JOURNAL OF ENVIRONMENTAL SCIENCES 59 (2017) 39– 47

Available online at www.sciencedirect.com

ScienceDirect

www.elsevier.com/locate/jes

Street floods in Metro and possible solutions

Alfredo Mahar Lagmay1,2,⁎, Jerico Mendoza2, Fatima Cipriano2, Patricia Anne Delmendo2, Micah Nieves Lacsamana2, Marc Anthony Moises2, Nicanor Pellejera III 2, Kenneth Niño Punay2, Glenn Sabio2, Laurize Santos2, Jonathan Serrano2, Herbert James Taniza2, Neil Eneri Tingin2

1. National Institute of Geological Sciences, University of the Philippines, 1101, Philippines 2. Nationwide Operational Assessment of Hazards Phil-LiDAR 1 Flood Modelling Component, UP NIGS, Quezon City 1101, Philippines

ARTICLE INFO ABSTRACT

Article history: Urban floods from thunderstorms cause severe problems in due to road Received 14 December 2016 traffic. Using Light Detection and Ranging (LiDAR)-derived topography, flood simulations Revised 1 February 2017 and anecdotal reports, the root of surface flood problems in Metro Manila is identified. Accepted 6 March 2017 Majority of flood-prone areas are along the intersection of creeks and streets located in Available online 10 March 2017 topographic lows. When creeks overflow or when rapidly accumulated street flood does not drain fast enough to the nearest stream channel, the intersecting road also gets flooded. Keywords: Possible solutions include the elevation of roads or construction of well-designed drainage Flood modeling structures leading to the creeks. Proposed solutions to the flood problem of Metro Manila may LiDAR avoid paralyzing traffic problems due to short-lived rain events, which according to Japan Urban flooding International Cooperation Agency (JICA) cost the Philippine economy 2.4 billion pesos/day. © 2017 The Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences. Published by Elsevier B.V.

Introduction Apart from devastating floods like those spawned by Tropical Storm Ondoy in 2009 (Lagmay et al., 2010) and the typhoon- Metro Manila is located on an isthmus between the Manila enhanced southwest monsoon rains in 2012, 2013 (Lagmay et al., Bay and Laguna de Bay. The entire region is composed of one 2014) and 2014, more frequent floods caused by short-lived major catchment called the Marikina River Basin, which thunderstorms are also a problem. Once parts of the road covers 535 km2, and eight smaller, river sub-basins, which network are blocked by floods, traffic develops and paralyzes the cover 683 km2 that drain directly into Manila Bay and Laguna entire city. According to JICA, traffic jams due to thunderstorm- de Bay. The Marikina, Pasig, San Juan and Tullahan rivers related flashfloods costs PhP 2.4 billion a day from wasted serve as the main outlets for a network of tributaries of the gasoline and lost economic productivity (Rodis, 2014). Marikina River Basin and smaller catchments of Metro Manila Flashfloods are traditionally blamed on the loss of infiltra- (Fig. 1). Highly urbanized and populated by almost 12 million tion due to urban concrete, a century-old drainage system, residents (Cox, 2011), the metropolis lies on one of the widest and clogged streams. This study analyses nuisance floods floodplains in the Philippines. caused by brief, heavy downpours. It identifies other factors to

⁎ Corresponding author. E-mail address: [email protected] (A.M. Lagmay).

http://dx.doi.org/10.1016/j.jes.2017.03.004 1001-0742/© 2017 The Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences. Published by Elsevier B.V. 40 JOURNAL OF ENVIRONMENTAL SCIENCES 59 (2017) 39– 47

Fig. 1 – Metro Manila natural drainage. (A) Location of Metro Manila, (B) administrative boundaries of component cities and (C) watersheds and tributaries. find relatively inexpensive solutions to flood-generated traffic where V is the average velocity in m/sec, h is the flow depth in problems. meters, and i is the excess rainfall intensity in mm/24 hr. Other variables are slope (S), acceleration due to gravity (g), pressure

