Documentation

This section describes how some of the input data sets are treated during the placement of the data into the CAGEX grid system. This section will probably be continually under construction, as changes are made and other data is incorporated. A list of subjects is below, followed by descriptions of how each parameter or set of parameters fits into the CAGEX system. Some of this stuff is repeated in the DATA DESCRIPTION page. It is hoped that the user of CAGEX will be able to glean enough information from this page so they will be able to reproduce what we did in their own radiative transfer codes. If you still have unanswered questions, send us an e-mail request for information, and we'll update the page and respond to you as needed.

SOUNDING INPUTS

Pressures

See the Vertical Resolution section of the GRID SYSTEM page for a detailed look at the selection of pressures used in CAGEX.

Temperatures

Temperature processing was split up into two groups: upper and lower levels:

Upper Level Temperatures

Upper level temperatures were obtained from NMC, and are a TOVS product. Vertical resolution is 9 levels, .4, 1, 2, 5, 10, 30, 50, and 70 mb. These are input as 65 x 65 degree polar stereographic grid, which was interpolated into a 145 x 37 degree equal-angle grid using a canned routine. This grid was then interpolated spatially into the two CAGEX grids described above, using temperature as the interpolated parameter. The 20 mb level, the only level to be interpolated here, was interpolated using T with respect to (hereafter wrt) the natural log of pressure. Following the vertical interpolation, the temperatures was interpolated temporally (from once/day to the required 18 half-hour timesteps).

Lower Level Temperatures

Lower level temperatures were obtained from NWS sounding data, which was interpolated into a 14 x 10 1 degree grid using a Barnes interpolation scheme, part of the McIdas package. The vertical resolution of this data is 12 levels: surface, 925, 850, 700, 500, 400, 300, 250, 200, 150, 100 mb, and the tropopause. These data are similar to the dataset used in the determination of cloud properties. The temporal resolution of these input temperatures is only twice-daily! The tropopause and surface values are inserted into their proper locations within the standard level soundings. A pressure profile is created with 124 levels, 100 of these below 700 mb. The temperatures are interpolated with respect to these (logs of) pressures yielding 124 temperatures. Next, the twice-daily temperatures are interpolated (using T now) into the 18 timesteps required. Following this is a horizontal interpolation which places the 1 degree data into the CAGEX grids (using T). Finally, the desired levels below 700 mb are interpolated using ln(T) with respect to P, from the 100 levels previously created.

The reason the vertical interpolation is split up is this: I needed the same number of levels in each grid for the horizontal interpolations. I couldn't use the original data for this because of problems which would arise with the surface and tropopause values, so I made everything with the same number of levels, horizontally interpolated, and then vertically interpolated the lowest 100 levels to yield the final temperature profile, which does not contain the same number of levels in each grid.

The twice-daily nature of the input temperatures means that the diurnal cycle of temperature increase near the surface is largely left out! This causes problems in the calculation of downwelling longwave fluxes. Calculations of DLF using the MAPS sounding product, show a significant improvement over those calculated with the core soundings.

Humidities

Humidity processing was also split up into two groups: upper and lower levels:

Upper Level Humidities

Inputs were McClatchey profiles for midlatitude summer and winter. These humidities were converted to ln(specific humidity) [ln(SH)] and were interpolated with respect to pressure into the desired upper levels (.4 - 300 hPa). These two profiles were then averaged, and distributed spatially and temporally into the CAGEX grid system. Hence all upper-level humidities are the same.

Lower Level Humidities

Lower-level humidities were obtained in the same manner as lower-level temperatures. However, the vertical extent of the humidities extend to only 300 mb, and do not include a tropopause value. Inital data has humidities in the form of dewpoints. These humidities were interpolated in a manner similar to temperature interpolation. The natural log of specific humidity was interpolated with respect to pressure in the vertical, and specific humidity in the horizontal and temporal grids.

Ozone

Ozone is an SBUV2 product and was obtained from NMC. The ozone comes in two forms: a total column ozone in Dobson units (DU) and mass mixing ratios for 6 upper-atmosphere levels: .4, 1, 2, 5, 10, and 30 mb. The mixing ratios were converted to DU, and these were subtracted from the column values. The subtraction yields the lower-level ozone amount, which was distributed vertically into the desired profile, weighted using the McClatchey ozone standard atmosphere. All interpolations were done using mass mixing ratio with respect to pressure. Spatially, ozone was interpolated from a 65 x 65 degree polar stereographic grid into the CAGEX grids. Temporal interpolation went from once-daily to the 18 timesteps.

