Soil temperature is one important environmental factor affecting soil formation and plant growth, especially in the alpine environment. The studies on plateau soil temperature may help researchers understand how soil heat is conducted during the freezing and thawing process as well as how fragile the ecosystem of an alpine region is. However, so far the studies on soil temperature of alpine regions are not so helpful and instead affect proper characterization of the variation of soil temperature, because they are often conducted in fields different in vegetation, topography, latitude and longitude, etc. This study was laid out in a typical shrub meadow area of the Lhasa River Valley on a mountain slope uniform in natural conditions, including vegetation type, slope degree and aspect. The slope is covered dominantly with Rhododendron primuliflorum, making the total vegetation coverage up to 82%. The soil on the slope is of the type of sub-alpine shrub meadow soil, slightly acidic. Nine monitoring points were distributed over the slope from elevation of 4 000 m to 4 800 m with a gradient of 100 m; and in each point 2 temperature recorders (The Onset HOBO Company of USA; The type is U23-003; Operation range Internal sensors: -40°C to 100 °C) were placed, each with 4 temperature sensors placed at the depth of 5, 10, 20and 30 cm, separately. Data were collected once an hour from October 5, 2013 to September 15, 2014. As the monitoring points at 4 000 m, 4 400 m and 4 700 m were damaged owning to unknown causes, only the data collected from the points at 4 100, 4 200, 4 300, 4 500, 4 600 and 4 800 m were analyzed with the typical statistical method for characteristics of the variation of soil temperature with elevation and soil depth gradients. Results show that (1) in a year, the daily mean soil temperature within the 0 ~ 30 cm soil layer followed a cosine function curve, fluctuating within the range from -9.05 °C to 14.21 °C and averaging 2.94 °C; it rose at a rate of 0.11 °C d^-1 and fell at a rate of -0.19 °C d^-1, and it displayed a rising trend for 106 days and a declining trend for 73 days, and remained frozen for 147 days in a year; (2) the soil temperature followed a quasi-sine curve in daily variation, fluctuating within the range of 2.8 °C; it varied sharper in summer than in fall; the rising trend of soil temperature lasted shorter than the declining trend did: and especially in winter, the former lasted for only 4 hours, the shortest among the four seasons; but in summer it did for 8 hours and peaked at 17:00 and the soil temperature bottomed at 10:00 within a day; (3) the annual mean soil temperature decreased with rising elevation, at a rate of -0.63 °C (100 m^-1), and the phenomenon was more obvious in summer than in the other seasons, with the rate being -0.76 °C (100 m^-1) and the least in fall, being only -0.37 °C (100 m^-1); variation coefficient of the soil temperature increased with rising elevation, which indicates that soil temperature varies sharply and intricately with elevation; (4) the annual mean soil temperature was a power function of soil depth; with increasing depth, soil temperature varied less and less, the difference between soil layers in temperature became narrower and narrower, and the occurrence of peak and bottom values was delayed; during the temperature rising period, soil heat transferred downward, while during the temperature declining period, it did reversely; during the period when the soil was frozen, variation of soil temperature tended to be uniform regardless of soil depth; however, freeze-up of the top soil layer lasted the longest, for about 149 days; and soil heat transfer at 20cm in depth was relatively steady during the freezing-thawing process; and (5) in the sight of diurnal variation of soil temperature relative to soil depth, soil temperature descended in variation range and peak value with increasing soil depth, but ascended in bottom value; and in soil layers below 10cm, the occurrence of peak and bottom values of soil temperature was delayed by 2 h (10 cm^-1).