1. Methods Table 1 – Metro Manila Development Authority list of flood-prone places in Metro Manila. The Metro Manila Development Authority (MMDA) released Street name City a list of flood-prone areas in the National Capital Region 1. Espana–Antipolo–Maceda Manila (Table 1), verified by accounts collected from photographs 2. P. Burgos (City Hall) Manila posted in social media. 3. R. Papa, Rizal Avenue Manila Crowd-sourced data (Fig. 2a) were overlaid on a 100-year 4. Buendia Extension–Macapagal Avenue Manila rain return flood-hazard map (Fig. 2b, NOAH, 2013). 5. Buendia–South Superhighway (northbound) Manila – LiDAR-derived topography was used to create profiles of the 6. Buendia South Superhighway (southbound) Manila 7. Osmeña– (northbound) main roads in these areas, as well as profiles of the road sides. A 8. Makati Makati Roces Street and CP Garcia Avenue in the University of the 9. Don Bosco Makati Philippines (UP) were also examined. Field work was also 10. EDSA Pasong Tamo, Magallanes Makati conducted to check the drainage crossing the streets in those 11. West Service Road, Merville Paranaque areas. 12. East Service Road–Sales street Muntinlupa Floods were simulated in FLO-2D GDS PRO using the St. Venant 13. McKinley Road Taguig equations for continuity and momentum (Eqs. (1) and (2)) and 14. C-5 Bayani Road Taguig 15. C-5–BCDA Taguig the finite-difference scheme to compute flood velocities: 16. C-5 Bagong Ilog Pasig – ∂ðÞVh ∂ðÞh 17. EDSA SM Megamall Mandaluyong þ ¼ i ð1Þ 18. EDSA–Camp Aguinaldo Gate 3 Quezon ∂ðÞx ∂ðÞt 19. Quezon Ave.–Victory Ave./Biak na Bato Quezon 20. NLEX–Balintawak Cloverleaf Quezon V 1 ∂ðÞV ∂ðÞV 21. North Avenue fronting Trinoma Mall Quezon ∂ðÞh g g S ¼ S− − − ¼ 0 ð2Þ 22. EDSA–North Avenue Quezon ∂ðÞx ∂ðÞx ∂ðÞt 23. Philcoa area Quezon JOURNAL OF ENVIRONMENTAL SCIENCES 59 (2017) 39– 47 41

Fig. 2 – Flood-prone areas (a) plotted by netizens, and (b) overlain on a flood map.

