ANALYZING URBAN FLOOD DISASTERS IN EMERGING MEGACITIES USING EARTH OBSERVATIONS

.................................................................................................................. ii ACKNOWLEDGMENTS ............................................................................................ iii PREFACE ..................................................................................................................... iv TABLE OF CONTENTS ............................................................................................... v LIST OF TABLES ....................................................................................................... vii LIST OF FIGURES .................................................................................................... viii CHAPTER 1 .................................................................................................................. 1 CHAPTER 2 .................................................................................................................. 3 Abstract .......................................................................................................................... 4

Economic betterment is the root cause of the global phenomenon of rural to urban migration (Cohen, 2003). This mass movement of populations has been a vital part of the urbanization process from ancient times and continues till now (Lall, Selod, & Shalizi, 2006). Accepting the challenge of accommodating the increased population as part of the global urbanization processcities are turning to megacities. A strong association is seen between population growth and land cover change (Dewan & Yamaguchi, 2009) while this urban expansion and growth lacks proper planning in necessary infrastructural development (Akanda & Hossain, 2012).
Urbanization is causing drastic changes in city layout and water infrastructures that are less resilient to natural disasters such as urban flooding. Land use land cover change (K. A. Aderogba, 2012;K. Aderogba, Oredipe, Oderinde, & Afelumo, 2012;Odunuga, 2008), population growth, topography of an area, alteration in precipitation pattern and intensity, inadequate urban planning, and arbitrary solid waste disposal (Adeloye & Rustum, 2011;Lamond, Bhattacharya, & Bloch, 2012) are some of the reasons behind urban flooding. Among those, we selected two major factors, one anthropogenic and one natural, behind chronic urban flooding in emerging megacities: (i) land use land cover changeanthropogenic factor and (ii) changes of intensity and frequency of precipitationnatural factor.
The objective of this MS thesis is to uncover and understand the relative importance between (a) land use land cover change and (b) intensity and frequency change of precipitation for urban flooding. We found that LULC change is a principal driving 1 Introduction

Background
Urban floods are an increasingly frequent and damaging environmental disasters across the globe. Due to high population growth, rapid urbanization and unplanned expansion is continuing in many regions of the planet and resulting in changes in land cover and land use (Dewan & Yamaguchi, 2009). In many developing regions of the world, this urban expansion and development lacks proper urban and regional planning and has led to large concentrations of substandard housing settlements with inadequate water, sanitation, and drainage infrastructure (Akanda & Hossain, 2012).
As a result, a large portion of the world's urban dwellers has become vulnerable to natural disasters, especially during floods.
Changes in land use in many of these emerging megacities have exasperated hydrological processes and resulting flood events. Hydrological modifications from increased urbanization impact infiltration and evaporation at both temporal and spatial scales (Ali, Khan, Aslam, & Khan, 2011). Hence, runoff generation and flow patterns are altered, resulting in changes in the recurrence and severity of flooding (Ali et al., 2011). An increase in the volume of rainwater runoff and a decrease in natural storm water retention areas are also the consequences of illegal encroachment and development of catchment areas. Lack of proper solid waste management and illegal dumping of bigger populations also decrease the drainage capacity of natural canals.
Land subsidence is also increasing at alarming rates in many megacities due to the unplanned extraction of groundwater (The World Bank, 2011). The gravitational capacity of natural drainage channels is hampered by land subsidence, which has 6 added more vulnerability to flooding and increased the risk of coastal flooding as well.
The disaster risks in coastal cities are thus much greater when above factors are combined with sea-level rise and intense rainfall (The World Bank, 2011).
Change of land use is not only a physical process of transforming one land use to another but also is linked to the alteration of the social, political, economic, and cultural orientation of any society (Pangaribowo, 2018). Conversion of agricultural land to non-agricultural land is increasing to meet the land and housing requirements of growing populations, which has an impact on economy, society, and environment as well. Socio-economic factors, i.e., higher land price near the urban areas, the opportunity of diverse livelihoods, and chances of high income in urban areas, subsequently, have an impact on the land use change processes (Larasati & Hariyanto, 2018).
In this study, Houston Metropolitan Statistical Area, Texas, Mexico City, Mexico, Jakarta, Indonesia and Dhaka, Bangladesh have been chosen as example cities of urbanization and associated hydrological and land cover changes. These four major cities are all prone to chronic urban flooding problems, but each is chosen from four difference economic groups: a developed (United States), upper middle income (Mexico), and lower middle-income (Indonesia), and a recently graduated lower middle income from a least developed economy (Bangladesh). These countries have been selected to effectively compare the evolution of these trends and correlate the changes in each city's individual development contexts. 7

