RESEARCH ARTICLE Assessing the effects of using high-quality data and high-resolution models in valuing flood protection services of mangroves Pelayo Menéndez ID1*, Iñigo J. Losada1, Saúl Torres-Ortega ID1, Alexandra Toimil1, Michael W. Beck2 1 Environmental Hydraulics Institute (IHCantabria), Universidad de Cantabria—Avda, Isabel Torres, Parque Cientı́fico y Tecnológico de Cantabria, Santander, Spain, 2 Institute of Marine Sciences, University California Santa Cruz, Santa Cruz, CA, United States of America a1111111111 a1111111111 * menendezfp@unican.es a1111111111 a1111111111 a1111111111 Abstract The rate of change on coastlines is accelerating from climate change and coastal develop- ment. Coastal flooding is a particularly pressing and increasing problem, which affects hun- OPEN ACCESS dreds of millions of people and damages trillions of US$ in property. Scientists, practitioners Citation: Menéndez P, Losada IJ, Torres-Ortega S, and managers must be able to quickly assess flood risk and identify appropriate adaptation Toimil A, Beck MW (2019) Assessing the effects of and risk reduction measures often with limited data and tools, particularly in developing using high-quality data and high-resolution models countries. To inform these decision-making processes, we identify how sensitive flood risk in valuing flood protection services of mangroves. and adaptation analyses are to changes in the resolution of data and models. We further do PLoS ONE 14(8): e0220941. https://doi.org/ 10.1371/journal.pone.0220941 these comparisons in the context of assess the benefits of an ecosystem-based approach for risk reduction. There is growing interest in these ecosystem-based approaches as cost Editor: Marco Fusi, KAUST University, SAUDI ARABIA effective measures for adaptation and risk reduction. We assess flood risks from tropical cyclones and the flood risk reduction benefits provided by mangroves in Pagbilao (the Philip- Received: October 9, 2018 pines). Then, we also compare risks and risk reduction (benefits) using different quality data Accepted: July 27, 2019 and models, to identify where to invest in in new modeling and data acquisition to improve Published: August 20, 2019 decision-making. We find that coastal flood risk valuation improves by using high resolution Copyright: © 2019 Menéndez et al. This is an open topography and long time series of data on tropical cyclones, while flood reduction benefits access article distributed under the terms of the of mangroves are better valued by using consistent databases and models along the whole Creative Commons Attribution License, which process rather than investing in single measures. permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability Statement: All relevant data are included within the paper and its Supporting Information files. The input and output data for the Introduction flood risk model are available at: https://osf.io/ Coastal flooding effects are expected to enhance considerably during the 21st century, due to s5v7k/ (DOI 10.17605/OSF.IO/S5V7K). three main reasons. First, development in the coastal zone, that has led to an increasing num- Funding: This work was supported in part by the ber of people and property located in coastal floodplains [1]. Second, the increase in intensity World Bank WAVES Program (WealthnAccounting of extreme storms [2,3], such as recent tropical cyclones in 2017 (e.g., Franklin, Harvey, Irma, and the Valuation of Ecosystem Services) and by the Spanish Ministry of Economy and Innovation Katia, José and Marı́a), that will result in devastating consequences to people and property [4] (currently named as Ministry of Science, and could accentuate with Sea Level Rise [5,6]. Third, the coastal ecosystem loss [7,8] reduces Innovation and Universities), within ECOPROOF the protection capacity of coastal areas to climate hazards and increases flood risks. PLOS ONE | https://doi.org/10.1371/journal.pone.0220941 August 20, 2019 1 / 14 High-quality data and models in valuing flood protection of mangroves project (BIA2014-59718-R), RISKOADAPT project As flood risks increase, there is a growing interest in the understanding of how natural ecosys- (BIA2017-89401-R) and the FPI grant (BES-2015- tems (e.g. coral reefs, mangroves and salt marshes), damp waves and reduce flood levels [9]. This 074343). is an important service to coastal communities and should be valued to inform policies for sustain- Competing interests: The authors have declared able development, disaster risk reduction and environmental conservation. Mangroves are partic- that no competing interests exist. ularly relevant for risk reduction for many tropical nations [10–12], but they are being lost at an alarming rate. Given the increase in coastal risks, decision-makers must respond quickly, and they often have limited information, particularly for ecosystem-based solutions, that lead them to make wrong decisions, such as relying on traditional measures rather than prioritizing over green alter- natives, even when data clearly show their limitations in effectiveness and cost [13,14]. Assessing flood risks requires key data, numerical models and statistical tools to assess flooding consequences to people and property. Consequently, the access to high quality and time-space homogeneous data is a growing need for decision-makers and coastal communities to accurately assess risk and value adaptation measures. The available input data, numerical models and statistical tools will likely decide the geographical scale at which any flood risk assessment analysis could be addressed (global, national, regional or local) [15]. While global (national and regional included) approaches are best suited for screening assessments identify- ing hotspots and supporting first national ranking of ecosystem services, local studies are appropriate for specific service valuation, risk reduction or