building,
Camp Springs, MD on March 4, 1997. In attendance at that meeting were the
following committee members: Akira Kasahara, T. N. Krishnamurti, Greg Mandt, Jan
Paegle, and Donald Norquist, who agreed to serve as chairperson. The committee
heard presentations from several of the NPOESS OSSE project investigators, and from
a representative of the NPOESS Integrated Program Office. A brief summary of each
of the presentations follows.
Steve Lord and Eugenia Kalnay of NCEP/EMC provided an introduction and
overview. They stressed an NCEP need for the best observational data from the future
satellite system, to put the data into use quickly after launch, and to maintain neutrality
and objectivity in instrumentation selection as the sensor suite is designed. They expect
the data will serve their current operational areas of both mesoscale and global
modeling. Current data assimilation methods use data as observed. This necessitates a
knowledge of the ``forward operator'' for each sensor type. This function acts on
model variables to convert meteorological background (usually, short-term forecast)
fields to the form of the variables actually measured by the sensors. Also needed are
the adjoint (that is, the transpose) of the linearized forward operator and a specification
of the observational error covariance. These products are usually the responsibility of
the sensor designer. Using the sensor data as observed rather than performing a
pre-processing retrieval of the meteorological quantities has resulting in major positive
impacts for current satellite sensor types in numerical weather analyses and predictions.
Also, executing the forward operator on the background fields can give a preview of
what the actual measured radiances from the sensor should look like. Adaptive
observations is a name given to a strategy to potentially overcome local uncertainty in
weather forecasts. An ensemble of forecasts are executed from sequentially altered
initial conditions, and a large spread in forecast results is obtained in a certain region.
It may be possible to use this information to alter the sampling of satellite data in this
region (or upstream from it) to better measure the pre-forecast conditions and introduce
these ``adapted observations'' into the forecast's initial conditions. In any case, any
new observing systems must have measurement accuracies better than the accuracy of
the background fields, in order to have a positive impact on the forecasts.
Steve Mango of the NPOESS Integrated Program Office spoke next. He presented an
overview of the NPOESS program, and a second presentation on the Integrated
Operational Requirements Document (IORD). In addition to giving the committee a
history of the program to date, Steve presented a number of topics that were of
particular relevance to the OSSE project. First, he stressed that he expected NPOESS
OSSEs to shed light on the best orbital designs and sampling strategies. Second,
though the basic sensor types needed to obtain the required parameters listed in the
IORD have been specified, he expects that the OSSEs will determine the level of
impact each sensor type should have in numerical weather prediction. This may help in
deciding which specific instruments to actually deploy on the operational satellite
system. This in turn can result in significant cost savings in avoiding unnecessary
instrument design and fabrication. Finally, an expected payoff from the OSSEs is a
determination of level of impact on weather forecasting from various sensor types and
combinations of sensor types. This knowledge should provide guidance in setting the
ultimate prioritization of the sensor packages. This in turn should influence overall
payload design.
Bob Atlas of NASA then gave a review of the critical issues involved in OSSEs. He
listed the possible objectives of any OSSE as determining impact on numerical weather
analyses and forecasts, evaluating data assimilation methods, and design of sensor
studies. The currently planned OSSEs will initially consider only the first and last of
these objectives, while relying on the current NCEP operational data assimilation
methods. Later in the project, data assimilation methods developed by NASA and the
Naval Research Laboratory may be used to study assimilation method impacts. Bob
stated that two major OSSE design issues that can affect OSSE realism are the use of a
forecast model in the OSSE that is similar or identical to that used to generate the
nature run, and the assumptions used in simulating observations. The goal is to achieve
as much realism in the OSSEs as possible, so that impacts obtained will be very
representative of those actually experienced once the candidate observing system data is
actually used in the data assimilation system.
John Derber of NCEP/EMC gave a short talk on the NCEP data assimilation system.
