JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 92, NO. C5, PAGES 5489-5495, MAY 15, 1987 Convection Lines on a Heated Bottom Plate at Large Rayleigh Number NOBUYUKI TAMAI AND TAKASHI ASAEDA Department of Civil Engineering, University of Tokyo, Japan The microstructureof thermal convection at large Rayleigh number was studied experimentally. Sheetlike plumes generated at the heated bottom plate are composed of severalthermals generated successively. The projection of the sheetlike plumeson the bottom plate is called convection lines.On the other hand,in the area surrounded by sheetlike plumes wheredescending compensational flow prevails, the conduction layer always remains in the vicinityof the bottom,within whichexcess temperature is still high. The heat supplied from the bottom is transported horizontallyby weak flow in the conduction layer toward the convection lines where rising sheetlike plumescarry the converged heat into the main bodyof water.Fluctuation of the convection lines wasinvestigated. It is foundthat phenomenalogically, the merging of thermals causes much stronger sheetlike plumes and accelerates the formationof large- scale convection. INTRODUCTION Previous studies on the structure of thermal convection at large Rayleigh number are classified into two types. One type is concerned with the initial stage, where the convection just formed in still water after the onset of heating is studied. In another category the characteristicsof the steady state are studied. In both types of studies, it is observed that many plumes are generated near the heated bottom and that they maintain the convection. As for the starting process of convection, Foster [1965, 1969] measured the temperature distribution near the heated bottom or the cooled surface and the time required for the generation of convection. Similar studies wereperformed by Mollendorf and Ajiniran [1984] and Tamai and Asaeda [1984a] (this paper will be referred to hereafteras paper 1). These studies confirmed that the first appearance of plumesis determinedby the collapseof the conductionlayer formed beforehand on the heated bottom. The characteristics of the plumes at this initial stageare relatively clear, sincethe flow is originally quiescentand the formation of the plumes can be visualized clearly. The structure of the plumes generated after sufficient time elapsesfrom the onset of convection has not been understood well up to now, sincethey are formed in a complicatedturbu- lent flow. Howard [1964] proposed a theoretical model of the generation of plumes in the turbulent stagebased on the con- cept of the repeated collapse of the conduction layer. Howard's model is confirmed with good agreement in the frequency of the generation of plumesfor this stage[Sparrow et al., 1970]. Spangenberg and Rowland [1961] visualized the flow struc- ture near the cooling surface due to evaporation and found that a plume is generated from a plunging line distributed over the surfaceand has a sheetlikeshape. Later, Katsaros et al. [1977] and Katsaros [1978, 1980] measured the horizontal temperature and found that the number of plumes at some level decreased with the distance from the surface. Krishnamufti and Howard [1981] visualized plumes in a large-scalecellular convection by the bottom heating experi- ment and found that the plumes have an important role in the Copyright 1987 by the American Geophysical Union. Paper number 6C0597. 0148-0227/87/006C-0597505.00 formation of the large-scaleconvection. A similar plume was also found in the experiments by Fanaki [1971] and Krishna- murti and Howard [1981]. However, the structure and the role of that plume have not yet been sufficientlyclarified. In a previous paper [Tamai and Asaeda, 1984b] (this paper will be referred to hereafter as paper 2) the authors reported that the plume has a double structure; that is, a sheetlike plume is composedof several thermals generated by the col- lapse of the conduction layer. However, the fluctuation of the positions of thermal generation has an important role in the formation of large-scale convection as well as the propertiesof turbulence of the convection. Hence the purpose of this study is to clarify the microstructure of heat transport by the sheet- like plume and the fluctuation of the position of generation of the thermals. EXPERIMENTAL PROCEDURES Details of the experimental apparatus have been described in paper 2. Briefly, a heat-insulated tank 90.0 cm by 90.0 cm in section and 70.0 cm deep was utilized. The bottom is made of aluminum plate which is capable of homogeneously supplying a constant heat flux from below. The temperature was mea- sured by means of four sets of thermistor linearizers located at heights of 2.0, 4.0, and 7.0 mm and at middepth. In order to avoid disturbancecreated by a lower sensor, the upper sensor was displaced by about 1 mm horizontally from the nearest lower sensor. Tap water with a depth of 10 cm usually and at most 15 cm was used in experiments,so that the width/depth ratio was kept at more than 6 and in most cases 9. The flow structure and the velocity fields were obtained by using sus- pended aluminum particles of 10-3 cm diameter projected by a horizontal or vertical slit beam of light. The water temper- ature was set between 10øC and 25øC, which resulted in the Prandtl number Pr with values from 8.7 to 6.0. Since the experimentswere performed under conditions of constant heat flux from the bottom, the following heat-flux-typeRayleigh number is defined as o•Fgd'* n• - (•) pctc2v where F is the heat flux from the bottom; c• the thermal ex- pansioncoefficient; g the gravitational acceleration; p the den- sity of water; c the specific heat of water; v the kinematic viscosity of water; tcthe thermametricdiffusivity of water; and d the water depth. 5489