R Ecoa,b,*, TM Herreroa, F Llanesa,b, J Brionesa, CM Escapea,b, JJ Sulapasa,b, JAM Galanga,b, IJ Ortiza,b, JM Sabadoa,b, H Ariolaa, N Iraptab, AMF Lagmaya,b
aNationwide Operation Assessment of Hazards, U.P. NIGS, C.P. Garcia Ave., U.P. Diliman, Quezon City, 1101 Philippines
bNational Institute of Geological Sciences, University of the Philippines, C.P. Garcia corner Velasquez street, U.P. Diliman, Quezon City. 1101 Philippines
*Corresponding author: Email address: firstname.lastname@example.org
Rains brought by Typhoon Koppu generated debris flows in the towns of Bongabon, Laur and Gabaldon, Nueva Ecija province. Most of the rainfall was concentrated in the mountains of Sierra Madre as seen from TRMM and the clusters of landslides. Materials from these landslides then converged in the mountain stream network and were remobilized as debris flows. Numerous houses and structures were destroyed. However, there were no reported deaths as residents were able to evacuate due to timely warning from authorities.
On 17 October 2015, Typhoon Koppu made landfall in Casiguran, Aurora and subsequently traversed Luzon island, Philippines (Figure 1A). At this point, the Joint Typhoon Warning Center had upgraded Koppu to a Category-4 typhoon based on the Saffir-Simpson hurricane wind scale, packing ten-minute sustained winds of 185 km/hr . As the typhoon crossed the Sierra Madre mountains bordering the eastern part of Luzon, it delivered intense rains of up to 500 mm as measured by rain gauges over a span of 48 hours. As it traversed the central Luzon plains, before its exit to the West Philippine Sea, rains totaling 400 mm were recorded in the Zambales and Cordillera mountains (Figure 2). Upon reaching the West Philippine Sea, the typhoon reversed course towards northern Philippines where it gradually weakened into a tropical depression before it dissipated.
Several flooding incidents were reported in Bongabon, Laur, and Gabaldon, Nueva Ecija  at the foothills of Sierra Madre and within the basins of the Coronel River. Numerous houses and several bridges were damaged and destroyed. Field investigation revealed that much of the damage in these municipalities were caused by debris flows, fast moving water-laden slurries of soil, rocks and other materials . A similar event happened in the early morning of 4 December 2012 when Typhoon Pablo spawned debris flows that overwhelmed Brgy. Andap, New Bataan, Compostela Valley resulting in more than 500 fatalities . In contrast, the Nueva Ecija debris flows did not claim any lives because of the timely warnings and appropriate actions of the residents.
The northwest-southeast trend of the Coronel River in this area is the morphological expression of the Philippine Fault that formed steeply-sided foothills with the narrow strip of flatlands in between. This basin has become the depositional site for several alluvial fans which became areas where small villages were built up. During the high rainfall event, debris flows spread over most of the alluvial fans bringing in materials such as meter-sized boulders, sand, mud, and uprooted trees.
Alluvial fans are fan-shaped depositional geomorphic features located downslope of watersheds [5, 6]. These landforms are formed through aggradation of sediments transported via steep mountainous streams emerging to a gentle slope downstream. We refer to this point as the alluvial fan apex, and this is where we define the point of origin of alluvial fans. Lateral coalescences of several unconfined alluvial fans form a bajada. Natural hazards such as flooding and debris flows are linked to alluvial fan formation. Debris flows are important primary processes that delivers the volume of material to alluvial fans.
This study aims to characterize the debris flow events in order to better understand the underlying factors that contributed to the hazards identified in the three municipalities of Nueva Ecija. It also discusses the factors that led to the successful evacuation of communities in harm’s way. Only by thoroughly describing the debris flow event and the actions taken by the people can we begin to properly manage this type of hazard elsewhere.
Satellite remote sensing and field investigation was conducted in the municipalities of Laur, Gabaldon and Bongabon to identify new landslides and investigate the alluvial fans where debris flows occurred. Remotely sensed data facilitated focused assessment of the debris flows and targeted areas for field validation.
