It has become easy to accumulate and to deliver scientific data by the evolution of computer technologies. The GEO Grid project has collected global satellite images from 2000 to present, and the amount of the collection is about 150 TB. It is required to generate new values by integrating satellite images with heterogeneous information such as Web contents or geographical data. Using GEO Grid satellite images, some researches detect feature changes such as earthquakes, fires and newly constructed building. In this paper, detections of feature changes from time series satellite image are referred to as events, and we focus on events about newly constructed buildings. Usually, there are articles about such newly constructed buildings on the Web. For example, a newly started shopping center is usually introduced in a news report, and a newly constructed apartment is often on the lips of neighboring residents. So, we propose an event detection system that extracts candidate events from satellite images, collects information about them from the Web, and integrates them. This system consists of an event detection module and a Web contents collection module. The event detection module detects geographical regions that have differences with elevation values between two satellite images which are temporally different. The expressions of regions are translated from latitudes/longitudes to building names by using an inverse geocoder. Then, the contents collection module collects Web pages by querying names of buildings to a search engine. The collected pages are re-ranked based on temporal information which is close to event occurrence time. We developed a prototype system. The result of evaluation showed that the system detected some information of building construction events with appropriate web contents in Tsukuba, Japan.