Climate is closely related to human life, food security and ecosystems. Forecasting future climate provides important information for agricultural production, water resources management and so on. In this paper, historical climate data from 1962–2001 was used at three sites in Tianjin Baodi, Tianjin and Tanggu districts as baseline and the model parameters were calibrated by the Long Ashton Research Station Weather Generator (LARS-WG). 2m-temperatures in 2011–2020 were verified under two scenarios, representative concentration pathway (RCP) 4.5 and RCP8.5 in different atmospheric circulation models with optimal minimum 2m-temperatures at the three sites. From 2031–2050, Tianjin will be using more moderate minimum 2m-temperatures in future simulations. Support vector machines (SVM) were used to optimize the simulated data to obtain more accurate future maximum and minimum 2m-temperatures for the three sites. The results showed that the determinant coefficient of LARS-WG simulation was 0.8 and SVM optimized determinant coefficient was 0.9 which greatly improved the prediction accuracy. The minimum and maximum future 2m-temperatures optimized under European Community Earth System Model (EC-EARTH) were relatively low and the same future 2m-temperatures optimized under Hadley Centre Global Environment Model Earth System (Had-GEM2-ES4) were high especially in the RCP8.5 scenario which simulated 2051–2070 climate. The SVM optimization showed that the maximum and minimum 2m-temperatures were in general agreement with the original simulation values.

Rationale of Study – With the development of information and communication technologies, colleges and universities worldwide have set up smart classrooms. This study aims to quantify the driving factors influencing student academic performance in a smart classroom environment.Methodology – A meta-analysis method is used to quantitatively analyse smart classrooms' impact on higher education students' academic performance, covering seven courses, such as foreign languages and computers, and 15 papers from four countries. The label encoding scheme is used to convert the categorical variables into numerical variables, and the linear regression model is constructed to analyse the influence of each variable on students' academic performance.Findings – The results show that smart classrooms positively impact the academic performance of higher education students with an impact value of 1.18, and the impact is more pronounced for operational courses such as dentistry and computer science.Implications – A smart classroom environment dramatically improves students' learning effect, which is significant to students' autonomous learning.Originality – This paper is based on an original study on the impact of smart classrooms in higher education in four countries.