Open Water Journal Volume 4 Issue 2 Special Issue: 2017 CUAHSI Article 4 Hydroinformatics Conference

2017 Large-Scale Flood Inundation Modeling in Data Sparse Environments using TanDEM-X Terrain Data Joseph L. Gutenson US Army Engineer Research and Development Center, [email protected]

Michael L. Follum US Army Engineer Research and Development Center, [email protected]

Alan D. Snow US Army Engineer Research and Development Center, [email protected]

Mark D. Wahl US Army Engineer Research and Development Center, [email protected]

Follow this and additional works at: https://scholarsarchive.byu.edu/openwater

BYU ScholarsArchive Citation Gutenson, Joseph L.; Follum, Michael L.; Snow, Alan D.; and Wahl, Mark D. (2017) "Large-Scale Flood Inundation Modeling in Data Sparse Environments using TanDEM-X Terrain Data," Open Water Journal: Vol. 4 : Iss. 2 , Article 4. Available at: https://scholarsarchive.byu.edu/openwater/vol4/iss2/4

This Article is brought to you for free and open access by the All Journals at BYU ScholarsArchive. It has been accepted for inclusion in Open Water Journal by an authorized editor of BYU ScholarsArchive. For more information, please contact [email protected], [email protected]. Large-Scale Flood Inundation Modeling in Data Sparse Environments using TanDEM-X Terrain Data

Cover Page Footnote This project was supported by the Deputy Assistant Secretary of the Army for Research and Technology through the Engineer Research and Development Center’s Military Engineering applied research work package title Austere Entry. It was also supported by the Geospatial Intelligence Directorate of the Marine Corps Intelligence Activity.

This article is available in Open Water Journal: https://scholarsarchive.byu.edu/openwater/vol4/iss2/4 Large-Scale Flood Inundation Modeling in Data Sparse Environments using TanDEM-X Terrain Data Joseph L. Gutenson1, Michael L. Follum1, Alan D. Snow1, Mark D. Wahl1 1US Army Engineer Research and Development Center

ABSTRACT

This paper demonstrates a workflow used by the U.S. Army Coastal and Hydraulics Laboratory to quickly gen- erate high-resolution flood inundation maps nearly anywhere in the world. Previous research illustrates that hindcast and forecast streamflow for nearly any river/stream in the world is possible by combining global runoff datasets from land surface models (LSMs) with hydrography datasets from hydrologically corrected global ele- vation datasets. Recent research has also shown that flood inundation estimates are possible through combining resulting streamflow estimates with high-resolution terrain and land cover data. This paper provides a detailed workflow of how near global flood inundation maps runoff data are generated using the HRES/LAND LSM data from the European Center for Medium Range Weather Forecasts; hydrography datasets from the Shuttle Ele- vation Derivatives at multiple Scales as well as derived from the World Wildlife Federation terrain data; Visual Navigation (VISNAV) land cover data; and the newly-collected TerraSAR-X add-on for Digital Elevation Mea- surements (TanDEM-X) digital elevation model (DEM) datasets. This current research discusses the inputs and outputs for this flood modeling methodology, limitations, and provides an example for Lionrock that hit in late August 2016.