adaptation projects implementa- tion and cost-benefit analysis. Unfortunately, many local decisions are based on low accurate methods and low resolution data [16]. Most assessments of the sensitivity of flood risk models focus on exploring the sensitivity of coastal flood risk against single elements, such as the reso- lution of Digital Elevation Model (DEM) [17,18] and assets (i.e. spatial distribution of people and property [19]). No studies have explored the sensitivity of flood models for assessing risk reduction benefits of ecosystem-based adaptation measures, flood risks in the presence of coastal ecosystems and to other combined data and methods [20]. Comparing flood protection service value calculated using different approaches and data- sets thus helps to quantify the order of magnitude of errors behind making direct use of simpli- fied approaches or low-resolution datasets for local applications when there is a lack of local specific datasets and economic or technical resources. The aim of this study is to provide guid- ance on where to invest in new modeling and data acquisition to improve assessments of flood risk and ecosystem-based adaptation measures. To answer this question, we carry out a sensi- tivity analysis of flood risks to variations in the number of tropical cyclones, coastal segmenta- tion, DEM resolution, flood methods and population data resolution. Each single element is individually tested and compared with the case of fully availability of high-resolution data and process-based models (Benchmark case). Methods and study site Methodology overview The workflow diagram (Fig 1) summarizes the process followed in this work to assess the sen- sitivity of flood risk and flood risk reductions to different sets of data and modeling tools. To quantify risks, we follow a four-step methodology based on the risk assessment and management framework of the IPCC (Intergovernmental Panel on Climate Change) [5]: Haz- ard analysis includes steps 1 and 2; Impacts are calculated in step 3; and Exposure, Vulnerabil- ity and Risk are all evaluated in step 4. This multi-step methodology has been applied by others [10,11,21] and the output of each step is briefly described below: 1. Pre-habitat modeling: Offshore waves and storm surge statistical distributions produced by tropical cyclones propagated to the habitat. PLOS ONE | https://doi.org/10.1371/journal.pone.0220941 August 20, 2019 2 / 14 High-quality data and models in valuing flood protection of mangroves Fig 1. Workflow diagram and results. Strategy for testing the sensitivity of flood risk and flood protection benefits of mangroves to different approaches, with the aim of informing decision-makers where the maximum benefit is gained with improving data and models. https://doi.org/10.1371/journal.pone.0220941.g001 2. Habitat modeling: The Total Water Level along the coast, namely Flood Height (FH), due to waves and surge propagation through mangrove fields. 3. Flood impacts: coastal flooding and local water depth. 4. Exposure, vulnerability and risk of flooding to land (km2), people (nº) and property (US$ millions of industrial and residential stock). In this study we assess flood risk to annual expected floods for land, people and property, with and without mangroves. We also assess annual expected flood reduction benefits of adap- tation measures to land, people and property. Then we evaluate sensitivity of estimates of flood risk and adaptation benefits to variations in five key data and modeling elements (storms, coastline segmentation, topography, flood methods and exposure data). We compare each individual variation (sensitivity test) and the lowest resolution set for the five elements (Baseline case), with the highest resolution set (Benchmark case). In the sensitivity tests, we keep everything at low-resolution except for one, the variable where we higher the resolution. Sensitivity tests. We ran sensitivity tests on all five variables (data and models). (a) Storms (Number of tropical cyclones): Improving tropical cyclone´s historical databases by using larger time series (S1 Fig, S2 Fig and S3 Fig). (b) Coastline segmentation (number of coastal cross- shore profiles): Increasing the number of coastal segments (S4 Fig). (c) Topography (DEM reso- lution): Measuring the effect of improving the DEM horizontal resolution (S5 Fig). (d) Flood method (Bathtub vs process-based): Comparing a stationary flood method based on hydraulic connectivity (bathtub method) with a process-based model (S6 Fig). (e) Exposure data (Popu- lation resolution): Calculating the effect of using high resolution people distribution data ver- sus coarse gridded population data (S7 Fig). On the one hand, we use as our Benchmark case a high-resolution assessment of flood risk for Pagbilao, simultaneously using the best set of databases and models from the five sensitivity test, to provide site-specific results, that could, for example, be used for local service valuation, adaptation projects implementation and cost-benefits analysis. On the other hand, we consider as our Baseline case the currently available global data and model of coastal flood risk, PLOS ONE | https://doi.org/10.1371/journal.pone.0220941 August 20, 2019 3 / 14 High-quality data and models in valuing flood protection of mangroves Table 1. Multi-step methodology to evaluate flood risks and flood reduction benefits of mangroves. BASELINE CASE BENCHMARK CASE � � (1) PRE-HABITAT MODELING Data Tropical cyclones: Historical IBTrACS [22] Tropical cyclones: Synthetic [24] Astronomical Tide (GOT) Astronomical Tide (GOT) Mean Sea Level Mean Sea Level Bathymetry global: GEBCO [23] Bathymetry global: GEBCO [23] Tools Clustering method: DMA [25] Clustering method: DMA [25] Offshore: Delft3D model (2D mesh at 5 km) Offshore: Delft3D model (2D mesh at 5 km) Nearshore: Delft3D model (2D mesh at 100 m) Nearshore: Delft3D model (2D