This paper presents a method for calculating reliability for captured images in mobile observation for natural environments. The need for natural environment observation is increasing to understand the global environment, predict environmental changes, and take measures. In our previous studies, we have designed and developed a prototype of a fast scene search system for massive video data streamed from environmental monitoring cameras on the digital earth platform. Checking the quality and reliability of observed data is very important process for subsequent appropriate analysis of observed data; however, such reliability is not always guaranteed in natural environment monitoring, although some methods have been proposed for evaluating the quality of images regarding blurry images or missing pixels. The feature of our method is to calculate the reliability score for each captured image in terms of temporal continuity, spatial continuity, and observed object validity. Our method makes it possible to calculate the reliability of observed images, even if the image appears to have no problems due to the reason that the observation target is in a normal state and the image quality is also normal. In the experiments using the prototype, we will confirm the feasibility of the proposed method.