High-accuracy and high-resolution land-cover datasets are crucial for city planning and sustainable development. In recent decades, Shijiazhuang City has experienced significant land use/cover changes resulting from economic development, population growth, and urban expansion. However, few studies have been reported on land-cover datasets over Shijiazhuang City, which has a complex topography and a heterogeneous landscape. In this study, single- and multi-temporal Landsat images over Shijiazhuang City were classified by random forest, support vector machine, and classification and regression tree classifiers based on 382 field survey samples; their accuracies were assessed through a comparison with two other land-cover datasets (GlobeLand30-2020 and GLC_FCS30-2020). Land-cover dynamics from 1988 to 2020 and greening trends from 2000 to 2020 were determined. The results show that the classification of multi-temporal images with spectral and phenological characteristics using random forest classifiers achieved the highest overall accuracy of 86.4% in comparison with 69.6 and 47.5% for GlobeLand30-2020 and GLC_FCS30-2020, respectively. From 1988 to 2020, the impervious surfaces and deciduous broad-leaved forest regions in the study area expanded, while irrigated cropland and shrubland areas decreased gradually. From 2000 to 2020, the normalized difference vegetation index (NDVI) of natural vegetation types in urban and mountainous areas significantly increased (p < 0.05), while the greenness of the entire study area and irrigated cropland regions exhibited no significant changes. In this paper, we provide useful information for research into city land-cover classification and assessment, along with ecological environment protection and planning.