Targeted Spectral Threshold Tests for Reducing Cloud Omission in the GOSAT-2 CAI-2 Cloud Mask

Abstract. Cloud detection is a common problem in optical remote sensing, but its consequences vary by mission. In GOSAT-2, missed clouds matter because they can affect greenhouse gas retrievals. This research introduces a targeted omission-recovery approach for the GOSAT-2 CAI-2 Level 2 cloud mask over water. Rather than replacing the operational mask, the method identifies and corrects its errors. True-positive and false-negative cloud footprints are mapped into two-dimensional spectral feature spaces proposed from native CAI-2 L1B ultraviolet, visible, and near-infrared radiances. Polynomial logistic regression was used to estimate the probability that a cloud footprint belonged to the detected-cloud group. A Kolmogorov–Smirnov test then compared false negatives at or above the weakest detected TP level with those below it, and the three feature spaces that separated these groups most clearly across granules were converted into probability-threshold tests. Finally, these tests were applied only to pixels labeled no-cloud in the original L2 mask, with additional controls such as sun-glint filtering. Among four overwater granules around Japan, the revised mask reduced cloud omission by 64–83 %, while overestimation increased by 0 to 4 percentage points. Cohen's kappa increased by 5–14 points, reaching 0.88 to 0.93. In the independent granule, the tests recovered 55.4 % of the clouds missed by the L2 mask while introducing false positives over only 2.07 % of cloud-free pixels. Also, an indicator semivariogram showed that recovering the missed clouds improves the mask most at the scales of edges and small and scattered clouds. These results show that missed clouds in CAI-2 contain recoverable spectral structure, which can be used to improve the operational CAI-2 cloud mask over the ocean.

Authors

Institutions

Publication Details

Published
2026-11-16
DOI
https://doi.org/10.5194/egusphere-2026-5309
Primary Topic
Atmospheric aerosols and clouds
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Targeted Spectral Threshold Tests for Reducing Cloud Omission in the GOSAT-2 CAI-2 Cloud Mask

Tsuneo Matsunaga, Mohammadreza Safabakhshpachehkenari
1 links
Atmospheric aerosols and clouds
preprint

Targeted Spectral Threshold Tests for Reducing Cloud Omission in the GOSAT-2 CAI-2 Cloud Mask

Tsuneo Matsunaga, Mohammadreza Safabakhshpachehkenari
preprint en

Abstract

Abstract. Cloud detection is a common problem in optical remote sensing, but its consequences vary by mission. In GOSAT-2, missed clouds matter because they can affect greenhouse gas retrievals. This research introduces a targeted omission-recovery approach for the GOSAT-2 CAI-2 Level 2 cloud mask over water. Rather than replacing the operational mask, the method identifies and corrects its errors. True-positive and false-negative cloud footprints are mapped into two-dimensional spectral feature spaces proposed from native CAI-2 L1B ultraviolet, visible, and near-infrared radiances. Polynomial logistic regression was used to estimate the probability that a cloud footprint belonged to the detected-cloud group. A Kolmogorov–Smirnov test then compared false negatives at or above the weakest detected TP level with those below it, and the three feature spaces that separated these groups most clearly across granules were converted into probability-threshold tests. Finally, these tests were applied only to pixels labeled no-cloud in the original L2 mask, with additional controls such as sun-glint filtering. Among four overwater granules around Japan, the revised mask reduced cloud omission by 64–83 %, while overestimation increased by 0 to 4 percentage points. Cohen's kappa increased by 5–14 points, reaching 0.88 to 0.93. In the independent granule, the tests recovered 55.4 % of the clouds missed by the L2 mask while introducing false positives over only 2.07 % of cloud-free pixels. Also, an indicator semivariogram showed that recovering the missed clouds improves the mask most at the scales of edges and small and scattered clouds. These results show that missed clouds in CAI-2 contain recoverable spectral structure, which can be used to improve the operational CAI-2 cloud mask over the ocean.

National Institute for Environmental Studies (JP)
Climate action
Atmospheric aerosols and clouds
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon