FORUM All-Sky Simulated Observations Database for Scene Classification and Fast Radiative Transfer Inversion

We collected atmospheric scenario data from January and July, 10th, 2021 at 12:00 UTC. The dataset covers the entire globe, using a regular grid with 2° increments in both latitude and longitude, for a total of 31,862 cases.

The data are distributed in two compressed archives:

january.tar.xz – data for January 2021

july.tar.xz – data for July 2021

Folder structure

Within each archive, all cases are organized into separate folders. Each folder corresponds to a specific geographical location, identified by its latitude and longitude in the folder name, for example:

N42.00W2.00

indicating latitude 42.00° N and longitude 2.00° W.

File contents (per location folder)

Each location folder contains the following files:

1. Spectral data

forum_spectrum.datContains the FORUM spectral information associated with the atmospheric scenario, with 4 columns:

First column: wavenumber from 100 to 1600.48 cm⁻¹ in 0.36 cm⁻¹ increments

Remaining columns: radiance values, NESR (Noise Equivalent Spectral Radiance ), ARA (Absolute Radiometric Accuracy) in W/(m2· sr · cm⁻¹)

2. Surface emissivity

e_CLAIM* (e.g. e_CLAIM_N80.00W114.00_11)Files containing surface emissivity data. Each file contains 2 columns:

First column: wavenumber from 50 to 1650 cm⁻¹ in 5 cm⁻¹ increments

Second column: emissivity values

3. Cloud condition and optical depths

clear or cloudy fileIndicates clear-sky or cloudy conditions and reports three optical depth values in 3 columns:

total ice cloud optical depth

total liquid cloud optical depth

total cloud optical depth

4. Soil type information

slt.datIncludes three soil type classifications:

ERA5 (ECMWF Reanalysis v5) soil type

dominant MODIS (Moderate-resolution Imaging Spectroradiometer) soil type

Huang soil type with the highest probability, as determined by [1]

5. Atmospheric profiles

prof_* (e.g. prof_20210110T120000_2853)Contains vertical atmospheric profiles.

First row: surface temperature in K and surface pressure in hPa 

Following rows: a matrix describing the vertical profiles of atmospheric gases across the atmospheric layers, in the following order and units:

Pressure, hPa

Temperature, K

H2O, g/kg MMR

CO2, ppv

O3, ppv

N2O, ppv

CO, ppv

CH4, ppv

SO2, ppv

HNO3, ppv

NH3, ppv

OCS, ppv

HDO, ppv

CF4, ppv

LWC (liquid water content), kg/kg

REL (liquid particle radius), um

IWC (ice water content), kg/kg

REI (ice particle radius), um

6. Data generation process

Meteorological data were sourced from the ERA5 reanalysis dataset, with additional gas concentrations (CO2, CO, CH4, NO2, HNO3, SO2) from the CAMS global Greenhouse Gas reanalysis. Remaining gases were taken from the Initial Guess 2 (IG2) database.

Surface emissivity was obtained from the Huang dataset and matched to each location and time using a tailored procedure incorporating the CAMEL database over land and Masuda modification over the sea to introduce variability [1].

Cloud properties, including ice and liquid water contents, were derived from ECMWF. Cloud particle sizes were modeled using the Wyser and Martin approaches. A scene was considered cloudy if the total optical depth at 900 cm⁻¹ exceeded 0.03.

Given computational constraints, optical depths and radiative transfer calculations were parameterized using the σ-F2N code [2]. Multiple-scattering effects were included via the Chou scaling method. CLAIM code [3] was then used for spectral convolution and to add instrument noise, using FORUM specifications with 0.36 cm⁻¹ resolution and NESR levels from Phase A instrument studies.

[1] L. Sgheri, C. Sgattoni, and C. Zugarini. “Determination of emissivity profiles using aBayesian data-driven approach”. In: Mathematics and Computers in Simulation 229(2025), pages 512–524. issn: 0378-4754. doi: https://doi.org/10.1016/j.matcom.2024.10.015.url:https://www.sciencedirect.com/science/article/pii/S0378475424004051[2] L. Sgheri et al. “The FORUM end-to-end simulator project: architecture and results”.In: Atmospheric Measurement Techniques 15.3 (2022), pages 573–604. doi: 10.5194/amt-15-573-2022. url: https://amt.copernicus.org/articles/15/573/2022/[3] G. Masiello et al. “The new σ-IASI code for all sky radiative transfer calculations in thespectral range 10 to 2760 cm-1: σ-IASI/F2N”. In: Journal of Quantitative Spectroscopyand Radiative Transfer 312 (2024), page 108814. issn: 0022-4073. doi: https://doi.org/10.1016/j.jqsrt.2023 .108814. url: https: //www .sciencedirect.com/science/article/pii/S0022407323003321