A LiDAR-based flood modelling approach for mapping rice … · 2018. 12. 3. · A LiDAR-based...

14
A LiDAR-based ood modelling approach for mapping rice cultivation areas in Apalit, Pampanga Luigi L. Toda a, c , John Colin E. Yokingco a, * , Enrico C. Paringit b , Rodel D. Lasco a, d a The Oscar M. Lopez Center for Climate Change Adaptation and Disaster Risk Management Foundation, Inc., Pasig City, Philippines b Phil-LiDAR 1 Hazard Mapping of the Philippines using LiDAR, National Engineering Center, University of the Philippines, Quezon City, Philippines c Crawford School of Public Policy, The Australian National University, Canberra, ACT, Australia d World Agroforestry Center, International Rice Research Institute, Los Ba~ nos, Laguna, Philippines article info Article history: Received 23 November 2015 Received in revised form 22 March 2016 Accepted 27 December 2016 Keywords: Flood modelling LiDAR GIS Rice cultivation areas abstract Majority of rice cultivation areas in the Philippines are susceptible to excessive ooding owing to intense rainfall events. The study introduces the use of ne scale ood inundation modelling to map cultivation areas in Apalit, a rice-producing municipality located in the province of Pampanga in the Philippines. The study used a LiDAR-based digital elevation model (DEM), river discharge and rainfall data to generate ood inundation maps using LISFLOOD-FP. By applying spatial analysis, rice cultivation zone maps were derived and four cultivation zones are proposed. In areas where both depth and duration exceed threshold values set in this study, varieties tolerant to stagnant ooding and submergence are highly recommended in Zone 1, where ood conditions are least favorable for any existing traditional lowland irrigation varieties. The study emphasizes that a decline in yield is likely as increasing ood extents and longer submergence periods may cause cultivation areas for traditional irrigated lowland varieties to decrease over time. This decrease in yield may be prevented by using varieties most suitable to the ooding conditions as prescribed in the rice zone classication. The method introduced in this study could facilitate appropriate rice cultivation in ood-prone areas. © 2017 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). 1. Introduction Climate change impacts such as ooding pose threats to food production and agricultural communities in Asia. Located in oodplains, majority of rice cultivation areas in the Philippines are susceptible to severe ooding due to excessive rainfall. The Philippines is one of the countries most exposed to typhoons and oods due to its location in the western rim of the Pacic Ocean. An estimated 20 typhoons pass through the Philippine Area of Re- sponsibility (PAR) every year (PAGASA, 2009). These typhoons are usually accompanied by heavy rains which can cause adverse ef- fects particularly ooding. There are a total of 18 major river basins in the Philippines, covering approximately 36% of the country's landmass (DENR). These river basins provide a natural source of irrigation favorable for agricultural activities, of which approximately 32% of the countrys population is involved in (The World Bank, 2012). Areas found within these river systems are regularly ooded com- pounded by past ood events (Zoleta-Nantes, 2000). Notwith- standing its susceptibility to ooding, many of the country's highly populated areas lie in the banks of river systems, Pampanga river basin included. The effect of climate change has increased the fre- quency and intensity of ooding events and is projected to have adverse effects on crop production particularly in subsistence sec- tors (IPCC, 2007). Flooding is considered a major problem partic- ularly for rice production in the Philippines. From 2007 to 2010 alone, an annual average of 23.76 million US dollars of damage to rice farming in the Philippines was attributed to ooding (Israel, 2012). Consequently, ood damage can lead to long term impov- erishment of affected individuals (Arnall, Thomas, Twyman, & Liverman, 2013). Farmers are particularly affected because in the long run ood risks affect the viability of crop cultivation as a source of livelihood (Gwimbi, 2009). With 12 million farmers dependent on rice cultivation for livelihood (Altoveros & Borromea, 2007), the need to develop strategies to adapt to changing ood conditions now becomes a great priority (Mitin, 2009). Although rice are cultivated in ooded elds, prolonged sub- mergence can cause serious damage (Ram et al., 2002). * Corresponding author. E-mail address: [email protected] (J.C.E. Yokingco). Contents lists available at ScienceDirect Applied Geography journal homepage: www.elsevier.com/locate/apgeog http://dx.doi.org/10.1016/j.apgeog.2016.12.020 0143-6228/© 2017 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Applied Geography 80 (2017) 34e47

Transcript of A LiDAR-based flood modelling approach for mapping rice … · 2018. 12. 3. · A LiDAR-based...

  • A LiDAR-based flood modelling approach for mapping rice cultivationareas in Apalit, Pampanga

    Luigi L. Toda a, c, John Colin E. Yokingco a, *, Enrico C. Paringit b, Rodel D. Lasco a, d

    a The Oscar M. Lopez Center for Climate Change Adaptation and Disaster Risk Management Foundation, Inc., Pasig City, Philippinesb Phil-LiDAR 1 Hazard Mapping of the Philippines using LiDAR, National Engineering Center, University of the Philippines, Quezon City, Philippinesc Crawford School of Public Policy, The Australian National University, Canberra, ACT, Australiad World Agroforestry Center, International Rice Research Institute, Los Ba~nos, Laguna, Philippines

    a r t i c l e i n f o

    Article history:Received 23 November 2015Received in revised form22 March 2016Accepted 27 December 2016

