How To Add Nws River Gauges Web Services To Openlayers

• Journal Listing
• Wiley-Blackwell Online Open
• PMC6686479

J Am Water Resour Assoc. 2016 Aug; 52(4): 950–964.

A High‐Resolution National‐Scale Hydrologic Forecast System from a Global Ensemble State Surface Model^†

Alan D. Snow, Research Civil Engineer, ⁱ Scott D. Christensen, Graduate Educatee, ⁱⁱ Nathan R. Swain, Graduate Student, ⁱⁱ E. James Nelson, Professor, ^two Daniel P. Ames, Professor, ^two Norman L. Jones, Professor, ² Deng Ding, Product Engineer, ³ Nawajish S. Noman, Lead Projection Engineer, ³ Cédric H. David, Scientist, ⁴ Florian Pappenberger, Scientist, ⁵ and Ervin Zsoter, Scientist ⁵

Alan D. Snow

¹ Coastal and Hydraulics Laboratory, U.s. Army Engineer Enquiry and Development Center, 3909 Halls Ferry Rd, Vicksburg, MS, 39180

Scott D. Christensen

² Department of Civil and Environmental Engineering, Brigham Immature University, Provo, UT, 84602

Nathan R. Swain

² Department of Civil and Environmental Engineering, Brigham Young University, Provo, UT, 84602

Eastward. James Nelson

^two Department of Ceremonious and Ecology Engineering, Brigham Young University, Provo, UT, 84602

Daniel P. Ames

² Department of Civil and Environmental Engineering, Brigham Young University, Provo, UT, 84602

Norman L. Jones

ⁱⁱ Section of Civil and Environmental Technology, Brigham Young University, Provo, UT, 84602

Deng Ding

³ Esri, Redlands, CA, 92373

Nawajish S. Noman

³ Esri, Redlands, CA, 92373

Cédric H. David

^iv Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA, 91109

Florian Pappenberger

^v European Center for Medium‐Range Weather condition Forecasts, Shinfield Park, Reading, United kingdom

Ervin Zsoter

⁵ European Centre for Medium‐Range Weather Forecasts, Shinfield Park, Reading, United kingdom

Received 2015 Jun 8; Accepted 2016 April 13.

Abstract

Alarm systems with the ability to predict floods several days in advance have the potential to benefit tens of millions of people. Accordingly, large‐scale streamflow prediction systems such every bit the Advanced Hydrologic Prediction Service or the Global Flood Sensation System are limited to fibroid resolutions. This article presents a method for routing global runoff ensemble forecasts and global historical runoff generated past the European Eye for Medium‐Range Weather Forecasts model using the Routing Application for Parallel computation of Discharge to produce high spatial resolution xv‐twenty-four hour period stream forecasts, approximate recurrence intervals, and warning points at locations where streamflow is predicted to exceed the recurrence interval thresholds. The processing method involves distributing the computations using reckoner clusters to facilitate processing of large watersheds with loftier‐density stream networks. In addition, the Streamflow Prediction Tool web application was developed for visualizing analyzed results at both the regional level and at the attain level of high‐density stream networks. The application formed part of the base of operations hydrologic forecasting service bachelor to the National Flood Interoperability Experiment and can potentially transform the nation's forecast ability by incorporating ensemble predictions at the nearly 2.7 million reaches of the National Hydrography Plus Version 2 Dataset into the national forecasting organisation.

Keywords: ECMWF, RAPID, Tethys Platform, CondorPy, HTCondor, CI‐Water, GloFAS, NFIE, flood prediction, streamflow prediction, forecast

Introduction

Catastrophic floods impact tens of millions of people each year and cause significant infrastructure damage. Global statistics for the period of 2004‐2014 indicate that more than 951 meg people were impacted past floods, over $324 billion in impairment occurred, and in that location were approximately 66,000 deaths (Guha‐Sapir et al., 2015). Improvements in flood forecasting and the ability to communicate actionable data to emergency responders accept a substantial lifesaving and budgetary do good (Pappenberger et al., 2015). Because of this, i of the kickoff priorities of the new National Water Centre in Tuscaloosa, Alabama ( http://www.nws.noaa.gov/oh/nwc/) is to appoint the academic community in the National Flood Interoperability Experiment (NFIE) (Maidment, 2015). NFIE aims to address several disquisitional science and technology questions including: (1) How can the National Hydrography Plus Version 2 (NHDPlus V2) dataset (Horizon Systems Corporation, 2011) be used to generate nationwide near‐real‐time hydrologic simulations at high spatial resolution? (2) Can such modeling atomic number 82 to improved emergency response and community resilience? (3) What is a sustainable path from inquiry to operations in terms of flood forecasting (CUAHSI, 2015)? This article begins to address these questions by presenting a computational forecast framework and a web‐based visualization application that has the potential to be a office of the national forecasting arrangement with near‐real‐time high‐resolution ensemble flood forecasts. This system forms part of the foundation from which the NFIE can work to accomplish the stated goals.

