Among key component threatening global biodiversity in future such as landuse alterations, climate changes can be primary stressors on stream ecosystem. Although many studies in the discipline of landscape genetics have tried to explore genetic consequences of oncoming climate change, most of them used anonymous genes even though majority of these genetic regions are just known to be “neutral” (no relation) to natural selection. In this study, we developed a model for projecting adaptive genetic shift of a stream caddisfly species under climate change scenarios, using a distributed hydrothermal model and 8 global climate models (GCM). We used non-neutral AFLP (Amplified Fragment Length Polymorphism) loci, which were statistically defined to be subject to natural selection, in the target species found at 45 sampling localities. We acquired bias-corrected air temperature and precipitation data of 3 RCP scenarios from the GCMs. Then, these climatic data were used to calculate annual metrics of current velocity, water depth and water temperature using the hydrothermal model. Models describing adaptive genetic variation, which is represented as linear response of allele frequency at each locus to environmental predictors, have been constructed based on current hydrothermal variables in a previous work. We used the model for estimating future ranges of the allele frequency along changing climate gradients. Finally, we projected spatial pattern of genetic diversity (heterozygosity) for the studied caddisfly species. Degraded pattern of genetic diversity was more notable under extreme climate scenarios when comparing with in current climate, depending on higher air temperature projected in the future.