Keywords global-scale; flood mapping; flood impact assessment; natural hazards

1.0 Introduction military with tactical and strategic information critical to mission planning and success (Wahl et al., 2016). 1.1 The Need for Global Inundation Capabilities 1.2 Current Methodologies and Technologies Both within the continental United States (CONUS) and outside CONUS (OCONUS) continental-scale At a global scale, modeling of flood inundation hydrologic forecasting has become possible (Snow, traditionally takes place in either quasi-real-time or 2015; Snow et al., 2016; Maidment, 2016). In addition, retrospective analyses. Most of the global flood stud- rapidly forecasting hydraulics and resulting inundation ies involve the use of remote sensing technologies. from these hydrologic forecasts is possible utilizing the An example of this is the National Aeronautical and AutoRoute software (Follum et al., 2016; Follum, 2013). Space Administration (NASA) Global Flood Mapping With the expansion of large-scale hydrologic and river- System, produced by a collaborative group of NASA and ine hydraulic capability, the United States military is bet- Dartmouth Flood Observatory researchers (Policelli et ter capable of understanding the physical conditions that al., 2017). Policelli et al. (2017) make use of NASA’s will occur in their theatres of operation. Understanding MODerate resolution Imaging Spectro-radiometer the hydrology and hydraulics OCONUS provides the (MODIS) satellite imagery, with a 250 meter hor- Open Water 2 izontal resolution. A major component of NASA’s in the field (Follum, 2013). AutoRoute is a hydrau- system is the Near Real Time (NRT) Global Flood lic model that can quickly estimate inundation at Mapping service (Nigro et al., 2014). Addtionally, continental scales. Prior research links AutoRoute the remote sensing is the driver for the Global Flood to RAPID to form AutoRAPID. AutoRAPID can Dection System (GFDS) that uses the Advanced efficiently generate inundation maps at large geo- Microwave Scanning Radiometer - Earth Observing graphic scales (Follum et al., 2016; Snow, 2016a). System (AMSR-E) instrument on board of the NASA As with most hydraulic models, DEM resolu- EOS Aqua satellite (Kugler and De Groeve, 2007). tion impacts the accuracy of AutoRoute solutions. A break from hindcast or realtime flood hazard esti- Traditionally, 30-meter horizontal resolution DEM mation is the forecasing framework, the Global Flood is the standard globally (Jarihani et al., 2015). At this Awareness System (GloFAS) (Alfieri et al., 2012). resolution, limited hydraulic modeling capability can GloFAS provides a gridded streamflow product at a occur, particularly in the upper reaches of a watershed. 10-km horizontal resolution. However, GloFAS does Thus, terrain data resolution often limits the ability to not operationally provide inundation and at the coarse model river hydraulics at global scales. A limitation of spatial resolution. The Snow et al. (2016) framework the accuracy of AutoRoute is the resolution available to routes the same runoff product that GloFAS utilizes. map the inundation. AutoRoute can estimate inunda- Snow et al. (2016) expanded their national model tion at the resolution provided by the DEM the model to the global scale by generating stream networks and is utilizing. However, DEMs with large horizontal res- catchments using the HydroSHEDS and HydroBASINS olutions and/or numeric precision limits AutoRAPID datasets (Lehner & Grill, 2013). HydroSHEDS is com- inundation estimate accuracy (Follum, 2013). prised of hydrologically corrected, 3-arc second (90 meter) Shuttle Radar Topography Mission (STRM) 1.3 Potential Impact of TanDEM-X DEM data (Lehner et al., 2013). The researchers take this terrain and generate subwatershed catch- Recent release of provisional or raw TanDEM-X ments and flowlines that are topologically connected. DEM products (Krieger et al., 2007) to the United ECMWF generates gridded reanalysis runoff using States Department of Defense (DOD) at a 12-meter the HTESSEL LSM with weather input from the ERA horizontal resolution (Boeer et al., 2014) improves Interim reanalysis dataset (Balsamo et al., 2015). In the capability to model riverine inundation at addition, ECMWF generates a 52-member ensemble nearly any location on Earth. In this research, LSM forecast with up to 15-day lead times (Alfieri et the authors detail the methodology by which they al., 2013) and a high resolution (HRES) forecast for model inundation on large spatial scales by com- up to 10-day lead times (ECMWF, 2016). The hydro- bining continental scale hydrologic simulations, the logic routing model, Routing Application for Parallel TanDEM-X DEM data, and the AutoRoute software. Computation of Discharge (RAPID) (David et al., Together these data and models can function to 2011), takes these stream networks and ECMWF quickly and effectively model inundation on large geo- hindcast and forecast runoff grids and routes the run- graphic scales in data sparse environments. The speed off through the stream network (Snow et al., 2016). of acquisition of the data and the runtime of the mod- AutoRoute is a one-dimensional (1D) hydrau- els can take less than one workday in many instances. lic model that solves Manning’s equation at specified cross-sections along a stream network. The inputs 2.0 Methods necessary to run AutoRoute are spatial data on stream Figure 1 illustrates the process by which the locations (i.e., flowlines or streamlines) with flow esti- authors generate forecast inundation maps using mates, a digital elevation model (DEM), and grid- global LSM forecasts, HydroSHEDS, VISNAV, and ded land cover characteristics. Global land cover TanDEM-X datasets. Each of these datasets rep- datasets are available from the VISNAV dataset at resents external data that the authors rely upon to 30-meter horizontal resolution globally (NGA, 2015). generate inundation maps. A series of Python scripts The authors use the VISNAV data to derive compos- and wrappers (AutoRoutepy (Snow, 2016a) and ite roughness values at each cross-section. The orig- RAPIDpy (Snow, 2016b)) process the data internally. inal intent of AutoRoute was to function as a means of determining fording routes for military vehicles Open Water 3

Figure 1. System processing flowchart, the authors develop internal processes and datasets. The authors access external datasets remotely.

