General circulation models (GCMs) that can simulate global climate are used to predict climate changes caused by an increase in atmospheric CO₂ concentration. However, the spatial resolutions of currently running models are rough with a resolution of about 3° to 6° in latitude/longitude. Thus, reducing the relatively large-scale climatic states that the GCM provides to smaller-scale ones is required to evaluate impacts of climate changes on agriculture and natural ecosystems at local and regional scales. We constructed a dataset, namely the mesh climate change data of Japan, using the inverse distance weighted interpolation against coupled atmosphere-ocean GCMs’ (A-O GCMs) experiment results under gradually increasing atmospheric CO₂ concentration. The A-O GCMs used in this article are ECHAM4/OPYC3 (Germany), CGCM1 (Canada), CSIRO-Mk2 (Australia), and CCSR/NIES (Japan). The dataset gives tha average climate change scenarios in Japan for every 10-year period over the next 100 years with a resolution of 7.5′ in longitude and 5′ in latitude (approximately 10 by 10 km). This article describes the construction method and contents of the dataset. In order to demonstrate the characteristics of the dataset, we examined the transient changes in spatial distribution of accumulated surface air temperature, accumulated precipitation, and mean short wave radiation during the months of May through September, corresponding with the major crop cultivation period in Japan.