Flooding represents a significant natural hazard in the United Kingdom, particularly in urban environments where rapid runoff increases risk. This study addresses four key research gaps identified in the literature. First, there is a lack of physically interpretable flood prediction models that incorporate warning lead-time assessment for the Kingston upon Thames sub-catchment. Second, the Hydrologic Engineering Center-Hydrologic Modelling System (HEC-HMS) has seldom been applied to temperate, urban catchments in Western Europe, particularly in south-east England. Third, calibration presents challenges when hydrological observations do not correspond to the scale of the modelled sub-catchment. Fourth, data-driven methods are challenging to implement and interpret in the absence of local observational data. The research comprised two main components. First, a systematic review following PRISMA guidelines screened 906 records and retained 54 peer-reviewed studies. This review identified the predominant modelling approaches for flood prediction and demonstrated that HEC-HMS is infrequently applied in comparable UK urban contexts. The findings also highlighted the importance of data sufficiency, spatial consistency, and model interpretability in selecting appropriate flood prediction methodologies. Second, a GIS-based hydrological case study was conducted for the 164.526 km² Kingston upon Thames sub-catchment. QGIS, SAGA GIS, and GRASS GIS were employed to derive hydrological data from LiDAR terrain, land cover, and geological information, which informed an HEC-HMS rainfall-runoff model. Model calibration yielded poor performance (NSE = −181.5; PBIAS = 1477%). Sensitivity analysis indicated that adjusting the Curve Number alone did not resolve the calibration issues. The primary challenge was the spatial mismatch between the modelled sub-catchment and the substantially larger 9,950 km² catchment represented by the available NRFA discharge data, complicating the calibration process. This study demonstrates the development of a reproducible GIS-based HEC-HMS framework and identifies its limitations in the absence of high-quality hydrological observations for the study area. The model’s estimated response timing indicates that hydrological modelling may support lead-time assessment; however, reliable early warning systems would require improved alignment of rainfall and discharge data, as well as further model calibration and validation.