It is widely believed that an experience in a forest environment can induce positive emotions for people. Current evidence that supports this argument mostly come from studies using self-reported scores on questionnaires as a source of data. As respondents have to be informed in advance, reported emotional scores can partially result from a participant's study awareness. To overcome this potential bias, a crossover study was conducted in Heilongjiang Forest Botanical Garden (forest) and the Central Street (urban) over two days of early summer at Harbin city, Northeast China. A total of 8792 photos with spontaneous facial expressions showing happy and sad emotions were randomly taken for pedestrians during the day. Young and middle-aged pedestrians in forests generally showed a higher level of least-square (LS) means of happy scores than those in urban settings. The dynamic expression of LS means of happy and sad scores both showed a decreasing trend. During the afternoon (14:00 16:00 pm), happy scores in urban settings suddenly declined and were significantly different from the forest setting. Regression analysis using the zero-inflated negative binomial model indicated that only in the morning (09:00 10:00 am) did the forest setting experience induce a positive effect on happy expressions in middle-aged females. Overall, our results revealed that facial expression scores will decline with time for both positive and negative emotions, but the decline can be mediated by the forest environment than urban settings. The higher level of positive expression scores for forest experiencers was the result of at least the sharper decline of positive emotion of people in the promenade, instead of the solely effect by an urban forest experience.