Using a pansharpened Landsat 8 OLI/TIRS image with 15 m pixel resolution, landslides were identified by analyzing their shape, location and spectral response. A post-event image dated 24 October was compared with a pre-event image dated 8 October to determine if the landslides occurred after the typhoon. Bands 5, 4, and 3 were displayed as red, green, and blue (RGB), respectively. This band combination shows a traditional color infrared image where red represents vegetation, with healthier vegetation being more vibrant. The post-event Landsat 8 image was used for image classification in order to delineate the river and flow deposits. The clouds were first identified by executing ISODATA unsupervised classification on the color infrared with emphasis on vegetation band combination (Bands 5,4,3). After the delineation of the clouds, they were masked out on all the bands except for the quality assessment band and Band 8 (panchromatic). After the cloud masking was done, training sites for supervised classification were selected using the land/water band combination (Bands 5,6,4). The sites were divided into 5 classes: river, flow, vegetated (elevated), vegetated (plain) and urban. Maximum Likelihood Classification was then applied on the land/water multi-band raster using the signature file of the training site. After the classification, the resulting raster was reclassified to eliminate the other classes except for the river and flow classes. The resulting image showed the uniform classification of the flow (be it the debris flow deposits or hyperconcentrated deposits) from the elevated regions to the plains near the urban areas. The source landslides on the mountainous region were also identified as flow. However, some of the cloud shadows which weren’t masked out were included in the flow class. On the other hand, the river class has a very sinuous and detailed delineation.
Alluvial fans were delineated in Esri ArcGIS 10 using a digital elevation model (DEM). Topographic features including contour lines, watershed, slope, and stream networks were automatically generated from the DEM. Alluvial fans were identified by slope topography from the fan apex to the fan lateral extent, as depicted by increasing contour line spacing located directly downstream of watersheds. Detected alluvial fans were delineated with polygons by outlining the visible fan pattern.
Fieldwork was conducted in areas identified from satellite imagery where debris flows occurred. The extent of the deposits were verified and the thickness was measured for volume computation. Cross-sections were observed to characterize the lithologies, structures and texture of the deposits.
Residents were interviewed to reconstruct the sequence of events during the typhoon which triggered the debris flows. Interviews were also focused to learn the manner by which the residents responded to the warnings given by civil authorities.
Impact marks on trees and houses were measured for run-up heights to approximate the flow velocity. We use the equationwhere v is the flow velocity, g is the acceleration due to gravity and h is the measured run-up height. This relationship assumes that all of the kinetic energy is converted to potential energy as the flow runs up vertically against an obstacle . It should be noted, however, that the values computed are only minimal values, as this formula does not take into consideration channel roughness or internal friction forces [8, 9].
3.1 Landslide inventory
Analysis of the post-Koppu satellite image revealed a significant number of landslides on mountainous slopes in the eastern border of the Nueva Ecija province (Figure 3). In the satellite image, recently scarred areas have more distinct coloration compared to the surrounding vegetated areas and have well-defined arcuate scarps. These were the sources of rock material, soil, and other sediments constituting the flows. There were 230 landslides identified from the Landsat image throughout the mountains of Aurora National Park, Sierra Madre, though the actual number may be much greater because other smaller-scale landslides were not detected given the limited resolution of the image. Out of these landslides, 11 were most likely pre-existing landslides reactivated during the downpour. Other landslides found in the 24 October Landsat image were obscured due to cloud cover present in the 8 October image, and it could not be determined if they are newly-formed or re-initiated.
3.2 Alluvial fans
Twenty-four apices make up six alluvial fans and one bajada along the stretch of mountains facing the Coronel River within the four municipalities. In this report, we refer to these alluvial fans based on the barangay or municipality of each fan area covered (Figure 3). In Bongabon, there are two alluvial fans: Bongabon and Pesa alluvial fans. The Bongabon alluvial fan covers a 26 km2 area fed by a 109 km2 catchment area and three stream networks. The Pesa alluvial fan has a 0.6 km2 area fed by a 5 km2 catchment area and a single perennial river. In Laur, there are two alluvial fans: Olivete fan and San Vicente fan. The Olivete fan has a 1 km2 area fed by a 7 km2 catchment area and a single river network. The San Vicente fan has a 4 km2 area fed by a 12 km2 catchment area with two river networks. In the municipality of Gabaldon, the alluvial fans identified were the Bagting and Tablang fan and the Gabaldon bajada. The Bagting fan has a 1.7 km2 area fed by an 8.5 km2 catchment area whereas the Tablang fan has a 0.9 km2 fan area and a 4.9 km2 catchment area. Each fan is fed by a single perennial river. In Gabaldon, coalescing alluvial fans named as the Gabaldon bajada was delineated in the plains between the Aurora Memorial National Park mountains and General Tinio mountains. The bajada has a 48.8 km2 depositional area, the largest in the three municipalities studied, and is supported by multiple catchments with a combined area of 440 km2 and an extensive river system.