Advances in geospatial information, atmospheric and weather information, hydrologic modeling, and calculating resources have led to an improved ability to make instream forecasts. There are several essential elements of the National Weather condition Service's (NWS) Hydrological Ensemble Forecasting Service (HEFS) that tin can be used as a model for a flood forecasting arrangement. The master elements include: (i) a meteorological ensemble forecast; (2) a hydrologic processor that inputs the meteorological data into hydrologic, hydraulic, and reservoir models; (3) a hydrologic ensemble postprocessor to account for total hydrologic uncertainty; and (4) an ensemble verification service to identify the skill and fault in the forecast (Demargne et al., 2014). In this article, we demonstrate a method to improve the spatial resolution of the hydrologic routing portion of a streamflow prediction system.

The NWS hosts a spider web‐based hydrologic prediction system known as the Avant-garde Hydrologic Prediction Service (AHPS). The predictions are created using the Advanced Weather Interactive Processing System (AWIPS) which consists of automated cuff data, satellite data, Doppler radars, weather observation stations, advanced figurer models, and super computers. The AHPS tin can display the forecasted streamflow, the forecasted flood level, the probability of flooding, and maps of the areas potentially affected past the flooding at many of the iii,600 forecast stations nationwide. These predictions tin can range from hours to months in accelerate (Mcenery et al., 2005; NOAA, 2015).

In addition, global atmospheric condition forecasts and hindcasts are available from multiple sources. Dr. David Maidment's group at the Academy of Texas is working on developing a high‐resolution forecasting arrangement using United States (U.S.)‐based models and datasets for NFIE (Salas et al., 2014; Maidment, 2015). Therefore, in this article, we focus on the global runoff datasets generated by the European Centre for Medium‐Range Weather Forecasts (ECMWF). The ECMWF global gridded runoff prediction dataset includes surface and subsurface runoff depth in meters derived from the Tiled ECMWF Scheme for Surface Exchanges over Land with a revised land surface Hydrology (HTESSEL) land surface model (Balsamo et al., 2009; ECMWF, 2011). Modeling and forecasting is intrinsically uncertain (Buizza et al., 2005; Pappenberger and Beven, 2006; Slingo and Palmer, 2011; Beven et al., 2015). The ECMWF produces Numerical Atmospheric condition Prediction (NWP) ensemble forecasts as a method for amend representing and communicating the uncertainty in the forecast (Beven and Cloke, 2012). The ECMWF likewise produces a dataset for historical runoff that is an output of the global atmospheric reanalysis ERA‐Acting and begins in 1979 and extends to the present with near‐real‐time updates (Dee et al., 2011).

The Global Alluvion Sensation System (GloFAS), developed by ECMWF and the Articulation Enquiry Centre of the European Commission, is a coupled hydro‐meteorological model that generates ensemble streamflow predictions for big‐scale river basins globally for up to 15 days in advance (Alfieri et al., 2013). The GloFAS grid cell size (0.i°) is also large for determining local impacts for watersheds smaller than 10,000 km²; hence, high‐resolution hydrologic, hydraulic, and flood touch models are required for more detailed forecasts capable of producing actionable information at the local level.

Although the AHPS provides national coverage and GloFAS provides global coverage, both are limited to a relatively coarse spatial resolution for streamflow predictions. The resolution gap tin be bridged by routing the ECMWF runoff predictions through the NHDPlus V2 stream network using the Routing Awarding for Parallel ComputatIon of Discharge (RAPID) model (David et al., 2011; David, 2013). This ECMWF‐RAPID integration generates streamflow predictions at a local scale corresponding to the U.Due south. NHDPlus V2 dataset. The ECMWF‐RAPID integration has the potential to be incorporated into the national forecast system to increase the resolution to virtually 2.7 million forecast points with predictions every bit an ensemble. Nevertheless, before this can occur, several improvements demand to occur in the electric current ECMWF‐RAPID organization such as initialization, calibration, and calculation reservoir routing. This organization has been developed for use at the NFIE Summertime Establish (Tuscaloosa, Alabama, summer 2015) where further testing, comparison, and awarding will continue to provide a pathway that transforms the spatial density of forecasting while incorporating ensemble forecasts that tin can better communicate uncertainties involved.

