# Minutes of the third International Radiative Transfer Modeling Workshop

============================================================================ ========== =========== ========== IRTMW01- 10/11 October, 2001 =========== ========== =========== ============================================================================ Chairmen: --------- Stefan Buehler Patrick Eriksson Joachim Urban Carmen Verdes ============================================================================ Participants ------------ Chalmers: Patrick Eriksson Carlos Jimenez Donal Murtagh IUP: Stefan Buehler Carmen Verdes Shreerekha T.R. Claudia Emde Viju O. John Nikolay Koulev Oliver Lemke Thomas Kuhn Jungang Miao DLR: Franz Schreier UBordeaux: Jo Urban KF Karlsruhe: Gerhard Kopp Fujitu FIP Corporation: Chikako Takahashi CRL: Yasuko Kasai Satoshi Ochiai NASDA/EORC: Sho Tsujimaru Uni Bonn: Ariane Thiele Kokugakuin Univ.: Kazuo Shibasaki ----------------------------------------------------------------------------- participating models: ARTS IUP/Chamlers 1A 1B 3down 3limb 3up CRL Satoshi Ochiai 1A 1B 3down 3limb 3up Moliere (MOLIERE-4.86) Jo Urban, UBordeaux 1A 1B 3down 3limb 3up DLR Franz Schreier 1A 1B 3down 3limb 3up SMILES Simulator "SS","FIP" Yasuko, Chikako 1A 1B 3down 3limb Sho's model "EORC" Sho 1A 1B 3down Karlsruhe "FZK" Gerhard Kopp 1A 1B 3down 3up ----------------------------------------------------------------------------- SEE THE VERY END FOR SUMMARY OF FINDINGS AND THINGS TO DO! %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% CASE 1A Session Species absorption %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% GENERAL ITEMS: ============== - Voigt line shape calculation should be compared for a specific case of one line. Test case will be defined by IUP. - partition functions for isotopes should be checked - it seems that for O2 the isotope has as strong lines as the main isotope - it has to be checked. In HITRAN96 is an error for the isotopic lines of O2. In HITRAN00 this error is removed for the sake of introducing other errors (H2O isotopes?). See comment in ARTS from Axel. - weak NO lines (150.19 and 150.57) should be omitted. - all models should include the (f/f_o)^2 in the lineshape - pressure shift should be included (only non-zero for HCl). Overview of absorption comparison ------------------------------------ 1) H2O22GHz ARTS-Moliere Moliere is higher in the wings CLO501GHz ARTS-Moliere good agreement O3625GHz ARTS-Moliere good agreement 2) H2O22GHz ARTS-CRL O3625GHz ARTS-CRL good agreement CLO501GHz ARTS-CRL good agreement 3) H2O22GHz ARTS-DLR problem >5% difference O3625GHz ARTS-DLR CLO501GHz ARTS-DLR 4) H2O22GHz ARTS-SS 10% difference in the wing O3625GHz ARTS-SS 10-25% diff. CLO501GHz ARTS-SS 5% diff. in the lower band side 5) H2O22GHz ARTS-FZK very large difference at the band ends!! O3625GHz ARTS-FZK 2% diff. CLO501GHz ARTS-FZK 15% diff. in the lower band side Detailed discussion of absorption comparison ARTS-Moliere ---------------------------------------------------------- H2O-22GHz -10e-3 offset in both band wings at higher altitudes PE: Voigt implementation maybe different CLO501GHz PE: Voigt implementation maybe different indicated by the features in the differences HNO3544GHz high altitudes diff. in the lower band end HCL626GHz asymmetric differences around the line center 1-2% level PE: center frequency shifted? Window150GHz At 150.19 and 150.57 there are lines missing in the ARTS "window" calculations CO576GHz strange feature around the line center !!LOOK into the pressure shift calculation in ARTS!! CLO major difference in the line center Detailed discussion of absorption comparison ARTS-CRL ---------------------------------------------------------- H2O22GHz diff not pressure depended CRL uses here Lorentz line shape ARTS VoigtKunz-quadratic without mirror line Window89GHz 30% diff. with some structure in the diff. O3110GHz line shape diff. seen, (f/f_o)^2 not in CRL model O2118GHz diff. in the band wings are a straight line also diff. in the line center Window150GHz lines not missing like in Moliere (-> Moliere added som lines?) HCN354GHz not special diff. CLO501GHz in the band wings are the largest diff. N2O502GHz at highest alt. 20% diff. in the line wing HNO3544GHz diff. in the lower band wing CO576GHz 8% diff at lower altitudes O3625GHz 0.5% diff. in the lower band wing HCL635GHz see pressure shift entry in the HITRAN catalog CLO650GHz 50% diff. in the band wings!! investigation of the Voigt implementation of CRL because this calculation is especially implemented by Satoshi himself, so one can compare this implementation with the standard calculations (e. g. Kuntz) Detailed discussion of absorption comparison ARTS-DLR ------------------------------------------------------ DLR is calculating on a wavenumber grid not on a frequency grid. FS: I guess this should not make any difference in the spectra sent to Bremen: indeed the cross sections have been calculated on a wavenumber grid, but have been saved on file as cross sections vs frequency, i.e. the x-axis has been scaled by 29.99792458 (actually the hitran line data are in wavenumbers, too, hence everybody had to convert to frequencies at some stage of the exercise!) General discrepancy: FS: suggests that the order of the spectra is simply wrong so that in the comparison in fact two spectra of different (p,T) are