mesh at 100 m) (2) HABITAT MODELING Data Bathymetry global: GEBCO [23] Bathymetry global: GEBCO [23] Bathymetry reefs: SeaWiFS [26] Bathymetry reefs: SeaWiFS [26] Mangroves 2010: WCMC [7] Mangroves 2010: DENR [27] Coral Reefs: UNEP-WCMC Coral Reefs: UNEP-WCMC � � Tools Cross-shore Profile tracer (2 km) Cross-shore Profile tracer (200 m) Delft3D model (1D mesh at 10m) Delft3D model (1D mesh at 10m) � � (3) FLOODING IMPACTS Data Topography: MERIT at 90 m [28] Topography: IFSAR at 5 m [30] Coastline: GSHH [29] Coastline: GSHH [29] � � Tools Flood method: Bathtub [21] Flood method: RFSM-EDA model [32] Reconstruction method: RBF [31] Reconstruction method: RBF [31] Extreme distribution: Pareto Poisson Extreme distribution: Pareto Poisson � � (4) EXPOSURE, VULNERABILITY AND RISK Data Population data: GPW at 1 km [33] Population data: WorldPop at 100 m Property data: GAR15 [34] Property data: GAR15 [34] Damage functions: HAZUS [35] Damage functions: HAZUS [35] Tools Downscaling people: from 1 km to 90 m Downscaling people: from 100 m to 5 m Downscaling property: from 5 km to 90 m Downscaling property: from 5 km to 5 m Annual Expected Function Annual Expected Function Key data and tools for assessing flood risk and risk reduction benefits. We show all the data and tools for the Benchmark and Baseline cases. We note with an asterisk (� ) the variables that we assessed in sensitivity tests. https://doi.org/10.1371/journal.pone.0220941.t001 simultaneously using the low-resolution set of the five sensitivity tests, to provide global or large regional risk assessment. We calculate the Error Rate Index (ERI) (Eq 1) to compare Benchmark case (EBenchmark) with the Baseline case and each sensitivity test (Ei). Sensitivity tests could either overestimate (ERI >0) or underestimate (ERI<0) Benchmark values. It allows us to compare sensitivities to data and models, with the aim of choosing the most efficient way to improve flood risk assess- ments from any global, national or regional scale. ERI ¼ 100 � ðEi EBenchmark Þ=EBenchmark ð1Þ Data and tools applied at each step of the methodology and for both, Baseline and Bench- mark cases, are summarized in Table 1. Study area We assess the coastal flood protection service provided by mangroves in Pagbilao (the Philip- pines), a municipality located in the southern part of Quezon Province in Luzon Island, the north coastline of Tabayas Bay. This study site was chosen because: first, its reasonable coast- line extension for local high resolution analysis (~20 km); second, the availability of high-qual- ity local data; third, the remarkable presence of mangroves [36], and fourth, its exposure to local extreme storms. Pagbilao covers 15,820 ha, whereof 4,560 ha are mangrove forests. It has a total population of 75,000 people of which 27,958 live in low-lying areas exposed to flood threats, and it accounts for US$45.06 million of property located in potentially flooded areas (US$16.28 PLOS ONE | https://doi.org/10.1371/journal.pone.0220941 August 20, 2019 4 / 14 High-quality data and models in valuing flood protection of mangroves million of industrial stock and US$28.78 million of residential stock). There has been a 43% decline in mangroves since 1950 mainly because of aquaculture, which has reduced mangroves to a narrow coastal band only hundreds of meters wide (Fig 2). The observed habitat decline over the last decades built a real concern in the conservation of natural resources [37] and management alternatives for the Pagbilao mangroves [38]. Coastal flooding in Pagbilao climate is mainly caused by tropical cyclones. The maximum observed surge in the offshore area of Pagbilao bay does not exceed 2 m and is produced by storms with a southeast-northwest track when crossing it. Significant differences are observed between Baseline (Fig 3A1 and 3A2) and Benchmark (Fig 3B1 and 3B2) cases, which illustrates the land area flooded by a 1-in-50-year storm with (Fig 3A1 and 3B1) and without (Fig 3A2 and 3B2) mangroves in Pagbilao Bay. In the Baseline case the flooding surface is underestimated (upper charts), while the Benchmark case signifi- cantly improves the quality (more pixel resolution) resulting in a larger flooded area. Results We assessed the sensitivity of each of the variables individually and in combination for assess- ing flood risk and adaptation measures (Table 2 and S8 Fig). In terms of risk, increasing DEM resolution (Test “c”) always overestimates land, people and property flooded. Meanwhile, using higher resolution flood models (Test “d”) leads to significant underestimations of flood risk (when coupled with low resolution data). In terms of risk reduction, we find that benefits to people and property are less sensitive than risks to single improvements in data and models. It highlights the importance of being consistent in databases resolution along the whole meth- odology (e.g. applying Baseline case) rather than investing in individual high-resolution data- bases or process-based models. We use the values given in Table 2 to calculate the error index, ERI (Eq.1), with the aim of assessing sensitivity that would most help improve risk assessments. Higher absolute ERI’s indicate greater errors and lower sensitivity. The first ranking table (Table 3) sorts each case according to land, people and property risk estimates. S1 Table is similar but expressed in terms of assessment of adaptation options. The DEM resolution is, by itself, the most efficient Fig 2. Mangrove cover in Pagbilao (The Philippines). Mangroves extent in 1950 and 2010 in Pagbilao municipality, zooming in four different areas with different mangrove covert development: (a) Mangroves retreat may be due to a change in the river sediment transport in the west side of the municipality, (b) mangroves density increment due to restoration policies, (c) mangroves conversion into aquaculture areas and (d) mangroves migration in