Methods

The goal of this piece of work was to produce higher spatial resolution streamflow predictions and provide an intuitive method for viewing the predictions. This was accomplished past developing a preprocessing method using ArcGIS tools to downscale both the ECMWF gridded runoff prediction dataset and ERA‐Interim reanalysis gridded runoff dataset to the catchment level to create input data for RAPID. The RAPID model is run for the period 1980‐2014 using the ERA‐Interim runoff to generate approximate historical streamflow and return menstruum data. Then, the RAPID model is run for each of the 52 ECMWF forecasts in the ensemble every 12 h using a distributed computational workflow. When the process repeats, future forecasts are initialized using the average of the forecasted streamflows from the previous forecast for each attain. Alert points are generated at the locations where the forecasted streamflow exceeds the estimated xx‐twelvemonth, 10‐year, and 2‐year return periods. Finally, a web app was created using Tethys Platform (Jones et al., 2014; Swain, 2015) to manage and visualize the high‐resolution stream forecasts and warning points for decision‐makers in a standardized, intuitive format. The application also incorporates AHPS predictions and U.S. Geological Survey (USGS) observed streamflows for comparison and validation and so that improvements can be fabricated in subsequent iterations of the system.

The ECMWF global runoff forecast ensemble and ERA‐Interim global reanalysis runoff dataset were used in this research. The ECMWF global runoff ensemble provides 52 separate predictions every 12 h that approximate cumulative runoff depths. The offset 51 predictions represent an ensemble of equally likely weather condition and are created at a lower resolution on a ~0.28‐degree filigree cell (upwardly to twenty-four hours 10, ~0.56‐degree filigree prison cell thereafter) with a six h accumulated runoff time step and a 15‐mean solar day lead time. The 52nd forecast is a deterministic "best judge" solution produced at a higher resolution with a ~0.14‐degree grid cell and a varying fourth dimension pace of accumulated runoff with a ten‐day lead time. The ERA‐Interim dataset used has a T511 grid (~39 km grid cell), has a daily time pace, and spans the years 1980‐2014.

The RAPID model was used to route the ECMWF runoff through the NHDPlus V2 stream networks. RAPID is an open source model used to route runoff of surface and groundwater inflow to rivers downstream with any density stream network (David et al., 2011). The NHDPlus V2 dataset combines the National Hydrography Dataset, the Watershed Purlieus Dataset, and the National Height Datasets (NED) and adds attributes that define stream social club and facilitates rapid stream network traversal and query (Horizons Systems Corporation, 2011). Reach routing with RAPID is based on the traditional Muskingum routing method which has two main parameters k and ten, where k is a storage constant with a time dimension and x characterizes reach properties that contribute to wave diffusion, is dimensionless, and is stable from 0 to 0.5 (Cunge, 1969).

The loftier‐resolution nature of the NHDPlus V2 stream network necessitates geoprocessing to convert the ECMWF runoff forecasts into the format required to route with RAPID. To this cease, a collection of gratis and open up source Python tools has been adult every bit geoprocessing tools for ArcGIS. The NHDPlus V2 dataset is conveniently available as an ArcGIS geodatabase. The ArcGIS tools are used to ready inflow to each reach in the stream network by converting the ECMWF model forecast from a gridded runoff depth to runoff volume using the NHDPlus V2 catchments (we will refer to this process every bit "downscaling"), and by generating other coincident inputs for the RAPID flood routing model to ensure smooth and efficient data transfer between models. The post-obit preprocessing operations are performed using the ArcGIS tools (version 10.three or greater):

1. Create the stream network connectivity file past traversing the NHDPlus V2 network and considering upstream and downstream connectivity.

2. Calculate the Muskingum parameters (g and x) based on stream lengths and flow moving ridge celerity input, and create the Muskingum parameter files for RAPID. The moving ridge celerity is the speed at which the h2o flow wave propagates in a river channel. The k parameter in the Muskingum method can exist computed based on the value of the flow wave celerity using the equationk =50/c, where L is the length of the river reach, and c is the celerity of the flow wave going through information technology. More information on the human relationship between flow wave propagation and the Muskingum method can exist found in Cunge (1969) or in David et al. (2011).

3. Create a weight tabular array past overlaying the NHDPlus V2 catchments on ECMWF runoff grids. The weight table describes the area (Ai) of the runoff grid prison cell (i) that overlays each catchment (Effigy1).

   Effigy 1

   

   Weight Table Derivation Analogy with Runoff Grid and Catchment (Step three).

4. Create the arrival file for the stream network by calculating the weighted average runoff volume from the ECMWF forecast at each fourth dimension footstep for the catchment that corresponds to each stream reach.

where V _j is the runoff book in gⁱⁱⁱ at time step j, A _i is the expanse of the catchment in m^two in grid cell i, and R _ji is the runoff depth in meters in the grid jail cell i at time step j, and due north is the number of grid cells that contribute runoff to the catchment.

