GRAPE
Global
Retrieval of ATSR Cloud Parameters and Evaluation
Introduction
The aim of the GRAPE project is to produce a global cloud dataset
using a state-of-the-art physical retrieval of the entire duration of the
ATSR-2 mission (aboard ERS-2). This dataset will be compared and contrasted
with existing climatologies (based on different instruments and very different
retrieval algorithms). The cloud parameters produced will allow a
direct comparison with model simulations from the ECMWF and the Met Office
Unified Model. In conjunction with water vapour estimates from the MLS
instrument aboard UARS, observational estimates of the total water vapour
budget will be possible.
Scientific Background
The importance of clouds to the radiation balance of the terrestrial
climate is well known. Accordingly, state-of-the-art climate models include
sophisticated parameterisations of the cloud variables. It has been known
for decades that climate simulations are very sensitive to changes in
these parameterisations. However, models can also introduce compensating
errors, that hide further sensitivities, for example, some GCMs differ
by almost an order of magnitude in their estimates of global ice water
path while having similar global top of the atmosphere (TOA) radiation
budgets. It is also well known that simulations aimed at predicting the
climate in the presence of increased CO2 are also
sensitive to the exact specification of the cloud parameterisations. Accordingly,
the community has put an imperative on the validation of these schemes
by confronting the model simulations with observations.
While high-resolution measurements obtained in experimental campaigns
are necessary to develop parameterisations, the evaluation of
model cloudiness requires comparison with global climatological data.
Such climatological comparisons can highlight specific areas of disagreement,
but do not always illuminate the reasons why the observations and models
disagree as other model problems can manifest themselves in the simulated
cloudiness. To some extent climate resolution simulations rely on parameterisation
development and validation carried out for Numerical Weather Prediction
(NWP) --- process studies at climate resolution are obviously much more
difficult to carry out in a meaningful way. For comparison with climate
models, observational studies restricted seasonally and/or spatially to
identify specific synoptic regimes are becoming more important.
A new cloud climatology based on ATSR-2
GRAPE intends to produce a new cloud database which will include the
following parameters:
Cloud Optical Depth
Cloud Phase
Cloud Particle Size
Cloud Top Pressure
Cloud Fraction
Cloud Water Path
along with associated error measurements (enabling the use of this
data in some form of data assimilation at a later date).
The database will be based on retrievals carried out using all
the available ATSR-2 data (from 1995 to present), and will be produced
using a method developed for Meteosat Second Generation SEVIRI measurements
and tested on ATSR-2 data (hereafter, the RAL algorithm). The RAL algorithm
is an application of Optimal Estimation (OE) to the cloud retrieval
problem: all available measurements and all available a priori information
are combined using fast radiative transfer and known error characteristics
to find the most probable values of all cloud parameters simultaneously.
The radiative transfer includes cloud, atmosphere and surface effects and
(currently) assumes a plane-parallel single layer cloud (the 'model cloud').
Application of OE is still a somewhat novel approach to cloud parameter
estimation. Traditional methods rely typically on sub-sets of measurements
used to independently derive sub-sets of parameters; for example,
visible (VIS) and near Infra-Red (NIR) measurements can be used to estimate
visible optical depth and drop size while an independent IR split window
can be used to estimate infrared optical depth and cloud temperature.
The OE method presents several advantages:
It is theoretically sound, based on radiative transfer calculations and
modelled errors in the input information.
It enables simultaneous use of all measurements, so that maximum information
is extracted from the data set. (E.g. optical depth is normally best estimated
from VIS channels as IR channels saturate very quickly, but in the extreme
case of thin cloud over land the IR information becomes more useful than
the VIS information which in compromised by the reflectance effects of
the underlying surface. The OE method ``knows'' this and weights measurements
appropriately.)
It can use a priori information easily and accounts correctly for errors
in this information.
Why produce a new cloud climatology?
Despite the apparent proliferation of cloud data, the information about
cloud properties is often limited to frequency information and optical
thickness along with environmental data (e.g. cloud top temperature and
pressure). While there is relatively good agreement in some cases between
sensors for gross measures of cloudiness (e.g. seasonal and zonal means)
there is still considerable disagreement in detail.
ATSR-2 data are available from 1995 until the present day, and the ATSR-2
instrument is exceptionally well calibrated. The RAL algorithm has been
validated in studies for Eumetsat and will be made effective enough for
the analysis of the entire ATSR-2 dataset under the auspices of the Cloudmap2
project. The 6+years of data will provide a comparable length dataset to
existing climatologies, and in conjunction with future data from the AATSR
instrument (due to fly on ENVISAT), a very long duration dataset is
possible. Analysis of current and future missions (e.g. MODISand MERIS)
may eventually provide good cloud parameters, but they will not be able
to measure in the 1990s!
One microphysical parameter of great interest to the global warming
community is the effective radius, re, of the droplets
in clouds. re can be derived from the cloud liquid
water (CLW) itself a prognostic variable in modern parameterisations.
The importance of re is that it can be directly employed
to couple the cloud parameterisation to the radiation, either via the cloud
optical depth alone or by the cloud optical depth in conjunction with albedo
and asymmetry parameters, all of which can be parameterised by a functional
dependence on re. Our retrieval method provides an
estimate of re (primarily based on the sensitivity
of the 1.6 and 3.7 micrometre reflection to the size of the particles in
the cloud), and this will be available wherever the retrieval is successful
(i.e. near global coverage for the mission duration).
Particular points of scientific interest to be covered will be the interhemispheric
difference in cirrus, and the quality of model predictions of cloud effective-radius
in specific case studies. One of the major points of advantage of the ATSR-2
retrieval will be an accurate assessment of cloud phase, a
notoriously difficult problem for model parameterizations. While high-resolution
in-situ campaigns are the best way to derive such parameterisations, it
is important that they are validated globally to ensure as much as possible
their accuracy in a wide-range of synoptic settings.
A rigourous comparison with numerical simulations should eventually
be carried out in the context of trying to understand particular experiments,
or to understand the climate behaviour of some specific model versions.
| 