3.3 Debris flow deposits
The debris flow deposits are found in rivers and alluvial fans whose network of drainage channels originate from the slopes of Sierra Madre with numerous landslides. The debris flow deposits fan out at the apices of alluvial fans and coalesce into the Coronel River (Figure 3). This river is about 30 km long from the source at General Tinio. It then merges with Pampanga River in Laur, which traverses the central Luzon and ends at Manila Bay. Sixteen flows were characterized from the high to moderate reflectance in the satellite imagery where 15 of them emerge out of the 24 apices. The areas with distinct flow features are found in Barangays Digmala (1), Calaanan (1), Bantug (1), San Fernando (1), San Vicente (1), Bagting (3), Calabasa (2), Ligaya (1), Tagumpay (1), Bagong Sikat (1), Bantug (1), and Pias (2).
Five out of 16 debris flow deposits (DFD) were investigated during fieldwork and were named after the barangays where the flows affected residents. These areas are in Barangays San Vicente, Bagting and Tagumpay. Field investigation showed boulder-rich deposits that were poorly sorted and reversely graded. Boulders as large as 6.5 m in diameter with an average of about 0.5 m were scattered over the deposit area (Figure 4). Rocks comprising the deposits were angular to sub-angular and were dominated by metamorphic rocks. Igneous and sedimentary rocks were also identified but in lesser quantities. The maximum measured thickness of the deposit was 3.5 m in Ligaya, 2 m in Bagting, 1.6 m in Calabasa and 3 m in San Vicente. The boulder-rich flows grade into cobble-to sand-sized sediment deposits with a thickness of 0.6 m as it merges with the Coronel River.
The San Vicente Debris Flow spread out west of its main river up to half of the whole alluvial fan where about 1.74 km2 of rice fields. Flooding at the village proper created new channels more than 5 m deep. In some areas, the material under the concrete road and houses were washed away causing them to collapse into the newly formed channels.
The Bagting Debris Flow was confined along the eastern portion of the fan. Its main river had dikes which were destroyed during the previous typhoon (Mujigae; Philippine name: Kabayan) on October 2, two weeks before Typhoon Koppu. The debris flow overtopped the dikes and swelled eastward where it scoured through part of a village 2 m above the original river level. A run-up height of 120 cm above the flow level was measured giving a velocity of 4.8 m/s at the flow edge.
The Calabasa Debris Flow diverted southward from the original river channel at the apex of its alluvial fan. The flow was directed towards Barangay Calabasa Proper where it damaged and swept away 42 houses. The mass of the flow reached 2 m in height as observed from the scour marks on several upright trees. A velocity of 5.9 m/s was calculated from the run-up height marked on intact houses along the central flow.
The Tagumpay Debris Flow reached flood heights of more than 10 m above the normal river height. Along the flood plains, 130 m upstream from Dupinga bridge, residents and resorts were inundated and buried in 5 m thick sand deposits. Along the valleys, the walls were scoured revealing the bedrock. The debris flow spread across half of the alluvial fan where it covered around 5.7 km2 of agricultural lands.
4.1. Debris flows
Debris flows are one of the most dangerous natural hazards in the world. These are slurries of water, rocks, soil, sand and other debris flowing like a liquid [3, 11] that is initiated from single or multiple collapses in the upper catchments of a mountain slope or volcanic edifice. It can be triggered by intense rainfall, eruption through a crater lake, pyroclastic surges joining streams, liquefaction of debris avalanches , and reservoir collapses . While it is being transported downstream, it may consist of single or multiple surges as it entrains loose sediments from erodible banks and gullies it passes through.
