Design of high redshift emission-line galaxy target selections for DESI-II

This is a repository for the code used for an emission-line galaxy (ELG) target selection for DESI-II using deep HSC wide grizy imaging and photometric redshifts from the COSMOS2020 catalog. The HSC wide imaging is the best proxy for desgining ELG samples that could be selected with early Large Synoptic Survey Telescope (LSST) imaging.

How to use the scripts in analysis folder:

Dependencies

• numpy
• pandas
• astropy
• matplotlib
• sklearn

What order to run cosmos2020 scripts:

1.) train_all_feature_random_forest_1.1_1.6
2.) plot_permutation_feature_importance_1.1_1.6
3.) train_best_feature_random_forest_1.1_1.6
4.) plot_color_color_best_feature_random_forest_prob_1.1_1.6
5.) plot_roc_curve_all_feature_1.1_1.6
6.) plot_roc_curve_best_feature_1.1_1.6
7.) plot_color_color_best_feature_random_forest_prob_selection_1.1_1.6
8.) plot_color_cut_selection_1.1_1.6
9.) plot_hist_color_cut_selection_1.1_1.6
10.) plot_hist_best_selections_1.1_1.6

Short description of pilot and truth samples:

Pilot and Truth refers to two different spectroscopic samples that were obtained by the Dark Energy Spectroscopic Instrument (DESI). The color cuts used to obtain the Pilot Sample were r - i < i - y - 0.19 and i - y > 0.35. This sample was used test a target selection of high redshift (1.1 < z < 1.6) ELGs using HSC wide imaging as a prototype for upcoming LSST imaging. The Truth Sample was obtained using broader color cuts than the current DESI ELG sample in order to study how imaging systematics affect DESI ELGs. This sample was also used to optimize a high redshift ELG selection using simple color cuts.

optimizecuts.py module:

Make sure this is in a directory that python knows the path of. This is a module which contains functions to calculate target density, redshift success rate, redshift range success rate, and net target density yield for both Pilot and Truth samples.
These functions require:

• g-fiber limiting magnitude
• shift to the r - i < i - y - 0.19 color cut
• shift to the i - y > 0.35 color cut
• i - z minimum

These are applied to the HSC wide catalog and a spectroscopic cross-matched catalog. This module also contains wrapper functions to be passed into the scipy.optimize.opt function in order to define a loss function to minimize.

Optimization of pilot and truth samples:

The optimize_selection.py scripts load in the HSC wide catalog and a cross-matched spectroscopic subset catalog to optimize selection cuts for selecting high redshift ELGs. After providing guesses for the optimal free parameter values and a weight for the wrapper defining our loss function, the optimized values along with the corresponding summary statistics will be printed out.