From 2000 to 2024, annual coral reef surveys were conducted around May–June in Kushimoto, Japan, by mixed teams of 3–18 volunteer recreational divers, professional divers, and marine biologists. The surveys followed the standard Reef Check methodology with local adaptations. Two fixed sites at different depths (3 m and 10 m) were established along the northeastern coast of Sō Island (Sōshima), Kushimoto, with a 100-meter transect line laid at each:

• Fish were surveyed using a 5 m × 5 m × 20 m rectangular prism method,
• Invertebrates (excluding corals) with a 5 m-wide belt transect, and
• Corals with a line-intercept method, recording the presence of live coral every 0.5 m to calculate percent cover.

The shallow site featured dense table coral (Acropora hyacinthus), while the deeper site had greater coral species diversity, dominated by Echinophyllia aspera. Target species included both globally standardized Reef Check indicators (specific fish, mollusks, crustaceans, and echinoderms) and locally selected Kushimoto species not overlapping with the global list. To help identify non-native southern species, species-level recording was added for some groups beyond standard Reef Check protocol.

Locally unique species were chosen from commonly observed native species that did not overlap with the global list, including 9 fish species, 3 mollusks, and several echinoderms. To monitor signs of climate change, particularly warming, the study focused on shifts in species distribution—such as declines in temperate species and increases in non-native southern species.

Long-term sea surface temperature data (since 1972) and the distance from Cape Shionomisaki to the Kuroshio Current’s axis were analyzed, using historical ocean current charts. A “Kuroshio proximity index” was created to quantify the current’s closeness, weighted by distance, and categorized into levels of strength.

Biological and environmental trends were analyzed using linear regression and Spearman’s correlation, excluding rare species (fewer than 50 individuals per site per year) from trend analysis due to low reliability.