The MAPS Soundings

The MAPS temperatures and humidities and the core temperatures and humidities were treated in a similar manner, with a one exception. Since the MAPS soundings do not include surface quantities, the temperatures and humidities are extracted from the ARM SMOS data set. When the MAPS soundings are included, they substitute for the lower level temperatures and humidities only. The upper levels remain the same. The only ozone difference between the two sets are the pressure levels on which the ozone is interpolated.

OTHER INPUTS

Solar Constant

The solar constant used by CAGEX is 1365.0 W/m**2, adjusted for the day of the year. This relationship is given by: solar constant= 1365.0 * r, where
r= 1 + a0 * sin ( ( 2 * pi/365.25 ) * ( nd - 365.25/4 ) ), where
a0 = 0.0167381 and nd = The number of days since January 4, 1958.

Solar Zenith Angle

Procedure for the calculation of solar zenith angles follows Iqbal (1983). CAGEX zenith angles were calculated by averaging the cosines of 31 zenith angles, calculated for each minute in the CAGEX 30-minute timestep. Temporally, the 31 zenith angles are centered on the CAGEX time. For example, the zenith angle for CAGEX timestep 1 (1409 GMT) was calculated using the 31 zenith angles from 1354 GMT to 1424 GMT.

Surface Albedo/Spectral Reflectance

The Fu & Liou radiation transfer code utilizes six spectral bands in it's parameterization of the short wave and near infra-red spectrum between 0.2 and 4.0 micrometers. They are: 0.2-0.7, 0.7-1.3, 1.3-1.9, 1.9-2.5, 2.5-3.5, 3.5-4.0 microns. In order to determine a set of spectral reflectivities to place in these bands it is necessary to find spectral measurements from various sources and try to match their results with these intervals. Unfortunately there are very few if any data that can be found that match these intervals one to one.

The method chosen was to use reflectances measured in similar bands and either interpolate or extrapolate these values to the Fu & Liou intervals. The weights used in the interpolation are determined by the low resolution radiation transfer model MODTRAN-3 calculated at the relatively high resolution of 50 inverse centimeters. Essentially, the incoming shortwave flux at the surface is integrated within the spectral bands of the known reflectances. Then, if these bands correspond to one of the Fu & Liou intervals, these values are used as weights for distributing the known reflectences across the Fu & Liou bands. For example, used in the CAGEX are the short grass/meadow spectral reflectences found in Briegleb et al. (1986) which are given for the spectral regions 0.2-0.5, 0.5-0.7, 0.7-0.85, and 0.85-4.0 microns. Hence, the first two of the Briegleb et al. intervals are used to interpolate to the region 0.2-0.7 microns. The second two numbers are used to interpolate a value for the 0.7-1.3 micron interval and the 0.85-4.0 reflectence is simply inserted into the last four Fu & Liou intervals.

A weighting factor was calculated by dividing the Briegleb et al. spectral reflectance by these spectral reflectances integrated over the Fu-Liou spectral bands.

Broadband surface albedos were calculated from ARM upward and downward-looking pyranometer data. Where missing, the monthly average of these surface albedos (for each temporal period) was substituted. The CAGEX spectral reflectance was determined by multiplying this broadband albedo by the Briegleb weighting factor for each spectral band.

Surface Emissivity

Surface emissivities are set equal to 1 for each of the 12 Fu-Liou bands. While there is some evidence to suggest we should decrease our emissivities somewhat, it is not done in release 1.

Skin Temperature

In the "core" dataset, the skin temperature is set equal to the sounding surface temperature. However, we believe a more accurate skin temperature can be found by utilizing tha ARM downward-looking pyrgeometer and applying the Stefan-Boltzmann law, assuming a surface emissivity of 1.0. Although this will result in a skin temperature which is too low for a non-black surface, this modification uses measured data which more accurately describes the change in temperature throughout the day. Since the core profiles are based on 12-hour soundings, there is not enough information to accurately describe the diurnal cycle. The CAGEX datasets includes some fluxes calculated using a pyrgeometer-based skin temperature.

