The term "small wetland" refers to wetlands of small scale, which can stabilize ecological systems in the process of long-term development. Against the backdrop of rapid urbanization, many small wetlands have been disappearing, often in groups in cities. Understanding the dynamic changes and their driving factors in small wetlands provides an important basis for the effective protection and management of small wetlands. In this study, we took Baohe District in Hefei City as the research area and analyzed the landscape pattern characteristics by interpreting remote sensing data through time (2006, 2010, 2014, and 2018). The spatial analysis method was used to determine the dynamic changes in small wetland landscapes over four study periods of 2006-2010, 2010-2014, 2014-2018, and 2006-2018. Based on a 300 m×300 m grid unit, boosted regression tree and geographically weighted logistic regression models were employed to identify the relationships between 13 predictive variables and the loss of small wetlands. The results showed that the total area of small wetlands decreased by 60.8% and the number of patches decreased by 60.5% during the period from 2006 to 2018. At the same time, the extent and complexity of the perimeter of the small wetlands decreased, the spatial distance between the small wetlands increased, and the distribution tended towards more discrete. The boosted regression tree model quantified the relative influences of the predictive variables and determined high-importance variables to further analyze the nonlinear relationships between the variables and the decline in small wetlands. In the early stage of urbanization, changes in the surrounding land-use type were the major driving factors in the loss of small wetlands. In the middle and late stages of urbanization, both the trend in large-scale urban sprawl slowed down, and the relative importance of other land-use type changes also declined to varying extents. The driving factors of patch area and slope on the loss of small wetlands gradually increased in importance. Construction land changes (14.4%), patch area of small wetlands (13.5%), dryland changes (11.1%), slope (10.1%), forestland changes (8.5%), and grassland changes (7.0%) were high-importance variables of small wetland losses from 2006 to 2018. The local and spatial influences of these high-importance variables were analyzed further by geographically weighted logistic regression using coefficients determined at each sample point. With the exception that the spatial visualization of the small wetland patch areas had no explanatory significance, the influences of the remaining variables varied with location, and their contributions also differed in magnitude and direction. This study provides a reference for the protection and management of small wetlands in rapidly developing urban areas.