compared. PE: pressure broadening different to ARTS, unit conversion problem? FS: will check this point Window89GHz in the line large diff. 25% O3625GHz in the line wing up to 8% Detailed discussion of absorption comparison ARTS-SS ----------------------------------------------------- VoigtKuntz line shape used in SS without the (f/f_o)^2 factor, conversion frm F77 to C made by Chikako herself H2O-22GHz 10% diff. in the lower band wing. Window89GHz good agreement O3110GHz diff. most in the line center 60% at 18 km O2118GHz line center 30% diff at 55 km altitude. Window150GHz specific lines show up in SS (300% diff. max.) which are not in ARTS (149.5, 150.3 GHz) especially at 55 km altitude. H2O183GHz At 183.8GHz a remarkable diff. of 150% , also difference at 182.7 and 182.2. Detailed discussion of absorption comparison ARTS-FZK ------------------------------------------------------ H2O22GHz with continuum calculated no real comparison possible Window89 diff. up to 80% O2118GHz in the line center diff. small but in the line wing very large diff. 100% VVW line shape function used in FZK. center line missing in FZK? CLO501GHz jump in the line wings symmetric around the line center at high altitude - why? N2O502GHZ same feature as for CLO501 at high altitude HNO3544GHz up to 60% diff. at lower band wing, 20% in the line range at higher altitudes. CO576GHz some sort of cut off in the line shape? Same feature as for the other species at high altitude. O3625GHz Same feature as for the other species at high altitude. CLO649GHz strong offset seen %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% CASE 1B Session continuum absorption %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% GENERAL ITEMS: ============== - problem with H2O isotopic lines around 550GHz. Compare with Moliere (Phillip Baron) and with JPL and HITRAN00. Detailed discussion of continuum absorption comparison ARTS-CRL -------------------------------------------------------------------- O2 included in CRL calculations CRL has simply taken the MPM89 model and changed the line catalog to MPM93. So the continuum term is still MPM89. No difference in CRL and SS Detailed discussion of continuum absorption comparison ARTS-Moliere -------------------------------------------------------------------- ask Phillip Baron about the strong isotopic H2O lines at 547 and 552 GHz. Error in the MPM93 but what is changed actually in Moliere? How to fix this error? Comparison Isotopic ratio is missing in MPM93. Outside of this, the two models are in very good agreement. Detailed discussion of continuum absorption comparison ARTS-EORC -------------------------------------------------------------------- MPM89 continuum used together with MPM93 lines, so directly comparable. Detailed discussion of continuum absorption comparison ARTS-FZK -------------------------------------------------------------------- MPM89 is used by FZK for this calculations so not directly comparison. Some sort of cut off in the FZK line shape? Detailed discussion of continuum absorption comparison FZK-EORC -------------------------------------------------------------------- large differences, but not sure where they come from. %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% CASE 2 Session species absorption comparison %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% GENERAL ITEMS: ============== - dry air continua to be checked in ARTS too high! (-->TKS) - pressure shift check in ARTS, Verdandi (definitely an error), Moliere. Is there a sign for the pressure shift data in the HITRAN catalog? Check. - HNO3@544GHz error in HITRAN96 - check - ARTS O2118 line check which model is actually used for this comparison in arts? - it seems to exist a "line shape problem" between the different implementations. So this basic stuff has to be in agreement before further calculations are performed. Detailed discussion ARTS-Moliere -------------------------------------------------------------------- H2O22GHz 5% global difference outside of the line center range. peak diff. at 21.95 and 22.3 GHz of 25% and 10%, respectively. Window89 large difference, in general 400%!! -> WHY? H2O183GHz line centers are shifted due to HITRAN-JPL difference. HCN354GHz ok CLO501GHz ok O3625GHz at 625GHz large diff. HCL625GHz features at 626 and 626.2 GHz, due to additional lines or isotopic ratios, pressure shift? Detailed discussion ARTS-EORC -------------------------------------------------------------------- H2O22GHz line shape differences? different continuum seen in the millimeter region differences between JPL and HITRAN catalogs seen Detailed discussion ARTS-FZK -------------------------------------------------------------------- dry air continuum differences might cause the large differences in the window regions. ARTS should be checked. line shift seen, but sure which implementation is correct. This has to be checked. line wings perhaps differently modeled. %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% CASE 3 Session Radiative transfer intercomparison limb case %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% GENERAL ITEMS: ============== - see if there is an error in the interpolation scheme in ARTS. Detailed discussion ARTS-Moliere -------------------------------------------------------------------- f=495-510GHz differences in the line center regions are seen