    Keywords:Flood modellingLiDARGISRice cultivation areas

    a b s t r a c t

    Majority of rice cultivation areas in the Philippines are susceptible to excessive flooding owing to intenserainfall events. The study introduces the use of fine scale flood inundation modelling to map cultivationareas in Apalit, a rice-producing municipality located in the province of Pampanga in the Philippines. Thestudy used a LiDAR-based digital elevation model (DEM), river discharge and rainfall data to generateflood inundation maps using LISFLOOD-FP. By applying spatial analysis, rice cultivation zone maps werederived and four cultivation zones are proposed. In areas where both depth and duration exceedthreshold values set in this study, varieties tolerant to stagnant flooding and submergence are highlyrecommended in Zone 1, where flood conditions are least favorable for any existing traditional lowlandirrigation varieties. The study emphasizes that a decline in yield is likely as increasing flood extents andlonger submergence periods may cause cultivation areas for traditional irrigated lowland varieties todecrease over time. This decrease in yield may be prevented by using varieties most suitable to theflooding conditions as prescribed in the rice zone classification. The method introduced in this studycould facilitate appropriate rice cultivation in flood-prone areas.© 2017 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND

    license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

    1. Introduction

    Climate change impacts such as flooding pose threats to foodproduction and agricultural communities in Asia. Located infloodplains, majority of rice cultivation areas in the Philippines aresusceptible to severe flooding due to excessive rainfall. ThePhilippines is one of the countries most exposed to typhoons andfloods due to its location in the western rim of the Pacific Ocean. Anestimated 20 typhoons pass through the Philippine Area of Re-sponsibility (PAR) every year (PAGASA, 2009). These typhoons areusually accompanied by heavy rains which can cause adverse ef-fects particularly flooding.

    There are a total of 18 major river basins in the Philippines,covering approximately 36% of the country's landmass (DENR).These river basins provide a natural source of irrigation favorablefor agricultural activities, of which approximately 32% of thecountry’s population is involved in (The World Bank, 2012). Areas

    found within these river systems are regularly flooded com-pounded by past flood events (Zoleta-Nantes, 2000). Notwith-standing its susceptibility to flooding, many of the country's highlypopulated areas lie in the banks of river systems, Pampanga riverbasin included. The effect of climate change has increased the fre-quency and intensity of flooding events and is projected to haveadverse effects on crop production particularly in subsistence sec-tors (IPCC, 2007). Flooding is considered a major problem partic-ularly for rice production in the Philippines. From 2007 to 2010alone, an annual average of 23.76 million US dollars of damage torice farming in the Philippines was attributed to flooding (Israel,2012). Consequently, flood damage can lead to long term impov-erishment of affected individuals (Arnall, Thomas, Twyman, &Liverman, 2013). Farmers are particularly affected because in thelong run flood risks affect the viability of crop cultivation as asource of livelihood (Gwimbi, 2009). With 12 million farmersdependent on rice cultivation for livelihood (Altoveros& Borromea,2007), the need to develop strategies to adapt to changing floodconditions now becomes a great priority (Mitin, 2009).

    Although rice are cultivated in flooded fields, prolonged sub-mergence can cause serious damage (Ram et al., 2002).

    * Corresponding author.E-mail address: [email protected] (J.C.E. Yokingco).

    Contents lists available at ScienceDirect

    Applied Geography

    journal homepage: www.elsevier .com/locate/apgeog

    http://dx.doi.org/10.1016/j.apgeog.2016.12.0200143-6228/© 2017 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

    Applied Geography 80 (2017) 34e47

  • Submergence can cause mechanical damage to the plants, but moreimportantly it inhibits desirable bio-chemical processes of theplant-like photosynthesis (Jackson & Ram, 2003; Ram et al., 2002;Setter et al., 1997). Multiple approaches to address these impacts offlooding to rice have been put in place. One of these is the devel-opment of flood-tolerant varieties (Septiningsih et al., 2009) thatexhibit desirable physiological traits helpful for the plant’s survival(Setter et al., 1997) and are embodied in the sub1-a gene (Bailey-Serres et al., 2010). The sub1 gene has been isolated and bred intorice varieties (Khanh, Linh, Linnh, Ham, & Xuan, 2013) includingthose available in the Philippines such as the IR64 sub1(International Rice Research Institute (IRRI) 2009) and NSIC Rc 194(PhilRice, 2010).

    Submergence-tolerant rice varieties can survive up to twoweeks of complete submergence without any adverse effects onproduction (IRRI, 2009). However, while tolerance to submergenceis a desirable quality, other traits should be considered for adaptiveagricultural strategies. For instance, farmers have other trait pref-erences such as yield and panicle quality (Manzanilla et al., 2010).Also, depending on the flooding conditions, for instance those thatlast for more than one week and with depth of one meter or more,would require varieties that are tall or have stems that rapidlyelongate (Mackill et al., 2010).

    In order to optimize the use of submergence tolerant varieties,the flooding conditions should be identified and mapped. Byidentifying the flood extent and duration in rice cultivation areas,potential damage to property and/or crops can be estimated andother countermeasures can be developed (Chau, Holland, Cassels,&Tuohy, 2013; Wang, Colby, & Mulcahy, 2002; Thieken, Merz,Kreibach, & Apel, 2006). Flood conditions are usually character-ized through flood inundation modelling and are then displayedthrough flood hazard maps. The hazard categories used in thesemaps are a function of the product of depth and extent (Baky,Zaman, & Khan, 2012) and/or velocity (Daffi, Otun, & Ismail,2014; Kourgialas & Karatzas, 2014; Purwandari, Hadi, & Kingma,2011; Toriman et al., 2009). These hazard maps are then used formitigation and preparedness purposes such as hazard assessments(Daffi et al., 2014; Purwandari et al., 2011), flood propagation andhazard estimation (Kourgialas & Karatzas, 2014), risk management(Baky et al., 2012), and flood management (Toriman et al., 2009).These hazard maps can take into consideration possible damageand potential risks to agriculture (Daffi et al., 2014; Kourgialas &Karatzas, 2014; Baky et al., 2012) but may not necessarily suggestadaptation options. Flood modelling has been used to estimatepotential impact on agricultural land through either mapping floodextents in different rain return scenarios (Chau et al., 2013) orcombining extent with flood duration (Hamdani & Kartiwa, 2014).Flood models have also been used in combination with risk man-agement to specifically optimize rice planning (Samantaray,Chatterjee, Singh, Gupta, & Panighry, 2015) using flood extent,duration, and depth to map suitable environments for specific ricevarieties.

