A lack of crop disaster datasets has limited the exploration of the influence of small-scale disasters on crops. Because disasters are often defined on the basis of human impact, disaster databases may underestimate the effect of disasters on crop production. Additionally, the resolution of such databases is insufficient for evaluating the effects of disasters on small areas. In this study, crop disaster and daily weather datasets covering the period from 2003 to 2022 in Taiwan were developed. Total 9,245 damage records from 233 observations of crop disasters were mined from the Report on Crop Production Loss Caused by Disasters of Taiwan. Daily weather data were collected from weather stations. Entire crop disaster information including multiple disasters, crops, and affected regions was stored in crop disaster dataset. All datasets were cleaned up and refined to enhance their quality, and characteristics such as disaster and crop classification were added to enhance the applicability of these datasets. These datasets can be used to determine the relationship between disaster type and crop production losses.