Stream flowlines generated from HydroSHEDS data The Korean Peninsula (Figure 2) is an area with are based on a 5 km2 contributing upstream watershed HydroSHEDS and TanDEM-X datasets. The authors threshold. Each RAPID simulation generates a nefcdf have developed flowline and catchment networks for the output file that contains a time series of discharge esti- Korean Peninsula. The researchers demonstrate how the mates. The ECMWF HRES product was chosen and combination of AutoRAPID and TanDEM-X data can the maximum discharge from the RAPID simulation effectively model inundation across the entire country. is fed into AutoRoute. The authors copy the max- The authors evaluate the inundation mapping system imum discharge estimate for each flowline into the using the that hit the northeastern attributes of the flowline shapefile. The maximum -dis of North Korea from August charge and flowline data are fed into AutoRoute with 29-31, 2016. Figure 2 illustrates the location of North the VISNAV and TanDEM-X datasets. AutoRoute Hamgyong Province. Typhoon Lionrock caused par- then generates inundation maps from these inputs. ticular devastation to the North Korean people around the , along the border of North Korea, China, and in the North Hamgyong Province. 3.0 Example The United Nations (U.N.) estimates that the event killed 133 people, 395 were missing, and 107,000 peo- 3.1 Typhoon Lionrock ple were displaced (BBC, 2016; Chae and Birsel, 2016). Open Water 4 3.2 Remote Sensing Products from Typhoon tions in Figure 2. The flow utilized in this effort is Lionrock derives from the maximum flow RAPID generates from the ECMWF HRES LSM forecasts for August From inspection by the authors, the inland 28-September 7, 2016. The authors route the forecast impacts of Typhoon Lionrock were not captured in runoff with RAPID using the RAPIDpy framework 3-day composite experimental data from NASA’s (Snow, 2016b). This method was initially introduced MODIS Near Real Time (NRT) Global Flood in Snow et al. (2016). The authors performed this Mapping service (NASA, 2017). The GFDS, Version exercise for most of North Korea as a demonstration. 2, does not capture the impacts of this event in the North Hamgyong Province either (GDACS, 2014). 4.0 Discussion

3.3 AutoRAPID/TanDEM-X Results Capturing the lower order streams is now possible with the 12 meter TanDEM-X terrain product. Though Figure 2 illustrates the location of several small vil- the horizontal resolution of the TanDEM-X is a major lages that inundated using the AutoRAPID/TanDEM-X improvement in global scale DEM resolution, sim- methodology during the Typhoon Lionrock event. ilar resolutions (1/3 and 1 arc second) have proven Figure 3 is a local aerial image in comparison to esti- to underestimate inundation extent by appreciable mated inundation extents. The numbers to the left amounts (Sanders, 2007). In addition, the TanDEM-X of each image correspond to the inundation loca- data still lack bathymetric data, which will likely lead

Figure 2. Map of the Korean Peninsula, North Hamyong Province, and Inundation Locations in Figure 3. Open Water 5

Figure 3. Inundation surrounding resulting from 10-Year streaflow event. to cause vertical inundation extent to exceed observed that these locations are near the Tumen River, where conditions. The lack of bathymetric data also pre- journalists report significant damage and impacts to cludes the use of raster inundation grids that pro- local residents (BBC, 2016; Chae and Birsel, 2016). vide estimates of depth within the inundation extent. Notable areas in which the AutoRoute inunda- From this exercise, the method we demonstrate indi- tion extent fails with current data are the locations cates that the spatial resolution the authors research is in which TanDEM-X needs additional refinement. an improvement over the NASA NRT and GFDS plat- Figure 4 demonstrates that the raw nature of the forms. However, the difficulty prevails in validating the current TanDEM-X product that the Department simulation results. The authors have been unsuccessful of Defense has access to. The raw TanDEM-X prod- in finding a validation dataset. This issue is persistent uct requires smoothing where large water bodies are in many locations across the world. In many cases, the found (Wendleder et al., 2013). Without smooth- authors are only able to validate in the data dense areas ing, artificial peaks and valleys appear within wet in CONUS. Thus, while Figure 3 demonstrates the portions of the channel. In addition, bridges and improved ability to estimate inundation globally, the other water crossings are present in the channel that authors are unable to validate these are locations were further limit inundation mapping. These errone- damage occurred. In general, the authors can assert Open Water 6

