We present a Sensor Web browser with an efficient spatio-temporal data loading mechanism as a client-side application in GeoCENS project. The same way the World Wide Web needs a web browser to load and display web pages, the World Wide Sensor Web needs a Sensor Web browser to access distributed and heterogeneous sensor networks. However, most existing Sensor Web browsers are just mashups of sensor locations and base maps that do not consider the scalability issues regarding transmitting large amount of sensor readings over the Internet. While caching is an effective solution to alleviate transmission latency and bandwidth problems, a method for efficiently loading spatio-temporal sensor data from Sensor Web servers is currently missing. Therefore, we present LOST-Tree, a new spatio-temporal structure, intended to be the sensor data loading component on a Sensor Web browser. By applying LOST-Tree, redundant transmissions are avoided and consequently enables efficient loading with cached sensor data.