TCOM-H2O: TOMCAT CTM and Occultation Measurements based daily zonal stratospheric H2O profile dataset [1991-2021] constructed using machine-learning.

Methodology: The TOMCAT simulation was conducted at a T64L32 resolution, consistent with previous work by Dhomse et al. (2021, 2022), covering the period from 2000 to 2024. These simulations utilized ERA-5 reanalysis data. H2O Profile Processing and Bias Correction Collocated H2O profiles are organized into five distinct latitude bins: NH polar: 90∘N - 50∘N NH mid-lat: 20∘N - 70∘N Tropics: 40∘S - 40∘N SH mid-lat: 70∘S - 20∘S SH polar: 90∘S - 50∘S Initially, differences between TOMCAT and satellite measurements (primarily ACE-FTS data) are calculated for each zonal bin across 51 height levels (ranging from 10,km to 60,km). Note that TOMCAT may not accurately capture H2O evolution post-HTHH eruption due to the sparse spatial coverage of ACE measurements, which limits training data. Separate XGBoost regression models are then trained for these H2O differences at each height level within a given latitude bin. These trained models are subsequently used to estimate H2O bias corrections for all daytime TOMCAT grids (9132 days), specifically sampled at 1:30 PM local time at the equator. This yields grid-specific bias corrections that are applied to the original TOMCAT profiles. Height-resolved H2O profile data are then interpolated onto 28 standard pressure levels (from 300,hPa to 0.1,hPa), using pressure levels directly from the TOMCAT grids. For overlapping latitude bins, values are averaged to ensure smoother fields near boundary regions. We acknowledge the inherent dry biases in the original TOMCAT H2O profiles, largely because the TTL entry mixing ratios are based on a simplistic sinusoidal seasonal cycle, which omits the H2O enhancement contributed by tropical convective clouds. Data Files The dataset includes two files containing daily mean zonal mean H2O profiles: zmh2o_TCOM_hlev_T2Dz_2000-2024_V1.1.nc: Contains height level data (10,km to 60,km). zmh2o_TCOM_plev_T2Dz_2000-2024_V1.1.nc: Contains pressure level data (300,hPa to 0.1,hPa). Reference Publication This methodology, incorporating only ACE-FTS data and various minor algorithmic developments, is based on the following publication: Dhomse, S. S. and Chipperfield, M. P.: Using machine learning to construct TOMCAT model and occultation measurement-based stratospheric methane (TCOM-CH4) and nitrous oxide (TCOM-N2O) profile data sets, Earth Syst. Sci. Data, 15, 5105–5120, https://doi.org/10.5194/essd-15-5105-2023, 2023.