    Flood inundation models entail the use of flood depth, extent,duration and flow velocity based on advanced algorithm and high-quality data (Samantaray et al., 2015). Flood modelling softwaresuch as Lisflood, HEC-RAS, and MIKE typically require rainfall,discharge, topography, and Digital Elevation Model (DEM) data togenerate flood models. The accuracy of the models depends on thequality of input data and the quality of DEMs are considereddirectly proportional to the reliability of the models (Brandt & Lim,2012). To increase the accuracy of flood models, high resolutionDEMs from sources such as LiDAR are highly favored (Sole, Giosa,Nol�e, Medina, & Bateman, 2008). LiDAR is a remote sensing tech-nology that uses rapid laser pulses to map out the surface of theearth and can be used to create high resolution terrain and surface

    DEMs. LiDAR has been widely used because of its vertical accuracyof ± 15 centimeter and spatial resolution of up to 1 meter. ThePhilippines has recently obtained airborne LiDAR technology(Paringit, Fabila,& Santillan, 2012) tomapmajor river basins, with amuch higher degree of accuracy than what is currently available.

    Apalit is one of the rice producing municipalities of Pampangawhich suffer from prolonged inundation and has a relatively flatterrain due to its location within a flood plain. For the purpose ofthis study, LiDAR surface models were used in the flood simulationin the municipality of Apalit. Other topographic mapping technol-ogies have difficulty in capturing flat terrains (Sole et al., 2008), butLiDAR is able to produce highly accurate flood models due to itscapability to capture gentle terrain changes. The aim of this study isto develop LiDAR-derived flood models to map rice cultivationareas based on depth and duration. The maps allow the identifi-cation of suitable rice varieties for each rice cultivation map. Thestudy therefore provides a tool in determining appropriate areas forrice cultivation and in selecting and developing suitable rice vari-eties based on different flooding condition sets.

    2. Study area

    The study was conducted in Apalit, one of the 20 municipalitiesin the Province of Pampanga (Fig. 1). Apalit is located in the southeastern portion of the province and has a total area of 6147 ha withan estimated population of 101,537 (NSCB, 2010). The municipalityhas an estimated 3799.95 has of agricultural land, as derived fromdigitized orthophotos. The rice-producing barangays (villages) areBalucuc, Calantipe, Cansinala, Capalangan, Paligui, Sampaloc, SanJuan, San Vicente, Sucad, Sulipan, and Tabuyuc. Barangay Colganteand portions of barangay Paligui practice aquaculture throughfishponds.

    The topography of the municipality is characterized by largelyflat and low with slopes ranging from 0 to 3% and an elevation notexceeding 20 m (Municipal Government of Apalit, 2000). It is alsoalluvial, dominated by streams and rivers attributed to its locationwithin the Pampanga floodplain. The Pampanga River runs throughthe mid-eastern portion of the municipality and is a source ofirrigation in the municipality (Fig. 1). Its cropping cycle runs be-tween the dry season from November to December and the wetseason from June to August. An average of 644 mm of rain isexperienced during the wet season. The riverine morphology andflat terrain makes the municipality favorable for rice cultivationwhile rendering it prone to flooding during the wet season. ThePinatubo eruption have also significantly altered flooding condi-tions (Siringan & Rodolfo, 2003). Flooding is largely characterizedby surface water inundation worsened by heavy rainfall during thewet season. This is as verified from the focus group discussion(FGD) and reports from Pampanga River Basin Flood Forecastingand Warning Center (2012).

    3. Data used

    3.1. LiDAR DEM

    The airborne LiDAR data used for the study areawas acquired bythe University of the Philippines Disaster Risk and ExposureAssessment for Mitigation (UP-DREAM) Program from October toDecember 2012. The data was processed to produce a 1-meterspatial resolution for the Digital Surface Model (DSM) and theDigital Terrain Model (DTM) (Fig. 2) with a point density of two tothree points per square meter. The LiDAR point cloud data wasprocessed and refined using Terra Scan (Terrasolid, Inc.) to elimi-nate stray points and normalizing elevations throughout the flightpaths.

    L.L. Toda et al. / Applied Geography 80 (2017) 34e47 35

  • The final DEM used for the model was resampled from the 1-meter resolution DSM into a 5-meter resolution DEM to fit the

    flood model resolution limit of 12 million cells. The resampledmodel does not affect the resolution accuracy since the land use of

    Fig. 1. Location map of Pampanga.

    Fig. 2. LiDAR digital terrain model (DTM) with elevation values and outline of Pampanga River.

    L.L. Toda et al. / Applied Geography 80 (2017) 34e4736

  • the municipality is largely agricultural and does not greatly affectflood behavior as opposed to urban areas (Tamiru& Rientjes, 2005).

    3.2. Rainfall and river discharge data

    The Rainfall Intensity-Duration-Frequency (RIDF) analysis datafor the flood models and the hypothetical, extreme validationmodel were obtained from the Philippine Atmospheric, Geophys-ical and Astronomical Administration (PAGASA). The RIDF datawere spatially interpolated across the region through the theissenpoligon technique in ArcGIS 10.2. Port Area rain gauge station wasselected.