the east side of Pagbilao. Reprinted from ArcGIS Online maps under a CC BY license, with permission from Esri, original Copyright 2018 Esri (Basemaps supported by Esri, DigitalGlobe, GeoEye, Earthstar Geographics, CNES/Airbus Ds, USDA, AEX, Getmapping, Aerogrid, IGN, IGP, swisstopo, and the GIS User Community). https://doi.org/10.1371/journal.pone.0220941.g002 PLOS ONE | https://doi.org/10.1371/journal.pone.0220941 August 20, 2019 5 / 14 High-quality data and models in valuing flood protection of mangroves Fig 3. Flood map comparison in Pagbilao. Coastal flooding produced by 1-in-50 years tropical cyclone in Pagbilao. Comparison between Baseline case (with and without mangroves) and Benchmark case (with and without mangroves). Reprinted from ArcGIS Online maps under a CC BY license, with permission from Esri, original Copyright 2018 Esri (Basemaps supported by Esri, DigitalGlobe, GeoEye, Earthstar Geographics, CNES/Airbus Ds, USDA, AEX, Getmapping, Aerogrid, IGN, IGP, swisstopo, and the GIS User Community). https://doi.org/10.1371/journal.pone.0220941.g003 way of reducing errors to flood risk to land and people. However, improving the DEM had rel- atively high errors for assessing impacts of flooding to property. Overall, as it is shown in Table 3, using the Baseline case is not always the worst option, which highlights that there are sensitivity tests, that, on their own, do not bring any improvement to the analysis and must be combined with other improvements so that it is worth investing resources and time in their use. We also assess the most efficient way of valuing risks in the presence of mangroves (S2 Table) versus in the absence of mangroves (S3 Table). We not only studied each single improvement of data and models individually, but also any possible combination of cases. S4 Table lists all the ERI indices of each existing combination, so that we can identify the most efficient way of combining datasets and modeling methods to reduce the error of flood risk estimates. PLOS ONE | https://doi.org/10.1371/journal.pone.0220941 August 20, 2019 6 / 14 High-quality data and models in valuing flood protection of mangroves Table 2. Flood risk and adaptation benefits assessment. LAND FLOODED PEOPLE AFFECTED PROPERTY DAMAGE (Hectares) (Nº) (US$ mill) With Without Benefit With Without Benefit With Without Benefit Benchmark case (high-resolution) 320 350 30 3898 4510 612 4.23 5.05 0.82 Baseline case (low-resolution) 139 216 77 696 1268 572 1.22 1.94 0.72 Test A: Nº of storms 182 256 74 1053 1701 648 1.68 2.43 0.75 Test B: Nº of profiles 137 214 77 683 1271 588 1.20 1.95 0.75 Test C: DEM resolution 340 380 40 5954 6653 699 7.80 8.95 1.15 Test D: Flood method 21 35 14 299 523 224 0.57 0.82 0.25 Test E: Exposure resolution (Pop) 139 216 77 736 1323 587 1.22 1.94 0.72 Flood risk (with and without mangroves) and risk reduction benefits to land, people and property. The benefits provided by mangroves are the difference in flooding with and without mangroves. https://doi.org/10.1371/journal.pone.0220941.t002 For a better understanding of the effect that each single improvement in datasets and meth- ods has, we analyze, one by one, the sensitivity of flood risks to each case: Sensitivity to the number of storms (tropical cyclones) Longer time series of storms (i.e. 1,000 years of synthetic tropical cyclones) results in better accu- racy in flood predictions than shorter time series (i.e. 71 years of historical IBTrACS dataset) (S1 Fig and S2 Fig). For example, by increasing the number of tropical cyclones, risk assessment in presence of mangroves to land, people and property increases by 43 ha (+31%), 357 people (+51%) and US$ 0.46 million (+38%) with respect the Baseline case (calculated from Table 2). Additionally, we better estimate floods in no-mangrove scenarios (47% of average error) than in mangrove pro- tected coastlines (59% of average error) if using synthetic tropical cyclones (S2 and S3 Tables). Sensitivity to the number of profiles Coastal impacts (land flooded) and risks (people affected and property damaged due to coastal flooding) are slightly underestimated if moving from 2 km to a 200 m longshore segmentation. Table 2 shows that the area flooded, people affected and property loss in presence of man- groves decreases by 2 ha (-1.5%), 13 people (-1.8%) and US$ 0.02 million (-1.6%) with respect to the Baseline case. Sensitivity to DEM resolution Overall predictions of flood risk to land and people was the most sensitive to changes in DEM. In presence of mangroves, coastal risk assessment increases by 201 ha (+144%), 5,258 people (+755%) and US$ 6.58 million (+539%) if using high resolution DEM (IFSAR 5m) instead of global DEM (MERIT 90 m). We observe in Table 3, that local high resolution DEM reduce errors with respect to the Baseline case when valuing land flooded (+7.41% vs -47.42%) and people affected (+50.13% vs -77.01%), but not property damages based on 5 km resolution dataset (80.81% vs -66.37%), due to the abrupt differences between the DEM and property data resolution (5 m and 5 km respectively). Sensitivity to flood method The use of a process-based flood model (RFSM-EDA), instead of the bathtub method, results in lower estimates of risk with respect to the Baseline case: -118 hectares (-85%), -397 people PLOS ONE | https://doi.org/10.1371/journal.pone.0220941 August 20, 2019 7 / 14 High-quality data and models in valuing flood protection of mangroves Table 3. Ranking table for valuing risks, based on ERI index. RISK LAND PEOPLE PROPERTY Rank Sensitivity test ERI Rank Sensitivity test ERI Rank Sensitivity test ERI 1 DEM res. +7.41% 1 DEM res. +50.13% 1 Nº storms -56.08% 2 Nº storms -34.99% 2 Nº stoms -67.63% 2 Baseline -66.37% 3 Baseline -47.42% 3 Exposure res. (Pop) -75.89% 3 Exposure res. (Pop) -66.37% 4 Exposure res. (Pop) -47.42% 4 Baseline -77.01% 4 Nº profiles -66.51% 5 Nº profiles -48.02% 5 Nº