∂ðhÞ ð∂ðVÞÞ ð ∂ðVÞÞ gradient ∂ðxÞ , and the local ∂ðtÞ and convective V ∂ðtÞ Five sites have bridges (Appendix A): Philcoa, R. Papa, C-5 accelerations. These are solved using the finite-difference Bagong Ilog, Osmeña–Skyway, and Don Bosco. The street at scheme to get the velocity across the boundaries in eight Philcoa stands 3.8 m above the creek bottom with a rectangular, potential flow directions of every grid element. 2.37 × 4.4-m culvert perpendicular to the road and two circular, The simulations used 1 × 1 m LiDAR-derived elevation data. 1 m-diameter culverts parallel to the road. R. Papa is 1.38 m The floodplains were delineated into catchment areas based on above the creek bottom. In C-5 Bagong Ilog, the street is the flow direction and accumulation. Manning's coefficient of 3.92 m above the stream bottom. Along Osmeña–Skyway is a 0.03 was assigned to streams, which is the normal value for 22.3-m bridge 3.34 m below street level. A stream with its bed main channels (Chow, 1959), and 0.15 to the floodplains which 4.5 m below South Expressway in the Don Bosco area is are predominantly concrete. Inflow and outflow nodes were drained by a parallel 4 × 2.5-m drainage structure. Eight places assigned based on where the water flows in from the upper do not seem to have drainage networks, which could be masked watershed and out through the main stream channel. by overlying concrete. Rainfall is simulated as a non-point source carrying water throughout the model. Once flood-prone areas were identified 2.2. Roads in UP Diliman from the 100-year flood-hazard map, higher-resolution simu- lations in sub-basins of concern were conducted for short- Both UP creeks are headwaters of a drainage network. A Roces lived thunderstorms. An hour of rainfall with intensities of 30– is 4 m and CP Garcia is 1.25 m above the banks (Fig. 3). In CP 70 mm/hr was used to simulate thunderstorms. Observations, Garcia, there appears to be a bigger channel, where the road is road profiles and flood simulations revealed the causes of street lower in elevation than the bank of the creek. While it is built flooding and indicated appropriate solutions. up above the lowest portion of the channel, it is still below the main channel banks and remains susceptible to flooding (Fig. 3). The streets each have two 1 m-diameter culverts. 2. Results A Roces is sufficiently elevated to avoid flooding even when the creek swells, as during Typhoon Ketsana in 2009. 2.1. Intersection of creeks and streets In contrast, CP Garcia is flooded and impassable even during short-lived torrential thunderstorms. Here, flood depths reached The flood-prone areas list and flood-hazard maps show floods 1 m during Ketsana, and are around 0.9 m during brief at intersections of streets and creeks, and at ponded areas thunderstorms. such as Padre Burgos in Manila (Fig. 2). Road profiles reveal that they follow topographic lows and are not significantly 2.3. Thunderstorm flash-flood scenarios elevated from roadsides and creek banks (Appendix A). R. Papa is lower than the roadside and lies below the tops of Floods during short-duration thunderstorms block traffic, as three stream segments. EDSA-North Avenue floods at its exemplified by four sites in the MMDA list: Bayani Road, lowest portion, which follows the topography of a 250-m wide, Taguig City, Victory Avenue, , R. Papa 1-m high channel. Street, Manila, and EDSA–North Avenue (Fig. 4). 42 JOURNAL OF ENVIRONMENTAL SCIENCES 59 (2017) 39– 47

Fig. 3 – Profiles of (a) A Roces Street and (b) CP Garcia Avenue. Transect a–a′ is brown in the profile; transect b–b′ is green. (c) A Roces and CP Garcia creek during (d) summer and (e) a short-lived thunderstorm.

Bayani Road starts to flood knee-deep after 40 mm of rain, the roads get flooded. Roads that follow topographic lows, and floods are waist-high when rain delivers 70 mm. Victory even if they are above the creek's bank, should have enough Avenue starts to flood gutter-deep at 40 mm, knee-deep after clearance or freeboard to accommodate creek swelling 60 mm, and tire-deep after 70 mm. R. Papa begins to flood (Fig. 5). Small streams may oftentimes be dry or nearly dry, gutter-deep across three sections with 50 mm of rain and but they, too, can swell and overtop their banks (Fig. 4). reaches half-tire-deep with a 70 mm rain. EDSA–North Avenue Roads above creeks, therefore, must be designed and begins to flood at 40 mm and reaches tire-level with intensities constructed in the same manner as roads that cross large of 60 mm. Snapshots of simulations for the other 18 MMDA rivers. flood-prone areas are provided as supplementary files in the In the Philippines, flood-control design is governed by Appendix. technical standards based on flood-frequency expressed by return periods which are based on the size of the catchment area, importance of the project area, and economic viability 3. Discussion (JICA, Japan International Cooperation Agency and DPWH, Department of Public Works and Highways, 2003). Ideally, Metro Manila floods are classified into those that endanger discharge values are calculated through runoff analysis. When people, and those that merely cause traffic jams. Not all of available, annual maximum-flood data are analyzed as a more Metro Manila is flood-prone. Floods generate traffic jams only convenient alternative. In the absence of these two data sets, in specific areas and may be eliminated with relatively the return period is determined using rainfall data. A Rainfall low-cost solutions. Intensity Duration Frequency (RIDF) curve is utilized in calcu- Many flood-prone areas are where streets and creeks lating discharge for catchment areas below 20 m2 using the intersect; others are areas where water accumulates. The Rational Formula Method (Kuichling, 1889; Viessman and roads in UP campus present two contrasting scenarios: A Lewis, 1995; Eq. (3)): Roces hardly ever floods whereas CP Garcia gets inundated Q ¼ c i A ð3Þ during severe weather as well as short-lived thunderstorms (Fig. 3). These show that the construction of roads relative to where Q is the discharge in cm, c is the runoff coefficient, i is the elevation of intersecting creeks affects whether or not rainfall intensity in mm/hr and A thedrainageareainm2. JOURNAL OF ENVIRONMENTAL SCIENCES 59 (2017) 39– 47 43

