-------------------------------------------------------------------- (c) Lazaros Oreopoulos 5/99 Last modified 10/99 -------------------------------------------------------------------- The cloud field in this example is a simple 1-d "step cloud" consisting of 32 columns (pixels) along the x-direction the first 16 of which have an optical depth of 2 while the remaining have an optical depth of 18. The size of the field is set to 0.5 km, so that all columns have a width of dx=0.5/32 km. The geometrical thickness of the cloud (along z-axis) is set to 0.25 km everywhere (flat cloud) and the extinction coefficient is assumed homogeneous. The cloud has infinite extent in the y-direction. Obviously, this is an academic/pedagogical example where the main interest lies in examining model behaviour around the region of optical depth transition. The cloud field can be obtained by downloading the file tau_field. A sample Fortran code that reads the cloud field follows: c----------------------------------------------------- c Read in step cloud optical depth field parameter (nx=32) real tau(nx) open (1, file='tau_field', status='old') do ix=1,nx c Read from low to high x-coordinates read (1, '(f7.3)') tau(ix) enddo close(1) c------------------------------------------------------ Description of experiments -------------------------- ASSUMPTIONS: 1) No atmospheric effects 2) Periodic boundary conditions (cloud field is repeated an infinite number of times along the x direction) 3) Henyey-Greenstein phase function (PF) with g=0.85 4) Black (zero reflectance) surface 5) Single scattering albedos w0=1 and w0=0.99 (with the same PF and extinction field). SOLAR GEOMETRY: 1) Sun at (SZA of) zero degrees 2) Sun at 60 degrees and 0 azimuth (Sun shining from low x coordinates) This makes a total of 2x2=4 experiments: Exp.# 1: SZA = 0, w0 = 1 2: SZA = 60, w0 = 1 3: SZA = 0, w0 = 0.99 4: SZA = 60, w0 = 0.99 Required output --------------- 1) Mean and higher order moments of: albedo (R) transmittance (T) absorptance (wherever applicable) (A) net horizontal flux (H) defined as 1-R-T-A nadir reflectivity (Iu) defined as pi*Nu/(F*mu0) where Nu is the upward exiting radiance at 0 degrees, mu0 is the cosine of the solar zenith angle, and F=1 is the incident solar flux oblique reflectivities (Io) pi*No/(F*mu0) where No is the upward exiting radiance at 60 degrees zenith and 0 and 180 degrees azimuth angles (i.e. two oblique reflectivities are requested) zenith transmissivity (Id) defined as pi*Nd/(F*mu0) where Nd is the downward exiting radiance at 180 degrees !!! This quantity is requested only when SZA=60!!! Higher order moments are calculated as sum_i(x_i-x_mean)^k/N, k=2,6 and N=32, i=1,N. The accuracy of all output means should be at least 0.001 (feel free to perform calculations of higher accuracy, and whenever possible, please document your uncertainty values). All the above quantities (except A and H) are registered at cloud boundaries (cloud top for albedo and reflectivity and cloud base for transmittance and transmissivity). Directional tolerance (angular width of radiation "pencil") for reflectivity and transmissivity is left up to the user, but must be documented and provided when the results are submitted. You are asked to submit a file summarizing statistics of the output fields, a file summarizing pixel-level and errors to the mean, as well as files containing the fields themselves. Plan also for a file that will include CPU time for each experiment. In order to evaluate code performance as opposed to hardware performance we'll need to normalize with your machine's SPECfp_base95; you can look up your computer at http://www.specbench.org/osg/cpu95/results/cfp95.html, click at the HTML link for it and pull out the SPECfp_base95 number. Please submit the SPECfp_base95, and the CPU and "wall clock" (real elapsed) times for each experiment. For UNIX platforms CPU time can be obtained by using the time or timex commands; see your man pages for specific options. Formats and file name conventions can be found below. ************************** How to submit your results ************************** ------------------------Name convention------------------------ a) Fields of radiative quantities ================================= Create a separate file for each experiment and output field. Use the following name convention: I3RC_RQ_case_exp#.inst[n] where: i) "RQ" is the radiative quantity contained the file. RQ takes the following values: RQ=R (reflectance) RQ=T (transmittance) RQ=A (absorptance) RQ=H (net horizontal flux) RQ=Iu (nadir reflectivity) RQ=I601 (reflectivity at 60 view, 0 azimuth) RQ=I602 (reflectivity at 60 view, 180 azimuth) RQ=Id (zenith transmissivity) ii) "case" is the cloud field case For phase I the cloud field cases have been assigned the following numbers: case=1 step cloud case=2 MMCR-retrieved cloud case=3 Landsat-retrieved cloud Thus, for this field, put case=1 iii) "exp#" is the experiment number as listed above; valid number are 1-4 for this case iv) "inst" is the four-letter code that has been assigned to each institution participating in the experiment. The codes are listed in filename_TBD . v) "[n] is an index number following the institution whenever there are more than one participant or codes from the same institution. There is is no number for institutions with one participant and one code. We have already notified participants that have index numbers. ****Examples**** 1) The nadir reflectivity field of experiment 3, step cloud case submitted by a participant affiliated with institution AESC should have the following filename: I3RC_Iu_1_3.AESC 2) The reflectance field of experiment 2, step cloud case submitted by the 2nd participant of institution UMBC should be put in the following file: I3RC_R_1_2.UMBC2 b) Statistics of radiative quantities ===================================== Create a separate file for each experiment. This file will contain the statistics of all radiation quantities. Use the following convention: I3RC_fname_case_exp#.inst[n] where fname="stats" (without the quotes) and the rest is as before. c) Mean and pixel-level errors ========================================= Create a separate file for each experiment. This file will contain the mean pixel-level error and the error to the mean. Use the following convention: I3RC_fname_case_exp#.inst[n] where fname="errors" (without the quotes) and the rest is as before. d) Computing performance ======================== Create a single file for the case. This file will contain CPU time information for each experiment. Use the following convention: I3RC_fname_case_inst[n] where fname="CPER" (uppercase, without the quotes) and the rest is as before. ------------------------Output format------------------------- 1) For each field create a file using output format as the one produced by the following Fortran code --------------------------------------------------------------- parameter (nx=32) real R(nx) open (11, file='I3RC_R_1_1_UCOL1', status='unknown') do ix=1,nx write (11, '(f10.4)') R(ix) enddo