of up to 8K. probably the interpolation of the abs. coeff is the main source of the difference %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% CASE 3 Session Radiative transfer intercomparison downlooking case %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% GENERAL ITEMS: ============== interpolation problem, conversion problem and ground model reflection are the major three different points. Set the ground emissivity to one for avoiding reflection to get a better setup for further comparisons: - sigma = 1 - output should be in `true' (Planck) brightness temperature Detailed discussion of comparison without sensor ARTS-Moliere -------------------------------------------------------------------- 80-200GHz Moliere has no reflection but emission of the ground, ARTS has both terms. Detailed discussion of comparison without sensor ARTS-CRL -------------------------------------------------------------------- 80-200GHz CRL has no reflection but emission of the ground, ARTS has both terms. Detailed discussion of comparison without sensor Moliere-CRL -------------------------------------------------------------------- 80-200GHz diff. of 1K at 89 channel, 3K at 150 GHz and 4K at 183 GHz. Difference of RT in the optical thick region. Conversion to brightness temperature is done by Rayleigh-Jeans in the case of CRL and Planck in the case of Moliere. PE: suggests that the difference is due to interpolation SB,VJ: the conversion from Planck to Rayleigh-Jeans is the cause for the difference SB: another thing must be different since the difference can not only explained by the conversion. Detailed discussion of comparison with sensor ARTS-Moliere ----------------------------------------------------------- no real difference compared to the subcase without the spectrometer Detailed discussion of comparison with sensor Moliere-CRL ----------------------------------------------------------- difference in the sensor model compared to the subcase without the spectrometer The sensor model should average both sidebands! %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% CASE 4 Session RT comparison %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% GENERAL ITEMS: ============== Because we have shown that in both absorption and RT are remarkable differences, the case4 comparison will be of no use. So this case will be redone after the mentioned differences are clarified. Minutes of the Third International Radiative Transfer Modeling Workshop 2001 =============================================================== Minutes of 11 October 2001 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Special Session: recalculations %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% General Items ============= 1.) Patrick ----------- in ARTS the interpolation is actually done linearly in pressure and not in log(p) as thought yesterday (in one place in the RT calculation). Different results are seen for different interpolation schemes for sparse grids. So one has to calculate the integral on a dense grid to avoid differences due to different interpolation schemes. Franz is doing his interpolation with the quadrature method ("trapez" rule or splines) for the calculation of the optical depth. For sparse altitude grids one should see differences to ARTS. Franz: Added comments --------------------- Interpolation and quadrature (= numerical evaluation of definite integrals) are different things. I am using standard methods of numerical mathematics to solve the integrals showing up in the radiative transfer problem, i.e. beer's law in an inhomogeneous atmosphere (here just one molecule) T(v,s) = exp [-integral_0^s a(v,s') n(s') ds'] where v is wavenumber or frequency a is the absorption coefficient n is the molecular number density s is distance along line of sight and the 'schwarzschild equation' (ignoring a source/background term) I(v) = integral B(T(s),v) [dT(v,s)/ds] ds So in general a quadrature rule approximates integral_a^b f(x) dx = sum w_i f(x_i) where w_i and x_i are the nodes and weights which depend on the quadrature rule actually used. For example, trapez quadrature is simply integral_x0^xN f(x) dx = sum 0.5 (x_i-x_{i-1}) [f(x_i) + f(x_{i-1})] The quadrature rules I am using to solve beer and schwarzschild are trapez, method of overlapping parabolas, and a method based on a spline representation of the integrand. These methods work for arbitrarily spaced abscissas=nodes (in contrast, the standard simpson would require the x_i to be spaced equidistantly). "My" nodes are defined by the altitude grid points where p, T, and gas densities are given. Thus in case of up or downlooking, the path variable s along the line of sight is identical to the altitude z (maybe except for its orientation), and in case of general slant pathes s is just z with a factor cos(angle). For limb its a little more complicated, certain altitudes show up twice along the line of sight, i.e. to the left and right of the tangent point. Anyway, the integrals are evaluated exactly at these grid points s_i, and there is no interpolation required to solve these integrals. (However, as mentioned in Delmenhorst, an interpolation will be made in case the tangent height is between two altitude grid points.) Except for this special case of intermediate tangent points, interpolations are used in mirart at just two occasions: If atmospheric data are read from different files (each with its own altitude grid), the data are interpolated to one common grid. And: As cross sections are calculated on a wavenumber (frequency) grid which is unique for each level (p, T) and molecule, the cross sections are interpolated to common (i.e. the finest) grid before the radiative transfer calculation (abs. coeff., transmission, radiance) is started. I should also mention that there is no interpolation done for finite field of view simulations (again with the above mentioned exception). 