    To produce the discharge for the flood scenarios, RIDF analysisdata gathered by PAGASA and water-level readings from ASTI wereused as inputs in the 2D HEC-HMS flood models of Pampanga forthe different rainfall-runoff scenarios. Table 1 shows peak values ofoutflow and rainfall at the Sto. Ni~no station for the 5, 25 and 100year rain return periods.

    3.3. Land cover and roughness coefficient

    Land cover was manually classified and digitized at a scale of

    1:1250 from a combination of orthophotos obtained by UP-DREAMProgram from October to December 2012 and stitch maps takenfrom Google Earth™ (Fig. 5). Land cover classes were assigned withappropriate roughness coefficients based on Chow (1959) andstandardized by Brunner (2010).

    4. Methodology

    4.1. Framework for rice cultivation area mapping (Fig. 3)

    4.1.1. Flood model developmentThe development of the flood models for the Pampanga River

    Basinwas done using LISFLOOD-FPmodel (version 5.9.6) developedby the School of Geographic Sciences of the University of Bristol.LISFLOOD-FP is capable of integrating 1D and 2D parameters into asingle model and can depict not only riverine flooding, but alsosurface flooding. For 1D parameters, the study used LISFLOOD-FP'sdiffusive solver. The diffusive solvermodels the behavior of the flowof water within a river or stream over time by running thedischarge values through the channel and then computing the ef-fects to water depth within the channel in relation to the channel'scross-section, friction values, and gravity. When the river depth

    Table 1Peak values of the Sto. Ni~no outflow.

    RIDF Period Total Precipitation (mm) Peak Rainfall (mm. in 10 min) Sto. Ni~no Peak Outflow (m3/s) Time to Peak

    5yr 223.5 29.7 4591.2 2 days, 1 h, 12 min25yr 328.8 41.4 7646.8 2 days, 2 h, 50 min100yr 415.6 51.0 10,210.4 2 days, 6 h, 10 min

    Source 1 Philippine Astronomical, Geophysical and Atmospheric Administration (PAGASA) and Disaster Risk and Exposure for Assessment of Mitigation (DREAM) Program.

    Run LISFLOOD using inputs to produce time series of

    water depths and total inundation time outputs

    Stack time series images and convert to .img file

    Rice cultivation zone map

    Validate model results using historical flood maps and focus group discussion

    Obtain flood modelling inputdata: rainfall-discharge, LiDAR DEM,

    roughness coefficients, river cross-section

    Calculate mean depth by using stacked time series .img file as input in compute mean

    statistics tool in ENVI

    Input mean depth raster and total inundation time output in

    Python program to produce zone map

    Clip zone map extent to rice cultivation extent

    t

    d

    Fig. 3. Framework for rice cultivation area mapping.

    L.L. Toda et al. / Applied Geography 80 (2017) 34e47 37

  • reaches the bank, water flows to the adjacent cells within the DEM.The process is described in the 1D diffusive wave equation (1)(Bates, Trigg, Neal, & Babrowa, 2013) where Qx is volumetric flowrate at a given time t in the x Cartesian direction, A is the crosssectional area of flow, h is the water depth, g is gravity, n is theManning's friction value, R is the hydraulic radius of the river sec-tion, t is time, and x the distance in the x Cartesian direction whichcan also predict backwater effects. The acceleration solver was usedfor the 2D parameters which calculated flows between cells as afunction of friction and water slopes, and local water acceleration.

    vQxvt|{z}

    local acceleration

    þ vvt

    Q2xA

    !|fflfflfflfflfflffl{zfflfflfflfflfflffl}

    convective acceleration

    þ gA vðhþ zÞvx|fflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflffl}

    water slope

    þ gn2Q2x

    R4=3A|fflfflffl{zfflfflffl}friction slope

    (1)

    Input data for the model included the rainfall and dischargedata, raster DEM, river profile, channel and floodplain friction, andmodel time step. Model simulation time was approximated basedon discharge duration. The discharge simulation time was identi-fied at 900,000 s (equivalent to 10.4 days). The input data wereconverted to text formats with specific file extensions that the

    model would read based on the parameter file. The data re-quirements and formats are described in Table 2. The model has thecapacity to generate time-series output images of inundationextent and depth, maximum depth, hazard maps, flow velocity, andtotal inundation time. For the purpose of this study, time seriesoutput images were used along with the total inundation time. Atotal of 1500 time series images were produced for each rain returnscenario.

    4.2. Model validation

    Flood maps showing observed flood events and their extentswere used to subjectively validate the resultant flood models bycomparing actual flood accounts from key informant interviews(KIIs) and focus group discussion (FGD). Observed flood maps wereused for visual comparison with the resultant flood models. Thesemaps were gathered from MODIS satellite images acquired onDecember 4 and 9, 2004 and were digitized using ArcGIS 10.2. Inparticular, model simulated for Habagat 2013 (south-westmonsoon rain) (Fig. 4) event was generated using LISFLOOD-FP forcomparison and validation purposes.

    Table 2Model data requirements, format and description.

    Data requirement Format Description of format (based on Bates et al., 2013)

    Parameter file .par contains the information necessary to run the simulation including file names and locations and the main model and run controlparameters.

    Rainfall .rain specifies a time-varying rainfall rate.Discharge .bdy specifies time varying boundary conditions associated with varying discharge values per time step in the river.Raster LiDAR DEM .dem.ascii consists of a 2D raster array of ground elevations in ARC ascii raster formatRiver profile .river river characteristics described in terms of its width, Manning's n friction coefficient and bed elevationRoughness

    coefficients.n.ascii channel and floodplain friction - based on Chow (1959) and standardized by Brunner (2010) in a 2D array of friction values in ARC ascii

    raster format.Model time step (seconds) user defined time step

    Fig. 4. Simulated maximum flood depth of south-west monsoon rain (Habagat) in August 2013.

