Quantitative evaluation of plant evapotranspiration effect for green roof in tropical area: A case study in Singapore

The efficacy of existing evapotranspiration (ET) models commonly used for hourly thermal performance simulation of green roof remains unclear, especially for the tropical climate. To address the issue, field experiment was conducted to quantify the ET rate of four plant species in Singapore. Results showed that the average daytime rate of ET ranged from 198.4 g m(-2) h(-1) to 320 g m(-2) h(-1), while the average nighttime rate of ET ranged from 18.7 g m(-2) h(-1) to 25.5 g m(-2) h(-1). The percentage of daytime ET accounting for solar radiation ranged from 51.4% to 62.7%. The hourly ET rate predicted by three types of physical models were compared to the measured ET rate. It was found that the water vapor diffusion model had the best prediction performance, while the energy balance model had the worst prediction performance. Considering the complexities of the water vapor diffusion model, fifteen artificial neural network (ANN) models using multi-layer perceptron regressor were developed and evaluated. It was found that the ANN models had a better average prediction performance compared to water vapor diffusion models. Conclusions drawn from this study could provide a reference for accurate modelling of thermal and hydrological performance of green roof in tropical area. (C) 2021 Elsevier B.V. All rights reserved.

Automated summary

The efficacy of existing evapotranspiration (ET) models for green roof thermal performance simulation in tropical climates remains unclear, prompting a field experiment in Singapore to quantify ET rates of four plant species. The study found varying daytime and nighttime ET rates for different plants, with artificial neural network (ANN) models demonstrating better prediction performance compared to water vapor diffusion models. These findings can guide more accurate modeling of thermal and hydrological performance of green roofs in tropical regions.