Evaporation from impervious surfaces plays a vital role in the catchment water cycle. Exploring the spatiotemporal variation patterns and influencing mechanisms of impervious surface evaporation at the catchment scale can improve the understanding and evaluation of the evaporation process. This study downloaded 0.5 m resolution images of the Baiyangdian catchment (BYD) from Google Earth and used deep learning to identify impervious surfaces. This was used to revise impervious surfaces of the China land cover dataset in 1985 and 1990–2020. Potential evaporation (PET) from three types of impervious surfaces (roofs, ground affected or not affected by the building height) was calculated by modifying the parameters of the Penman–Monteith equation, and daily precipitation and water-storage capacity of impervious surfaces were taken into account to estimate impervious surface evaporation (E). The results showed that E values of the three types of impervious surfaces were between 72.1 and 178.2 mm/year and all exhibited the spatial distribution of high in the northwest and low in the southeast of the BYD in 1980–2020. Compared with that in 1980, in 2020, the cumulative evaporation (EAP) increased by 134.4%. The ratio of EAP to cumulative precipitation ranged from 3.0% to 6.9%, increasing significantly in a fluctuating manner. The increments in precipitation days and impervious surface area played a major role in the increase of EAP, and the decrease in precipitation was the fundamental reason for the increase in the proportion of impervious surface evaporation and water resource pressure in the BYD. Excluding the continuous evaporation from the remaining water can likely lead to underestimating the impervious surface evaporation. This study provides an efficient and reasonable novel approach for calculating impervious surface evaporation in long series and large-scale catchments. 相似文献