Standards dramatically advance streamflow and flood forecasting in US and elsewhere – Part 3 of 5


In Part 1 of this series, I introduced the U.S. Open Water Data Initiative (OWDI) and the National Flood Interoperability Experiment (NFIE). In Part 2, I described the data and modeling frameworks for NFIE that produced the first high-resolution streamflow forecast at continental-scale of the United States, in near-real-time. Part 3 (this segment) is about a number of related projects that are adapting this technology for use in other countries. Part 4 (next segment) will be about the underlying data exchange standards that made this possible (OGC WaterML 2.0 and netCDF-CF), and Part 5 (final segment) will be about the evolution of WaterML 2.0 to TimeseriesML. 

Contributors to this segment include Jim Nelson, Brigham Young University; Ahmad Tavakoly, USACE-ERDC; and Michael Follum, USACE-ERDC

National Water Observations: Publishing, Discovery, and Access

First, we’ll look at publishing national water observations over the web. This part of the story really started in 2012, the year WaterML 2.0 Part 1 – Timeseries was adopted as an international standard by the Open Geospatial Consortium (OGC). WaterML 2.0 is a conceptual model and an XML encoding for water observations (e.g., river discharge or streamflow, stream depth, and other characteristics) as time series. Hydrology professor David R. Maidment at the University of Texas at Austin, one of the driving forces behind WaterML 2.0, started looking for ways of encouraging international use of this data standard among national water agencies.  One of his first steps was to build a GIS database of stream gage locations, using data from the U.S. Geological Survey (USGS) National Water Information System (NWIS), the Global Runoff Data Center (GRDC) in Germany, and other sources developed by graduate students at UT Austin. An interactive map of stream gage locations provided a visual catalog or index to all stream gages, enabling a user to quickly find those of interest. Clicking on a gage location in the map generated a popup window with links to the time series data and hydrograph for that gage, see Figure 1. 

For outreach, I added this map to the Global Earth Observation System of Systems (GEOSS) registry, and it became discoverable and accessible from anywhere. [1] [2] [3]

Figure 1

Figure 1. Finding Streamflow Data from a Web Map of Gages

However, we did not have a program to keep this kind of map up to date, so wanted to stimulate national agencies to host their own water data and maps online. Following this initial experiment, I led the Water Societal Benefit Area (SBA) theme activity for the GEOSS Architecture Implementation Pilot (AIP) for the next 3 years, with the help of Professors Jim Nelson and Dan Ames from Brigham Young University (BYU). With AIP-6 in 2013 [4] [5] [6]  and AIP-7 in 2014 [7] [8] , we had successes in Italy, New Zealand, Canada, France, Belgium, Taiwan, and a number of countries in Latin America. 

While many organizations and people were involved in these, I want to highlight the exceptional work of Dr. Silvano Pecora, hydrologist in the Po River basin, northern Italy. Dr. Pecora ported the CUAHSI Hydrologic Information System (HIS) central data server, together with many standardized and interoperable facilities developed in the context of an international community, to run in Italy. Following on this, the Italian Ministry of Environment (ISPRA) promoted this system to be used nationwide. The ISPRA HIS now has all of Italy’s stream gages (and rainfall and temperature monitors) registered and publishing in real-time, and is also registered in GEOSS. These advancements brought for the first time in Italy’s history the ability to understand the national water situation on demand. As a result of this project, Dr. Pecora was appointed as a technical representative to the World Meteorological Organization (WMO) Commission for Hydrology (CHy), where he has conducted outreach and development to promote water data publishing by other countries. [10]  The WMO Hydrological Observing System (WHOS) is now online, largely through Dr. Pecora’s efforts. He’s gone from being a local hydrologist to international technical diplomat in just a couple years, what an achievement!