The ArcGIS preprocessing workflow Steps 1‐3 produces static files, so they but need to be performed one time for each watershed that is incorporated in the system. The resulting inflow file from Step 4 is a Network Common Data Course (NetCDF) file compatible with RAPID and contains the incremental runoff divers at each time step of the forecast for each catchment in the watershed.

Computational Forecast Framework

Earlier running the forecasts, the ERA‐Acting information is downscaled and then routed in RAPID from 1980 to 2014 to produce 35 years of daily streamflow estimates. From this data, using a unproblematic Weibull distribution (Benson, 1962) with the partial elapsing serial method, estimates for render periods are generated for each of the reaches in the continental U.Southward. Due to fourth dimension and computational constraints, the NFIE Mississippi Region return period data were non generated in this study.

The overall downscaling and routing process in prediction mode uses a parallel computational forecast framework illustrated in Figureii. In Footstep 1 the ECMWF forecast ensemble is retrieved. The downscaling and RAPID routing are performed in Footstep 2 and the simulations are distributed across multiple machines in order to decrease computation time. The computation tin can exist summarized every bit two processes: Step 2a the weight table preprocessed earlier past the Esri tools is used to downscale the ECMWF runoff forecast and Step 2b the forecasted runoff is routed through the reaches using RAPID. From the average of all of the forecasts generated from Step 2, in Step 3 the RAPID streamflow initialization file is created for the side by side prediction to and then be used in Footstep 2b. In Step four, the render period data generated from ERA‐Interim information are used to generate alarm points where the average or ane standard deviation above the average of forecasted period from Step ii at each reach exceeds the render period. In Step v the instream forecasts and alarm points are deposited in a CKAN data store. CKAN is an open up source information management portal that streamlines the process sharing and publishing data (Open Knowledge Foundation, 2013). Finally, in Step 6, a Tethys Platform web app downloads the forecasts and warning points from CKAN to display them to the user.

Schematic of ECMWF‐RAPID Downscaling and Routing Process in Prediction Mode.

The ECMWF ensemble runoff forecasts are published every 12 h, introducing a constraint on the computation time of the downscaling and routing process. If the unabridged process were to require an execution fourth dimension longer than 12 h, computations on updated datasets would be delayed, causing a lag. To prevent this state of affairs on the big national‐scale dataset, efficient computation methods are required to downscale the forecasts in a timely manner and enable the system to run operationally.

The method used to improve computational efficiency was to distribute the computations between estimator processors. The distribution method can be practical to a single server with multiple processing cores or to a cluster of computers with shared processing cores as would be available via cloud services such as Amazon Web Services (AWS) or Microsoft Azure. In this study, both a local compute cluster and an AWS compute cluster, each with 52 cores, were used to perform computations. Using this compute cluster, it was possible to simultaneously downscale each of the 52 ECMWF runoff forecasts in the prediction ensemble for each private watershed. HTCondor and a custom Python library named CondorPy ( https://pypi.python.org/pypi/condorpy) were used to distribute the computations in Step ii in the workflow. HTCondor is a batch‐scheduling and resource direction software that distributes jobs to computing resource based on resource availability (Buyya et al., 2013). CondorPy interfaces with HTCondor and is used to facilitate programmatic chore creation and submittal.

Streamflow Prediction Tool Web App

The last product of the downscaling and routing procedure is 52 NetCDF files, one for each member of the ensemble forecast, that contain the predicted hydrographs for each reach of the watershed. This information needs to be communicated to end‐users in a comprehensible format so that it tin can inform conclusion‐makers and the public at large (Pappenberger et al., 2013). We addressed this need past developing a web application or web app chosen the "Streamflow Prediction Tool" using Tethys Platform. The spider web app medium is an effective way to share new developments in water resources modeling, information science, and decision support. Tethys Platform bridges the divide that prohibits many water resources scientists and engineers from developing spider web apps by providing (1) a suite of free and open up source software that addresses the unique data and computational needs common to h2o resource web app development, (2) a Python software development kit for incorporating the functionality of each software element into spider web apps, and (3) a customizable web site that can be used to deploy the finished web apps. Amidst the software projects included in the Tethys Platform are GeoServer, 52 North WPS, PostgreSQL with PostGIS, OpenLayers, Google Maps, Highcharts JS, and HTCondor (Jones et al., 2014; Swain, 2015).

Nosotros designed the Streamflow Prediction Tool to consume the ensemble stream forecast produced in Stride 2 and the warning points produced in Stride 4 from the CKAN information store in Step 6 as shown in the workflow in Figure2. This allows the computation and visualization to operate independently, offer flexibility in deployment of the system. The web app automatically maintains a cache of the most recent weeks' worth predictions and warning points on the server to facilitate faster access to the data. The netCDF4‐python and NumPy Python modules are used to extract, compile statistics, and analyze the stream forecast prediction ensemble.