Fig. 4 – Inundation scenarios for short-lived thunderstorms. Flood depths are classified based on the MMDA flood gauge. (Ankle ≤ 0.1 m, gutter ≤ 0.201 m, half-knee ≤ 0.255 m, half-tire ≤ 0.331 m, knee ≤ 0.484 m, tires ≤ 0.661 m, waist ≤ 0.941 m, chest ≤ 1.144 m).

Fig. 5 – Cartoon of inundation of a road with a design that follows the topography as it crosses the creek and where the road is elevated relative to the creek banks. Blue circles are culverts. 44 JOURNAL OF ENVIRONMENTAL SCIENCES 59 (2017) 39– 47

Culverts and lateral drains, understandably, are not The hydrograph (Fig. 7) reaches its low after peak flow at designed for unusually big floods. Short-lived thunderstorms 1 hr and 20 min then rises again. Water draining into the (i.e. >/30 mm/hr), however, flood the streets as well. culvert causes the initial rise; the succeeding rise is caused by The theoretical discharge of the creek intersecting the road flow from farther reaches of the watershed. in Philcoa was computed using the Rational Method (Eq. (3)). The absence of information regarding subsurface drainage Philcoa is one of the places in the MMDA list of flood-prone in Philcoa was addressed through field work. In the simulation, areas. The calculations for 1 hr of rain with intensity of a 1 m × 1 m drain connects directly with the creek, mimicking 70 mm/hr yielded a value of 29.95 cm. This is still below the the unexposed lateral drainage. 87 m3/sec discharge capacity of the culvert obtained with In the model, even if the culvert can handle the discharge Manning's equation (Gauckler, 1867; Manning, 1891; Chanson, of surface flow from heavy downpour, knee-high flooding still 2004) (Eq. (4)): occurs for 24–30 min, then subsides until it ends after another 30 min. Flow into the topographic low is faster than into the 1 2 1 Q ¼ A R3 S2 ð4Þ creek. Inasmuch as the creek never fills during short-lived n thunderstorms even when the culvert is partially clogged where A is its inner cross-sectional area in m2, n is the with garbage, the problem is really not the culvert, but the Manning's coefficient, R is the hydraulic radius in m, and S is impervious low-lying road. the stream bed slope obtained through LiDAR topography. A Roces Ave. in UP Diliman, provides a possible solution: a Concrete box culverts have a Manning's n of 0.012–0.015. The hydraulic structure under the street large enough to accommo- latter was used for a more conservative computation. date the volume of water, which is equivalent to the area Despite the capacity of the culvert, field interviews revealed covered by the flood multiplied by the height of inundation, that street floods happen even during brief thunderstorms. The which for Philcoa is only knee-high. The street surface must culvert filled only during Typhoon Ketsana, showing that other have steel mesh drains to accommodate the runoff that factors play important roles in generating thunderstorm- accumulates almost instantaneously in the topographic low. related floods. Ideally, the subsurface water retention area should drive flow One and a half hour simulations using 70 mm/hr rainfall into the creek. Constructing it below the street, if structurally were conducted to determine those factors. The rain was feasible, avoids right-of-way problems. distributed over the first 60 min of the simulation. Flooding The proposed solution is simulated. Since the Flo-2D model begins at 33 min and peaks at 39 min, then slowly decreases cannot simulate flow below the street, an alternative model was and drains fully at 90 min. Peak discharge at the culvert inlet generated. A DEM (Digital Elevation Model) with open strips happens at 57 min when the flood along the road has paralleling the street and sloping towards the creek was used to already started to wane, much later than the street flood simulate subsurface flow. The simulation result shows that all (Fig. 6). the flood water was confined in the strips, which means that the