profiles -77.15% 5 DEM res. +80.81% 6 Flood method -91.72% 6 Flood method -90.37% 6 Flood method -85.14% A comparison of sensitivity of estimates of risk to 5 different factors and to the Baseline case. Baseline case in also ranked to show which elements improve risk estimates (ranked above the Baseline) and which do not (ranked below the Baseline). Risk is assessed as flooding of land (left), people (mid) and property (right). https://doi.org/10.1371/journal.pone.0220941.t003 (-57%) and US$ -0.65 million (-53%). These differences observed in Table 2 lead to an average error increment of 23% (Table 3) with respect to the Baseline case. However, if we combine the RFSM-EDA model with high resolution topography data (IFSAR 5 m), the error in risk assessment to land, people and property would be reduced to -36%, +40% and +67% respec- tively (S4 Table, row 14, column 1: “Flood”, column 2: “People” and column 3: “Property”). Sensitivity to population resolution Using high-resolution population distribution (100 m WorldPop) as an alternative to low reso- lution data (1 km GWP) does improve the estimation of people affected by coastal flooding by 40 people (+6%), but not enough to reach the predicted 3,898 people by the Benchmark case (Table 2). Discussion Given the growing risks of people and property to flooding and the need to quickly decide on effective risk reduction solutions, we identified the key factors that would most help improve upon existing global data or poor data sites and models for applications at national and poten- tially site-specific levels. The assessment of flood risk to land and people is most sensitive to resolution of the topo- graphic data. That is, you would benefit the most from making improvements in topography but also in the storm data (ranked #2 in both cases). Meanwhile, estimates of risks to property are most sensitive to the quality of the storms database. Consequently, improving DEM resolu- tion is the most effective way of improving risk to land and people, but storms data may be the most important overall for improving risk estimates. We believe the results are particularly sensitive to storms data because, since we analyze damage in a statistical way (i.e. annual expected function), results are highly dependent on the number of elements considered. Increasing the number of tropical cyclones reduces uncertainty in the extreme value distribu- tion analysis and improves overall annual expected risk estimates. Conversely improvements of the flood method had the least improvement in all the risk assessments. That is, in many instances using a bathtub model is a fine approximation given what is available in the other databases. In terms of assessing benefits from adaptation options, the model results were most sensi- tive to changes in the number of coastal segments to estimate people and property benefits, and changes in DEM and flood method to estimate land flooded benefits. These findings are different from the flood risk assessment because benefits are less sensitive than risks to PLOS ONE | https://doi.org/10.1371/journal.pone.0220941 August 20, 2019 8 / 14 High-quality data and models in valuing flood protection of mangroves database and model improvements and, consequently, to better estimate them, it is more important to use consistent databases and models along the whole process rather than invest- ing in single measures. However, it is sometimes effective to properly combine different elements to achieve the best results. For instance, using process-based flooding model (i.e. RFSM-EDA) underestimate flooding unless combined with high-resolution elevation data. We expect these results to be broadly applicable to many areas but there are some particu- lars of the Pagbilao case that are important: (1) Pagbilao has experienced many cyclones. The effectiveness of increasing the number of tropical cyclones is more noticeable in coastal areas with high cyclone activity because it allows to generate a wider range of synthetic tropical cyclone intensities and tracks, improving the statistical analysis of extreme value distribution by narrowing confidence bands, especially for high return period storms (see S1 Fig and S2 Fig). (2) Pagbilao has a relatively simple and homogeneous coastline. That is why using 200 m spaced cross-shore profiles, rather than 2 km, do not significantly improve risk estimations in Pagbilao. This findings should not be projected to other regions with longshore morphology variability, different mangrove species and more complex distribution of coastal assets, which are probably more sensitive to the increase of coastal segmentation. (3) Pagbilao does not have available high-resolution socioeconomic data. The abrupt downscaling required to rescale the global 5 km grid data of property to 5 m (the same than high-resolution DEM) leads to an spa- tial misallocation of coastal exposure and the consequent overestimation of the capital loss. This fact reduces the efficiency of using high-resolution DEM. (4) Pagbilao is a local area with short coastline (~20km of coastline), where using process-based models (e.g. RFSM-EDA) are computationally affordable. However, it is no longer applicable spatial scales larger than 100 km because its time-consuming pre-processing mesh generation. Further methodological limitations come up when valuing flood risk at different locations. First, calculating annual expected values of risk and benefits requires large sample data to reduce errors. However, historical datasets do not always provide enough information. Sec- ond, one-dimensional propagations neglect two-dimensional processes such as longshore cur- rents or waves diffraction, missing some energy losses, that could be relevant in longshore varied coastal areas. Third, using 1 km bathymetry data is a high limitation to wave hydrody- namics modelling, especially in coral reef and mangroves