Cloud Placement

The cloud products are already properly situated spatially and temporally. The placement of clouds vertically into the CAGEX grid system is done as part of the radiative transfer pre-proccesing. The placement method is outlined here so the CAGEX user can reproduce our method if the desire.

Cloud height parameters are given in kilometers above mean sea level (MSL). The elevation of the ARM Central Facility is approximately 315 meters above MSL, according to the ARM data sets. For the CAGEX grid, an elevation map is used to place cloud heights relative to the surface. It should be noted here that a discrepancy exists between the central gridbox elevation in this elevation map and the 315 meters stated by ARM. The elevation difference is only 23 meters however, and the ARM value is trusted. Since the elevation map (a product of interpolation) is used to represent the entire gridbox rather than the ARM Central Facility only, we use the elevation map value for CAGEX.

First, the tropopause is found by searching the profile for the lowest temperature. Working down from the tropopause, the top of the cloud is placed in the first layer whose midpoint temperature is greater than that of the cloud top. Next, the cloud bottom is found by applying the same method to the cloud bottom temperature. If an inversion exists, the cloud bottom is placed in the layer above the inversion. Once the cloud is placed, the liquid water/ice water content is distributed throughout each layer containing clouds by weighting the water content based upon the thickness of the layer.

Aerosols

This section describes the creation of the CAGEX version 1.1 release aerosol database. The database consists of three parameters: optical depth, single-scattering albedo, and asymmetry factor. All three parameters are distributed spatially (in the CAGEX 3x3 .3 degree grid), temporally (26 days of April 1994, 18 half-hourly time periods), vertically (CAGEX vertical profile, 46 or 47 layers), and spectrally (18 Fu-Liou spectral bands). Spatial distribution was accomplished by mapping, although vertical profiles will be different spatially due to different relative humidities. Temporal distribution was done by linearly interpolating missing column optical depths. Visual inspection of the input column optical depths determined the feasibility of temporal interpolation.

I. INPUT PARAMETERS

Input data for this exercise is from three sources:

1) Aerosol column optical depths, for 5 spectral bands (.412, .498, .606, .663, and .856 microns) were measured at the ARM SGP CART site, during clear-sky days in April 1994. This data comes to us courtesy of Joe Michalsky, and were measured using a multi-filter rotating shadowband radiometer (MFRSR).

2) Aerosol extinction coefficients, single-scattering albedos, and asymmetry parameters, as functions of wavelength and relative humidity, were taken from tables contained in d'Almeida et al. (1991).

3) An expression for describing the distribution of scattering coefficients as a function of height was obtained from Spinhirne (1993).

II. DESCRIPTION OF THE PROCESS

1) Spinhirne's expression for vertical scattering distribution was used to calculate scattering coefficients from 0 to 17000 meters in 1 m increments. The high resolution is necessary because these coefficients must be averaged in the CAGEX layers to produce a layer scattering coefficient (the distribution is exponential, so simple linear interpolation is not accurate)

2) d'Almeida tables were placed into the Fu-Liou shortwave spectral domain by energy-weighting each d'Almeida value by the energy contained in that wavelength, as determined by MODTRAN-3. For longwave, the values were simply averaged into the Fu-Liou spectrum.

3) The Michalsky aerosol column optical depths (MCtaus) were placed into the first two Fu-Liou shortwave bands, using energy weighting for band 1, and mapping for band 2, since there was only one available MCtau for band 2 and four for band 1.

4) Vertical distributions of the d'Almeida tables were calculated by interpolating their values based on the relative humidities of each CAGEX layer.

5) Non-normalized extinction coefficients were calculated for the first two Fu-Liou bands by dividing the Spinhirne layer-averaged scattering coefficients by the d'Almeida RH dependent single-scattering albedos.

6) Non-normalized layer optical depths were calculated for the first two Fu-Liou bands by multiplying the extinction coefficients calculated in (5) by the layer thickness.

7) Layer optical depths were calculated by weighting the non-normalized layer optical depths by the Michalsky column optical depth/non-normalized column optical depth ratio, for the first two Fu-Liou bands.

8) Layer optical depths were calculated for the remaining Fu-Liou bands by multiplying the layer optical depths for Fu-Liou band 2 by the ratio of each bands extinction coefficient with band 2's extinction coefficient, both obtained from the d'Almeida tables.