    L.L. Toda et al. / Applied Geography 80 (2017) 34e4738

  • 4.2.1. Focus group discussion and key informant interviewsAn FGD was conducted with farmers not only to validate the

    flood maps with their experiences on flooding but also to elicitinformation about local agricultural adaptation practices and vari-etal types used. Twenty-five among the invited thirty six farmersacross all barangays (villages) took part in the FGD. These farmersrepresented nine of the 11 barangays that were mostly agricultural,namely Colgante, Capalangan, Calantipe, San Juan, Cansinala, Tab-uyuc, Sampaloc, Balucuc, and Sucad. Despite lack of representationfrom two barangays, the answers were treated collectively at themunicipal level, not per barangay. Activities included introductionof FGD team and participants, overview of the topic, explanation ofthe FGD process and ground rules, and discussion guided by semi-structured questions. Specific information gathered included (1)historical accounts on flood duration and extent; (2) agriculturaladaptation practices; (3) cropping patterns/calendar; and (4) ricevarieties being used with respect to characteristics, yield, cost andcultivation area coverage.

    In addition to the FGD, semi-structured KIIs were also con-ducted on multiple visits to obtain information on actual/observedflood records, flood impacts on agriculture, existing agriculturaladaptation practices and land use changes, among others. To gatherthese information, the Municipal Agricultural Officer (MAO), theMunicipal Planning Officer (MPO) and a fewMunicipal AgriculturalTechnologists (MATs) were chosen for the KIIs because of the depthof their knowledge and experience on the subject on which theywere interviewed.

    4.2.2. Estimation of existing rice cultivation extentThe rice cultivation area (Fig. 6) for Apalit was digitized from

    interpretation of land use (Fig. 5) from orthophotos that were takenwith the LiDAR data. This was used in conjunctionwith the rice areamaps obtained from IRRI using MODIS images obtained from 2000to 2012.

    4.3. Classification of cultivation areas for rice varieties

    ArcGIS 10.2 was used to overlay the map layers of rice cultiva-tion area and the extents of inundation time and mean depth togenerate the zone maps. The total area affected by inundation wascalculated by clipping the flood extent given total inundation timeand mean depth. The average depth over the whole inundationtime provides a picture of howdepthwould affect the rice crop overtime. Mean depth was calculated by stacking thewater depth resultproduced from LISFLOOD-FP as raster time-series images and thenthe image processing software ENVI™ version 4.7 computed theaverage of those images. Total inundation time and mean depthwere integrated using a Python script to produce a raster repre-senting the different depth-duration conditions (Table 3) and wasthen converted into polygons and overlain with the rice cultivationarea. The conditions used in the zoning were based on the depththreshold specified for shallow water rice in Mackill et al. (2010)and the duration threshold based on the short duration submer-gence specified in Salam, Biswas, and Rahman (2004).

    5. Results and discussion

    5.1. Flood inundation models

    Fig. 7 shows the west-east elevation (section A-B) profile from

    Fig. 5. Land use classification map of Apalit municipality.

    L.L. Toda et al. / Applied Geography 80 (2017) 34e47 39

  • the approximatemid-point of the river. It illustrates howPampangaRiver is planked by alluvial deposits that act as natural levees whichimpered overflow from the Pampanga River. Increased river runoffin the river partly elucidates why no significant overflow wasobserved during rainfall events as confirmed during the FGDs. Theinundationmodels indicate increase in affected areas for both floodinundation time and mean depth with rain-return period (Figs. 8and 9). Of particular interest are the expanded coverage of floo-ded areas located in the southwestern portion with increasing re-turn periods. The apparent absence of overtopping in theinundation models of Pampanga River is consistent with the ac-counts obtained from the FGD.

    5.2. Local rice varieties

    There are a total of seven rice varieties that farmers usethroughout the year depending on the season, one of which (IR64)is submergence-tolerant. IR64 was introduced and adopted bymajority of the farmers in the 1990s, but is no longer being grown

    due to its low pest tolerance. In contrast, NSIC Rc 150 has a highyield and good eating quality apart from its tolerance to submer-gence. However, it fell out of usage due to age and thus was dis-continued and replaced with other varieties by local farmersaccording to our interview (Manzanilla, personal communication).With its unintended floodresistant characteristic, NSIC Rc150 mayprove to be a better substitute to IR 64 due to its lowcost, high yield,and short days to maturity. This enables farmers particularly in thebarangays of Balucuc and Calantipe and some portions of Tabuyucand Cansinala to maximize their resources and have the opportu-nity to harvest at the most three times a year.

    5.3. Observed flood conditions

    Contrasting flooding conditions were found between theeastern and western portions of the municipality. The easternportion has greater flood depths, but recedes faster. Two of thebarangays, Calantipe and Balucuc, and some portions of Tabuyuc,Capalangan, and Cansinala do not experience detrimental flooding(Fig. 10) attributed to their comparatively higher elevation as dis-cussed in the FGDs. Barangays in the western portion have shal-lower flood depths, but have very long submergence periods. Theworst case is found in the Barangay of Colgante, where observedflooding lasts between 4 and 7 months. The severe flooding con-ditions are also seen as a result of the conversion from agriculturallands to fish ponds of a portion of barangays Paligui and Colgante.According to the FGD participants, the continuous outflow of laharfrom Mt. Pinatubo and the sea-level rise originating from ManilaBay are major contributors to the constricted stream system whichengender the current flooding conditions. It is estimated that

    Fig. 6. Extent of areas planted to rice in Apalit, Pampanga.

    Table 3Zone value matrix representing the flood depth and duration conditions.

    Zone Conditions

    Depth (cm) Mean total duration (days)

    1 �20 �72 �20 �73 �20

  • around 1500 has or 25% of Apalit will be affected by a two-meterrise in sea level (PEMSEA, 2012).