environments due to the inability to capture the level of detail required to model coastal ecosystems processes. For that reason, we increase the accuracy in the vegetated shallow areas using both, local specific bathymetry of ecosystems [26] and parameterized profiles to correct bathymetric errors (S9 Fig). In brief, for local high-resolution flood risk analysis, we must consider how to combine databases and methods to obtain the best possible result with the less effort and resources expenses, before making directly use of all the high-resolution databases available. Supporting information S1 File. Sensitivity tests description: S1_File_Sensitivity_tests_description. (DOCX) S1 Fig. Sensitivity Test A: Offshore HS distribution in Tabayas Bay (Pagbilao). Offshore maximum significant wave height extreme distribution produced by (A) historical tropical cyclones and (B) synthetic tropical cyclones. Black circles represent the most probable value of HS. The solid line represents the best fit adjustment of the most probable values of HS. Dashed lines represent the 95% confidence interval of the analytical extreme value distribution. (TIF). (TIF) PLOS ONE | https://doi.org/10.1371/journal.pone.0220941 August 20, 2019 9 / 14 High-quality data and models in valuing flood protection of mangroves S2 Fig. Sensitivity Test A: Offshore SS distribution in Tabayas Bay (Pagbilao). (A) Offshore maximum storm surge produced by historical tropical cyclones. (B) Offshore maximum storm surge produced by synthetic tropical cyclones. Black circles represent the most probable value of SS. The solid line represents the best fit adjustment of the most probable values of SS. Dashed lines represent the 95% confidence interval of the analytical extreme value distribution. (TIF). (TIF) S3 Fig. Sensitivity Test A: TWL pre-habitat and flood height extreme distributions with and without mangroves in Tabayas Bay (Pagbilao). (A) TWL pre-habitat. (B1) Flood Height distribution produced by historical tropical cyclones (solid line) and synthetic tropical cyclones (dashed line) in case of preserving the 2010 mangrove cover. (B2) Flood Height distri- bution produced by historical tropical cyclones (solid line) and synthetic tropical cyclones (dashed line) in case of losing mangroves. (TIF). (TIF) S4 Fig. Sensitivity Test B: Coastline segmentation in Pagbilao. (a) Example of 2 km spaced cross-shore profiles. (b) Example of 200 m space cross-shore profiles. Reprinted from ArcGIS Online maps under a CC BY license, with permission from Esri, original Copyright 2018 Esri (Basemaps supported by Esri, DigitalGlobe, GeoEye, Earthstar Geographics, CNES/Airbus Ds, USDA, AEX, Getmapping, Aerogrid, IGN, IGP, swisstopo, and the GIS User Community). (TIF). (TIF) S5 Fig. Sensitivity Test C: Digital elevation model comparison in Pagbilao. (a) General view of Pagbilao bay. (b) Global SRTM 30 m resolution model. (c) MERIT DEM at 90 m resolution, obtained from SRTM by filtering out the vegetation height. (d) Local high resolution IFSAR DEM (5 m resolution). All the figures have been labeled between 0 and 20 m height. Reprinted from ArcGIS Online maps under a CC BY license, with permission from Esri, original Copy- right 2018 Esri (Basemaps supported by Esri, DigitalGlobe, GeoEye, Earthstar Geographics, CNES/Airbus Ds, USDA, AEX, Getmapping, Aerogrid, IGN, IGP, swisstopo, and the GIS User Community). (TIF). (TIF) S6 Fig. Sensitivity Test D: Flood method comparison in Pagbilao. (a) 1-in-50-year flooding in the presence of mangroves calculated with the bathtub method. (b) 1-in-50-year flooding in the presence of mangroves calculated with the RFSM-EDA model. Reprinted from ArcGIS Online maps under a CC BY license, with permission from Esri, original Copyright 2018 Esri (Basemaps supported by Esri, DigitalGlobe, GeoEye, Earthstar Geographics, CNES/Airbus Ds, USDA, AEX, Getmapping, Aerogrid, IGN, IGP, swisstopo, and the GIS User Community). (TIF). (TIF) S7 Fig. Sensitivity Test E: Population datasets in Pagbilao and Lucena. (a) 1 km-resolution data GPW. (b) 100 m-resolution data Worldpop. Reprinted from ArcGIS Online maps under a CC BY license, with permission from Esri, original Copyright 2018 Esri (Basemaps supported by Esri, DigitalGlobe, GeoEye, Earthstar Geographics, CNES/Airbus Ds, USDA, AEX, Get- mapping, Aerogrid, IGN, IGP, swisstopo, and the GIS User Community). (TIF). (TIF) S8 Fig. Annual expected risks and benefits. (A1) Annual Expected Flooding with mangroves (light grey) and without mangroves (dark grey), calculated following the Baseline case, Benchmark case and each sensitivity test. (A2) Annual Expected Risk in terms of people affected by coastal PLOS ONE | https://doi.org/10.1371/journal.pone.0220941 August 20, 2019 10 / 14 High-quality data and models in valuing flood protection of mangroves flooding with mangroves (light grey) and without mangroves (dark grey), calculated following the Baseline case, Benchmark case and each sensitivity test. (A3) Annual Expected Risk in terms of property damaged by coastal flooding with mangroves (light grey) and without mangroves (dark grey), calculated following the Baseline case, Benchmark case and each sensitivity test. (B1) Annual Expected Flooding reduction due to the presence of mangroves, calculated following the Baseline case, Benchmark case and each sensitivity test. (B2) Annual Expected Benefits in terms of people protected by mangroves, calculated following the Baseline case, Benchmark case and each sensitivity test. (B3) Annual Expected Benefits in terms of property protected by mangroves, cal- culated following the Baseline case, Benchmark case and each sensitivity test. (TIF). (TIF) S9 Fig. Cross-shore profile parameterization. Parameterized cross-shore profile typical from coral reef and mangroves regions. Where “h” values represent dater depth or topographic ele- vation, “B”, “L”, “W” and “D” represent horizontal