Figure 4. Example of the raw nature of the TanDEM-X data provided to the DOD. Notice that the channel is irregular with large peaks and valleys. Further, the relief in this location is minimal. ous elevation readings limit the inundation mapping The TanDEM-X data are a valuable resource to that is possible with the raw TanDEM-X product. large-scale inundation mapping. However, the data Follum (2013) and Follum et al. (2016) offer a good are proprietary. Thus, the price of this data limits their discussion of limitations with the AutoRoute modeling potential use in developing nations. However, the algorithms. Follum et al. (2016) discusses how a lack authors and their organization routinely support both of terrain in a location may impact inundation results the United States Armed Forces and relief organiza- because of the 1D nature of the hydraulic simulations. tions when disasters are imminent. Thus, as members A lack of terrain may play a role in the failure to inun- of the DOD, the authors are able to support opera- date locations pictured in Figure 2 and Figure 4. Thus, tions with this methodology, even when those nations AutoRoute will produce areas where the channel is dis- impacted do not have access to the necessary data. continuous or the inundation profile is jagged, both of which occur in Figure 4. Figure 3, on the other hand, 5.0 Conclusions and Future Work depicts a location that is in a mountainous with defined channel locations. Figure 3 depicts a much more The combination of ECMWF runoff, RAPID, realistic inundation extent produced by AutoRoute. AutoRoute, and the TanDEM-X terrain dataset offers the ability to generate a forecast of inundation at nearly any Open Water 7 location in the world. The method is quick and useful (Abdallah et al., 2013). Both SWOT and space-borne for a number of military and humanitarian applications. LiDAR are experimental technologies at this juncture. In some cases, the resolution of the resulting inundation is higher than existing tools and methodologies can References provide. However, verifying the results of these simula- tion is difficult given the lack of observed information. Abdallah, H., Bailly, J. S., Baghdadi, N. N., Saint-Geours, N., & Fabre, F. (2013). Potential of space-borne LiDAR sensors The AutoRAPID/TanDEM-X method is not with- for global bathymetry in coastal and inland waters. IEEE out issues. Because TanDEM-X data are raw, the cur- Journal of Selected Topics in Applied Earth Observations rent inundation products can be discontinuous and and Remote Sensing, 6(1), 202–216. https://doi.org/10.1109/ inaccurate at many locations, such as those found in JSTARS.2012.2209864 Figure 2 and Figure 4. As processing of the TanDEM-X Alfieri, L., Burek, P., Dutra, E., Krzeminski, B., Muraro, D., dataset ensues, usability within the framework the Thielen, J., & Pappenberger, F. (2013). GloFAS-global ensem- ble streamflow forecasting and flood early warning. Hydrology authors provide will increase. However, TanDEM-X and Earth System Sciences, 17(3), 1161–1175. https://doi. data are proprietary, which will limit implementa- org/10.5194/hess-17-1161-2013 tion of this methodology to those who have access to Balsamo, G., Viterbo, P., Beljaars, A., Hurk, B., Hirschi, M., the data. Further, implementation of the AutoRoute Betts, A. K., & Scipal, K. (n.d.). A Revised Hydrology for the and RAPID offer limited accuracy in areas of mini- ECMWF Model: Verification from Field Site to Terrestrial mal terrain because of the methodology’s 1D nature. Water Storage and Impact in the Integrated