The Streamflow Prediction Tool provides an intuitive user interface that allows the easy lookup and visualization of results (Figure3). GIS visualization of the stream network and other spatial layers was accomplished through a coupling of GeoServer 2.7.0 spatial data publishing and OpenLayers 3.2.1 spatial mapping systems. Stream layers are served as an Open Geospatial Consortium Spider web Feature Service (OGC‐WFS) and the other spatial layers are served past GeoServer as OGC Web Map Services (OGC‐WMS) (Michaelis and Ames, 2012; Open Geospatial Consortium, 2015a, b). OpenLayers is used to query GeoServer using OGC‐WFS and OGC‐WMS and display the layers in an interactive map. On the map, clicking on a reach volition look up the forecast for that reach.

Screenshot of Streamflow Prediction Tool App.

The ensemble forecast is summarized and displayed on a Highcharts plot ( world wide web.highcharts.com), which provides interactive visualization. In the plot, the blackness line represents the high‐resolution forecast and the greenish bands represent the uncertainty of all 52 members of the prediction ensemble. Where USGS stream gages exist, that serve observed information, the data are added to the chart for preceding days on selected stream reaches, which tin be particularly useful for evaluating functioning when viewing older forecasts for which now the observed data exists. If available, AHPS stations provide streamflow predictions or observed data for some of the reaches. The observed USGS cuff information for the station will announced as a bluish dashed line and the AHPS station information will announced as a purple dashed line on the plot. Additionally, the estimated render menstruum information is also shown on the chart with colored bands. The xanthous band represents the range of flows betwixt the 2‐yr and x‐year return catamenia streamflows. The reddish band represents the range of flows betwixt the x‐year and 20‐year return period streamflows. And the regal band represents flows exceeding the 20‐year render period streamflow. Many of these elements are demonstrated in the chart in Figurethree.

The app demonstrates a novel arroyo to displaying high‐density stream networks, using stream order to dynamically load streams based on the zoom level, like to how the pyramiding technique is applied to high‐resolution images. When the zoom level is set to the full extent of the watershed, only reaches with higher stream orders are displayed as illustrated past Effigy4a. A mid‐range zoom level volition upshot in reaches with mid‐range stream orders beingness added to the display as shown in Figure4b. On the last zoom level, reaches of all stream orders are displayed as shown in Figureivc.

Stream Network Zoom Levels with Dynamic Stream Densities: (a) High Stream Order, (b) Mid‐Level Stream Club, (c) Complete Stream Network.

The app also presents a new method for displaying high‐resolution warnings at both an overview and at the level of each individual river achieve. Information technology facilitates display at the NFIE Region scale by combining the points within a close proximity together and representing the group as a unmarried icon with a count of the number of points included. Additionally, it divides the alert points into 3 main groups corresponding to the peak flow of the boilerplate (represented by larger triangles) or the standard deviation above the average (represented by smaller triangles) of the ensemble forecast that exceeded the return catamenia. The warning with the highest return menses is the but i generated for the reach if any warnings be. Xanthous triangles represent exceedance of the 2‐year menses, ruddy triangles correspond exceedance of the 10‐year flow, and royal triangles stand for exceedance of the 20‐year flow. Examples are shown for the warnings generated for each return catamenia during the May 2015 flooding in Texas in Figurev.

Warning Points for Texas Floods for May 2015: (a) 2‐Year Warnings, (b) x‐Year Warnings, (c) twenty‐Twelvemonth Warnings.

Validation

Setup

We used the watersheds from the NFIE to validate the performance of the downscaling and routing computation framework and the Streamflow Prediction Tool app. We performed timing tests using a local compute cluster every bit well as using the Amazon Spider web Service computing cloud. A map of all of the NFIE regions is shown in Figure6.

National Flood Interoperability Experiment (NFIE) Regional Divisions.

Computational Forecast Framework

Nosotros compared the computational functioning of computing using a local compute cluster and the Amazon Spider web Services (aws.amazon.com). Additionally, we estimated the results computing serially based off of the individual computation times as a means of measuring the time saved using distributed calculating.

It is apparent from Table1 that the employ of the distributed computing significantly reduces computation time in all cases. In fact, distributed computing is essential for meeting the 12‐h (43,200 due south) constraint for an operational arrangement that covers the entire U.Southward.