distances and “m” values represent bottom slope. (TIF). (TIF) S1 Table. Ranking table for valuing benefits, based on ERI index. A comparison of sensitiv- ity of estimates of mangrove benefits (risk reduction) to 5 different factors and to the Baseline case. Baseline case in also ranked to show which elements improve benefits estimates (ranked above the Baseline) and which do not (ranked below the Baseline). Ranking table to prioritize the best-practice case of valuing mangrove´s protection capacity in three different ways: Flood reduction (left), people protected (mid) and total property benefits (right). (DOCX). (DOCX) S2 Table. Ranking table for valuing flood risks in mangrove presence, based on ERI index. Ranking table to prioritize the best-practice case of valuing flood risks in presence of man- groves to 5 different factors and to the Baseline case. Baseline case in also ranked to show which elements improve risk estimates (ranked above the Baseline) and which do not (ranked below the Baseline). Ranking table to prioritize the best-practice case of valuing mangrove´s protection capacity in three different ways: Flood reduction (left), people protected (mid) and total property benefits (right). (DOCX). (DOCX) S3 Table. Ranking table for valuing risks in mangrove absence, based on ERI index. Rank- ing table to prioritize the best-practice case of valuing flood risks in absence of mangroves to 5 different factors and to the Baseline case. Baseline case in also ranked to show which elements improve risk estimates (ranked above the Baseline) and which do not (ranked below the Base- line). Ranking table to prioritize the best-practice case of valuing mangrove´s protection capacity in three different ways: Flood reduction (left), people protected (mid) and total prop- erty benefits (right). (DOCX). (DOCX) S4 Table. Error Rate Index (ERI). Error Rate Index (ERI) calculated at any intermediate case. (DOCX). (DOCX) Acknowledgments We gratefully acknowledge support from the World Bank WAVES Program (Wealth Account- ing and the Valuation of Ecosystem Services) and by the Spanish Ministry of Economy and PLOS ONE | https://doi.org/10.1371/journal.pone.0220941 August 20, 2019 11 / 14 High-quality data and models in valuing flood protection of mangroves Innovation (currently named as Ministry of Science, Innovation and Universities), within ECOPROOF project (BIA2014-59718-R), RISKOADAPT project (BIA2017-89401-R) and the FPI grant (BES-2015-074343). Authors are grateful to the useful contributions provided by Sheila Abad and Javier Garcı́a Alba (IHCantabria). Author Contributions Conceptualization: Pelayo Menéndez, Iñigo J. Losada. Data curation: Pelayo Menéndez. Formal analysis: Pelayo Menéndez, Saúl Torres-Ortega, Alexandra Toimil. Funding acquisition: Iñigo J. Losada, Michael W. Beck. Methodology: Pelayo Menéndez, Iñigo J. Losada. Project administration: Pelayo Menéndez, Iñigo J. Losada. Resources: Iñigo J. Losada, Saúl Torres-Ortega, Michael W. Beck. Software: Pelayo Menéndez, Saúl Torres-Ortega, Alexandra Toimil. Supervision: Alexandra Toimil. Validation: Pelayo Menéndez. Visualization: Pelayo Menéndez. Writing – original draft: Pelayo Menéndez, Iñigo J. Losada, Michael W. Beck. Writing – review & editing: Pelayo Menéndez, Iñigo J. Losada, Alexandra Toimil, Michael W. Beck. References 1. Hinkel J, Lincke D, Vafeidis AT, Perrette M, Nicholls RJ, Tol RSJ, et al. Coastal flood damage and adap- tation costs under 21st century sea-level rise. Proc Natl Acad Sci. 2014; 111: 3292–3297. https://doi. org/10.1073/pnas.1222469111 PMID: 24596428 2. Emanuel K. Increasing destructiveness of tropical cyclones over the past 30 years. 2005; 436: 686– 688. https://doi.org/10.1038/nature03906 PMID: 16056221 3. Elsner JB, Kossin JP, Jagger TH. The increasing intensity of the strongest tropical cyclones. 2008; 455: 2–5. https://doi.org/10.1038/nature07234 PMID: 18769438 4. Rahmstorf S. Rising hazard of storm-surge flooding. 2017; 114: 11806–11808. https://doi.org/10.1073/ pnas.1715895114 PMID: 29078412 5. IPCC. Summary for policymakers. In: Climate Change 2014: Impacts, Adaptation, and Vulnerability. Part A: Global and Sectoral Aspects. Contribution of Working Group II to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Field, C.B.,. Cambridge, United Kingdom and New York, NY, USA; 2014. 6. Reguero BG, Losada IJ, Dı́az-Simal P, Méndez FJ, Beck MW. Effects of climate change on exposure to coastal flooding in Latin America and the Caribbean. PLoS One. 2015; 10: 1–19. https://doi.org/10. 1371/journal.pone.0133409 PMID: 26177285 7. Spalding M, Kainuma M, C L. World atlas of mangroves. A collaborative project of ITTO, ISME, FAO, UNEP-WCMC. London, UK: Earthscan; 2010. 8. Polidoro BA, Carpenter KE, Collins L, Duke NC, Ellison AM, Ellison JC, et al. The loss of species: Man- grove extinction risk and geographic areas of global concern. PLoS One. 2010; 5. https://doi.org/10. 1371/journal.pone.0010095 PMID: 20386710 9. Spalding MD, Mcivor AL, Beck MW, Koch EW, Möller I, Reed DJ, et al. Coastal ecosystems: A critical element of risk reduction. Conserv Lett. 2014; 7: 293–301. https://doi.org/10.1111/conl.12074 10. Menéndez P, Losada IJ, Beck MW, Torres-Ortega S, Espejo A, Narayan S, et al. Valuing the protection services of mangroves at national scale: The Philippines. Ecosyst Serv. 2018; 34: 24–36. https://doi. org/10.1016/j.ecoser.2018.09.005 PLOS ONE | https://doi.org/10.1371/journal.pone.0220941 August 20, 2019 12 / 14 High-quality data and models in valuing flood protection of mangroves 11. Losada IJ, Menéndez P, Espejo A, Torres S, Dı́az-Simal P, Abad S, et al. The global value of man- groves for risk reduction. Technical Report. Berlin; 2018. https://doi.org/10.7291/V9DV1H2S 12. Narayan S, Beck M, Reguero B, Losada I, van Wesenbeeck B, Pontee N, et al. The benefits, costs and effectiveness of natural and nature-based coastal defenses. Pap Submitt to PLOS. 2016; 13. Sutton-Grier AE, Gittman RK, Arkema KK, Bennett RO, Benoit J, Blitch S, et al. Investing in natural and nature-based infrastructure: Building better along our coasts. Sustain. 