Assessing transgenerational thermal responses in copepods using integrated U-Net-based morphometric and growth models

Global warming threatens zooplankton across generations, yet quantifying transgenerational thermal responses remains labor-intensive. We developed and validated a U-Net-based morphometric model pipeline integrated with non-linear growth models to quantify body-size dynamics of Mesocyclops woutersi across temperatures and generations. Trained on 1786 annotated images, the model achieved high segmentation accuracy (R2 = 0.99, IoU = 0.90, Dice = 0.94) and strong agreement with manual measurements (R2 = 0.974, MAE = 0.02, RMSE = 0.044; n = 300), substantially reducing processing time. Body length and volume were fitted to four growth models under 24, 28 and 32°C across three generations. The Hernandez-Llamas & Ratkowsky model best described length dynamics (R2 = 0.91–0.94, competitive AIC, narrow confidence intervals). Copepods at 28°C developed faster and reached larger sizes than at 24°C, indicating optimal conditions. In contrast, the third generation at 32°C showed delayed development without significant changes in size or growth coefficients, suggesting thermal stress primarily altered developmental timing. Integrating machine learning–based morphometrics with growth modeling framework provides a scalable, objective approach for transgenerational copepod research under climate warming.