Observational Requirements for Modeling of Global...
51
important. But this too is a potential focus for a global satellite climate monitoring effort, if
suitably precise remote measurements of aerosol and cloud properties can be made.
It is also true that existing field programs have not yet fully exploited their potential to contribute to climate research. In part this is due to the surprising lack of interaction among
climate modeling, satellite retrieval, and aircraft observation experts to date. This is a problem
of spatial scale, from the cloud microphysical scale to the satellite pixel scale to the GCM
grid scale, a range of many orders of magnitude. It is also a problem of temporal scale, from
the several hour duration of a typical aircraft flight to the monthly time scale on which most
GCM parameters are studied. To bridge the scale gaps, all parties must move in directions
they traditionally resist. Modelers must understand the behavior of their GCMs on synoptic
time scales. Data from field programs, which do not represent climatologies, can nonetheless
be compared by modelers to the frequency of occurrence of similar phenomena in the GCM,
and those instances that are relevant can be analyzed in more detail to see how well they
correspond to what occurs in the real world. Field program workers in turn need to consider
the domain-averaged characteristics of their observations and statistical relationships between
observed parameters, to link to the satellite and GCM scales. Availability of field experiment
data in useful, readable form is also a chronic problem causing these data to be underutilized
by modelers.
Satellite data sets must in turn provide enough information to permit a modeler to diagnose
the same variable in the same way in the model (and modelers in turn must more often "fly
the satellite" over their GCMs to create the most instructive comparisons possible). For some
satellite data, information may exist at a variety of spatial scales, but resolution is degraded
to limit the volume of the data set. A better approach would be to create several regional
data sets at full resolution for limited time periods (e.g., the canonical seasonal months) to
supplement global data at lower resolution. For example, Landsat images at sub-kilometer
resolution have proven valuable as a complement to ISCCP data at 5 km resolution (Wielicki
and Parker, 1992).
c) Climate change monitoring data
Ultimately, we will not know with certainty whether GCM predictions of climate change are
realistic until unambiguous evidence of long-term trends in a variety of climate parameters
can be observed. To date, there is a useful record of surface temperatures over much of the
globe spanning more than a century, and several regional trends of cloud cover have been
noted over a period of a few decades (cf. Henderson-Sellers, 1992). But many of the data one
would like to use to detect trends are not adequate for the purpose. For example, there is a
fairly long record of radiosonde specific humidity at various locations around the world, but
different nations use different instruments with different accuracies, and both instrument types
and reporting methods have changed over time, making deduction of trends a particularly
difficult undertaking (Elliott and Gaffen, 1991). Satellite data sets in principle can be used
for monitoring, but in practice, for many parameters there are gaps in the record, and for
others, calibration uncertainties make it impossible to detect a trend when one satellite replaces
another.
If climate monitoring is to be viable, one must first ask what parameters one wishes to monitor,
and whether it is possible to detect changes of the magnitude that we expect to occur over
the next few decades. To understand the sensitivity of climate to external perturbations,
one needs to know the dominant forcings and feedbacks. Greenhouse gas concentrations are
adequately measured in situ (except for ozone, whose vertical profile is important and could be
monitored with a sufficient density of ozonesondes), so the primary forcing monitoring need is
for tropospheric aerosol optical depths. The relevant feedbacks we need to monitor are changes
51
important. But this too is a potential focus for a global satellite climate monitoring effort, if
suitably precise remote measurements of aerosol and cloud properties can be made.
It is also true that existing field programs have not yet fully exploited their potential to contribute to climate research. In part this is due to the surprising lack of interaction among
climate modeling, satellite retrieval, and aircraft observation experts to date. This is a problem
of spatial scale, from the cloud microphysical scale to the satellite pixel scale to the GCM
grid scale, a range of many orders of magnitude. It is also a problem of temporal scale, from
the several hour duration of a typical aircraft flight to the monthly time scale on which most
GCM parameters are studied. To bridge the scale gaps, all parties must move in directions
they traditionally resist. Modelers must understand the behavior of their GCMs on synoptic
time scales. Data from field programs, which do not represent climatologies, can nonetheless
be compared by modelers to the frequency of occurrence of similar phenomena in the GCM,
and those instances that are relevant can be analyzed in more detail to see how well they
correspond to what occurs in the real world. Field program workers in turn need to consider
the domain-averaged characteristics of their observations and statistical relationships between
observed parameters, to link to the satellite and GCM scales. Availability of field experiment
data in useful, readable form is also a chronic problem causing these data to be underutilized
by modelers.
Satellite data sets must in turn provide enough information to permit a modeler to diagnose
the same variable in the same way in the model (and modelers in turn must more often "fly
the satellite" over their GCMs to create the most instructive comparisons possible). For some
satellite data, information may exist at a variety of spatial scales, but resolution is degraded
to limit the volume of the data set. A better approach would be to create several regional
data sets at full resolution for limited time periods (e.g., the canonical seasonal months) to
supplement global data at lower resolution. For example, Landsat images at sub-kilometer
resolution have proven valuable as a complement to ISCCP data at 5 km resolution (Wielicki
and Parker, 1992).
c) Climate change monitoring data
Ultimately, we will not know with certainty whether GCM predictions of climate change are
realistic until unambiguous evidence of long-term trends in a variety of climate parameters
can be observed. To date, there is a useful record of surface temperatures over much of the
globe spanning more than a century, and several regional trends of cloud cover have been
noted over a period of a few decades (cf. Henderson-Sellers, 1992). But many of the data one
would like to use to detect trends are not adequate for the purpose. For example, there is a
fairly long record of radiosonde specific humidity at various locations around the world, but
different nations use different instruments with different accuracies, and both instrument types
and reporting methods have changed over time, making deduction of trends a particularly
difficult undertaking (Elliott and Gaffen, 1991). Satellite data sets in principle can be used
for monitoring, but in practice, for many parameters there are gaps in the record, and for
others, calibration uncertainties make it impossible to detect a trend when one satellite replaces
another.
If climate monitoring is to be viable, one must first ask what parameters one wishes to monitor,
and whether it is possible to detect changes of the magnitude that we expect to occur over
the next few decades. To understand the sensitivity of climate to external perturbations,
one needs to know the dominant forcings and feedbacks. Greenhouse gas concentrations are
adequately measured in situ (except for ozone, whose vertical profile is important and could be
monitored with a sufficient density of ozonesondes), so the primary forcing monitoring need is
for tropospheric aerosol optical depths. The relevant feedbacks we need to monitor are changes