Objectives
Among many responsible driving factors, an area's topography, changes in land use and land cover (K. A. Aderogba, 2012;Odunuga, 2008), changes in precipitation intensity and frequency, urbanization and population growth, defective urban planning, and arbitrary solid waste disposal (Adeloye & Rustum, 2011;Lamond et al., 2012) are some important factors behind urban flooding. Among these, we investigate two major driving factors, one natural and one anthropogenic, behind the evolution of urban flooding in this study: (i) land use and land cover change, and (ii) intensity and frequency of precipitation. The goal of the study is to assess the relative importance to anthropogenic (land use land cover change) and natural (precipitation frequency and intensity) to urban flooding vulnerability and determine the strength and role of these drivers in the context of the four growing regions. The study covers the time period from 1979 to 2017, with a two-decade period  of overlapping availability of ground and Earth Observations (EO) of precipitation and land use land cover data.

Study Area
In the United States, flooding is regarded as the number one among all natural disasters in terms of frequency as 28 out of 60 natural disasters were flood related between year 1980 to 2004 (Fang, Safiolea, & Bedient, 2006). Houston, Texas, ranked as the fifth largest metro area in the U.S. with a population over 7 million, is chronically vulnerable to large flooding disasters. The city is flood prone due to its close proximity to the Gulf of Mexico, the strong nature of Gulf Coast rainfall, rapid 8 urban growth resulting in more paved areas and roadways, the presence of clayey soils reducing infiltration, and high runoff rates along with mild slopes (Fang et al., 2006).
Mexico City is located on the basin of Mexico, a lake basin, which is around 2260 meter above MSL. The city is surrounded by large mountains on three sides (Ochoa, Quintanar, Raga, & Baumgardner, 2015). This area had a large number of lakes and wetlands until the 1500s, and were subsequently drained and filled after the Spanish Conquest. The land cover of the city was a combination of shrubs and deciduous vegetation along with willows and pines on mountains before urbanization took place (Torres-Vera, Prol-Ledesma, & García-López, 2009). The growth of the city over the last 50 years can be divided into two groups: planned urban area for the middle and upper class population and unplanned urban areas near the periphery of the city for the poor and immigrants (Torres-Vera et al., 2009). Urbanization has intervened mostly in central and northern parts of the city whereas southern part is a blend of conserved forests, agricultural lands, wetlands and grasslands (Zambrano, Pacheco-Muñoz, & Fernández, 2018). The total urbanized area of the city consists nearly 20 million 9 people (Quintana-Belmares et al., 2018). A long history of illegal settlement and lack of demarcated land use between the center and the suburbs of Mexico City (Platt, 2010) led to unplanned urban development. Unceasing urban expansion along with climate change intensify spatial and temporal extent of flooding (Eakin et al., 2017). It has flooding history in the year of 1976of , 1979of , 1982of , 1987of , 1989of , 1990of , 1992of , 1994of , 1998of (Tellman et al., 2018. Most precipitation is observed between May and September with a variation in the northern and southern parts of Mexico City. The average annual precipitation in the southern areas is 1,200 mm, which is 600 mm in the northern areas (Romero Lankao, 2010).