    5.4. Rice cultivation at risk

    Rice cultivation in Apalit is at risk as low-lying barangaysexperience prolonged submergence that extend up to four monthsduring the wet season (June to August). During the FGD, farmersstated that heavy rainfall was the main cause of submergence ofrice fields. In a typical year, floods affect between 50% and 80% ofrice cultivation areas across barangays. Based on the cultivationzone maps (Fig. 11), increasing trend in flood extents of varyingflood depths is observed for 5-year to 100-year rain-return periods.There will also be an increase in submergence period as moreintense rains occur more frequently.

    Generally, the varying flooding conditions between the easternand western barangays may indicate the differences in their agri-cultural practices and adaptation strategies. The eastern barangayssuch as Calantipe and Balucuc are able to harvest up to a maximumof three times a year because flooding does not affect them whilebarangays on the western side have reduced their cropping season

    to only once each year. Farmers in Colgante and Paligui have evenresorted to converting their lands to fishponds due to recurringfloods.

    5.5. Cultivation zone classification

    Cultivation areas were classified into four zones (Fig. 12) basedon four flooding conditions. Zone 1 is suitable for flood-tolerantvarieties and those tolerant to stagnant flooding. Depth and dura-tion in this zone exceed the threshold values set in this study andare least favorable for existing traditional lowland irrigation vari-eties. Between 10 and 21% of current cultivated land should growonly varieties that can withstand flooding conditions beyondshallow depth and 7-day flood duration limit (Fig. 13). On the otherhand, there is a decreasing pattern of extent for traditional lowlandirrigation varieties, which means more areas might need to adoptvarieties with known submergence tolerance traits that last longerthan seven days and in depths greater than 20 cm. Areas whichmight need of these varieties are located in the south easternportion mainly in Capalangan and Tabuyuc. Sucad may also adoptthese varieties as the area for Zone 3 decreases in the barangay per

    Fig. 7. (a) Elevation map with (b) cross-section of Apalit along the section A-B.

    L.L. Toda et al. / Applied Geography 80 (2017) 34e47 41

  • Fig. 8. Mean depth for (a) 5-, (b) 25- and (c) 100-year rain-return periods.

    L.L. Toda et al. / Applied Geography 80 (2017) 34e4742

  • rainfall scenario.Accounts from the FGD indicate that inundation could extend up

    to several months. The need for rice varieties also tolerant tostagnant water for weeks to several months regardless of depth

    tolerance is crucial. A decline in yield will be likely as cultivationareas for traditional lowland irrigated varieties may decrease overtime due to expanding flooding and submergence. This decrease inyield may be prevented by planting varieties most suitable to the

    Fig. 9. Total inundation time for (a) 5-, (b) 25- and (c) 100-year rain-return periods.

    L.L. Toda et al. / Applied Geography 80 (2017) 34e47 43

  • environment as prescribed in the rice zone classification. However,more than one of the conditions can ensue in any particular flood-prone environment. Therefore, it is desirable to develop varietiesthat have a combination of tolerance traits for flood-prone areas.

    Traditional (local) varieties aremostly grown in themunicipalitydespite flood risks. Adoption of submergence-tolerant varietieshowever presents some drawbacks considering their cost and lowtolerance to pest. As some cultivation areas already suffer stagnantinundation of at least a month, medium to deepwater rice varietiesmight thrive for any flood scenario. However, more than one of theconditions can ensue in any particular flood-prone environment.Therefore, it is desirable to develop varieties which possess acombination of tolerance traits for flood-prone areas whenpossible. The potential re-adoption of IR64 and sustainability of useof other submergence-tolerant varieties will be critical in thedevelopment and accessibility of flood-resistant rice cultivars astheir adoption will depend on their productivity, yield, cost andsocio-economic factors. Alongside the development of low-costhigh-yielding flood-tolerant rice varieties, it is crucial that accessto sustainable agricultural services is in place. This, however, willalso depend on the availability of developed flood-tolerantcultivars.

    6. Conclusion

    The study has demonstrated the use of LiDAR technology forproducing fine scale flood inundation maps to classify cultivationareas for certain rice varieties, the results of which could facilitateagricultural adaptation and help strengthen food security in Apalit,a rice-producing municipality in Pampanga that is generally char-acterized by surface flooding. Using LISFLOOD-FP model, inunda-tion depth and duration conditions were simulated fromintegrating river discharge and rainfall data and LiDAR-based DEM.These conditions were chosen as they are crucial in the survival andproductivity of rice varieties. The inundation models provide an

    illustration of the variation in the extent of flood depth and dura-tion across rice cultivation areas given a rainfall scenario. Based onthemodels, there is a potential increase in extent of flood depth andduration as rain-return periods increase. The study found that adecline in yield is likely as cultivation areas for traditional lowlandirrigated varieties may decrease over time due to increasing depthand longer submergence periods. This decrease in yield may beprevented by growing varieties that are more suitable to theenvironment as prescribed in the rice zone classification.

    Consequently, four zones are proposed. The flood-tolerant va-rieties and those tolerant to stagnant flooding are recommended inZone 1 where both depth and duration exceed the threshold valuesset in this study, meaning flood conditions are least favorable forany existing traditional lowland irrigation varieties. More than oneof the flooding conditions can ensue in any particular flood-proneenvironment, which makes it is desirable to develop rice varietiesthat possess a combination of tolerance traits for flood-prone areas.Apalit is naturally lowland so medium to deepwater rice varietiesmight thrive for any flood scenario as some parts of the munici-pality already suffer stagnant inundation of at least a month duringextreme rainfall events.

