Nat Hazards (2012) 63:305–323 DOI 10.1007/s11069-012-0153-1 ORIGINAL PAPER An evaluation of the impacts of land surface modification, storm sewer development, and rainfall variation on waterlogging risk in Shanghai Xiaodan Wu • Dapeng Yu • Zhongyuan Chen • Robert L. Wilby Received: 7 November 2011 / Accepted: 15 March 2012 / Published online: 3 April 2012 Ó Springer Science+Business Media B.V. 2012 Abstract Despite continuing efforts to upgrade the urban storm sewer system since the late 1950s, the City of Shanghai is still vulnerable to persistent rainstorm waterlogging due to excess surface runoff and sewer surcharge, which frequently cause significant damage to buildings and disruption to traffic. Rapid urbanization and associated land cover changes are the major factors contributing to waterlogging. However, it is unclear to what extent changes in rainfall variability over the past few decades are also involved. This paper investigates the combined impacts of land use and land cover change, storm sewer development, and long-term variations in precipitation. Evidence of persistent waterlog- ging is presented first. We then give an account of land surface modifications during the process of urbanization and the development of the city’s urban storm sewer system. Statistical analysis suggests that the increase in runoff coefficient due to conversion of lands from agricultural to industrial, commercial, and residential uses is a major factor driving greater waterlogging risk. In particular, historical analysis of aerial photographs reveals the rate and extent of modification to river networks in the past few decades. The natural drainage network has shrunk by 270 km, significantly reducing the city’s capacity to transport excess surface flow. In line with other studies, we find no significant overall trends in annual rainfall totals (at Baoshan and Xujiahui). However, seasonal and monthly rainfall intensities have increased. At the daily scale, we find that compared to pre-1980s: (i) there has been an increase in the number of wet days with precipitation exceeding 25 mm (Heavy Rainfall) and decrease in those below 25 mm and (ii) the number of consecutive wet days with precipitation maximum and average exceeding the threshold known to cause waterlogging shows an increasing trend. Since rainfall intensity is expected X. Wu Department of Geography, East China Normal University, Shanghai 200062, China D. Yu (&) Á R. L. Wilby Department of Geography, Centre for Hydrological and Ecosystem Science, Loughborough University, Leicestershire LE11 3TU, UK e-mail: [email protected] Z. Chen State Key Laboratory of Estuarine and Coastal Research, East China Normal University, Shanghai 200062, China 123 306 Nat Hazards (2012) 63:305–323 to increase under climate change, this could further compound the impacts of land use changes and place even greater pressure on the existing storm sewer system. Keywords Waterlogging Á Flood risk Á Urban drainage Á Land use change Á Precipitation extremes Á Climate change 1 Introduction Rainstorm waterlogging due to excess surface runoff and insufficient capacity of storm sewer systems is one of the most frequent and serious natural hazards in many big cities around the world, especially those subject to frequent torrential rainfall. Such events can pose problems for cities undergoing rapid urbanization where land surface characteristics have been modified in ways that favour increased surface runoff. The volume and flow rate may exceed the capacity of existing storm sewer systems, leading to more frequent sur- charging, surface flooding, property damage, and traffic disruption (e.g. Butler et al. 2007). Although inundation depths brought about by waterlogging are generally shallower compared to those caused by fluvial or coastal floods, the damage and disruption can be substantial. It is not only during urbanization that a city could be adversely affected by waterlogging. Rainstorm waterlogging may also pose widespread problems for developed cities with well-established sewer systems. For example, it was reported that the majority of the 2007 floods in the UK originated from overloaded rainstorm sewer in developed areas, and among those buildings affected, about 25 % were built during the last 25 years (Pitt 2008). Recent research on urban waterlogging has focused on three aspects: (i) assessing waterlogging risks, particularly in terms of damages to buildings (e.g. Huang et al. 2008; Shi et al. 2010), and in some cases disruption to traffic (e.g. Gao et al. 2009); (ii) modelling the extent and magnitude of waterlogging either within a GIS environment (e.g. Boyle et al. 1998; Zhang et al. 2005; Chen et al. 2009) or using 1D–2D coupled numerical models (e.g. Hsu et al. 2000; Mark et al. 2004; Schmitta et al. 2004); or (iii) evaluating impacts of land use and land cover change on waterlogging risks (e.g. Quan et al. 2010), sometimes in combination with downscaled climate change scenarios (e.g. Xia 1990; Zhu 1990, 1999; Deng 1998; Jin 2000). Some risk assessments have focused on the impacts of waterlogging on traffic