Table 1

Results of Computation Fourth dimension Based on Surface area and the Number of Reaches

Watershed Name	Surface area (sq km)	Number of Reaches	Compute Time (seconds)
			Serial	Amazon Spider web Services	Local Cluster
NFIE Souris‐Red‐Rainey Region	213,488	29,053	7,343	49	141
NFIE Rio Grande Region	564,840	55,854	9,083	64	175
NFIE New England Region	169,445	65,858	10,906	63	210
NFIE Texas‐Gulf Region	464,493	66,373	10,417	68	200
NFIE Great Basin Region	367,058	96,269	8,589	43	165
NFIE Great Lakes Region	324,434	104,645	15,873	115	305
NFIE Mid‐Atlantic Region	277,755	125,398	16,900	109	325
NFIE California Region	421,995	140,759	22,367	164	430
NFIE Colorado Region	660,454	187,010	28,105	210	540
NFIE Pacific Northwest Region	814,493	231,806	24,325	180	468
NFIE Southward Atlantic‐Gulf Region	675,734	360,175	41,083	319	790
NFIE Mississippi Region	3,302,913	1,242,008	316,930	1,558	half dozen,095

The computation time vs. the number of reaches from Tableone is shown in Effigy7 with polynomial lodge 2 trend lines having coefficients of determination of 0.99. As expected, the computing fourth dimension increases with the size of the computing problem (i.eastward., the number of river reaches). Note that the good fit with a second order polynomial order suggests that the solving procedure might include two inner loops. This is likely due to the default RAPID option for solving linear systems (an iterative Richardson method) being used in this report. Culling approaches using noniterative solvers (David et al., 2015) could assist in decreasing computing time for this application.

Computation Time vs. Number of Reaches.

This is particularly true of the local cluster, where the rate of computation of the Mississippi region is half that of the Colorado region. The slowdown is non likely caused by the processing algorithms, considering the Amazon cloud and local compute cluster curves practise non slow at the aforementioned charge per unit. In that location are various factors at play, only the slowdown may exist caused by differences in hardware or perhaps overhead of the computing environments. The efficiency of the system running on the local cluster could be improved significantly by dividing the Mississippi watershed into two or three watersheds with 400,000‐600,000 reaches each.

Streamflow Prediction Tool Web App

The GIS visualization capabilities of the Streamflow Prediction Tool app successfully displayed all of the NFIE regions. The loading fourth dimension is relatively fast, averaging around 7 due south to load. Displaying large high‐density stream networks is currently express to stream networks divers past the NHDPlus V2 dataset equally it has the stream social club defined, making dynamic display of stream networks possible.

Potential NFIE Comparison

From May 24 to May 28 of 2015 major flooding occurred in the Austin, Texas region. In this section, we will demonstrate how participants at NFIE can compare the forecasted results from ECMWF‐RAPID with USGS stations using an example at the reach with COMID 5781369 in Onion Creek (coordinates 30.175347, −97.656148) and USGS station 08159000.

Four days before the offset of the flood event, the ECMWF‐RAPID forecast shows that there is an event that will occur equally shown in Figure8. However, the predicted boilerplate of the magnitude and timing of the event is significantly off. In addition, the two events seem to be merged into ane outcome. Nevertheless, the loftier‐resolution ensemble seems to capture the timing of second event, though the peak flow is overestimated. Due to the under prediction of the magnitude of both storms, but a minor 2‐year warning was developed every bit shown by the small yellow triangle in the center of the stream in Figureix.

Forecast Comparison Beginning May 20, 2015 (four days earlier flood).

Onion Creek Alarm Signal Produced, May twenty, 2015.

Then, at two days out, the ECMWF‐RAPID forecast hateful is starting time to show two distinct events as shown in Figure10. However, while the timing and magnitude of the events are closer, they are still predicting alluvion peaks lagging behind what really happened with peaks captured only in the upper extremes of the prediction. Like to the forecast from twenty May, only a small two‐year alert was adult as shown by the small yellow triangle in the heart of the stream in Figure11.

Forecast Comparison Beginning May 22, 2015 (ii days before flood).

Onion Creek Warning Signal Produced, May 22, 2015.

On the day of the offset of the overflowing event, the forecast more closely aligns with the timing and magnitude of the event. Notwithstanding, for both peaks, the mean predicted streamflow is significantly below the actual streamflow as shown in Figure12. Because the predicted series that was a standard deviation above the average had a peak flow above the 20‐year threshold, a small purple warning point was generated. Too, because the acme menstruation of the average series was higher up the 2‐twelvemonth threshold a large yellow triangle was too generated. Both triangles are shown in the center of the stream in Figure13.

Forecast Comparison Beginning May 24, 2015 (day of overflowing).

Onion Creek Alarm Bespeak Produced, May 24, 2015.

As expected, the case study illustrates that the predictions go more authentic, the closer to the time of the flooding event. It is clear that the system needs further evaluation and comeback to accurately capture events over the broad stream network. Even so, the framework does provide participants at NFIE and others with the ability to begin providing a more widespread evaluation of the value of this downscaled, high‐density hydrologic forecast.