    The flood models and the resultant rice zone maps weredesigned to take advantage of LiDAR data sets generated throughthe efforts of Disaster Risk and Exposure Assessment for Mitigation(DREAM) Program of the University of the Philippines and theDepartment of Science and Technology (DOST). While the findingsof the study relies heavily on models, the need for their validationwith observed flood information prompted the conduct of FGD. TheFGD was also necessary to obtain relevant information on impactsof flood on rice cultivation, current rice adaptation strategies, andvarietal types used. Although LiDAR technology offers more accu-rate topographic detail for specific purposes such as flood inun-dation estimation, its accessibility and availability present somedrawbacks for studies that cover larger scale areas where floodingconditions are critical to rice cultivation. Alternatively, free and

    Fig. 10. Map of observed flood depths accounted during FGD.

    L.L. Toda et al. / Applied Geography 80 (2017) 34e4744

  • Fig. 11. Recommended varietal types and characteristics per cultivation zone.

    L.L. Toda et al. / Applied Geography 80 (2017) 34e47 45

  • accessible but scale-appropriate DEMs should be maximized usingthe methodology employed in this study.

    Acknowledgments

    This study has been done as a part of a project titled ‘‘FloodInundation Mapping using LiDAR and GIS’’ of the Oscar M. LopezCenter for Climate Change Adaptation and Disaster Risk Manage-ment Foundation, Inc., the Philippines (The Oscar M. Lopez Center).The authors would like to acknowledge the Department of Scienceand Technology (DOST) and the University of the Philippines (UP)through the Disaster Risk and Exposure Assessment for Mitigation(DREAM) Program for the LiDAR and discharge data and forallowing us to use their flood modelling software; NAMRIA for theadministrative boundary; and PAGASA for the rainfall information.The authors would also like to acknowledge Mark Joshua Salvacionof DREAM Program who helped in executing the zone maps.

    References

    Altoveros, N., & Borromea, T. (2007). Country report on the state of plant geneticresources for food and agriculture-Philippines. Rome: Food and Agriculture Or-ganization of the United Nations (FAO).

    Apalit Local Government. (2000). Comprehensive land use plan.

    Arnall, A., Thomas, D., Twyman, C., & Liverman, D. (2013). Flooding, ressetlement,and change in livelihoods: Evidence from rural Mozambique. The AuthorsJournal Compilation, 1e21.

    Bailey-Serres, J., Fukao, T., Ronald, P., Ismail, A., Heuer, S., & Mackill, D. (2010).Submergence tolerant rice: SUB1's journey from Landrace to modern cultivar.Rice, 138e147.

    Baky, A. A., Zaman, A. M., & Khan, A. U. (2012). Managing flood flows for cropproduction risk management with hydraulic and GIS modeling: Case study ofagricultural areas in Shariatpur. APCBEE Procedia, 318e324.

    Bates, P., Trigg, M., Neal, J., & Babrowa, A. (2013). LISFLOOD-FP user manual. Bristol:School of Geographical Sciences University of Bristol.

    Brandt, S., & Lim, N. (2012). Importance of river bank floodplain slopes on the ac-curacy of flood inundation mapping. River Flow 2012: Volume 2. In Proceedingsof the International Conference on fluvial hydraulics, San Jos�e, Costa Rica, 5-7September 2012 (pp. 1015e1020). The Netherlands: CRC Press/Balkema (Taylor& Francis).

    Brunner, G. W. (2010). HEC-RAS river analysis system: Hydraulic reference manual.Davis, California: US Army Corps of Engineers Institute of Water Resources.

    Chau, N. V., Holland, J., Cassels, S., & Tuohy, M. (2013). Using GIS to map impactsupon agriculture from extreme floods in Vietnam. Applied Geography, 65e74.

    Chow, V. T. (1959). Open channel hydraulics. New York: McGraw-Hill.Daffi, R. E., Otun, J. A., & Ismail, A. (2014). Flood hazard assessment of river dep

    floodplains in North-Central Nigeria. International Journal of Water Resourcesand Environmental Engineering, 67e72.

    Department of Environment and Natural Resources (DENR). (n.d.): 18 major riverbasins. Retrieved from river basin control office: http://rbco.denr.gov.ph/index.php/18-major-river-basins.

    Gwimbi, P. (2009). Linking rural community livelihoods to resilience building inflood risk reduction in Zimbabwe. Journal of Disaster Risk Studies, 71e79.

    Hamdani, A., & Kartiwa, B. (2014). Mapping and economic impact analysis of floodon rice field citarum watershed, Indonesia. In 6th International Conference onflood management, (pp. 1e11). Sao Paolo, Brazil.

    International Rice Research Institute (IRRI). (2009). Scuba Rice: Breeding floodtolerance into Asia's local megarice varieties. Retrieved August 8, 2013, fromhttp://r4d.dfid.gov.uk/PDF/Outputs/IRRI/DFID_impact_case_study_SUB1rice_FINAL%5B1%5D.pdf.

    IPCC. (2007). Climate change 2007: Impacts, adaptation and vulnerability. UnitedKingdom: Cambridge University Press.

    Israel, D. (2012). Typhoons, floods, and, droughts: Regional occurrence and value ofdamages to rice farming in the Philippines. Policy Notes Philippine Institute forDevelopment Studies, (15).

    Jackson, M., & Ram, P. (2003). Physiological and molecular basis of susceptibility andtolerance of rice plants to complete submergence. Annals of Botany, 227e241.

    Khanh, T. D., Linh, H. L., Linnh, T. H., Ham, L. H., & Xuan, T. D. (2013). Rapid and high-precision marker assisted backcrossing to introgress the SUB1 QTL into theVietnamese elite rice variety. Journal of Plant Breeding and Crop Science, 5(2),26e33.

