Rising sea surface temperature (SST) due to climate change has increased thermal stress on coral reef ecosystems, frequently resulting in coral bleaching events. To facilitate routine monitoring of the spatiotemporal dynamics of SST, products from multiple satellite datasets have been developed in recent decades. However, most existing SST maps are designed for continental to global applications and typically have spatial resolutions of several kilometers, limiting their utility for monitoring local-scale coral conditions in spatially heterogeneous coastal environments. In this study, we integrated recent satellite data from the Global Change Observation Mission–Climate with other well-established satellite-derived SST products to generate daily SST maps at a spatial resolution of 250 m, covering the period 1985–2024. These downscaled SST maps were used to calculate degree heating weeks (DHW), a widely used metric for quantifying thermal stress. The resulting DHW values were employed as input features for machine learning-based prediction of coral bleaching events around Okinawa Island, Japan. The results showed that the downscaled DHW maps yielded higher predictive performance relative to benchmark products derived from existing lower spatial resolution SST datasets (~1 km). Furthermore, analysis of the downscaled dataset revealed interannual and seasonal patterns of DHW variability, providing a foundation for improved local-scale predicti on of coral bleaching in this region and identifying populations that are resistant to elevated seawater temperatures.
Mizuochi et al. (Thu,) studied this question.