Endpoint-separated modelling of environmental associations with <i>Noctiluca scintillans</i> occurrence probability, conditional positive abundance, and Bloom90 probability in the southern Black Sea
REGIONAL STUDIES IN MARINE SCIENCE, cilt.103, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 103
- Basım Tarihi: 2026
- Doi Numarası: 10.1016/j.rsma.2026.105514
- Dergi Adı: REGIONAL STUDIES IN MARINE SCIENCE
- Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, BIOSIS, Zoological Record
- Anahtar Kelimeler: Blocked validation, Cluster-aware inference, Environmental gradients, Heterotrophic dinoflagellate, Monitoring prioritisation, Türkiye, Upper-tail abundance
- Açık Arşiv Koleksiyonu: AVESİS Açık Erişim Koleksiyonu
- Karadeniz Teknik Üniversitesi Adresli: Evet
Özet
Noctiluca scintillans (Macartney) Kofoid & Swezy is a heterotrophic bloom-forming dinoflagellate whose dense aggregations may alter trophic interactions and nutrient cycling. This study examined environmental and temporal associations of N. scintillans using a seven-year fixed-station monitoring dataset covering winter and summer sampling at 21 stations from & Idot;& gbreve;neada to Hopa along the southern Black Sea coast during 2018-2024. An endpoint-separated modelling framework distinguished occurrence probability, conditional positive abundance, and Bloom90 probability. Bloom90 was defined as a dataset-specific upper-decile abundance endpoint corresponding to similar to 1744 ind. m(-3). Noctiluca scintillans was detected in 211 of 292 valid abundance observations (72.3%) and reached a maximum abundance of 66,751 ind. m(-3). Each 6.88 degrees C increase in temperature, equivalent to one complete-dataset standard deviation, was associated with a 91.0% reduction in occurrence odds (odds ratio = 0.090, 95% confidence interval 0.031-0.223). The occurrence-probability model retained good discrimination under leave-one-station-out blocked validation (area under the receiver operating characteristic curve [ROC AUC] = 0.860) but lower discrimination under leave-one-year-out blocked validation (ROC AUC = 0.780). Conditional positive abundance varied principally among years and showed limited out-of-sample performance. After environmental adjustment, summer sampling events had 13.7-fold higher Bloom90 odds than winter sampling events. In the complete dataset, Bloom90 probability was negatively associated with temperature and phosphate phosphorus and positively associated with Chlorophyll-a; the Chlorophyll-a association was not retained after exclusion of 2021. The Bloom90-probability model showed modest discrimination under leave-one-station-out blocked validation (ROC AUC = 0.698) and no meaningful discrimination under leave-one-year-out blocked validation (ROC AUC = 0.517). S2/K0 in the Istanbul Strait outflow sector had the highest observed Bloom90 frequency, but wide, overlapping station-level confidence intervals precluded precise spatial ranking. The endpoint-separated modelling framework therefore provides an explanatory basis for monitoring prioritisation but does not support temporally transferable forecasting of Bloom90 events.