2.) Gerhard ----------- In the FZK model there was a rudimentary Voigt function implemented which was switched on at hight altitudes. This caused the featured differences compared to ARTS. 3.) Satoshi ----------- Case 1A difference: there are still remarkable differences remaining Case 1B difference: Also. 4.) Jo ------- extra lines in Moliere two NO lines were taken into account by Moliere but not in ARTS. This was just a misunderstanding in the way to perform the comparison. Comparison of ARTS and Moliere for the Odin FM comparisons study. Then diff. is of the order of 5% or lower for the frequency range around 500 GHz. Moliere-Odin operational code differences are very small, negligible. %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Session: Case 3 RT intercomparison limb looking %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% General Items ------------- - specify the cosmic background (yes/no) more clearly in the comparison instructions - specify the Earth radius more clearly in the comparison instructions - specify the tangent altitudes for the limb cases more clearly in the comparison instructions - finer absorption grid calculations for the further comparisons - true Planck brightness temperature is the reference for the future comparisons ==> repeat the calculations with these new specifications Detailed discussion ARTS-Moliere -------------------------------- 495-510GHz differences of 10K at lower altitudes (->interpolation?) PE: do the calculation for a fine grid to see more clearly the difference The cosmic background is not included in Moliere, so in the windows the diff has an offset of the cosmic background. Detailed discussion ARTS-CRL -------------------------------- 495-510GHz differences up to 20K can be seen in the line wings, while in the line center the diff. is very low. The differences are all going into one direction. Perhaps the Earth radius is differently introduced since only the observation angles are provided for this comparison. Detailed discussion ARTS-DLR -------------------------------- 495-510GHz differences are very high, but the structures of the differences are so that one can assume a unit conversion problem or a mismatch of the tangent altitudes. Detailed discussion ARTS-SS -------------------------------- 495-510GHz differences are highest in the line center up to 20K, but in the line wing the difference is small. Comparing SS with Moliere shows also high differences in the line center but also some remarkable diff. in the line wings (offset). The cosmic background is not included in SS. %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Session: Case 4 RT intercomparison limb looking %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% General Items ------------- Omitted, sine basic problems have to be resolved first. %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Session: Case 3 RT intercomparison up looking %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% General Items ------------- - DLR recalculate the stuff with appropriate units of abs. coeff. - CRL recalculate the stuff with appropriate platform hight. - zero degree error in FZK. Detailed discussion ARTS-Moliere -------------------------------- 142GHz diff. (<0.4K) in the wings perhaps the cosmic background which is in ARTS but not in Moliere calculated. Additionally the interpolation can cause some of the difference. Detailed discussion ARTS-CRL -------------------------------- offset of 50K can be seen. Causes: the observation platform altitude was different in both calculations. Detailed discussion ARTS-DLR -------------------------------- abs. coeff. conversion was not done for the DLR input. DLR needs 1/cm and given were 1/m. Otherwise good agreement. However, the redone the calculations during the morning session (i.e. included the 1/m to 1/cm conversion in my code) and the comparison carried out during the lunch break showed consistent results withing a Kelvin Detailed discussion ARTS-FZK -------------------------------- diff. of 1K in the line center and approx. zero in the line wing for zero degree observation angle. All the other observation angles are in good agreement. Some interpolation differences in the line center region are seen %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Session: Case 3 RT intercomparison up looking sensor part %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% General Items ------------- Detailed discussion ARTS-Moliere -------------------------------- good agreement (diff in the line center 0.3K in the line wing 0.38K) %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Session: Case 4 RT intercomparison up looking with sensor %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% General Items ------------- - ARTS should calculate this case with a monochromatic frequency grid of just 1MHz spacing. Also the sensor has to be refined to smaller channel width in the line center. ==> finer frequency grids! - the specifications for this case should be more clear. Also which species should be included in th atmospheric scenario (H2O, N2, O2,etc.) Detailed discussion ARTS-Moliere -------------------------------- Cosmic background included in Moliere here. Diff 2K in the line center. One can see the different line data base in the plot because of the different line center frequency. One should JPL for this case and not HITRAN because of its more accurate center frequency. Detailed discussion ARTS-CRL ---------------------------- offset of approx. 