Figure 2

Figure 2. WMO Hydrological Observing System (WHOS)

Another national success was in New Zealand, through the efforts of Sean Hodges and Brent Watson from Horizons Regional Council (HRC) on the north island, and Dr. Jochen Schmidt, Chief Scientist for the National Institute for Water and Atmospheric Research (NIWA). Their efforts in publishing water data on the web for HRC led to OGC WaterML 2.0 becoming a recognized national standard within New Zealand within 2 years. Web publishing of water data is now being implemented throughout the country, following HRC’s approach. New Zealand also now has a public web site called Land Air Water Aotearoa (LAWA) providing access to water quantity and quality data throughout the country. Kudos to the NZ team! 

In 2014, Dr. Maidment’s team had reached a breakthrough in the U.S., forecasting the continental streamflow in high resolution and near-real time, as part of NFIE (see previous article in this series). Also during 2014, we connected with the Global Flood Awareness System (GloFAS), an activity of the European Commission’s Joint Research Centre (JRC) in Italy and the European Centre for Medium-range Weather Forecasting (ECMWF) in UK. GloFAS predicts probability of flooding up to 30 days in advance for many large river basins around the world. However, GloFAS predictions are not as high-resolution as we could generate with RAPID, and are too coarse to work well for smaller watersheds or island nations. For GEOSS AIP-8 in 2015 [9], we decided to conduct global outreach for NFIE, dubbing our activity in AIP-8 as “GFIE”.  The rest of this story is about that activity.

Global Forecasts: Predicting Rainfall and Streamflow 

The U.S. National Weather Service (NWS) predicts near-term rainfall just for the U.S. Most developed countries, and many developing countries, have similar national programs for predicting their near-term weather. But none of them have been predicting national-scale streamflow, as the NWS is starting to do in the U.S.  To bring NFIE to other countries, we are grateful to leverage the ECMWF two-week weather forecasts, for which they produce a 51-member ensemble of hydro-meteorology variables every 12 hours. Dr. Nelson at BYU, with significant contributions from Alan Snow (now at the U.S. Army Corps of Engineers) and using tools developed by Esri, has led the way to integrate Cedric David’s RAPID streamflow routing model with the ECMWF runoff ensembles, downscaled from the ECMWF forecast model grid to the catchment areas mapped for the country. Until recently, ECMWF output grid was 28km resolution for the ensemble and 14km for its high-resolution forecast. Starting in early March 2016, the ECMWF output grids are 16km for the ensemble and 9km for the high-resolution forecast. (By comparison, the US NWS forecasts had 3km resolution in 2015; this will be 1km resolution starting in 2016.)

Figure 3 illustrates a web application built using the Tethys Platform open source water resources web app tools developed at BYU. This uses rainfall and land-surface dynamics predictions from the ECMWF. The US NWS does not plan to use the ECMWF forecasts, but academically, we can apply this approach anywhere in the world. 

Figure 3

Figure 3. Tethys Streamflow Prediction Tool

In Figure 3, the colored triangles on the map represent potential flood levels based on flood return rates of 2 years (yellow), 10 years (orange) and 20 years (purple). These return rates are based on statistics derived by simulating flows for each river reach with RAPID going back to 1980. These simulated “historical” statistics are derived from the ECMWF Re-Analysis (ERA Interim) model, which estimates the 35-year history of every grid cell as a globally-consistent principle of measurement, see Figure 4. The streamflow time series outputs for each stream reach are available in netCDF-CF format, with plans to provide the output in WaterML in the future.  

In the charts of Figure 3 and Figure 4, the green area and dark green line (called ECMWF in the legend below the graph) represent the ensemble bounds and mean, refactored to the selected catchment and stream segment. The black line labeled ECMWF – High Res is for ECMWF’s high-resolution forecast. 