Jakarta, the capital of Indonesia is also highly vulnerable to flooding disasters. Since 1980, Jakarta has undergone a dramatic transformation due to massive landdevelopment projects (Padawangi & Douglass, 2015). In the last fifty years, Jakarta's population has increased from 2.7 million in 1960 to about 9 million in 2007 (Budiyono, Aerts, Brinkman, Marfai, & Ward, 2015). This drastic increase has resulted in rapid changes in land use (Verburg & Bouma, 1999). Urban areas have become denser and only one-third of the city's area remains green and unpaved (Padawangi & Douglass, 2015). Real estate developers have invested in large geographical areas to maximize profits, resulting in large-scale land development projects, shifting existing land surfaces to urban areas. Floods have become a common consequence of the significant increase in paved area. In Jakarta, devastating flooding disasters occurred in 1996, 2002, 2007 and 2013, which inundated about 40% of the city in 2007. Such massive development has also led to significant subsidence in the northern parts of the Jakarta metro area, where a number of neighborhoods often 10 experience coastal flooding and the old port area had to be protected by a seawall.
Flood risk has dramatically increased due to population growth and a subsidence rate of 10 cm/year in some areas (Brinkman & Hartman, 2008).
Dhaka, the capital city of Bangladesh is an example of unplanned urbanization. It is one of the most densely populated cities in the world, with the highest growth rate in unplanned settlements (Akanda & Hossain, 2012). The population of Dhaka has increased from 3.44 million in 1981 (Dewan & Yamaguchi, 2009)  August (Hossain, Fien, & Horne, 2018). The city was originally developed in floodfree high lands, but the recent occupation of low-lying riparian suburbs around the city has drastically increased the flood vulnerability of the people (Adikari, Osti, & Noro, 2010). Low lying lands, rivers, canal, and water bodies are increasingly being filled to construct new accommodations on lands that previously worked as natural drainage channels (Hassan & Southworth, 2017). Artificial drainage is also hampered due to poor design of drains and sewer networks, unplanned construction, and dumping of uncollected wastes on the roadside (Yasmin & Rahman, 2017). Wetlands also operate as a recharge source of groundwater storage and allow drainage of extra precipitation that may otherwise cause urban flooding (IRIN, 2012). Thus, their recession has made the city more vulnerable to larger flooding events (IRIN, 2012). In addition, there is an embankment surrounding the Dhaka city to protect from river flooding. During 11 monsoon, river water levels are often higher than the city's water level inside the embankment, which creates hindrance in drainage by gravity (Mark, Wennberg, Van Kalken, Rabbi, & Albinsson, 1998 imagery was used for 2017.

