Abstract: Synoptic weather background for Hainan was similar on 5-6 June 2016. However, the severe weather on 5 June was characterized by large gust wind and EF2 tornado, whereas flash flood dominated the next day (6 June). In order to study the causes for different types of strong convective weather in two consecutive days, conventional soundings and surface observations, intensive surface observations at 10-min intervals, Haikou Doppler weather radar observations and NCEP-GFS analysis data are used. The results are as follows. (1) The sounding at Haikou on 5 June shows that the relative humidity was 41%, the precipitable water (PW) was 49 mm and the lapse rate was 7.25℃/km in middle levels. This condition was favorable for the formation of strong downdraft and cold pool. On 6 June, the sounding shows that the atmosphere was moist and the environmental wind was weak. Thereby the outflow and the cold pool both were weak and the storm system moved slowly, leading to heavy rainfall. (2) The convections in these two days both occurred under weak synoptic circulation background. The difference is that the vertical wind shear from 0 to 3 km was stronger on 5 June, which contributed to the formation of the squall line. (3) Structural analysis shows that on 5 June, the convective storm was accompanied with quite strong gust outflow, fairly vertical wind shears and gust outflows from many single convective cells that merged and lifted together, which eventually led to the formation of squall line and bow echoes that lasted for 1.5 hours. However, the convective cell on 6 June was generally a conventional cell with lower center and weaker gust outflow, and its lasting time was equal to the duration of a general cell, even though many cells merged to form a quasilinear storm. (4) CIN in the first day was larger than that in the next day. Thereby stronger low level lifting is needed to trigger convection, storm only triggered and developed near the strong sea-breeze convergence line while the next day convection was triggered here and there due to high humidity at low level hereafter without CIN. (5) It was hard to predict the first day's tornado based on the environmental parameters for they were not similar with significant supercell tornado. It's impossible to issue tornado warning for the tornado occurred 3 minutes after the mesocyclone formed.