When a debris flow is deposited on an alluvial fan, it results in a significant reduction in the slope and the opening of the channel to a less-confined space . At deposition, the magnitude of the debris flow usually has become exponentially larger than its initiating collapses  and therefore has the capacity to inflict even more damages and loss of life. The infrequency of its occurrence and the limited local knowledge on this type of event pose a big threat to communities living within an alluvial fan [15, 16].
In the Philippines, the focus has primarily been on lahars, debris flows composed primarily of volcaniclastic materials . Previous workers have determined the debris flow initiation thresholds of Mount Pinatubo from the 1992 post-eruption lahars [18, 19] and of Mount Mayon from lahar-triggering rainfall events in the 1980s .
In recent years, the catastrophic event in Brgy. Andap in December 2012  has prompted ongoing research on nonvolcanic debris flows and other hazards that can occur on alluvial fans. Automated GIS-based mapping of alluvial fans have also been conducted in the Philippines and have so far identified more than 1,200 fans across the country. Alluvial fans in Gabaldon, Nueva Ecija that were previously identified by this method are part of the area that experienced extensive flooding and debris flows during the course of Typhoon Koppu.
4.2. Triggering event
Measurements from NASA’s Tropical Rainfall Measuring Mission (TRMM) recorded at least 600 mm of total rainfall in the Sierra Madre mountains from 13-20 October (Figure 5). The threshold for the multiple shallow landslides and subsequent debris flows in adjacent alluvial fans at the base of these mountains can be estimated using rainfall duration and intensities from available data. A global threshold for the initiation of shallow landslides and debris flows based from 73 landslide events from around the world was first established in 1980 . This was refined more than 25 years later after compiling a more extensive global database of 2,626 rainfall events . In the Philippines, previous workers have determined the debris flow initiation thresholds of Mount Pinatubo from the 1992 post-eruption lahars [18, 19, 22] and of Mount Mayon from 14 rain events that triggered debris flows from 1986-1989 .
For Typhoon Lando, four automated rain gauges (ARG) within a 40-km radius of Aurora National Park in the Sierra Madre mountains were verified to select the best available data for estimating the threshold. Rain values over a period of 96 hours from 16-19 October 2015 were checked for errors or missing data. The ARG in Alfonso Castañeda, Nueva Vizcaya stopped transmitting data at 8:15AM on 17 October and only resumed at 2:00AM on 18 October. Meanwhile, the ARG from Barangay Calabuanan in Baler recorded very little rainfall, with only an accumulated amount of 38 mm over the 96-hour period. In Dingalan, Aurora, the rain gauge in Dingalan Central School was able to record an accumulated rainfall of 155.24 mm in the same timeframe. However, it stopped transmitting data at 1:30AM from 18 October, rendering the data incomplete and unreliable.
The most reliable rainfall data was obtained from the ARG in Rizal Town Hall in Bongabon, Nueva Ecija. From 16-19 October, the ARG recorded an accumulated rainfall of 251 mm. Rain values from 17 -19 October was extrapolated to 4-hour average intensities and was plotted against its corresponding duration in the 72-hour timeframe (Figure 6). Global thresholds [20, 21], as well as the lahar threshold determined for Mayon Volcano , were graphed in the same figure to determine the best fit to characterize the debris flow event.Two differing accounts were collated from local residents regarding the time of occurrence of the Calabasa debris flow along the Gabaldon alluvial fan. According to the barangay captain of Calabasa, large amounts of water started to flow from the main channel that contributes to the fan from 12 midnight of 18 October. However, the boulder-rich debris flows did not arrive until 9AM of the same day. Another resident of the area described the boulder-rich debris flows to have started at 10:30AM. For simplification, the vertical dashed line on the graph on Figure 6 is placed at the 9AM plot, the earliest probable time of debris flow occurrence. In San Vicente, Laur, barangay council members described the debris flows in their area to have started at 8AM on 18 October. A separate vertical dashed line placed at the 8AM plot denotes the start of occurrence of boulder-rich debris flows in San Vicente.