Forecast System. https://doi.org/10.1175/2008JHM1068.1 The hydrologic forecast currently lacks a means to BBC. (2016). North Korea floods: Tens of thousands displaced. route river discharge through reservoirs. In addition, Retrieved June 29, 2017, from http://www.bbc.com/news/ the model fails to estimate reservoir levels because of no world-asia-37335857 existing reservoir routing scheme and a lack of bathy- Balsamo, G., Albergel, C., Beljaars, A., Boussetta, S., Brun, E., metric data for the reservoirs (Snow et al. 2016; Tavakoly Cloke, H., Dee, D., Dutra, E., Muñoz-Sabater, J., Pappenberger, et al., 2016). Thus, improvements in the ability estimate F., de Rosnay, P., Stockdale, T., and Vitart, F. (2015). ERA- Interim/Land: a global land surface reanalysis data set, bathymetry and provide reservoir routing will substan- Hydrol. Earth Syst. Sci., 19, 389-407, doi:10.5194/hess-19-389- tially improve the process of forecasting inundation. 2015. http://www.hydrol-earth-syst-sci.net/19/389/2015/hess- The current routing methodology implemented 19-389-2015.html within RAPID does not offer the ability for two-way Boeer, J., Steinbrecher, U., Bachmann, M., Schulze, D., & coupling of upland and coastal hydraulic processes. The Braeutigam, B. (2014). Overview and Status of TerraSAR-X / Simulation Program for River Networks (SPRNT) soft- TanDEM-X Long Term System Monitoring. In EUSAR 2014; 10th European Conference on Synthetic Aperture Radar; ware may provide a means of hydraulic routing of LSM Proceedings of (pp. 1–4). runoff as it solves the full Saint-Venant equations (Liu Chae, Y. H., & Birsel, R. (2016). North Korea flood- and Hodges, 2014). This would provide a river rout- ing kills 133, displaces 107,000: U.N. Retrieved June ing methodology that is fully physics-based, which can 29, 2017, from http://www.reuters.com/article/ conceptually link to coastal models like the ADvanced us-northkorea-floods-idUSKCN11I0SR CIRCulation Model (ADCIRC) (Luettich et al., 1992). David, C. H., Maidment, D. R., Niu, G., Yang, Z., Habets, F., & Eijkhout, V. (2011). River Network Routing on the NHDPlus Coupling of the upland and coastal hydrologic and Dataset. Journal of Hydrometeorology, 12, 913–934. https:// hydraulic processes using ADCIRC has been a topic of doi.org/10.1175/2011JHM1345.1 prior and ongoing research (Van Cooten et al., 2011; Durand, M., Andreadis, K. M., Alsdorf, D. E., Lettenmaier, D. P., Tromble et al., 2009). However, SPRNT has not been Moller, D., & Wilson, M. (2008). Estimation of bathymetric parallelized and will likely require additional refine- depth and slope from data assimilation of swath altimetry into ment before large scale implementation is possible. a hydrodynamic model. Geophysical Research Letters, 35(20), 1–5. https://doi.org/10.1029/2008GL034150 Results will also be enhanced with the addition ECMWF. (2016). Set I - Atmospheric Model high resolution of riverine bathymetry. One possibility is the use of 10-day forecast (HRES). Retrieved July 14, 2017, from https:// hydraulic geometry though these are not available on www.ecmwf.int/en/forecasts/datasets/set-i large scale outside the United States. Another possibil- Follum, M. L. (2013). AutoRoute Rapid Flood Inundation Model, ity is the use of remote sensing through the upcoming ERDC/CHL CHETN-IV-88. Surface Water Ocean Topography (SWOT) (Durand Follum, M. L., Tavakoly, A. A., Niemann, J. D., & Snow, A. D. (2016). AutoRAPID: A Model for Prompt Streamflow et al., 2008) mission or from space-borne LiDAR Estimation and Flood Inundation Mapping over Regional Open Water 8

