This repository contains a set of translation models (weights, training, and testing code) designed to sample model outputs in a way that is consistent with MODIS cloud droplet number concentration (Nd) retrievals. These tools enable meaningful comparisons between model simulations and satellite observations, as described in the study:
"Bias in spaceborne cloud droplet number concentration retrieval and new alternatives"
https://essopenarchive.org/doi/full/10.22541/essoar.175691401.16440458
Translates grid-averaged Nd (cm⁻³) at 900 hPa (1° resolution) to a MODIS-like Nd estimate, enabling direct comparison with satellite retrievals.
Translates in-cloud, cloud-base Nd (cm⁻³) to a MODIS-like Nd estimate, suitable for evaluation against MODIS observations.
MODIS-type retrievals are generated by applying the algorithm described in:
Grosvenor et al. (2018):
"Remote Sensing of Droplet Number Concentration in Warm Clouds: A Review of the Current State of Knowledge and Perspectives".
https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2017RG000593
to COSP-simulated MODIS output.
If you use this code or model in your work, please cite:
Aerenson, T., McCoy, D., Barahona D. (2026). Bias in spaceborne cloud droplet number concentration retrieval and new alternatives. (https://essopenarchive.org/doi/full/10.22541/essoar.175691401.16440458)