From the graph, it can be synthesized that the best fit curve for the threshold is the power function derived for Mayon Volcano. Both the San Vicente and Calabasa debris flow initiation plots occur well above the Guzzetti and Caine thresholds (Figure 6). This could be because the global thresholds were obtained from sample sizes that are too varied and generalized to accommodate a wider range of rainfall-induced shallow landslide and debris flow events. The database for both global thresholds were also derived from locations with completely different climates and geology from that of the Philippines. In Mayon, the primary reason for the higher threshold is the coarse, granular, and very porous nature of the volcaniclastic surface materials ; the similarity in climate and other local factors of the Mayon rainfall events must have contributed to it being the best fit for the Nueva Ecija debris flows. Obtaining more rainfall events that trigger different types of landslides in the area can generate a more accurate localized threshold for use in early warning and disaster mitigation in the future.
4.3. Disaster averted
On the afternoon of 15 October, Koppu rapidly intensified into a typhoon. Signal Number 1 was hoisted over the province of Nueva Ecija and other areas in the northern part of the Philippines. It was upgraded to a Super Typhoon on 17 October, causing light rains in some areas in the vicinity of Gabaldon, Nueva Ecija. That same afternoon, several residents–mostly women and children–evacuated as advised by the municipal administrator despite clear skies. That night, Gabaldon Mayor Rolando Bue called for forced evacuation as heavy rains started to pour well until the next day, with the exception of some male heads of household and barangay officials. DOST Project NOAH released a list of municipalities that will possibly receive more than 100 mm of rainfall within 24 hours per day for the next four days (17-20 October, Saturday to Tuesday) due to the typhoon, where 100 mm of rainfall in certain areas could cause landslides in susceptible locations. The municipalities of Bongabon, Gabaldon and Laur were on the list.
Various accounts from interviews state that the debris flow events happened between 9 AM and 12 Noon that day, bringing a mixture of water, sediment, and boulders to the settlements on the alluvial fans. Eyewitnesses in Barangay Calabasa, Gabaldon and San Vicente, Laur recalled that they initially witnessed floodwaters overflowing from the rivers, that later on turned into a bouldery flow. Seeing this, they promptly decided to retreat to safer grounds. Houses that were in the way of the slurry were completely washed out. Hectares of land, rice paddies, and vegetable farms were inundated and rendered uninhabitable, while a year-old bridge connecting the barangays of Bugnan and Bagong Sikat was destroyed with debris from the Sierra Madre mountain range. There were no recorded casualties in Bongabon, Gabaldon and Laur despite the widespread destruction brought by the debris flows.
When Typhoon Koppu struck the eastern side Luzon, it generated numerous landslides in the Sierra Madre mountains. Rocks, soil and other debris from these landslides converged into the mountain stream networks, and were remobilized as debris flows. These debris flows drained out to the alluvial fans of Bongabon, Gabaldon and Laur, all located at the footslopes of Sierra Madre.
Analysis of satellite imagery and field investigation revealed the extensive damages caused by debris flows houses, buildings and agricultural lands. Out of the 24 apices that make up the seven alluvial fans in the area, 16 were identified to have DFDs emerging out of them. The DFDs observed in the field ranged from boulder-rich deposits to mostly sandy hyperconcentrated flow deposits with thickness that range from 0.5 m to 3.5 m.
New Bataan, Bongabon, Laur and Gabaldon have very similar geophysical settings. All have alluvial fans, as well as active faults. The presence of the faults most likely loosened the rocks in the mountains, providing source materials for the debris flows. More importantly, all municipalities have settlements living within these alluvial fans, exposing them to flood and debris flow hazards. To date, there are more than 1,200 alluvial fans identified by Project NOAH. This means that people living within these alluvial fans are also exposed to similar hazards brought by Typhoons Pablo and Koppu. It is therefore important to assess the hazards in these areas so that mitigation measures can be put in place as soon as possible.
 NOAA, Prognostic reasoning for super typhoon 24w (koppu) warning nr 19, Online, accessed at: http://weather.noaa.gov/pub/data/raw/wd/wdpn32.pgtw..txt (2015).
 NDRRMC, Sitrep no. 17 re preparedness measures and effects of typhoon ”lando” (i.n. koppu), Online, accessed at: http://ndrrmc.gov.ph/attachments/article/2607/A SitRep No 17 re Preparedness Measures and Effects of Typhoon LANDO 24OCT2015 0600H.pdf (2015).