2018; 10: 1–11. https://doi.org/ 10.3390/su10020523 14. Reguero BG, Beck MW, Bresch DN, Calil J, Meliane I. Comparing the cost effectiveness of nature- based and coastal adaptation: A case study from the Gulf Coast of the United States. 2018; 1–24. https://doi.org/10.17605/OSF.IO/D6R5U 15. De Moel H, Jongman B, Kreibich H, Merz B. Flood risk assessments at different spatial scales. 2015; 865–890. https://doi.org/10.1007/s11027-015-9654-z 16. Laurice Jamero M, Onuki M, Esteban M, Billones-Sensano XK, Tan N, Nellas A, et al. Small-island com- munities in the Philippines prefer local measures to relocation in response to sea-level rise. Nat Clim Chang. 2017; 7: 581. https://doi.org/10.1038/nclimate3344 17. Leon JX, Heuvelink GBM, Phinn SR. Incorporating DEM uncertainty in coastal inundation mapping. PLoS One. 2014; 9: 1–12. https://doi.org/10.1371/journal.pone.0108727 PMID: 25250763 18. Hinkel J, Lincke D, Vafeidis AT, Perrette M, Nicholls RJ, Tol RSJ, et al. Coastal flood damage and adap- tation costs under 21st century sea-level rise. Proc Natl Acad Sci. 2014; 111: 3292–3297. https://doi. org/10.1073/pnas.1222469111 PMID: 24596428 19. Mondal P, Tatem AJ. Uncertainties in Measuring Populations Potentially Impacted by Sea Level Rise and Coastal Flooding. 2012; 7. https://doi.org/10.1371/journal.pone.0048191 PMID: 23110208 20. Rohmer J, Cazenave A, Van De Wal R, De Winter R, Pedreros R, Balouin Y, et al. Environmental Modelling & Software Evaluating uncertainties of future marine fl ooding occurrence as sea-level rises. 2015; 73: 44–56. https://doi.org/10.1016/j.envsoft.2015.07.021 21. Beck MW, Losada IJ, Mendendez P, Reguero BG, Dı́az-Simal P, Fernandez F. The global flood protec- tion savings provided by coral reefs. Nat Commun. 2017; https://doi.org/10.1038/s41467-018-04568-z PMID: 29895942 22. Knapp KR, Kruk MC, Levinson DH, Diamond HJ, Neumann CJ. The international best track archive for climate stewardship (IBTrACS) unifying tropical cyclone data. Bull Am Meteorol Soc. 2010; 91: 363– 376. https://doi.org/10.1007/978-90-481-3109-9_26 23. Weatherall P, Marks KM, Jakobsson M, Schmitt T, Tani S, Arndt JE, et al. A new digital bathymetric model of the world’s oceans. Earth and Space Science (AGU Publications). 2015. pp. 1–15. https://doi. org/10.1002/2015EA000107 24. Nakajo S, Mori N, Yasuda T, Mase H. Global stochastic tropical cyclone model based on principal com- ponent analysis and cluster analysis. J Appl Meteorol Climatol. 2014; 53: 1547–1577. https://doi.org/10. 1175/JAMC-D-13-08.1 25. Camus P, Mendez FJ, Medina R, Cofiño AS. Analysis of clustering and selection algorithms for the study of multivariate wave climate. Coast Eng. 2011; 58: 453–462. https://doi.org/10.1016/j.coastaleng. 2011.02.003 26. Robinson JA, Feldman GC, Kuring N, Franz B, Green E, Noordeloos M, et al. Data fusion in coral reef mapping: working at multiple scales with SeaWiFS and astronaut photography. Proceedings of the 6th International Conference on Remote Sensing for Marine and Coastal Environments. 2000. pp. 473– 483. Available: http://eol.jsc.nasa.gov/newsletter/DataFusionInCoralReefMapping/ 27. Long JB, Giri C. Mapping the Philippines’ mangrove forests using Landsat imagery. Sensors. 2011; 11: 2972–2981. https://doi.org/10.3390/s110302972 PMID: 22163779 28. Yamazaki D, Ikeshima D, Tawatari R, Yamaguchi T, O’Loughlin F, Neal JC, et al. A high-accuracy map of global terrain elevations. Geophys Res Lett. 2017; 44: 5844–5853. https://doi.org/10.1002/2017GL072874 29. Wessel P, Smith WHF. A global, self-consistent, hierarchical, high-resolution shoreline database. J Geophys Res Solid Earth. 1996; 101: 8741–8743. https://doi.org/10.1029/96JB00104 30. Li X, Photogrammetrist C. Accuracy of airborne ifsar mapping. 31. Camus P, Mendez FJ, Medina R. A hybrid efficient method to downscale wave climate to coastal areas. Coast Eng. 2011; 58: 851–862. https://doi.org/10.1016/j.coastaleng.2011.05.007 32. Jamieson SR, Wright G, Lhomme J, Gouldby BP. Validation of a computationally efficient 2D inundation model on multiple scales. In: Klijn F, Schweckendiek T, editors. Comprehensive flood risk management: research for policy and practice. Rotterdam: Taylor & Francis Group; 2012. pp. 121–122. 33. Center for International Earth Science Information Network—CIESIN—Columbia University. Gridded Population of the World, Version 4 (GPWv4): Population Count. [Internet]. Palisades, NY: NASA Socioeconomic Data and Applications Center (SEDAC); 2016. https://doi.org/10.7927/H4X63JVC PLOS ONE | https://doi.org/10.1371/journal.pone.0220941 August 20, 2019 13 / 14 High-quality data and models in valuing flood protection of mangroves 34. UNISDR. Making Development Sustainable: The Future of Disaster Risk Management. Global Assess- ment Report on Disaster Risk Reduction. International Stratergy for Disaster Reduction (ISDR). 2015. 9789211320282 35. Scawthorn C, Flores P, Blais N, Seligson H, Tate E, Chang S, et al. HAZUS-MH Flood Loss Estimation Methodology. II. Damage and Loss Assessment. Nat Hazards Rev. 2006; 7: 72–81. https://doi.org/10. 1061/(ASCE)1527-6988(2006)7:2(72) 36. Bravo DR. Restoration and management of the Pagbilao Mangrove Genetic Resource Area. Proc Eco- tone IV, Ecol Manag Mangrove Restor Regen East Southeast Asia. 1996; 190–213. 37. Primavera JH, Esteban JMA. A review of mangrove rehabilitation in the Philippines: successes, failures and future prospects. Wetl Ecol Manag. 2008; 16: 345–358. 38. Janssen R, Padilla JE. Preservation or conversion? Valuation and evaluation of a mangrove forest in the Philippines. Environ Resour Econ. 1999; 14: 297–331. https://doi.org/10.1023/A:1008344128527 PLOS ONE | https://doi.org/10.1371/journal.pone.0220941 August 20, 2019 14 / 14