Impact of climate change on the Vrana Lake surface water temperature in Croatia using support vector regression
Željka Brkić, Ozren Larva
Abstract
The case study presents Vrana Lake on Cres island in Croatia, which is the largest freshwater resource on the Mediterranean islands. It is used for the public water supply of the residents and tourists who inhabit this island during the summer months. Using the Support Vector Regression (SVR), the influence of future climate changes on the lake surface water temperature (LSWT) was analysed. Input data were monthly air temperatures (AT). Model training and validation were based on measured LSWT and AT in the period 1981–2022. Expected LSWTs for the period 2023–2070 under the RCP8.5 emission scenario were forecast based on climate modelling data of monthly AT for the period 1971–2070. The results showed that the applied SVR model can effectively forecast monthly LSWTs, which was confirmed by a correlation coefficient of approximately 0.99 between the measured and simulated LSWTs. Root mean square errors were lower than 1 °C. The LSWT warming trend in 2023–2070 is expected to be lower than that observed in 1981–2022, and will vary from 0.2 °C dec -1 –0.3 °C dec -1 . The largest increase in LSWT can be expected in spring while the increase in LSWT will be the smallest in the summer months. In approximately 50 years, the expected LSWT could be higher by an average of 1.2 °C. • The SVR is a suitable tool for predicting lake surface water temperature (LSWT). • The largest increase in LSWT can be expected in the spring season. • The smallest increase in LSWT can be expected in the summer months. • In approximately 50 years, the expected LSWT will be higher by 1–1.4 °C. • Knowledge about LSWT was enhanced.