Precipitation Data
We collected precipitation data (TRMM_3B42_Daily) from Tropical Rainfall

Method
(a) Detecting changes in LULC using Landsat Images: For Landsat 5 images, bands 1 to 5 were stacked to a single layer; and for Landsat 8 images, another layer was created by stacking bands 2 to 6. As we compared images in this study, we stacked bands of Landsat images with similar wavelengths (μm). A subset was created with each stacked layer according to the Area of Interest (AOI) and then unsupervised image classification with 40 classes is done. With the help of ERDAS IMAGINE 2016 software, each class was geo-referenced with Google Earth image of that particular period and assigned to a specific land cover: urban areas, vegetative cover, waterbodies, barren land, sand filled areas, future housing projects, and forest land. This helps to visualize the changes in land use/land cover over time. In the post classification process, urban areas were used as masks to detect other land 13 uses that were transformed to urban land use and the area calculation was done in acres.
(b) Creating land use/land cover change index: The land use/land cover change index was created on the basis of infiltration capacity in this study. The lands which have more infiltration or drainage capability and transferred to paved areas are categorized with the highest index. The determination of the land-use index is shown in Table 1.

(c) Non-parametric Mann-Kendall (MK) Trend Test
To determine the monotonic increasing or decreasing trend of climatological variables, non-parametric Mann-Kendall (MK) test (Mann, 1945) & (Kendall, 1955) ( (Yu,Zou,14 & Whittemore, 1993); (Douglas, Vogel, & Kroll, 2000); (Singh, Kumar, Thomas, & Arora, 2008)) is highly used due to its accommodating ability of missing values (Gajbhiye, Meshram, Mirabbasi, & Sharma, 2016). In this trend test, the null hypothesis (H0) is there is no monotonic trend in the precipitation over time and the alternative hypothesis (HA) is there is a monotonic trend (increasing or decreasing) available in precipitation over time. In any rainfall trend analysis, outliers will be there due to extreme rainfall events. These outliers have less impact (Birsan, Molnar, Burlando, & Pfaundler, 2005) on the result of this MK test as its statistics is based on positive or negative sign rather than any value (Gajbhiye et al., 2016). Here, "modifiedmk" package of RStudio software is used to determine the Mann-Kendal Trend and Sen's slope. We assume that the rainfall time series is independent. (1) where and are sequential data for the ith and jth terms, sign is the signum function, and n is the sample size.
The statistic S is nearly Gaussian when n = 18 with the mean E(S) and variance Var(S) of the statistic S given by If there is tie in the dataset, then Var (S) has to be adjusted and becomes The variable q and tp are the number of tied groups and number of data values in the pth group, respectively. The standardized statistic (Z) for one-tailed test of the statistic S is given as follows:

Sen's Slope
The magnitude of the trend change can be identified by a slope estimator , which was first proposed by Sen (Sen, 1968) and then extended by Hirsch (Hirsch, Slack, & Smith, 1982). is the median of overall all possible combinations of pairs for the whole dataset. The magnitude of trend was calculated predicted by the Sen's slope estimator with the slope of all data pairs was computed as follows: Where x j and x i are considered as data values at time j and I (j > i) correspondingly.
The median of these N values of T i is represented as Sen's estimator of slope. Sen's estimator is computed as 16 when N is odd, and it is considered as when N is even. At the end, Q med is computed by two-sided test at 100 (1 − α) % confidence interval, and then a true slope can be obtained by the non-parametric test.
A positive value of Q i indicates an upward or increasing trend, and a negative value gives a downward or decreasing trend in the time series.  (Hayes, 2000) into soil wetness measurements. The theoretical probability (Bonaccorso, Cancelliere, & Rossi, 2015) of occurrence of each interpretation derived from normal probability density function (Guhathakurta, Menon, Inkane, Krishnan, & Sable, 2018) is also given below:    Extreme growth of urbanization is observed in Jakarta between the year 1997 to 2017.
During this 20 years, urban areas have increased from 41% to 60% of the total area, which clearly reflects the haphazard urban expansion pattern. Apart from urban areas,         Year 1992Year , 1994Year , 1998Year , 2001Year , 2003Year , 2015

Validation of the LULC Analysis
The validation of the LULC analysis for Houston for year 2007 is done using 2006 National Land Cover Dataset (NLCD). The comparison of the total area and percentage is given below.

Urban Flood Risk Maps
The risk index calculated from equation (9) (Grineski et al., 2015). A qualitative check on the two maps shows partial validation of our risk calculation approach, while there is room for improvement for areas that are prone to flooding due to distinct elevation changes such as rivers, streams, and bayous.