Figure 4

Figure 4. ERA Interim Backcast Applied to Streamflow

In May 2015, Dr. Nelson of BYU and Angelica Gutierrez of the National Oceanic and Atmospheric Administration (NOAA) International Working Group teamed to organize and conduct an outreach workshop in Cartagena Colombia, under the auspices of AmeriGEOSS and CIEHLYC (Comunidad Para la Informacion Espacial e Hidrografica para Latinoamerica y el Caribe, see the November 2015 Newsletter and the CIEHLYC Webinar Series). This workshop was attended by about 80 representatives from Latin America. About 20 of these were involved in water data management for 14 major watersheds in 4 countries (mostly Colombia). Dr. Nelson led a hands-on workshop to help them download and learn to use CUAHSI’s HydroDesktop and HydroServer.  Participants mapped their own home watersheds, stream reaches, catchments, and many stream gages. These were all registered in the Tethys Streamflow Prediction app server at BYU (see Figure 5), and are continuing to produce streamflow predictions from the ECMWF runoff forecasts, every twelve hours. 

Figure 5

Figure 5. Selection of Watersheds Modeled for Streamflow Prediction in Latin America, May 2015

AutoRAPID: From Streamflow Routing to Flood Inundation

The last part of this story is to take the next step after forecasting streamflow rates, to predict and map flood inundation areas. The U.S. Army Corps of Engineers (USACE), led by Michael Follum and Dr. Ahmad Tavakoly in the Engineering Research and Development Center (ERDC), have linked RAPID outflow with the AutoRoute flood inundation model to produce operational high-resolution flood inundation maps (as shapefiles) over large extents.  The combined approach (referred to as AutoRAPID) has been deployed in several regions around the world (e.g. Bulgaria, England, Niger River basin, etc.) and tested against observed flood data within the U.S. [11] The AutoRAPID model shows a capability in preparing for oncoming flood events, such as Typhoon Koppu that flooded parts of the Philippines in October 2015. [12

Figure 6 illustrates the workflow. The left side of this diagram shows the streamflow simulation component (the RAPID model), and right side represents the flood mapping estimation component (AutoRoute). The outputs of AutoRAPID are (a) time series of streamflow and (b) the flood inundation map, as shown in Figure 7.  This figure shows deployments of the model to the Maritsa and Sava River Basins in Europe to simulate the most extreme flow event in each river reach using the ERA-Interim dataset (1979-2015). 

Figure 6

Figure 6. AutoRAPID Modeling Framework

AutoRAPID provides a reasonable first-order approach in estimating high-resolution flow and flood inundation at the regional to continental scale.  The accuracy of the flood results is most dependent on the accuracy of runoff products (inputs to the RAPID model) and on the resolution, accuracy, and precision of the elevation data.  The flood mapping component (AutoRoute) tends to perform well in areas defined by mid-to-high topographic relief with well-defined channels.  Flood mapping results show some errors in urban, coastal, and topographically flat areas.  Further testing and modifications to the model are currently underway.  In areas where the AutoRAPID model does not perform well, such as coastal settings, connections to higher-fidelity models are being made.  Although AutoRAPID has shown to be a useful tool, different approaches to simulate flood inundation at the regional scale, and inter-comparison of different models, are currently under development. The AutoRoute model and RAPID integration are planned for release to the public via github.

Figure 7

Figure 7. AutoRAPID Results

For more information… 

Both the BYU team and the USACE-ERDC team are continuing development of the tools mentioned for global outreach. Please contact the principals if you’re interested in learning more. Further advancements in streamflow and flood inundation mapping are expected during the NFIE Summer Institute 2016. Results of this will be presented at the CUAHSI Biennial Meeting, July 25-27, 2016. Check back at the NFIE Summer Institute link for updates. 

A larger collaboration is also underway, in which GloFAS is a key component: the Global Flood Partnership (GFP), an activity of the Global Disaster Alert and Coordination System (GDACS). From its website, “GDACS is a cooperation framework between the United Nations, the European Commission and disaster managers worldwide to improve alerts, information exchange and coordination in the first phase after major sudden-onset disasters.” Further, “The Global Flood Partnership is a cooperation framework between scientific organisations and flood disaster managers worldwide to develop flood observational and modelling infrastructure, leveraging on existing initiatives for better predicting and managing flood disaster impacts and flood risk globally. GFP is hosted as an Expert Working Group by GDACS.” These are important resources and networking channels for sharing practices, data and methods.  


