Convection velocity of temperature fluctuations in a natural convection boundary layer

•For the first time, two-point temperature measurements are carried out in a turbulent natural convection boundary layer.•The two-point correlations of temperature fuctuations revealed that the thermal structures have a uniform convection speed in the first half of the thermal boundary layer and the convection velocity scales as Ra1/4, where Ra represents the Rayleigh number.•Shadowgraph measurements are in excellent agreement with the results from the two point measurements and provided a visualisation of the turbulent structures in the flow.•A physical interpretation is proposed based on the experimental observations, which is consistent with the physical model put forth by Wells and Worster JFM 609 111–137 (2008). Measurements of instantaneous temperature are carried out in a natural convection boundary layer (NCBL) developing over a vertical hot plate using a combination of an infra-red sensor, three cold-wire probes and a Resistance Temperature device (RTD). The aim of this study is to establish experimental procedures for high spatial and temporal resolution temperature measurements, and use that information to measure the convective speed of thermal structures in the flow. The evolution of NCBL along the hot plate is compared against existing empirical models. Statistics of mean and fluctuating temperature are in excellent agreement with the previously reported experimental data for a turbulent NCBL. The boundary layer thickness is observed to grow as y0.45, where y is the vertical distance along the heated plate. Separate experiments are conducted by varying the heat input to investigate its effect on convection velocity (Uc) of thermal structures in the flow. It is found that Uc remained nominally constant in the region 0 ≤ x ≤ δT/2 (δT is the thickness of thermal boundary layer) and decreased in a log-linear fashion in the outer region of NCBL. The results of Uc are found to be consistent with the shadowgraph measurements carried out in the same test rig, and the normalised convection velocity, Uc*=UcδTν is found to depend on Rayleigh number as Uc*∝Ra0.82, where ν is the kinematic viscosity of air. Importantly, these experimental observations are in agreement with the three layer model put forth by Wells and Worster (2008).