Discussions and Conclusion
With the expansion of urbanization, vegetated soils convert to impervious surfaces that increase storm water flow and decrease both infiltration and natural storage (Wheater & Evans, 2009). Higher vegetative cover facilitates higher infiltration rate and quantity (Loch, 2000). The rate and magnitude of infiltration are dependent on the type, duration and intensity of precipitation, initial soil moisture content, soil type, evaporation, vegetation coverage, and terrain slope (G. Zhang, Qian, Wang, & Zhao, 2014). The soil composition of Houston is mainly the combination of fine sandy loam and clay, which has poor draining capacity (Muñoz, Olivera, Giglio, & Berke, 2018).
Like Houston, the soil profile of Central Jakarta consists of alluvial clay in the form of soft to stiff (Hsiung, Yang, Aila, & Ge, 2018) and Dhaka city is a blend of Pleistocene clayey soils and Holocene clayey and sandy soils (Rahman, Kamal, & Siddiqua, 2018). As infiltration capacity of the clayey soil is less than that of sandy soil due to its smaller pore size, it is understood to be one of the main reasons that cause urban flooding in our study areas. In addition, rain on barren land compacts the upper layer of soil, creating hindrance in infiltration and causing excess runoff. Therefore, increasing amount of barren lands in these megacities are also responsible for urban flooding. According to Manning's equation, the velocity of the storm water flow is indirectly proportional with the roughness of the land surface (Leopold, Wolman, & Miller, 2012). Therefore, increasing paved smooth surfaces amplify the storm water flow more than any natural rough surface (Jacobson, 2011 (Landers, 2017). As reducing impervious layer is not easy inside cities, Low Impact Development (LID) practices can be helpful in reducing the excess runoff. These practices are used to manage storm water at the source by providing permeable pavements, bio-retention areas, and creating intermittent impervious surface (Damodaram et al., 2010). These could be potential remedies to decrease the heavy runoff due to impervious layers.
Due to the combined sewer system in Mexico City, volume of wastewater after heavy rainfall increases immensely. In base flow conditions, the waste water volume is 45 m 3 /s, which increases to 300 m 3 /s in peak flow conditions (Siemens, Huschek, Siebe, & Kaupenjohann, 2008). The pipe network of the combined sewer system is complex due to large difference in pipe diameter (0.30 m to 3.05 m). The system also generates sediment (Jiménez, Méndez, Barrios, Salgado, & Sheinbaum, 2004), which hampers the flow and creates more flood risk eventually for the city. The authority 64 extracts 0.85 Mm 3 of sediments (Jiménez et al., 2004) from the system every year and disposes as landfill, but cannot cope with the heavy rate of sedimentation in combined sewers.
Imprudent storm water drainage (Padawangi & Douglass, 2015) of Jakarta is another reason that is responsible for repetitive flood occurrence. Besides, insufficient finances to develop institutional capabilities, regulatory framework is also responsible for this situation (Kartez & Lindell, 1987). Changes of land ownership and extensive land development projects have influence on urban flooding (Walker, Whittle, Medd, & Walker, 2011). East flood canal project, the World Bank funded project named Jakarta Urgent Flood Mitigation Project / Jakarta Emergency Dredging Initiative (JUFMP/JEDI), proposed sea wall project is expected to be helpful in decreasing the flood risk in urban Jakarta (Padawangi & Douglass, 2015).
Apart from natural factors, stormwater management of these cities is also responsible for the situation. For example, Dhaka has only 30% and 38% coverage of sewerage and storm water systems, respectively (World Bank: BD: Dhaka Water Supply & Sanitation Project, 2017). Many areas of Dhaka have local combined sewer facilities, which cannot accommodate the excess runoff due to high-intensity precipitation. Surface runoff goes to underground sewer networks through catch pits.
Inadequate intake capacity of catch pits or insufficient drainage capacity of sewer pipes cause surface flooding, which can contribute to urban flooding (Mark, Apirumanekul, Kamal, & Praydal, 2001).
The result of the study shows strong evidence that land use and land cover change (LULC) and insufficient water and drainage infrastructure development is mostly accountable for urban flooding with moderate impact from precipitation alteration.
Urban flood can occur any time but the frequency of occurrence is higher in wet periods of any area. The rainfall trends in wet periods of Houston, Mexico City and Dhaka are negative and for Jakarta, it is positive. It implies that land use land cover change is the main driving factor behind urban flooding in Houston, Mexico City, and Dhaka. For Jakarta, both factors are equally important for urban flooding.
Before approving any area as urban area, planners should test soil characteristics, which play a vital role in infiltrating floodwater and excess runoff. Accuracy metrics for LULC change analysis should be added in future analysis. In addition to protecting as much land as possible to preserve natural hydrological and drainage characteristics, installation of high capacity pumping stations, accommodating Low Impact Development (LID) practices should be incorporated at planning and implementation levels. Natural canal excavation to increase capacity, reclaiming illegally filled canals, separate sewage and storm water drainage system, and provision of retention basins and rainwater harvesting can further reduce the intensity of urban flooding conditions in developing cities. Strict law enforcement is also required in order to track and stop the illegal landfilling of the natural drainage system. Proper zoning is necessary to stop haphazard urbanization. As the world is rapidly urbanizing, steps to identify and