Satellite Cloud Measurements

International Satellite Cloud Climatology Project

The most commonly used cloud dataset in climate studies is the ISCCP (Rossow and Schiffer 1999). Originally intended to provide a cloud climatology and data for studying cloud processes at synoptic scales, ISCCP uses narrowband radiance data obtained from weather satellites to retrieve cloud fraction and other cloud properties in 280 km grid boxes around the globe every three hours. A radiative transfer model has been applied to these cloud data along with observed stratospheric aerosol and GCM-derived tropospheric aerosol to produce broadband SW and LW fluxes at the top of the atmosphere, the surface, and several levels in between (Zhang et al. 2004). It is interesting to note that interannual and decadal variations in the ISCCP flux dataset have some resemblance to those reported by ERBS (Figure 2.4), considering that they are independently derived datasets.

ISCCP faces much greater difficulty in maintaining homogeneous and stable data over time than do the ERB investigations since it makes use of a long series of weather satellites that were not designed to maintain onboard calibration or interface with other satellites. The ISCCP processing intercali-brates geostationary satellites (the main source of data) using a polar orbiter that flies below each of them, which itself must be calibrated over time. One problem involves calibrating successive polar orbiters so that artificial changes in global cloud properties do not appear at the transition between satellites, as was the case in the original ISCCP C-series dataset (Klein and Hartmann 1993). Since the surface of the Earth happens to be more radiatively stable than any current satellite-observing system, the processing of the ISCCP D-series

73 TO

14 7

14 7

1985 1987 1989 1991 1993 1995 1997 1999 Time (year)

Figure 2.4 Time series of longwave, shortwave, and net radiation anomalies averaged over 20°N to 20°S from the ERBS Nonscanner WFOV Edition3_Rev1 (red), ISCCP FD (blue), HIRS Pathfinder OLR (pink), and AVHRR Pathfinder ERB (green) data records. Anomalies are defined with respect to the 1985-1989 period (from Wong et al. 2006).

1985 1987 1989 1991 1993 1995 1997 1999 Time (year)

Figure 2.4 Time series of longwave, shortwave, and net radiation anomalies averaged over 20°N to 20°S from the ERBS Nonscanner WFOV Edition3_Rev1 (red), ISCCP FD (blue), HIRS Pathfinder OLR (pink), and AVHRR Pathfinder ERB (green) data records. Anomalies are defined with respect to the 1985-1989 period (from Wong et al. 2006).

dataset included the adjustment of satellite measurements such that global mean surface temperature and reflectance (including the effects of tropospheric aerosol) did not change over time (Brest et al. 1997). Although this method substantially reduced inhomogeneities in the ISCCP time series, it obviously precludes the use of the data to investigate changes in the global surface properties. Moreover, the method will introduce an artifact into the cloud data to the extent that the Earth's surface has changed in the global mean. Nonetheless, it is nominally possible to examine changes in global cloud properties, and ISCCP reports that a general decrease in global mean total cloud amount has occurred over the past couple of decades (Figure 2.5), due primarily to a decrease in low-level cloud amount at low latitudes.

Despite the multiple layers of calibration and adjustment in ISCCP, this trend and other variations in global mean cloud amount appear to be entirely spurious. The pattern of correlation between the global mean time series and the time series in each grid box resembles the circular geostationary satellite fi elds of view rather than any geophysical pattern, and the spatial pattern of local trends becomes strongly negative near the edges of the geostationary satellite fields of view. Evan et al. (2007) attribute this to a systematic change in the average satellite view angle over time. A geostationary satellite sees pixels near the outside boundary of its view area with a greater slant path than it sees pixels in the center of its view area (i.e., near-nadir). At visible wavelengths, thin low-level clouds appear optically thicker for greater slant path and thus are easier to detect by the threshold methods employed by ISCCP. Since only three geostationary satellites were available at the beginning of the ISCCP record and five geostationary satellites were available at the end, locations that were once seen at high satellite view angle now were seen at low satellite view angle. This means that very optically thin clouds that were once detected were no longer detected, and reported cloud amount consequently decreased.

The view angle artifact has less impact on the calculated ISCCP short- and longwave radiative fluxes than it does on cloud amount because the artificial reduction in cloud amount is compensated by a corresponding artif cial enhancement of cloud optical thickness. Moreover, the clouds that are no longer identified are those that are closest to the threshold of detection and thus least radiatively important.

Even when view angle artifacts associated with changes in the number and position of geostationary satellites are statistically removed, other apparently spurious variability remains. For example, coincident variations in cloud amount are seen across the entire view area of a geostationary satellite (Norris 2000b). There is no reason why cloud anomalies in opposite hemispheres and seasons (and often land vs. ocean) should be so closely correlated unless there were some unidentified artifacts in the satellite measurements or application of incorrect calibration (e.g., Norris and Wild 2007).

Year

Local correlation with global mean time series

Year

Local correlation with global mean time series

Local linear trend (%-amount per decade)

Figure 2.5 Top: time series of ISCCP global mean total cloud amount. Middle: correlation between the global mean time series and the time series at each grid box. Bottom: local linear trend in total cloud amount for each grid box.

Figure 2.5 Top: time series of ISCCP global mean total cloud amount. Middle: correlation between the global mean time series and the time series at each grid box. Bottom: local linear trend in total cloud amount for each grid box.

Additional Cloud Datasets

Another cloud dataset with a multidecadal record is the AVHRR pathfinder atmosphere (PATMOS) dataset (Jacobowitz et al. 2003). PATMOS cloud and radiation properties are derived from the AVHRR instrument on polar-orbiter weather satellites. Although they do not suffer from shifts in geostationary satellite view angle, as does ISCCP, the nominally sun-synchronous polar orbiters contributing to PATMOS experience substantial orbital drift during their tenure, causing the local equatorial crossing time to occur ever later in the day. This means that the diurnal cycle of cloud properties gets aliased into a long-term trend (i.e., if more clouds occur in the morning than in the afternoon, it will look like cloud amount has declined over time). PATMOS corrects for diurnal drift by statistically regressing out the long-term trend, which must be repeated for each successive satellite in the series. Removing trends in this manner limits the use of these data for long-term climate studies since any real cloud trend may get removed as well. A better method would be to observe the diurnal cycle from a geostationary satellite and then adjust the diurnal drift based on that. PATMOS time series also exhibit extremely large variations that appear to be related to transitions between satellites (Figure 2.4).

A third multidecadal cloud dataset is based on the high resolution infrared radiometer sounder (HIRS) on polar-orbiting weather satellites (Wylie et al. 2005). The HIRS dataset operates by comparing radiances from several infrared channels that view different layers of the atmosphere, and it is especially good at detecting high-level optically thin clouds. Like PATMOS, HIRS suffers from orbital drift and inhomogeneities at satellite transitions. The time series of tropical outgoing LW radiation derived from HIRS (Mehta and Susskind 1999), however, does appear to agree with the ISCCP flux dataset and ERBS time series (Figure 2.4).

0 0

Post a comment