A model involving meteorological factors for short- to medium-term, water-level predictions of small- and medium-sized urban rivers

With the increase in extreme weather, cities, especially those with small- and medium-sized urban rivers with protected areas smaller than 200 square hectares, are experiencing significantly more flood disasters worldwide. Heavy snowfall and rainfall can rapidly overflow these rivers and cause floods due to the unique geographic locations and fast runoff and confluence speeds of the rivers. Therefore, it is particularly important to accurately predict the short- to medium-term water levels of these rivers to reduce and avoid urban floods. In the present work, a particle swarm optimization (PSO)-support vector machine (SVM) water-level prediction model was constructed by combining PSO and SVM and trained with meteorological data from Wuhan, China, and water-level data from the Yangtze River. The PSO-SVM model is able to lower the mean square error (MSE) of the prediction results by 70.47% and increase the coefficient of determination (R-2) by 7.02% compared with the SVM model alone. The highly accurate PSO-SVM model can be used to predict river water levels in real time using hourly weather and water-level data, thereby providing quantitative data support for controlling urban floods, managing water project construction, improving response efficiency and reducing safety risks.