This is the system that will be used to perform the bulk of the OSSEs in the project. It
is expected to be representative of the type of operational system in use in 10 years
when the first NPOESS data is available. It is a global system based on a
three-dimensional variational method, in which all sensor data is considered
simultaneously. The heart of the system is its process of minimizing a so-called ``cost
function'', which describes the differences represented by the observed and background
states of the atmosphere. This involves the transformation of all background fields into
the form of the observed variables. These transformations constitute the forward
operators mentioned earlier. While for some observation types this process is trivial,
for others the physics of the relationships between the meteorological variables and the
instrument-sensed quantities must be known. A major benefit of the OSSE to the
operational weather prediction community is the preparation of the data assimilation for
the time when the data actually does become available.
Tom Kleespies of NOAA/NESDIS discussed advanced satellite sounder models.
Central to the design of such models is the understanding of the relationships between
the quantities sensed and the meteorological variables used in the numerical weather
prediction models. Tom presented work he and colleagues have been conducting on
the developing of forward models to relate meteorological profiles of temperature and
humidity with passive satellite-sensed radiance data.
Dave Emmitt of Simpson Weather Associates presented a review of space-borne lidar
observation simulation design, and its effect on past observing system simulation
experiments. Lidars would measure wind components as active signal returns are
reflected from aerosols or cloud particles. He stressed the limitations on the vertical
extent of the lidar-sensed profiles imposed by clouds, but stated that winds could be
obtained between breaks in cloud in partly cloudy scenes. Clouds were not explicitly
predicted in the nature runs used in past observing system simulation experiments, so
they had to be diagnosed from the nature fields and then corrected using real cloud
analyses. Investigation of the newly acquired ECMWF nature run will be conducted to
determine the extent to which realistic clouds are represented in the data sets.
Steve Lord reported that unpacking of the nature run and simulated observation data
recently acquired from the ECMWF has been begun. He indicated that the task
appears to be going smoothly. Evaluation of the nature run and simulated observations
will begin presently. Project investigators will compare the acquired data sets with
analyses based on real data to examine the physical realism and the types of errors
imposed on the simulated observations. He and other team members expressed concern
over the sufficiency of the 30-day duration of the nature run, especially given the fact
that it was begun directly from a real analysis, rather than allowed to evolve its own
climatology first. Additional sources for nature runs and simulated data will be
considered if it is deemed necessary to represent longer durations or other times of the
year.
Issues and Recommendations
After hearing the above presentations of the objectives, research plans, and future
milestones of the OSSE project, the committee members agreed that this is a very
important endeavor. Moreover, the committee recognized that the participating
scientists are well qualified to make significant contributions to advance the state of the
art in atmospheric data assimilation as well as to assist in the design of atmospheric
measurement instruments when this task is completed. Although the committee
members discussed the matter of funding of the OSSE project only briefly, all felt that
successful experiments will yield great savings in large expenditures for sensor systems.
Since the process of preparing to deploy the NPOESS sensor suite will task place over
a long period of time, it is important that the funding agency as well as the OSSE
research team maintain a long term view while working to meet short term objectives.
The single most important issue in the NPOESS OSSE project is the value of the effort
to the NPOESS/IPO. The IPO must be convinced that the funding of this project will
provide substantial guidance toward the selection, design, and deployment of the
eventual sensor suite. In order to assure the responsiveness of the OSSE research team
to what the IPO needs from them, we suggest that the IPO state the specific questions
that they need to have answered and objectives that need to be met through this project.
Defining the essential questions and objectives could serve to keep the OSSE research
team focused on what the IPO sees as important and allow us as a committee to judge
progress toward these objectives. In response, the OSSE research team should state the
format of how the questions will be answered and objectives met, with a proposed
schedule for the answers and fulfilled objectives. The advisory committee would be
willing to play a role in this exchange if the IPO and OSSE research team deem such
participation to be helpful.
