Lidar Simulation Model: GADS Cloud Model

GADS Cloud Fields

The cloud information (percentage and location) for a T213-based GADS is obtained directly from the T213 nature run.

For a T106-based GADS, the AGM cloud model (Emmitt and Wood, 1995) is based on the Slingo cloud parameterization scheme (Slingo, 1987). The Slingo approach provides distinctions between high and mid-tropospheric stratiform clouds, convective clouds with and without anvil cirrus, low level clouds driven by weak vertical motion or inversion capped moist boundary layers.

Convective Cloud

The convective cloud is inferred from 3 hour integrated precipitable water from the T106 meteorological profiles. A critical threshold value of 0.14 mm/day must be met for a convective cloud to be present. If the threshold is met, the convective cloud base amount is empirically derived. A limit of 80% is set for the convective cloud amount.

cc = 0.2473 + 0.1258 * cppt

where

cc - the base layer convective cloud amount (%)

cppt - the integrated precipitable water (mm/day).

The top of the convective cloud layer is a function of the base layer convective cloud amount and the tropopause height. The cloud top is limited at the tropopause.

cc_top = (cc + 0.2) * TH

where

cc_top - the convective cloud top (km)

cc - the base layer convective cloud amount (%)

TH - the tropopause height (km).

The convective cloud coverage between cloud base and cloud top is defined as 25% of the base layer convective cloud amount.

If the top of the convective cloud is above the 400 mb layer and the integrated precipitable water more than 3.4 mm/day, then an anvil is defined. All anvil clouds are considered to be thick cirrus layers. The anvil cloud amount (%) is defined as

cc_anv = 2 (cc - 0.3)

High Non-Convective Clouds

All non-convective high clouds are derived as a function of relative humidity from the ECMWF T106 meteorological profile. A high layer cloud is only derived when the tropopause height is higher than the 400 mb layer. The AGM evaluates the T106 relative humidity profile to find the highest value which is used to compute a relative humidity threshold.

RH_thr = (RH_hgh - 0.8)/(1.0 - 0.8)

where

RH_thr - the high relative humidity threshold (%)

RH_hgh - the high relative humidity (%).

If the relative humidity threshold is greater than 0%, then high level cloud cover (%) is estimated as follows

HC = (Rh_thr)²

The cloud is considered to be thin cirrus. See the Cloud Optical Property section for discussion on how this percent high cloud is used to provide variability in cloud optical depths.

Middle Non-Convective Clouds

All non-convective middle clouds are derived as a function of relative humidity from the ECMWF T106 meteorological profile. If there was a convective cloud or a high layer cloud, the AGM drys out the T106 relative humidity profile.

RH = RH * (1.0 - CC)

where

CC - either the convective or high layer cloud cover (%).

Like the high cloud algorithm, the AGM finds the highest relative humidity in the profile and computes the relative humidity threshold. If the relative humidity threshold is greater than 0%, then middle layer cloud cover (%) is estimated as follows

Low Non-Convective Clouds

The estimate of low level non-convective clouds is based upon two parameters: vertical velocity and the potential temperature profile. From vertical velocity, the AGM finds the layer with the largest negative vertical velocity and computes the critical relative humidity for the layer. If the vertical velocity is less than 0.1, the cloud cover is defined as

LC = (Rh_thr)²

else

LC = (RH_thr)² * (-10 * VV)

where

VV (m/s) - the vertical velocity for the layer.

The potential temperature is used only if there was no cloud cover from the vertical velocity method. Potential temperature lapse rates are computed for every sublayer between 1000 mb and 700 mb as follows

Q_lr= -6.67 * DT/ DP

where

Q_lr - the potential temperature lapse rate

DT - the change in temperature over the layer

DP - the change in pressure over the layer.

