wsxgshqk/Temperature_Humidity_Retrieval_Below_Clouds

This repository provides a sample dataset, trained model checkpoints, and normalization parameters to support reproducibility and easy testing of the code for retrieving planetary boundary layer (PBL) temperature and humidity profiles below clouds using AIRS–MODIS–ERA5 collocated data.

The full dataset used in the study is nearly one hundred gigabytes and cannot be hosted directly. Therefore, we provide a compact demo dataset (10,000 samples) that preserves the original data structure and allows users to run the code easily.

1. demo_10000_samples.npz

A lightweight demo dataset containing the first 10,000 collocated AIRS–MODIS–ERA5 samples. This file includes: MODIS interpolated radiances (input images), AIRS geometry information (e.g., VZA), Surface elevation, Cloud fraction (CF), Cloud-top pressure (CTP), ERA5 temperature and specific humidity profiles

2. good_channels_wf_freq.csv

AIRS channel metadata, including selected channels and their corresponding weighting-function peak frequencies, used for channel selection and analysis.

3. norm_params_*.npz

Stored normalization statistics (mean and standard deviation) for each input and output variable. These files are required to correctly normalize inputs and denormalize model outputs when running inference.

4. model_trained_on_01040710_select100_peak_same_dist_addTQ_four_output_*.pth

Models configurations trained with only selected AIRS channels or with fused AIRS & MODIS observations.

5. model_four_months_corr_selected_CTP_CF_TQ_VZA_BLH_Revised_SE_RH_60Lat_logSP_90per_*.ipynb

A jupyter notebook to read the datasets and model configurations to retrieve the temperature and specific humidities, together with evaluations and visualizations.

A simple code to load the demo dataset:

import numpy as np

data = np.load("demo_10000_samples.npz", allow_pickle=True)

X = data["X"] cloud_fraction = data["cloud_fraction"] cloud_top_pressure = data["cloud_top_pressure"] airs_lat = data["airs_lat"] airs_lon = data["airs_lon"] airs_vza = data["airs_vza"] airs_sza = data["airs_sza"] surf_pressure = data["surf_pressure"] blh = data["blh"] temp_full = data["temp_full"] humi_full = data["humi_full"] surf_elevations = data["surf_elevations"]