to Continental Extents. JAWRA Journal of the American Sanders, B. F. (2007). Evaluation of on-line DEMs for flood Water Resources Association, 80523, 1–20. https://doi. inundation modeling. Advances in Water Resources, 30(8), org/10.1111/1752-1688.12476 1831–1843. https://doi.org/10.1016/j.advwatres.2007.02.005 GDACS. (2014). Global Flood Detection System - Version 2. Snow, Alan D., “A New Global Forecasting Model to Produce Retrieved June 28, 2017, from http://www.gdacs.org/floodde- High-Resolution Stream Forecasts” (2015). All Theses and tection/searchareas.aspx Dissertations. 5272. Jarihani, A. A., Callow, J. N., McVicar, T. R., Van Niel, T. G., & http://scholarsarchive.byu.edu/etd/5272 Larsen, J. R. (2015). Satellite-derived Digital Elevation Model Snow, Alan D., Scott D. Christensen, Nathan R. Swain, James (DEM) selection, preparation and correction for hydro- Nelson, Daniel P. Ames, Norman L. Jones, Deng Ding, dynamic modelling in large, low-gradient and data-sparse Nawajish Noman, Cédric H. David, F. P. (2016). A High- catchments. Journal of Hydrology, 524, 489–506. https://doi. Resolution National-Scale Hydrologic Forecast System org/10.1016/j.jhydrol.2015.02.049 from a Global Ensemble Land Surface Model. Journal of the Krieger, G., Moreira, A., Fiedler, H., Hajnsek, I., Werner, M., American Water Resources Association, in press(4). https:// Younis, M., & Zink, M. (2007). TanDEM-X: A satellite doi.org/10.1111/1752-1688.12434 formation for high-resolution SAR interferometry. IEEE Snow, Alan D. (2016a). AutoRoutePy: 2.1.0. Zenodo. http://doi. Transactions on Geoscience and Remote Sensing, 45(11), org/10.5281/zenodo.54974 3317–3340. https://doi.org/10.1109/TGRS.2007.900693 Snow, Alan D., Whitaker, Timothy, & Christensen, Scott D. Kugler, Z., & Groeve, T. De. (2007). The Global Flood Detection (2016b). RAPIDpy: 2.3.4. Zenodo. http://doi.org/10.5281/ System. Ispra, Italy. Retrieved from http://www.unisdr.org/ zenodo.159576 files/9622_LBNA23303ENC002.pdf Tromble, E., Kolar, R., Dresback, K., Hong, Y., Vieux, B., Luettich, Lehner, B., Verdin, K., & Jarvis, A. (2013). HydroSHEDS R., Gourley, J., Kelleher, K., Van Cooten, S. (2009). Aspects Technical Documentation Version 1.2. Retrieved from http:// of Coupled Hydrologic-Hydrodynamic Modeling for Coastal www.hydrosheds.org Flood Inundation. In M. L. Spaulding (Ed.), Proceedings Lehner, B., Grill G. (2013): Global river hydrography and net- of the 11th International Conference on Estuarine and work routing: baseline data and new approaches to study the Coastal Modeling (pp. 724–743). Seattle, Washington: world’s large river systems. Hydrological Processes, 27(15): American Society of Civil Engineers (ASCE). https://doi. 2171–2186. Data is available at www.hydrosheds.org. org/10.1061/41121(388)42 Liu, F., & Hodges, B. R. (2014). Applying microprocessor anal- Van Cooten, S., Kelleher, K. E., Howard, K., Zhang, J., Gourley, ysis methods to river network modelling. Environmental J. J., Kain, J. S., … Spence, L. (2011). The CI-flow project: A Modelling and Software, 52, 234–252. https://doi. system for total water level prediction from the summit to the org/10.1016/j.envsoft.2013.09.013 sea. Bulletin of the American Meteorological Society, 92(11), Luettich, R. A., Westerink, J. J., Scheffner, N. W. (1992). ADCIRC: 1427–1442. https://doi.org/10.1175/2011BAMS3150.1 An Advanced Three-Dimensional Circulation Model for Wahl;, M., Follum, M., Snow, A., & Tavakoly, A. (2016). Shelves, Coasts, and Estuaries. Washington, D.C. Retrieved Developing Hydrologic Awareness. The Military Engineer, from http://oai.dtic.mil/oai/oai?verb=getRecord&metadata- 108(700). Retrieved from http://themilitaryengineer.com/ Prefix=html&identifier=ADA261608 index.php/staging/item/556-march-april-2016 Maidment, D. R. (2016). Open Water Data in Space and Time. Wendleder, A., Wessel, B., Roth, A., Breunig, M., Martin, K., Journal of the American Water Resources Association, 52(4), & Wagenbrenner, S. (2013). TanDEM-X water indication 816–824. https://doi.org/10.1111/1752-1688.12436 mask: Generation and first evaluation results. IEEE Journal NASA. (2017). NRT Global Flood Mapping. Retrieved June 28, of Selected Topics in Applied Earth Observations and 2017, from https://floodmap.modaps.eosdis.nasa.gov/index. Remote Sensing, 6(1), 171–179. https://doi.org/10.1109/ php JSTARS.2012.2210999 NGA. (2015). VISNAV - Landuse \ Landcover. Retrieved June 29, 2017, from https://www.arcgis.com/home/item. html?id=bbeb2585b4704b67bf81d1744acbc55d Nigro, J., Slayback, D., Policelli, F., & Brakenridge, G. R. (2014). NASA / DFO MODIS Near Real-Time ( NRT ) Global Flood Mapping Product Evaluation of Flood and Permanent Water Detection. Retrieved from https://floodmap.modaps.eosdis.nasa.gov/documents/ NASAGlobalNRTEvaluationSummary_v4.pdf Policelli, F., Slayback, D., Brakenridge, B., Nigro, J., Hubbard, A., Zaitchick, B., Carroll, M., and Jung, H. (2017). The NASA Global Flood Mapping System: Final Draft. In V. Lakshmi (Ed.), Remote Sensing of Hydrological Extremes (pp. 47–64). Switzerland: Springer. https://doi. org/10.1007/978-3-319-43744-6