disrup- tion and damage to buildings. For example, Shi et al. (2010) investigated the potential impacts of rainstorm waterlogging on old-style residence exposure risks in Shanghai based on distributed water depths derived from a series of scenario-based simulations using a waterlogging simulation model. Han et al. (2006) assessed the risks of storm waterlogging on traffic in the City of Tianjin using a coupled 1D/2D unsteady flow model partitioned on an irregular grid. These studies demonstrate the potential magnitude of waterlogging damage to buildings and traffic infrastructure. The advent of computational methods for simulating both subsurface pipe flow within the urban storm sewer system and runoff on the urban surface has enabled research into the extent of waterlogging (e.g. Hsu et al. 2000; Yin et al. 2006; Dong and Lu 2008). Studies typically couple a 1D model of flow through storm sewer system and 2D model of surface inundation, with a treatment of flow exchange at the common boundary (such as man- holes). For example, Hsu et al. (2000) used a 1D model of drainage flow (SWWM) and a 2D diffusion-based solution of the Saint-Venant equations for surface runoff, taking into account pumping stations at the outlets of the sewer system. The model was verified 123 Nat Hazards (2012) 63:305–323 307 against inundation observations collected during a 28-h rainstorm event in 1998 over the City of Taipei, simulated with recorded rainfall data and a 120-m resolution DEM. The model was subsequently used to predict waterlogging risks based on precipitation sce- narios. Other studies have also been undertaken to investigate the causes of waterlogging. For example, land use changes associated with urbanization are often considered to be one of the major contributory factors (e.g. Veldkamp and Verburg 2004; Quan et al. 2010). There has been relatively little research into interactions between land surface modi- fication, development of urban storm sewer systems, variation of local rainfall regimes, and waterlogging characteristics at the city scale. In particular, precipitation is thought to be the major factor affecting waterlogging in urban locations as it is the primary source of surface runoff in the absence of fluvial or tidal surges. Impacts of precipitation on waterlogging have been predominately investigated via frequency analysis, with synthetic rainfall hy- etographs (generated from historical data) linked to indicators of waterlogging. This is especially important since rising concentrations of greenhouse gases are expected to increase the intensity of precipitation at the regional level (IPCC 2007). Over the last few decades, numerous studies have been carried out to evaluate the impacts of global and regional climate warming on precipitation regimes (Zhao et al. 2009; Kioutsioukis et al. 2010). However, most have focused on relatively large regions and timescales that are unlikely to reveal potential links between precipitation and local waterlogging. Therefore, in the context of climate variability and change, there is a need to investigate the rela- tionship between daily to multi-day precipitation characteristics and occurrence of waterlogging. High-intensity precipitation events have become more frequent in South China (Zhai et al. 2005). Shanghai, a major city in the region, has experienced frequent waterlogging problems, and there is growing public awareness of the associated damage and disruption. Hence, the authorities are keen to better understand the underlying drivers. This paper evaluates three interrelated factors that affect waterlogging in the City of Shanghai, namely land surface modification, changes in the rainfall regime, and development of the drainage system during the course of urbanization over the last six decades. Section 2 describes the study site, data availability, and methods of data processing and analysis. This is followed by the results and discussion in Sect. 3. Conclusions are drawn, and research questions for future studies are raised in Sect. 4. 2 Methodology 2.1 Study site The City of Shanghai has developed on the floodplain of the Huangpu River, which originally comprised of many tributaries. The city has undergone rapid development during the past few decades and is now the most populated in China. The administrative divisions of Shanghai are shown in Fig. 1a, with the inner-city districts shown in Fig. 1b. Lying on the west coast of the East China Sea, Shanghai has a northern subtropical monsoon climate, with an annual average precipitation of 1,122 mm and an annual mean temperature of 15.8°C. With the influence of the East Asian monsoon and frequent typhoons plus high flows from the upstream catchments of the Huangpu River and storm surges from the East China Sea, the city is vulnerable to coastal and fluvial flood risk. With the construction of flood defences along the coast (Fig. 2a) and major watercourses 123 308 Nat Hazards (2012) 63:305–323 Fig. 1 a Location of the study area; and b extent and magnitude of waterlogging after the 10 June 1999 rainfall event in the inner-city area Fig.