Abstract: According to the external temperature characteristics of greenhouse crop growing season in Gaoyi County during 2008-2010, 2008, 2009 and 2010 were defined as the warmer, the colder and the normal year type, respectively. Based on the microclimate temperature observation data per 10min during the greenhouse crop growing season in Gaoyi County from 2008 to 2010, the arithmetic mean value calculated by every 10min observation data in a day was regarded as the actual mean temperature. The arithmetic mean values calculated by every 20min, 30min, 1 hour, 2 hours, 3 hours, 4 hours observation data respectively were regarded as the daily mean temperature. The mean absolute error(MAE), standard deviation(SD) and the accumulated temperature relative error(R) in the observation period between the arithmetic mean values calculated by every 20min, 30min, 1 hour, 2 hours, 3 hours, 4 hours observation data respectively and the actual daily mean temperature were analyzed and the significance was test by t-test. The suitable method and frequency of temperature observation in greenhouse was confirmed by the index of P<0.05. MAE, SD and R were applied for evaluating the calculation effect of these methods. The smaller was the MAE, SD and R, the better was the effect. The results showed, (1)when the observation interval was less than 4 hours, the daily average temperature calculated was not significantly different with the actual，MAE≤0. 19℃，SD≤0.24℃, R≤1.1%，so the method was suitable for greenhouse environment; Moreover, the shorter of the observation interval and the higher of the observation frequency, the better was the effect of the average temperature calculated. When the observation interval was 4 hours, the accuracy of the results has certain risks. (2)The daily mean temperature calculated by six compositions of four times observation data was significantly different with the actual, so the method should not be used in observation experiment. (3)The daily mean temperature calculated by the maximum and the minimum temperature was significantly higher than the actual value, therefore, the calculation method should not be used in observation experiment. (4)The daily mean temperature calculated by the improved method of using the 7：00, 10：00, 13：00, 16：00, 19：00 and 23：00 observation data was not significantly different with the actual，MAE≤0.24℃，SD≤0.30℃，R≤1.4%, and the observation time in night was also reduced, therefore, the method can be used in the actual production. (5)Overall, at least six times in a day observation, the daily mean temperature calculated may be representative of the actual mean temperature.

Key words: Solar greenhouse, Temperature, Observation frequency, The calculation effect, Paired t-test