Assessment and Classification of Cloud Coverage Using K-Means Clustering Algorithm for the Sentinel-3 LST Data: A Case Study in the Fujairah Region

Authors: Manar Ahmed Mohammed Alblooshi, Sirajul Huda Kalathingal, Shaher Bano Mirza , Fouad Lamghari Ridouane

Abstract: Clouds have a significant impact on the planet's energy balance, climate, and weather. They serve as the primary temperature regulator and function as a blanket to absorb thermal energy or longwave radiation. The present study estimates the percentage of rainfall clouds within a 100-kilometer radius of Fujairah City on the Gulf of Oman using image processing based on machine learning and digital image processing. The data for 9 months starting from January 2022 to October 2022 has been retrieved from the Copernicus satellite data component through the Sentinel 3 LST F2 channel. K-mean cluster analysis has been used to validate the accuracy of an algorithm which is applied to determine cloud cover, with a precision rate of 99.9% for clear weather and 95.5% for overcast conditions. The findings indicate that most of the rainy clouds were observed during the months of January and July. The remaining duration of the year exhibits a reduced occurrence of these clouds. Beginning in February, the region of interest experiences cloud cover accompanied by precipitation subsequent to the month of January. Similarly, the month of July exhibited cloud covers with moisture. Throughout the year, dry clouds are observed with moderate coverage percentages. However, there are no observations of any of these clouds during the months of May and December. In summary, automated systems for observing clouds in the atmosphere are a valuable method for detecting cloud cover and predicting climatic patterns in diverse geographical locations. Keywords: Cloud Coverage, LST, Land Surface Temperature, K-Mean Clustering, Sentinel-3, Copernicus, UAE, Fujairah

Assessment and Classification of Cloud Coverage Using K-Means for Sentinel-3 LST in Fujairah

Citation / ReferenceAlblooshi, M.A.M., Kalathingal, S.H., Mirza, S.B., & Ridouane, F.L. (2023). Assessment and Classification of Cloud Coverage Using K-Means Clustering Algorithm for the Sentinel-3 LST Data: A Case Study in the Fujairah Region. American Journal of Remote Sensing, 11(2), 32–35. doi:10.11648/j.ajrs.20231102.11

Abstract

Clouds influence global climate, weather, and energy distribution. This study estimates rainfall-cloud percentages within a 100 km radius of Fujairah using Sentinel-3 LST F2 channel data from January to October 2022. A K-means clustering algorithm was applied to classify cloud types with 99.9% accuracy for clear weather and 95.5% for overcast conditions. Results show that January and July recorded the highest rainy-cloud frequencies, while May and December had almost no cloud presence. Automated cloud-monitoring systems prove valuable for predicting climatic patterns.

Method Summary

Study Area & Data

• Region: 100 km offshore around Fujairah, UAE.

• Data Source: Copernicus Sentinel-3 LST F2 channel.

• Timeframe: 9 months (Jan–Oct 2022).

• Tools: Python, Google Earth Engine.

Cloud Categories

Sentinel-3 LST imagery contains five main cloud types:

1. Null area (no data)

2. Non-cloud area

3. Cloud with moisture

4. Rainy clouds

5. Dry clouds

K-Means Clustering

• Cluster count k = 6.

• Colour-based pixel segmentation after standardisation with a colour bar.

• Euclidean distance for cluster assignment; iterated until stable centroids obtained.

• Pixel percentages computed for each cloud class.

Key Findings

Rainy Clouds

• Highest: January (92.7%)

• Secondary peak: July (47.2%)

• Moderate: March (24.15%), April (9.23%), August (2.96%)

• None detected: February, May, June, September

Moisture Clouds

• Highest: July (97.9%)

• High: February (74.6%)

• Moderate: January, March, April, June, August

• None: May, September

Dry Clouds

• Present in most months, especially January and July

• Moderate coverage (>20%) in: Jan, Feb, Mar, Apr, Jun, Jul, Aug

• None detected: May, September

Accuracy

• 99%+ accuracy for clear skies

• 95% accuracy for cloudy/overcast imagery

Conclusion

Machine learning significantly enhances cloud-coverage monitoring. This study demonstrates that K-means clustering applied to Sentinel-3 LST data can reliably classify cloud types over Fujairah, supporting improved forecasting and climate pattern analysis. The approach provides a robust framework for automated cloud detection in diverse geographic regions.