SCIAMACHY NO regression fit MCMC samples

SCIAMACHY mesosphere NO data and regression model samples

SCIAMACHY mesosphere daily zonal mean NO data and Markov-Chain Monte-Carlo samples from the regression coefficient distributions, derived from and for use with the sciapy regression module.

This data set contains the following files:

• NO_regress_output_pGM_Lya_ltcs_exp1dscan60d_km32_float32.nc, NO_regress_output_pGM_Lya_ltcs_exp1dscan60d_km32_float64.nc - samples from the regression coefficient distributions (single and double precision)
• NO_regress_quantiles_pGM_Lya_ltcs_exp1dscan60d_km32.nc - the 0.1, 2.5, 16, 50, 84, 97.5, and 99.9 percentiles of the sampled distributions
• scia_nom_dzmNO_2002-2012_v6.2.1_2.2_akm0.002_geomag10_nw.nc - the SCIAMACHY daily zonal mean NO data
• sciapy_regress_tutorial.ipynb - example ipython notebook

MCMC Samples

The files NO_regress_output..._float32.nc and NO_regress_output..._float64.nc contain MCMC samples of the model as single and double precision floats. The file NO_regress_quantiles....nc contains the 0.1, 2.5, 16, 50, 84, 97.5, and 99.9 percentiles of the sampled distributions and is provided for convenience. The files contain the following parameters:

• kernel:log_sigma, kernel:log_rho - the "strength" and "lengthscale" of the Matérn-3/2 Gaussian Process kernel
• mean:offset:value - the constant offset of the NO model in [\(10^6\) cm\(^{-3}\)]
• mean:Lya:amp - the Lyman-\(\alpha\) coefficient of the mean model in [\(10^6\) cm\(^{-3}\) / Lyman-\(\alpha\)]
• mean:GM:amp - the geomagnetic coefficient (AE) in [\(10^6\) cm\(^{-3}\) / nT]
• mean:GM:tau0 - the constant lifetime of the geomagnetic lifetime in [d]
• mean:GM:taucos1, mean:GM:tausin1 - cosine and sine amplitudes of the yearly geomagnetic lifetime variation in [d]

Daily zonal mean NO data

The model was trained on the SCIAMACHY mesosphere NO dataset, binned into 10° geomagnetic latitude bins using the provided gm_lat variable and using the standard error of the mean as data uncertainties. The data are uploaded as scia_nom_dzmNO_2002-2012_v6.2.1_2.2_akm0.002_geomag10_nw.nc and were prepared by running (after installing sciapy):

bash> scia_daily_zonal_mean.py -g -b'-90:90:10' -o <daily_zonal_mean_NO.nc> </path/to/SCIAMACHY_NO_NOM_orbits_20??_v6.2.1.nc>

Regression sampling

The samples were generated by running the following command:

bash> python -m sciapy.regress <daily_zonal_mean_NO.nc> --proxies Lya:<Lyman-alpha_file.dat>,GM:<AE_file.dat> -A <altitude> -L <geomag_latitude_bin> -w 14 -b 800 -p 1400 -F \"\" -I GM --fit_annlifetimes GM --positive_proxies GM --lifetime_scan=60 --lifetime_prior exp -k -K Mat32 -O0 -m "nom_pGM_Lya_ltcs_exp1dscan60d_km32" -P