============ Readme1.txt ============ -------------------- file contents ---------------------------------------- bigRun25.mat / contains the LTS radiance profiles computed by smartG ====================================================================== dim1=25600= number of cases dim2=101= altitude grid 0:1:100 km (rem Aerosol is defined on z=0:1:50 km) dim3=8 wavelengths 300,315,351,525,600,675,745,1020 nm Name Size Bytes Class Attributes ALB 25600x1 204800 double case albedo BIGI 25600x101x8 165478400 double 1st Stokes (I) vector (relative to spectral solar irradiance) BIGQ 25600x101x8 165478400 double 2st Stokes (Q) vector BIGU 25600x101x8 165478400 double 3st Stokes (U) vector CAE 25600x5 1024000 double 5 PCA coeff for aerosol number density F0m 1x101 808 double average ozone volume mixing ratio (vmr) [ppm] IRANA 25600x1 204800 double case random indices IRANA(1)=23383 => ALB(23383) was used REF 25600x1 204800 double aerosol effective radius (10^-6 m) SAA 25600x1 204800 double solar azimuth angle [°] at tangent point SIG 25600x1 204800 double aerosol lognormal width SZA 25600x1 204800 double solar zenith angle [°] at tangent point n0 101x1 808 double average air number density [cm^-3] naerCell 2x6 68384 cell cell{1}= all (55) climatological aerosol logprofiles (=log(n_aer(z) [cm^-3])) cell{2}= mean climatological aerosol logprofile cell{3}= climatological aerosol logprofile eigenvectors cell{4}= PCA coefficients of cell{1} cell{5}= aerosol altitude domain [km] cell{6}= index of used altitudes ozCell 1x10 1305960 cell cell{1}= std of F0m [ppm] cell{2}= F0m [ppm] cell{3}= MONTH (101 alt x 18 [-85°:10°:85°] x 12 months) ozone climatology: month bins cell{4}= N (101 alt x 18 [-85°:10°:85°] x 12 months) ozone climatology: air density [cm^-3] cell{5}= OZvmr (101 alt x 18 [-85°:10°:85°] x 12 months) ozone climatology: ozone vmr [ppm] cell{6}= P (101 alt x 18 [-85°:10°:85°] x 12 months) ozone climatology: pressure [hPa] cell{7}= PHI (101 alt x 18 [-85°:10°:85°] x 12 months) ozone climatology: latitude bins [°] cell{8}= T (101 alt x 18 [-85°:10°:85°] x 12 months) ozone climatology: temperature [K] cell{9}= U (101 x 101) ozone vmr PCA eigenvectors (columwise) cell{10}= Z (101 alt x 18 [-85°:10°:85°] x 12 months) ozone climatology: altitude bins [km] ozList 1x25600 204800 double linear index of ozone profiles Name Size Bytes Class Attributes LOKI 25600x8 1638400 double LOKQ 25600x8 1638400 double LOKU 25600x8 1638400 double PI 2x61x8 7808 double PQ 2x61x8 7808 double PU 2x61x8 7808 double br25Ana2.mat / contains the statistical analysis of the LTS radiance profiles computed by smartG ================================================================================================= for safety, LTS cases are screened against possible outliers, before training the neural network. Only cases lying between 5% and 95 % percentiles will be retained. Name Size Bytes Class Attributes LOKI 25600x8 1638400 double logical index for screened cases Stokes I LOKQ 25600x8 1638400 double logical index for screened cases Stokes Q LOKU 25600x8 1638400 double logical index for screened cases Stokes U PI 2x61x8 7808 double 5 % and 95 % percentiles for Stokes I PQ 2x61x8 7808 double 5 % and 95 % percentiles for Stokes Q PU 2x61x8 7808 double 5 % and 95 % percentiles for Stokes U ozSet12.mat / summary of the LTS ozone profiles ================================================ Name Size Bytes Class Attributes A 25600x20 4096000 double 20 first coefficients (related to the PCA in bigRun25.mat) n0 101x1 808 double arbitraty air number density [cm^-3] average profile used for all vmr conversions noz 25600x101 20684800 double all ozone number density [cm^-3] profiles ozCell 1x10 1305960 cell same as in bigRun25.mat vmr 25600x101 20684800 double all ozone vmr [ppm] profiles aerSet3.mat / summary of the LTS aerosol profiles =================================================== Name Size Bytes Class Attributes AaL 25600x5 1024000 double 5 first coefficients (PCA in bigRun25.mat) of aerosol number density [cm^-3] profiles naerCell 2x6 68384 cell same as in bigRun25.mat naerL 25600x51 10444800 double all aerosol number density logprofiles refRan 25600x1 204800 double all aerosol effective radius [1e-6 m] of lognormal PSD sigRan 25600x1 204800 double all aerosol distribution width of lognormal PSD airSet1.mat / summary of the LTS air profiles =================================================== Name Size Bytes Class Attributes Aair 25600x5 1024000 double 5 first coefficients (PCA in bigRun25.mat) of air number log-density [cm^-3] profiles airCell 1x10 1305960 cell same as for ozCell in bigRun25.mat with air log-densities insteasd of ozone vmr lair 25600x101 20684800 double all air log (number densities) buildBigXY21.m ================ This routine builds the input X state vector that generated any LTS case and the output Y measurement vector For the logical array "loks" of considered cases, X contains 35 elements X=[A(loks,1:npcOz) CAE(loks,:) Aair(loks,:) REF(loks) SIG(loks) SZA(loks) SAA(loks) ALB(loks)]; = 20 ozone cooeff + 5 aerosol number log-density coeff + 5 air number log-density coeff + effective radius + sigma (dist. width) + sza + saa+ albedo nnitrain19.m ============= this routine to train the ONNI neural network from the PCA analysis of Y to predict X. It is using dedicated matlab toolboxes (Deep Learning and Distributed Computing) to run the NVIDIA GPU card. It automatically generate the neural network model as a routine. nniOz3.m, nniOz3a.m =================== The ONNI routine produced by nnitrain19.m to be used for performance analysis. nniOz3a.m is restricted to channels 300,315,351,525,600,675,745 nm for comparison with OMPS