Mass Mapping Starter

Minimal, shareable weak-lensing mass-mapping starter repository.

Repository layout

• scripts/ — runnable command-line scripts
  • download_kappatng_runs.py
  • prepare_kappatng_subset.py
  • generate_shear_from_kappa.py
  • plot_bundle_outputs.py
  • prepare_noise_mask_sizes.py
• operators.py, noise_mask.py — core NumPy operators and noise/mask helpers
• data/ — reference assets and example outputs
• notebooks/simple_forward_model_demo.ipynb — short end-to-end notebook demo

Included example outputs

• data/kappa_subset_lp001_runs001-010_wlmmuq_384.npz
kappa.shape = (10, 384, 384)
• data/shear_subset_lp001_runs001-010_wlmmuq_384.npz
• data/kappa_subset_lp001_runs001-010_wlmmuq_256.npz
• data/shear_subset_lp001_runs001-010_wlmmuq_256.npz
• data/cosmos_noise_mask_384.npz
• data/cosmos_noise_mask_256.npz
• Plot set in data/plots/

Requirements

• Python 3.10+
• numpy
• h5py
• matplotlib (for plotting)

Quick start (384x384)

Run from repository root:

# 1) Download source runs (example: 10 runs from LP001)
python scripts/download_kappatng_runs.py \
  --dataset fullphys \
  --lp-index 1 \
  --start-run 1 \
  --n-runs 10 \
  --output-root data/kappaTNG_fullphys

# 2) Build kappa subset (wlmmuq-like redshift combination: z01..z40)
python scripts/prepare_kappatng_subset.py \
  --ktng-dir data/kappaTNG_fullphys \
  --lp-index 1 \
  --start-run 1 \
  --n-maps 10 \
  --redshift-mode wlmmuq \
  --crop-size 384 \
  --remove-mean \
  -o data/kappa_subset_lp001_runs001-010_wlmmuq_384.npz

# 3) Generate noisy/masked shear maps
python scripts/generate_shear_from_kappa.py \
  --kappa-file data/kappa_subset_lp001_runs001-010_wlmmuq_384.npz \
  --seed 42 \
  -o data/shear_subset_lp001_runs001-010_wlmmuq_384.npz

# 4) Plot diagnostics
python scripts/plot_bundle_outputs.py \
  --kappa-file data/kappa_subset_lp001_runs001-010_wlmmuq_384.npz \
  --shear-file data/shear_subset_lp001_runs001-010_wlmmuq_384.npz \
  --output-dir data/plots \
  --max-maps 10

To generate more than 10 maps, increase both --n-runs and --n-maps.

256x256 workflow

# Ensure 256 mask/noise asset exists (already included in this repo)
python scripts/prepare_noise_mask_sizes.py \
  --input-file data/cosmos_noise_mask_384.npz \
  --sizes 256 \
  --output-dir data

# Build 256 subset
python scripts/prepare_kappatng_subset.py \
  --ktng-dir data/kappaTNG_fullphys \
  --lp-index 1 \
  --start-run 1 \
  --n-maps 10 \
  --redshift-mode wlmmuq \
  --crop-size 256 \
  --remove-mean \
  -o data/kappa_subset_lp001_runs001-010_wlmmuq_256.npz

# Generate 256 shear subset
python scripts/generate_shear_from_kappa.py \
  --kappa-file data/kappa_subset_lp001_runs001-010_wlmmuq_256.npz \
  --seed 42 \
  -o data/shear_subset_lp001_runs001-010_wlmmuq_256.npz

Noise model

Noise is added as complex Gaussian per pixel:

$$ n(x,y) = \left[\mathcal{N}(0,1) + i,\mathcal{N}(0,1)\right]\sigma_{\mathrm{noise}}(x,y). $$

std_noise is spatially varying and derived from COSMOS galaxy catalog statistics. In the original COSMOS construction:

$$ \sigma_{\mathrm{noise}}(x,y)=\sigma_e \frac{\sqrt{\sum_i w_i^2}}{\sum_i w_i}, $$

with sums over galaxies in each pixel. This is effectively galaxy-density / weight dependent.

Output format (generate_shear_from_kappa.py)

• kappa_true — float32, (n_maps, nx, ny)
• gamma_clean — complex64, (n_maps, nx, ny)
• gamma_noisy — complex64, (n_maps, nx, ny)
• noise — complex64, (n_maps, nx, ny)
• kappa_ks_e, kappa_ks_b — float32, (n_maps, nx, ny)
• std_noise — float32, (nx, ny)
• mask — bool, (nx, ny)
• extent — optional float32, (4,)

Notebook demo

Open:

notebooks/simple_forward_model_demo.ipynb

It demonstrates loading kappa maps, converting to shear, adding COSMOS mask/noise, and plotting results.