Fig. 6 – Simulation of flood at Philcoa showing the start of flooding along the road at 33 min, peak flooding along the road at 39 min, peak discharge at the culvert's inlet at 57 min, and full flood subsidence along the road at 90 min. JOURNAL OF ENVIRONMENTAL SCIENCES 59 (2017) 39– 47 45

Fig. 7 – Hyetograph and runoff hydrograph for the simulation. Peak rainfall occurred at 30 min. Peak discharge occurred at 57 min. and reached a low point 1 hr before beginning to rise again.

road will not flood (Fig. 8). This model can be used to determine Although only one example of discharge analysis was the appropriate surface infiltration of the street, and the depth presented, the findings could be applied to other places in and slope of the subsurface drain needed. Metro Manila. They can be used for other creek and street

Fig. 8 – Simulation result for the proposed solution for urban floods caused by short-lived but intense thunderstorms. The image shows all the flood water confined in the excavated strips with no flood on the road. Excavated strips as well as the creek culvert are underground, which means that the road surface is also not flooded. 46 JOURNAL OF ENVIRONMENTAL SCIENCES 59 (2017) 39– 47 intersections with culverts (e.g. Don Bosco) since they construction of a retention basin under the street big enough to have similar geometry. For ponded areas like in Padre accommodate the volume of water. The size of the retention Burgos in Manila and at EDSA Pasong Tamo, retention basin is equivalent to the simulated area covered by the flood basinscanalsobemadeunderthestreetandpumpedout along the street multiplied by the height of inundation. The after the thunderstorm. However, at EDSA Pasong Tamo, street above the proposed retention basin must have enough this type of intervention may be difficult as there is an infiltration capacity to accommodate the rapidly accumulated overpass in the area. High volume pumps are necessary to rainfall. The retention basin must also be designed to directly transfer floodwaters from this area into an adjacent drain into the nearest stream channel. The latter solution or its retention basin. equivalent may not completely address big floods spawned by severe weather events but may solve frequent street floods from short-lived thunderstorms. These proposed solutions may 4. Conclusions avert increased gas consumption and lost potential income from traffic jams during floods, which costs the Philippine Metro Manila's floods are compounded by many factors economy 2.4 billion pesos a day, a figure that may balloon to including encroachment of concrete surfaces, densification of 6 zbillion a day by the year 2030. buildings and residential areas, silting of riverbeds and canals, obstruction of waterways by informal settlers, clogging of floodways by garbage, narrowing of rivers due to development Acknowledgments on floodplains, draining and filling in of small rivers forcing more water into fewer channels, forest degradation, and This is to acknowledge the Philippine Council for Industry, reclamation of coastal land. Furthermore, humans have altered Energy, and Emerging Technology Research and Development the landscape in the metropolis which has grown rapidly but (PCIEERD). Department of Science and Technology (DOST) for with poorly planned urbanization. Since the 1970's, people have funding the Nationwide Operational Assessment of Hazards migrated from rural areas to Metro Manila increasing the (NOAH) program, particularly the DREAM Flood Modelling population from 4.9 million residents in 1975 to more than Component Project. 