Current limitations and potential improvements

Using the method and tools developed by this inquiry and the evaluations that can be performed through NFIE, the creation of additional tools and improvements to the electric current system is needed. The near pressing need is improving the streamflow initialization. The current method begins at zero flow for the very offset prediction and then initializes the adjacent prediction from the average of previously predicted flows. This method requires multiple forecast runs in social club for the model to "spin‐up" the streamflow in the rivers. Additionally, initializing from predicted streamflows may not render the best results for what the bodily streamflow volition exist. As such, data assimilation methods will demand to be incorporated into the model to provide better estimates for initial streamflow at the beginning of each prediction wheel. This tin can include incorporating real‐fourth dimension stream gage data or running the model with post‐processed reanalysis meteorological information to amend the streamflow predictions.

Additional known improvements include items such as methods for calibrating the model as well as a method for including reservoir releases in the modeling process. New tools could be created based on the loftier‐resolution stream forecasts. These new tools could involve ideas such as predictive overflowing alphabetize maps derived from the instream forecasts or improved analysis methods to determine how probable a overflowing will occur or when and how to warn the public of oncoming floods.

Conclusions

The creation of a overflowing alarm system that tin provide predictive information for floods days and even weeks in advance at a loftier spatial resolution at a national‐scale is within reach. The implementation developed by this research provides an of import contribution to the national aspirations to create such a system. By downscaling runoff forecasts generated by the ECMWF using Esri'due south RAPID toolbox and routing the runoff using the RAPID model, we were able to produce high‐density ensemble national‐scale stream forecasts. Even so, for this system to be fully functional, improvements in initialization and calibration as well equally the addition of reservoir operations need to be incorporated.

With Tethys Platform, we developed an interface to brandish the high‐density streamflow forecasts to determination‐makers that includes the power to compare to existing NWS forecasts and observations at USGS gages where these data streams be. This tool gives conclusion‐makers information from NetCDF datasets containing streamflow forecasts for hundreds of thousands of reaches, including a statistical summary of the potential streamflow up to 15 days in advance. Additionally, the tool displays warning points where the 20‐yr, 10‐year, and ii‐yr thresholds accept been exceeded in the predictions for each individual accomplish. These warnings make the catchment level predictions applicable at a national‐scale. As such, the app simplifies the data access and estimation for decision‐makers.

Acknowledgments

This inquiry is based upon work supported by the National Science Foundation under Grant No. 1135483. Cédric H. David was supported by the Jet Propulsion Laboratory, California Institute of Engineering, nether a contract with the National Aeronautics and Infinite Administration. We would also similar to express appreciation to Curtis Rae for his help in the GIS preprocessing of the NFIE regions.

Notes

Snow, Alan D. , Christensen Scott D., Swain Nathan R., James Nelson E., Ames Daniel P., Jones Norman L., Ding Deng, Noman Nawajish Southward., David Cédric H., Pappenberger Florian, and Zsoter Ervin, 2016. A High‐Resolution National‐Scale Hydrologic Forecast System from a Global Ensemble Land Surface Model. Journal of the American Water Resource Association (JAWRA) 52(4):950–964, DOI: 10.1111/1752-1688.12434 [CrossRef] [Google Scholar]

^†Newspaper No. JAWRA‐15‐0080‐P of the Journal of the American Water Resources Association (JAWRA).

Discussions are open up until six months from outcome publication.