Deep gratitude goes to those who were involved in development of these workflows, and review of this report: David R. Maidment, Silvano Pecora, Jochen Schmidt, Sean Hodges, Brent Watson, Cedric David, Florian Pappenberger, Peter Salamon, Jim Nelson, Dan Ames, Norm Jones, Nathan Swain, Alan Snow, Nawajish Noman, Ahmad A. Tavakoly, Michael L. Follum, Fernando R. Salas, Gonzalo Espinoza, and Tim Whiteaker. Various others contributed through CUAHSI, HydroShare, USACE, Esri, and other organizations mentioned in this article.   

David Arctur is a Research Scientist and Fellow at the University of Texas at Austin. He is a member of the EarthCube Leadership Council, and a senior participant in a number of EarthCube funded projects. EarthCube is a long-term, interdisciplinary NSF initiative to improve sharing of data and models within and across science domains. He also serves as Research/Academic Advocate for the Open Geospatial Consortium (OGC). See his blog at


  1. Fernando Salas, David Maidment, David Arctur. AIP-5 Water SBA: World Water Online, Linking People Everywhere with Water Data, Maps and Models. Presentation to GEO IX Plenary, Foz do Iguacu, Brazil, November 2012. URL: Video: . Back
  2. David Arctur, Matt Austin, Stefan Fuest, Perry Peterson. Water Scenario Engineering Report: GEOSS Architecture Implementation Pilot, Phase 5, February 2013. URL: Back 
  3. Open Geospatial Consortium, 2013. GEOSS Infrastructure Enhancements: Outcomes of GEOSS Architecture Implementation Pilot AIP-5. URL: Back
  4. David R Maidment, David K Arctur. GEOSS Water Services for Data and Maps: Report of the Water SBA for AIP-6. Presented to GEO X Plenary and Ministerial Summit, Geneva, January 2014. URL: Back
  5. David Arctur, Cinzia Alessandrini, Silvano Pecora, Jim Nelson, Peter Salamon. GEOSS Water Services for Data and Maps Engineering Report: GEOSS Architecture Implementation Pilot, Phase 6. March 2014. URL: Back
  6. Open Geospatial Consortium, 2014. GEOSS Infrastructure Enhancements: Outcomes of GEOSS Architecture Implementation Pilot AIP-6. URL: Additional materials from the Water SBA activity:  Back
  7. David Arctur, ed., AIP-7 GEOSS Water Services: Federating Water Resources Data Globally. Presented to GEO XI Plenary, Geneva, November 2014. URL: Video, Flemish Water Portal: Video, New Zealand Clients: Back
  8. Open Geospatial Consortium, 2015. GEOSS Infrastructure Enhancements: Outcomes of GEOSS Architecture Implementation Pilot AIP-7. URL: Additional materials from the Water SBA activity: Back
  9. David Arctur, ed., GEOSS Water Services: Publishing National & Global Streamflow and Flood Forecasts, Water Theme for GEOSS AIP-8. Presented to GEO XII Plenary and Ministerial Summit, Mexico City, November 2015. URL: Back
  10. World Meteorological Organization, 2015. WMO Data Operations and Management Global initiatives in hydrological data sharing URL: . Back
  11. Michael Follum, A.A. Tavakoly, J.D. Niemann, A.D. Snow, In Review. AutoRAPID: A Model for Prompt Streamflow Estimation and Flood Inundation Mapping over Regional to Continental Extents. Back
  12. Wahl, M., M.L. Follum, A.D. Snow, A.A. Tavakoly, 2016. Developing Hydrologic Awareness. The Military Engineer (700):65-66.Back

(See David Arctur's blog page.)