If the lapse rate is greater than zero, then the model tests upon relative humidity to compute the cloud cover. If the relative humidity is less than 60%, there is no cloud cover. If the relative humidity is greater than 60% and lower than the threshold relative humidity, then the cloud cover is

LC = Q_lr * (1 - (RH_thr - RH)/(1 - Rh_thr))

else

LC = Q_lr

Global Statistics and Empirical Adjustments

We expect that clouds (including subvisible cirrus) will be in the field-of-view (FOV) of a space-based lidar 70-80% of the time. This estimate is based upon the recently reported analysis of two years of HIRS data (Menzel et al., 1992), the cirrus climatology derived from SAGE data by Woodbury and McCormick (1986) and the Nimbus-7 global cloud climatology ( Stowe et al., 1989). Much of this cloud coverage is high cloud (above 400-500 mb) and is semi-transparent (~ 30-40%). Very thin or subvisual cirrus (ç <.07) is probably not detected by HIRS or Nimbus-7 but may be occasionally represented in the SAGE observations. Thus, we conclude that the occurrence of very thin cirrus is clearly underestimated in current climatologies.

Of particular interest to a space-based lidar program are the semi-transparent and optically thin clouds since they provide strong returns without full extinction (Emmitt and Wood, 1988). When one considers that the statistics given above are, in most cases, exclusive - i.e., they do not provide a good representation of coincident clouds at different altitudes, it is very likely that there are many occasions when there are multi-layers of thin clouds underlaid by opaque clouds.

The ECMWF T106 Nature Run provides accumulated convective precipitation (based upon Kuo's scheme) and the total cloud coverage. Unfortunately, layer-by-layer information on cloud cover is not provided. We have developed a means to use the basic concepts within the Slingo scheme to reproduce the layer-by-layer cloud statistics that we need for space-based lidar simulations.

We examined the distribution of clouds (over a 1 deg x 1 deg area) based upon the ECMWF total cloud coverage as a function of latitude. While the total coverage is quite reasonable and compares well with the Nimbus-7 statistics, the amount of midlevel cloud forecast for the tropics is considerably less than the 30-40% reported using the satellite data. Conversations with the National Meteorological Center's (NMC) personnel (Pan and Baker) suggest that this is an ongoing point for discussion and study with the modeling community suggesting that the interpretation of midlevel clouds in satellite imagery may be faulty.

Cloud Optical Properties

All opaque cloud backscatter values are preset in the AGM to be 1x10-06 m-1 sr-1 for 9 um and 1.86x10-5 for 2 um. We believe this value is properly conservative, since recent mid-layer cloud backscatter, measured with a lidar in the Antarctic, range from 1x10-06 to 1x10-04 m-1 sr-1 (Del Guasta et al., 1993).

Opaque cloud absorption coefficients are based upon Stephens (1979) as a function of liquid water content and cloud type as shown below.

For cirrus cloud layers, cirrus backscatter is based on Northeastern University's cirrus model and is a function of cirrus cloud temperature as shown below.

The cirrus cloud attenuation model is a modified version of an analytical AFGL cirrus algorithm found in the AFGL models Kneizys et al., 1996 ;Gallery et al., 1983), where

t = e^{-0.14 * L²}&

t - the cirrus transmittance

L - the cirrus cloud thickness.

Since the AGM is restricted to the coarse vertical resolution of the ECMWF Nature Run, SWA uses the cirrus cloud percentage as a surrogate for cloud optical thickness.

The major assumption is that while the Slingo model derives a percent cirrus cloud coverage (i.e., 30%) from an average relative humidity within a grid volume, it is just as reasonable to interpret a thickness tendency from the same fields. Instead of using the percent coverage as literally meaning that 30% of the grid has cirrus cloud and 70% is totally cloud-free, the AGM assumes that the whole grid area is covered by a cirrus cloud that has an optical thickness that scales to the percent coverage. The cirrus cloud attenuations is defined as

ci_att = 10^a * (CLD_% * 10)²

where

ci_att - the cirrus cloud attenuation

a - the AFGL cirrus attenuation coefficient for a 1 km thick layer

CLD_% - the cirrus cloud percentage cover.