 T. Takahashi, Estimation of potential debris flows and their hazardous zones: soft countermeasures for a disaster, Natural disaster science 3 (1) (1981) 57–89.
 A. Lagmay, R. Eco, J. Alconis, B. Salvio, What hit barangay andap, new bataan, compostela valley (initial assessment), Project NOAH Open-File Reports 1 (2013) 1–2, iSSN 2362 7409.
 F. Drew, Alluvial and lacustrine deposits and glacial records of the upperindus basin, Quarterly Journal of the Geological Society 29 (1-2) (1873) 441–471.
 W. B. Bull, The alluvial-fan environment, Progress in physical geography 1 (2) (1977) 222–270.
 A. T. Arguden, K. S. Rodolfo, Sedimentologic and dynamic differences between hot and cold laharic debris flows of mayon volcano, philippines, Geological Society of America Bulletin 102 (7) (1990) 865–876.
 J. E. Costa, Physical geomorphology of debris flows, in: Developments and applications of geomorphology, Springer, 1984, pp. 268–317.
 G. L. Gallino, T. C. Pierson, The 1980 polallie creek debris flow and subsequent dam-break flood, east fork hood river basin, oregon, Tech. rep., US Geological Survey, (1984).
 NASA, Typhoon koppu’s deadly philippine rainfall, Online, accessed at: http://pmm.nasa.gov/extreme-weather/typhoon-koppus-deadlyphilippine- rainfall (2015).
 R. M. Iverson, S. P. Schilling, J. W. Vallance, Objective delineation of lahar-inundation hazard zones, Geological Society of America Bulletin 110 (8) (1998) 972–984.
 A. Lagmay, K. Rodolfo, F. Siringan, H. Uy, C. Remotigue, P. Zamora, M. Lapus, R. Rodolfo, J. Ong, Geology and hazard implications of the maraunot notch in the pinatubo caldera, philippines, Bulletin of volcanology 69 (7) (2007) 797–809.
 M. Jakob, A size classification for debris flows, Engineering geology 79 (3) (2005) 151–161.
 E. L. King, H. P. Sejrup, H. Haflidason, A. Elverhøi, I. Aarseth, Quaternary seismic stratigraphy of the north sea fan: glacially-fed gravity flow aprons, hemipelagic sediments, and large submarine slides, Marine Geology 130 (3) (1996) 293–315.
 A. Welsh, T. Davies, Identification of alluvial fans susceptible to debris flow hazards, Landslides 8 (2) (2011) 183–194.
 T. R. Davies, M. J. McSaveney, et al., Principles of sustainable development on fans, Journal of Hydrology (New Zealand) 47 (1) (2008) 43.
 K. S. Rodolfo, A. T. Arguden, Rain-lahar generation and sediment delivery systems at mayon volcano, philippines.
 R. A. Arboleda, M. Martinez, C. Newhall, R. Punongbayan, 1992 lahars in the pasig-potrero river system, Fire and Mud: Eruptions and Lahars of Mount Pinatubo (1996) 85–102.
 N. M. Tuñgol, M. T. M. Regalado, Rainfall, acoustic flow monitor records, and observed lahars of the sacobia river in 1992, Fire and Mud: Eruptions and Lahars of Mt. Pinatubo, Philippines.
 N. Caine, The rainfall intensity: duration control of shallow landslides and debris flows, Geografiska Annaler. Series A. Physical Geography (1980) 23–27.
 F. Guzzetti, S. Peruccacci, M. Rossi, C. P. Stark, The rainfall intensity–duration control of shallow landslides and debris flows: an update, Landslides 5 (1) (2008) 3–17.
 C. J. van Westen, a. S. Daag, Analysing the relation between rainfall characteristics and lahar activity at Mount Pinatubo, Philippines, Earth Surface Processes and Landforms 30 (2005) 1663–1674. doi:10.1002/ esp.1225.
 E. Paguican, A. M. F. Lagmay, K. Rodolfo, R. Rodolfo, A. Tengonciang, M. Lapus, E. Baliatan, E. Obille Jr, Extreme rainfall-induced lahars and dike breaching, 30 november 2006, mayon volcano, philippines, Bulletin of volcanology 71 (8) (2009) 845–857.