Case 6: Broken Cumulus clouds observed by Terra over Brazil

Images of proposed test scene: Overview images, Terra observations, Terra data products

• Synopsis
• Detailed scene description
• Downloading scene parameters
• Description of experiments
• Output files
• Submission instructions
• Optional inverse experiments
• Future experiments

• Case 6 is based on a scene that was observed by the Terra satellite around 13:15 UTC on August 9, 2001.
• The scene covers a 68 km X 80 km area centered at 17.1°S and 42.1°W, in central Brazil.
• Stage 1 involves 1 km-resolution simulations of solar reflection at 0.67 µm wavelength, for the view directions of MISR AN, CA, and CF cameras (that is, for nadir and 60° viewing zenith angles).

• Input parameters are defined either at 1 km resolution or as constant values throughout the scene.
• Radiative transfer calculations should assume periodic boundary conditions.
• The solar zenith angle is 41° and the solar azimuth is 23° off the satellite track.


• Surface reflection is heterogeneous and anisotropic. Surface radiative properties are characterized at 1 km resolution using the easy-to-use Li-Sparse-Ross-Thick model. The fields of the three parameters required for this model were kindly provided to us by Yujie Wang and Alexei Lyapustin (both at UMBC GEST), based on MISR retrievals. The model is described, among others, in the manual of the SHARM code and in a paper by Lucht et al. (2000). A sample implementation of this model is available at the Boston University MODIS surface BRDF/Albedo web site MODIS user tools section, under the name "forward brdf model". (The code, written in C, is in the file "one_ref.c" of the tar-package.)


• Atmospheric absorption is assumed negligible at 0.67 µm.
• Rayleigh optical thickness is 0.0435; the representation of vertical profile of Rayleigh scattering is left to each participant. An article by Bodhaine et al. (1999) provides further details on Rayleigh optical thickness calculations.
• Aerosol optical thickness throughout the scene is assumed to be 0.05, the median value of MODIS retrievals in the scene. The aerosol is assumed to be uniformly distributed in the layer below the cloud base.
• Aerosol scattering phase function is provided in a text file, based on nearby Aeronet retrievals.


• Cloud base is assumed to be at 1 km altitude above ground.
• Cloud top altitude (above ground) is provided at 1 km resolution. Cloud top altitudes were estimated using MODIS 11 µm brightness temperature values.
• Cloud optical thickness is provided at 1 km resolution, using the operational Collection 4 MODIS cloud product.
• Cloud extinction coefficient is assumed to be vertically constant for each pixel.
• For Stage 1 simulations at 0.67 µm, cloud droplet effective radius is assumed to be constant. Droplet scattering phase function is provided in a text file, based on Mie calculations for a lognormal drop size distribution with 10 µm effective radius and with a standard deviation that is 0.3 times the mean radius.

• The 56 KB size combined file containing all required scene parameters can be downloaded here.
• This file contains 1km-resolution fields of all three surface parameters for the Li-Sparse-Ross-Thick model, cloud optical thickness, and cloud top altitude (above ground, in km), as well as aerosol and cloud scattering phase functions.
• The 1 km resolution fields are provided in ASCII text files that contain 68 rows and 80 columns.
• Scattering phase function text files contain two columns: one for the scattering angles (°) and the other for the corresponding phase function values.

• Experiment 1: Simulation of 0.67 µm nadir reflectance values at 1 km resolution.
• Experiment 2: Simulation of 0.67 µm reflectance values at 1 km resolution. Viewing zenith angle is 60°. Viewing azimuth is 0°, parallel to the columns in the scene. This simulation provides back scatter reflectances 23° off the solar plane.
• Experiment 3: Simulation of 0.67 µm reflectance values at 1 km resolution. Viewing zenith angle is 60°. Viewing azimuth is 180°, parallel to the columns in the scene. This simulation provides forward scatter reflectances 23° off the solar plane.

• Participants are requested to submit two files for each experiment: One containing the simulated reflectance values, and another containing the estimated absolute (not relative) uncertainty of these reflectances. Reflectance is defined as the radiance multiplied by pi and divided by the cosine of solar zenith angle and by the solar irradiance.
• All file names should start with "I3RC_Case6_", then include the experiment ID (e.g., "Exp2_"), a 5-digit model ID (including 4 digits for institution), and either "Refl.txt" or "UncR.txt" to identify reflectivity and uncertainty files. Using this convention, a sample file name is "I3RC_Case6_Exp3_UMBC1_Refl.txt".
• Submitted files should contain 68 rows and 80 columns of floating point numbers in ASCII text format.
• Results for oblique views (Experiments 2 and 3) should be registered at ground level.

• Files can be submitted via ftp to climate1.gsfc.nasa.gov. After anonymous login and changing to the directory called uploads, results can be deposited using the put or mput command. Please note that for security reasons the list of files in this directory cannot be displayed.
• Participants are requested to send an email to Tamás Várnai upon the submission of their results. The email should list the submitted files, include a brief description of the used model, indicate the interpretation of estimated uncertainties (e.g., whether they are at the widely used 68% confidence level), mention any other relevant information (e.g., number of photons used in Monte Carlo simulations), and provide the participant's contact information.
• Stage 1 submissions are kindly requested by August 15, 2006.

• In addition to the deterministic experiments listed above (in which scene properties are fully specified using operational Terra products), participants are also encouraged to perform optional experiments with the goal of creating more accurate representations of the test scene. In these experiments, participants can modify any surface or atmospheric properties so that the results of 3D simulations match Terra observations more closely for the modified than for the original scene.
• Participants are requested to submit the modified scene properties, the reflectance fields simulated using the improved scene, and the estimated uncertainties of simulated reflectances.
• A set of Terra observations that participants may want to use can be downloaded here. The 220 KB size combined file includes 1 km-resolution MODIS reflectances at Bands 1, 2, 6, 7, MODIS 2.1 µm effective cloud particle radius values, the difference between effective radii at 2.1 µm and 1.6 µm, MODIS latitude and longitude values, and co-located MISR reflectances at Band 3 for all 9 cameras.
• MISR images were co-located with MODIS data using a stereo matching algorithm that provided optimal matches at cloud altitude. As a result, surface features appear at different pixel locations for different MISR view angles.
• Because of the large size of ASTER images, ASTER Band 3 (0.8 µm) nadir reflectances and Band 14 (11 µm) brightness temperatures are provided in separate binary files. (ASTER Band 3 appears more useful than Band 2 because the Band 2 image is saturated over a large portion of clouds.) The 17.4 MB size Band 3 reflectance file contains 4200 rows and 4980 columns of 4-byte floating point numbers at 15 m resolution; the 2.2 MB size Band 14 brightness temperature file contains 700 rows and 830 columns of 4-byte floating point numbers at 90 m resolution. The ASTER files are not co-registered with the MODIS images; participants may perform such co-registration using the provided geolocation parameters.
• We will gladly provide additional Terra observations (e.g., additional wavelelengths, CERES radiances) to interested participants as needed. For additional observations, please contact Tamás Várnai.
• There is no submission deadline for optional experiments; submissions are welcome anytime.

• Future intercomparison experiments may include simulations for additional wavelengths (with absorption and/or emission), more complex atmospheric and/or cloud structures, and/or different sun-view geometries. In addition, future experiments may be designed to examine the causes of any significant model-to-model differences found in Stage 1. The list of new experiments will be finalized after the analysis of Stage 1 results.