    Kourgialas, N., & Karatzas, G. (2014). A hydro-sedimentary modeling system forflash flood propagation and hazard estimation under different agriculturalpractices. Natural Hazards and Earth System Sciences, 625e634.

    Mackill, D., Ismail, A., Pamplona, A., Sanchez, D., Carandang, J., & Septiningsih, E.(2010). Stress tolerant rice varieties for adaptation to a changing climate. Crop,Environment & Bioinformatics, 250e259.

    Manzanilla, D., Paris, T., Vergara, G. V., Ismail, A. M., Pandey, S., Labios, R. V., et al.(2010). Submergence risks and farmers' preferences: Implications for breedingSub1 rice in Southeast Asia. Agricultural Systems, 335e347.

    Mitin, A. (2009). Documentation of selected adaptation strategies to climate change inrice cultivation. Quezon City: East Asia Rice Working Group.

    National Statistical Coordination Board (NSCB). (2010). 2010 census of populationand housing. Quezon City, Philippines: National Statistics Office.

    Pampanga River Basin Flood Forecasting and Warning Center. (2012). Event: Ty-phoons “PEDRING” (NESAT) and “QUIEL” (NALGAE) September 26 to October 04,2011. San Fernando City: Department of Science and Technology.

    Paringit, E. C., Fabila, J. L. F., & Santillan, J. R. (2012). High-resolution digital elevationdataset derived from airborne lidar for flood hazard assessment and mappingapplications. In Proceedings of the 33rd Asian Conference on remote sensing(ACRS), Phuket, Thailand (in DVD).

    Partnerships in Environmental Management for the Seas of East Asia (PEMSEA).(2012). Integrating climate change and disaster risk scenarios into coastal land andsea use planning in Manila Bay. Quezon City, Philippines: Partnerships in Envi-ronmental Management for the Seas of East Asia (PEMSEA).

    Philippine Atmospheric, Geophysical and Astronomical Services Administration(PAGASA). (2009). Member report ESCAO/WMO typhoon committee. Philippineatmospheric, geophysical and astronomical services administration.

    PhilRice. (2010). Milestones 2010. Mu~noz, Nueva Ecija: Philippine Rice ResearchInstitute.

    Purwandari, T., Hadi, M., & Kingma, N. (2011). A GIS modelling approach for floodhazard assessment in part of Surakarta city, Indonesia. Indonesian Journal ofGeography, 63e80.

    Ram, P., Singh, B., Singh, A., Ram, P. P. N. S., Boamfa, A., & Singh, R. K. (2002).Submergence Tolerance in rainfed lowland rice: Physiological Basis and pros-pects for cultivar improvement through marker-aided breeding. Field CropResearch, 131e152.

    Salam, M. A., Biswas, P., & Rahman, M. A. (2004). Rice research and development inthe flood-prone ecosystem. Proceedings of the international workshop on plood-

    Fig. 12. Rice cultivation zone map for 5, 25 and 100-year rain-return periods.

    Fig. 13. Area percentages of zones per rainfall return period.

    L.L. Toda et al. / Applied Geography 80 (2017) 34e4746

  • prone rice systems in Gazipur. Bangladesh. Los Banos, Laguna (Philippines): In-ternational Rice Research Institute.

    Samantaray, D., Chatterjee, C., Singh, R., Gupta, P., & Panighry, S. (2015). Flood riskmodeling for optimal rice planning for delta region of Mahanadi river basin inIndia. Natural Hazards, 347e372.

    Septiningsih, E., Pamplona, A., Sanchez, D., Neeraja, C. N., Vergara, G., Heuer, S., et al.(2009). Development of submergence-tolerant rice cultivars: The Sub1 locusand beyond. Annals of Botany, 151e169.

    Setter, T., Ellis, M., Lureles, E., Ella, E., Senadhira, D., Mishra, S., et al. (1997). Phys-iology and genetics of submergence tolerance in rice. Annals of Botany, 67e77.

    Siringan, F., & Rodolfo, K. (2003). Relative sea level changes and worsening floods inthe western pampanga delta: Causes and some possible mitigation measures.Science Diliman, 1e12.

    Sole, A., Giosa, L., Nol�e, L., Medina, V., & Bateman, A. (2008). Flood risk modellingwith LiDAR technology. In Flood recovery, innovation and response (p. 600). WITPress.

    Tamiru, A., & Rientjes, T. (2005). Effects of LiDAR DEM resolution in flood modelling

    : A model sensitivity study for the city of Tegucigalpa. In Honduras. LaserScanning 2005 (pp. 168e173). International Society for Photogrammetry andRemote Sensing.

    The World Bank. (2012). Employment in agriculture (% of total employment).Retrieved from The World Bank http://data.worldbank.org/indicator/SL.AGR.EMPL.ZS.

    Thieken, A., Merz, B., Kreibach, H., & Apel, H. (2006). Methods for flood riskassessment: Concepts and Challenges. In International workshop on flash floodsin urban areas, (pp. 1e12). Muscat.

    Toriman, M., Gazim, M., Mokhtar, M., Mastura, S., Karim, O., & Aziz, N. (2009).Integration of 1-D hydrodynamic model and GIS Approach in flood manage-ment study in Malaysia. Research Journal of Earth Sciences, 22e27.

    Wang, Y., Colby, J., & Mulcahy, K. (2002). An efficient method for flood extent in acoastal floodplain using Landsat TM and DEM data. International Journal ofRemote Sensing, 3681e3696.

    Zoleta-Nantes, D. (2000). Flood hazards in Metro Manila: Recognizing commonal-ities, differences, and courses of action. Social Sciences Diliman, 60e105.

    L.L. Toda et al. / Applied Geography 80 (2017) 34e47 47