A derecho is a widespread, long-lived straight-line windstorm associated with a fast-moving mesoscale convective system. Due to their production of persistent extreme winds, derechos pose significant threats to human safety and property. It is thus critical to understanding the characteristics and underlying mechanisms of derechos and how climate change would affect them. By leveraging extensive observational datasets and cutting-edge deep learning techniques, we develop a novel high-resolution (4 km, hourly) derecho dataset that provides the detailed tracking and properties of derechos in the central-eastern United States from 2004 to 2021. Compared to previous studies focusing on derechos for earlier periods, our dataset has more derechos (31 vs. 6-21 per year), but similar derecho spatial distribution and seasonal variation patterns. The dataset indicates that derechos contribute ~20% of damaging gust reports in the central-eastern United States. Using self-organization map analysis, we distinguish six types of large-scale environments associated with derechos from the dataset. Warm-season derechos are accompanied by large-scale environments significantly different from cold-season ones. Using regional and global climate models with convection-permitting resolutions, we investigate the potential impacts of global warming on several selected derecho cases between 2012 and 2022 using the pseudo-global warming framework. Results show distinct responses of derechos to future warming: a derecho can become stronger or weaker or display no obvious changes under global warming.