It appears that the OSSE research team needs to clarify more fully with the
NPOESS/IPO which components of the converged satellite constellation configuration
will be constraints in the design of the OSSEs. According to Steve Mango, about the
only stated constraints to date are the required parameters as dictated in the IORD.
Because the IORD is considered a ``living document'', it can and probably will be
altered significantly as the process continues. The user community may desire access
to much of the raw data from the sensors, rather than environmental data records for all
sensors. There needs to be definitions given to what the raw data records will look
like, for each sensor type. The OSSE research team will have to set realistic
specifications for coverage of each sensor type in the design of the OSSEs. Then given
these specifications, the OSSEs will be run to evaluate impact.
The issue of quality control has implications for OSSE impacts. It is likely that some
of the OSSEs will include more than one candidate observing system. In conducting
such combined experiments, one must look at more than just the net impact. The
interaction between the contributions of the two or more systems must be evaluated.
For example, if one OSSE is run with system A, another with system B, and another is
run with system A + B, then the quality control (especially data rejection) and nature
of change to the background field should be studied by comparing the A + B
experiment with each of the solo system experiments. If, for example, the two or
more systems being used in conjunction oppose each other in contribution to the
analysis, that opposition must be seen through the interactions among the systems'
individual contributions in the OSSEs to realistically assess impacts of the candidate
systems.
A major issue in the performance of the project is maintaining, and not expanding, the
current scope. Their is a danger of the team overextending itself if too many
``interesting'' side studies are undertaken. The advisory committee recommends that
the team begin first with the evaluation of impact of the basic passive sensor suite and
active lidar based on each sensor's primary contribution to the experiments. For
example, the focus on lidar is how it can provide wind information to the data
assimilation process. If time permits after all primary experiments have been
conducted and evaluated, the team could consider the lidar's potential contribution of
some cloud and aerosol properties to the assimilation process. Another example of a
high interest side study is that of adaptive observations. While this topic is currently a
hot one, it is important that the experiments involving the standard expected data
sampling designs be conducted first. After all, it is in contrast to the standard satellite
data sampling patterns that adaptive observation sampling methods and their impacts
would be evaluated. In general, it is important to focus on what we know well early in
the OSSE project, and avoid setting up to run experiments for systems for which our
knowledge is more limited. The expected accuracy of the system, the level of its
expected impact, and how accurately it can be represented in the assimilation system
(that is, how well the forward operator is known), should determine the order in which
the observing system experiments should be done.
Limitations imposed by lack of computational resources, possible funding shortfalls,
and the fact that the majority of team members can only contribute part of their time to
the project was discussed. It was stressed that time will even be ``borrowed'' from
team members' colleagues that are not associated with the project. However,
discoveries made in the project will be of benefit to all operational work, and will
advance the state of research knowledge. With regard to the possible need for
computational resources, the advisory committee recommends contacting the
organizations that are primary sponsors of the NPOESS project. For example, the
Department of Defense has a high-performance computing network that might be
considered as a possible additional resource. On the topic of possible funding
problems, the advisory committee felt that they could better monitor and possibly
advocate for continued funding support if they were supplied with a cost estimate for
the current project statement of work. This would allow the committee to better
recommend where cuts in project effort could be made if funding cuts were actually
realized. Steve Lord was asked by the committee to provide such a cost estimate.
Next Meeting
The committee agreed with Steve Lord to schedule our next meeting to coincide with a point in the research team's efforts in which substantial progress could be reported. In fact, the committee felt that, rather than arbitrarily planning to meet at regular time intervals, we would meet at the suggestion and request of the OSSE research team. This approach is in keeping with the primary role envisioned for the committee, as stated in the introduction of this report. We agreed that the next meeting might take place in the Fall of 1997, subject to the perceived value of such a date to the OSSE research team as the date draws nearer. Of course, consultation with the committee members by phone, mail, fax, or e-mail at any time is an option that the research team should feel free to exercise.