Abstract The conservation of sweetfish (Plecoglossus altivelis) populations is important for river ecosystems in Japan from the viewpoint of fisheries resource conservation and fish community conservation. Despite this importance, only limited methods (e.g., observation and diving surveys) are used for surveying and evaluating sweetfish populations, which has limited the progress made by Japanese studies. Such studies have clarified the behavior of sweetfish during the day and in the short term, but they have not clarified the behavior at night and in the long term. We therefore aimed to quantitatively evaluate the behavior of sweetfish at night and in the long term by using individual-based models (IBMs). First, we developed an advanced telemetry system (ATS) that could continuously track the long-term behavior of a wild animal to which a radio transmitter was attached. We used the ATS to conduct a field survey and track the fish's behavior during the day and at night for 14 days. We analyzed basic characteristics such as home range, which averaged 6000m2. Sweetfish used mainly deep water and areas with slow currents. Seven days after release, the fish had moved upstream, and their home ranges were similar to those assessed previously. Second, we analyzed the basic spatial preferences of sweetfish (e.g., current velocity and water depth). Current velocity and water depth were calculated by hydraulic simulation. The fish used not only areas with high current velocity and shallow water but also those with low current velocity and deep water. These behavioral characteristics and spatial preferences have not been reported before. Finally, we used IBMs to model sweetfish behavior by inputting data from this study and our previous studies. The IBMs represented long-term behavioral characteristics. Our ATS data were essential for accurate behavioral representation by the IBMs. By using these approaches, we developed basic evaluation tools for species conservation. Our results indicated that interactions between measurement and simulation are needed to develop IBMs for evaluation and prediction aimed at conserving sweetfish populations.