2.5K and the line center frequency is slightly shifted and pressure broadening is might be different, different continua treatment? Detailed discussion ARTS-DLR ---------------------------- very good agreement, only in the line center a diff. of the order of 1K can be seen. Same spectroscopical input is used for ARTS and DLR, also Voigt is used in both models. But DLR uses the CKD continuum. Because the sensor with 25MHz frequency grid spacing (sparse in this case), so this can explain also some differences. Also the sensor is to broad, smaller channel width should be used. Detailed discussion DLR-FZK ---------------------------- Problem, oxygen is switched off in the calculations in case of FZK, so this explains the offset of some Kelvins. No sensor model included in FZK model, so the convolution is done by Carmen with her sensor model. %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Session: summary/discussion %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Case 1 Absorption ------------------ a. line shape functions - IUP will create a new case 0, where the Voigt function will be tested. Constant T=296K (to avoid the line strength temperature conversion) and for constant T=250K (for partition function implementation test), pressure changes, but VMR const., only one line. The differences should be below 10^-4. - The (f/f_o)^2 factor should be included in the lineshape. b. pressure shift - sign of the shift parameter (HITRAN ?, Verdandi ?) - implementation in the different models - The given catalog contains a shift for HCl, which should be taken into account in the simulations. c. errors in the line catalogs - O2 isotopic lines too strong in HITRAN96, also water vapor isotope problems are known in HITRAN00. - H2O-Isotopic lines 550GHz in MPM93, PE will ask P. Baron Franz: added comments --------------------- O2 isotopic lines too strong in HITRAN96 Birger and I had noticed these strong O2 isotope lines already when we analyzed the DLR heterodyne OH spectra in the far infrared. After comparsison with JPL O2 data we concluded that the isotope abundance factor is simply missing in the O2 isotope lines, so it is a very "systematic" error. d. line exclusion - exclude to NO lines from this comparison (see Moliere's extra lines) e. set the self line broadening term to the foreign broadening term if there is no information in HITRAN. f. target: below 0.1% Case 2 ------ - addressed to IUP: try to be realistic in the specifications for this case. (Also applies to the other models of course.) - also make a reasonable mix of the continua and line absorption terms in ARTS. Case 3 limb ----------- - viewing direction should be specified on a unique way (tangent altitude, Earth radius) - splitting: to determine the interpolation another calculation should be done with a very fine grid of 100m. Absorption and temperature will be provided on this fine grid by IUP. - units of the output should be in true Planck brightness temperature. - cosmic background should be implemented in the calculations, T=2.735K. - target: the differences for the fine grid comparisons should be below 1.0% and below 1K. Case 3 down-looking ------------------- - reflectivity neglected, the emissivity will be set to 1.0 to avoid discrepancies. - calculation should be done with a very fine grid of 200m. - units of the output should be in real Planck brightness temperature. - specifications of the sensor model: perfect double-sideband receiver. - 130-180 degree to be calculated (one side of the scan). Case 3 up-looking ----------------- - units of the output should be in real Planck brightness temperature. - cosmic background should be implemented in the calculations, T=2.735K. - one should go to higher zenith angles (0-80 degree), elevation angles between 10 and 90 degree. - more realistic width of the sensor channels (1MHz). - monochromatic frequency grid narrower that the sensor frequency grid. - no refraction. Case 4 ------ - Be as realistic as possible, this means also that each model should use its own optimized way of calculation. - line spectra - continua - refraction - cosmic background. - One should be go to higher zenith angles (0-80 degree), elevation angles between 10 and 90 degree. %%%%%%%%%%%%%%%%%%%%% Session ARTS overview %%%%%%%%%%%%%%%%%%%%% SB hints on ARTS free download. Comments are very welcome to improve the code and the portability. Next year next workshop at Bredbeck probably, then in summer. - Thank you for coming! ============================================================================ minutes written by Thomas Kuhn, iup, 10/11.10.2001 ============================================================================