11 million today. A survey by the National Housing Authority showed that by the early 1980s, a quarter of Metro Manila residents were informal settlers living in crowded shantytowns Appendix A. Supplementary data many along waterways. Further complicating the problem is ground subsidence. From 1978 to 2000, parts of Metro Manila Supplementary data to this article can be found online at sank by an amount ranging from 16 cm to 1.46 m. The probable http://dx.doi.org/10.1016/j.jes.2017.03.004. causes of subsidence are excessive groundwater extraction, soil compaction and tectonic movement, though more research is needed to fully determine the primary causes (Lagmay et al., REFERENCES 2010). Floods in the metropolis can be life threatening or a nuisance because of the traffic they generate. Traffic problems related to Chanson, H., 2004. The Hydraulics of Open Channel Flow. second urban surface floods from short-lived rain events can cost the ed. Butterworth-Heinemann, Oxford-UK, p. 630. Philippine economy P2.4 billion pesos a day. This study Chow, V., 1959. Channel Hydraulics Open. McGraw-Hill. identified the root causes of street floods from both severe Cox, W., 2011. The evolving urbanform:Manila,newgeography. weather events and brief but heavy downpour from thunder- Available: http://www.newgeography.com/content/002198- storms using 2D flood simulations. Street floods in Metro the-evolvingurban-form-manila (Accessed September 9, 2013). Manila do not happen everywhere but are mainly found at Gauckler, P., 1867. Etudes Th'eoriques et Pratiques Sur l'Ecoulement intersections of streets and creeks. Further analysis reveal that et le Mouvement des Eaux. C. R. Acad. Sci. 64, 818–822. street floods occur at the lowest portion of the street or JICA (Japan International Cooperation Agency), DPWH topographic lows. During severe weather events such as the (Department of Public Works and Highways), 2003. Ondoy, Habagat 2012, Habagat 2013 and Habagat 2014 floods, Manual on flood control planning. Project for the en- creeks in the 23 areas listed by MMDA as flood-prone, swelled hancement of capabilities in flood control and Sabo and overtopped its banks. This led to inundation of streets, engineering of the DPWH. Available: https://www.jica.go. jp/project/philippines/0600933/04/pdf/Manual_on_FC_ which closely followed the landscape down to the banks of the Planning.pdf (Accessed December 10, 2014). creeks. On the other hand, short-lived heavy downpour Kuichling, E., 1889. The relation between the rainfall and the (30 mm/hr) accumulate water in the low-lying areas faster discharge of sewers in populous districts. Trans. Am. Soc. Civ. than it can drain through the culvert along the creek. Streets in Eng. 20, 1–56. the MMDA list still get inundated even though the drainage Lagmay, A., Rodolfo, R., Bato, G., 2010. The perfect storm: floods – design of the culvert below the street has greater discharge devastate Manila. Earth 55, 50 55. Lagmay, A., Crisologo, I., Racoma, B., David, C., 2014. Volcanoes capacity than the theoretical discharge expected to flow magnify Metro Manila's southwest monsoon rains and lethal through the stream channel. In places where there are no floods. Front. Earth Sci. 2:60–105. http://dx.doi.org/10.3389/ creeks, floods are due to ponding of water in the low-lying areas. feart.2014.00036. Proposed solutions to street floods in many of the places in the Manning, R., 1891. On the flow of water in open channels and flood-prone list by MMDA include the elevation of roads and pipes. Trans. Inst. Civ. Eng. Ireland 20, 818–822. JOURNAL OF ENVIRONMENTAL SCIENCES 59 (2017) 39– 47 47

NOAH, 2013. Flood Hazard Map, Metro Manila. Department of globalnation.inquirer.net/113269/manilas-traffic-jamscost-57- Science and Technology (DOST) Available: http:// million-a-day. Accessed December 10, 2014. noah.dost.gov.ph/#/ 2013, (Accessed 2014). Viessman Jr., W., Lewis, G., 1995. Introduction to Hydrology. Rodis, R., 2014. Manila's Traffic Jams Cost $57 Million a second ed. John Wiley and Sons, p. 311. Day. Philippines: Philippine Daily Inquirer. Available: http://