Literature Cited

• Alfieri, 50. , Burek P., Dutra E., Krzeminski B., Muraro D., Thielen J., and Pappenberger F., 2013. GloFAS–Global Ensemble Streamflow Forecasting and Flood Early Warning. Hydrology and World System Sciences 17(three):1161–1175. [Google Scholar]
• Balsamo, G. , Beljaars A., Scipal G., Viterbo P., van den Hurk B., Hirschi 1000., and Betts A.Thou., 2009. A Revised Hydrology for the ECMWF Model: Verification from Field Site to Terrestrial Water Storage and Bear on in the Integrated Forecast System. Journal of Hydrometeorology 10(3):623–643. [Google Scholar]
• Benson, M.A. , 1962. Plotting Positions and Economics of Engineering Planning. Journal of the Hydraulics Sectionalization 88(6):57–71. [Google Scholar]
• Beven, Grand. , Cloke H., Pappenberger F., Lamb R., and Hunter N., 2015. Hyperresolution Information and Hyperresolution Ignorance in Modelling the Hydrology of the Land Surface. Science China Earth Sciences 58(one):25–35. [Google Scholar]
• Beven, Yard.J. and Cloke H.L., 2012. Annotate on "Hyperresolution Global State Surface Modeling: Meeting a Grand Challenge for Monitoring Earth's Terrestrial Water" by Eric F. Wood et al.Water Resources Enquiry 48(one), DOI: 10.1029/2011WR010982. [CrossRef] [Google Scholar]
• Buizza, R. , Houtekamer P., Pellerin 1000., Toth Z., Zhu Y., and Wei Chiliad., 2005. A Comparison of the ECMWF, MSC, and NCEP Global Ensemble Prediction Systems. Monthly Weather Review 133(5):1076–1097. [Google Scholar]
• Buyya, R. , Vecchiola C., and Selvi S.T., 2013. Mastering Cloud Computing: Foundations and Applications Programming. Morgan Kaufmann, Waltham, Massachusetts. [Google Scholar]
• CUAHSI , 2015. "National Alluvion Interoperability Experiment." The Consortium of Universities for the Advancement of Hydrologic Scientific discipline, Inc. (CUAHSI). https://www.cuahsi.org/NFIE, accessed Feb, 2015.
• Cunge, J.A. , 1969. On the Field of study of a Flood Propagation Computation Method (Musklngum Method). Journal of Hydraulic Research vii(2):205–230. [Google Scholar]
• David, C.H. , 2013. RAPID v1.iv.0, Zenodo, DOI: 10.5281/zenodo.24756. [CrossRef]
• David, C.H. , Famiglietti J.S., Yang Z.‐L., and Eijkhout V., 2015. Enhanced Fixed‐Size Parallel Speedup with the Muskingum Method Using a Trans‐Boundary Arroyo and a Large Sub‐Basins Approximation. Water Resources Research 51:1–25, DOI: ten.1002/2014WR016650. [CrossRef] [Google Scholar]
• David, C.H. , Maidment D.R., Niu 1000.‐Y., Yang Z.‐L., Habets F., and Eijkhout V., 2011. River Network Routing on the NHDPlus Dataset. Journal of Hydrometeorology 12(v):913–934. [Google Scholar]
• Dee, D.P. , Uppala Due south.M., Simmons A.J., Berrisford P., Poli P., Kobayashi S., Andrae U., Balmaseda One thousand.A., Balsamo G., Bauer P., Bechtold P., Beljaars A.C.M., van de Berg 50., Bidlot J., Bormann N., Delsol C., Dragani R., Fuentes M., Geer A.J., Haimberger L., Healy South.B., Hersbach H., Hólm E.V., Isaksen L., Kållberg P., Köhler G., Matricardi Grand., McNally A.P., Monge‐Sanz B.Chiliad., Morcrette J.‐J., Park B.‐K., Peubey C., de Rosnay P., Tavolato C., Thépaut J.‐North., and Vitart F., 2011. The ERA‐Interim Reanalysis: Configuration and Operation of the Data Assimilation System. Q.J.R. Meteorological Society 137:553–597, DOI: 10.1002/qj.828. [CrossRef] [Google Scholar]
• Demargne, J. , Wu L., Regonda S.G., Brown J.D., Lee H., He M., Seo D., Hartman R., Herr H.D., Fresch M., Schaake J., and Zhu Y., 2014. The Scientific discipline of NOAA's Operational Hydrologic Ensemble Forecast Service. Bulletin of the American Meteorological Society 95:79–98, DOI: 10.1175/BAMS-D-12-00081.i. [CrossRef] [Google Scholar]
• ECMWF (European Heart for Medium Range Weather Forecasts) , 2011. "IFS Documentation – Cy37r2 Operational Implementation 18 May 2011 ‐ Part Two: Data Absorption.
• Guha‐Sapir, D. , Beneath R., and Hoyois P., 2015. EM‐DAT: International Disaster Database. Université Catholique de Louvain, Brussels, Belgium.
• Horizons Systems Corporation , 2011. Weaving the National Hydrologic Geospatial Fabric. NHDPlus Version 2. (Apr 6, 2015).
• Jones, N. , Nelson J., Swain Due north., Christensen S., Tarboton D., and Nuance P., 2014. Tethys: A Software Framework for Spider web‐Based Modeling and Conclusion Support Applications In: Ames D.P., Quinn N.W.T., and Rizzoli A.E. (Editors), Proceedings of the 7th International Congress on Ecology Modelling and Software, June xv‐19, San Diego, California, ISBN: 978‐88‐9035‐744‐ii. [Google Scholar]
• Maidment, D. , 2015. A Conceptual Framework for the National Alluvion Interoperability Experiment. https://www.cuahsi.org/Files/Pages/documents/13623/nfieconceptualframework_revised_feb_9.pdf, accessed March 2015.
• Mcenery, J. , Ingram J., Duan Q., Adams T., and Anderson L., 2005. NOAA'southward Avant-garde Hydrologic Prediction Service: Building Pathways for Better Scientific discipline in H2o Forecasting. Bulletin of the American Meteorological Society 86:375–385, DOI: 10.1175/BAMS-86-3-375. [CrossRef] [Google Scholar]
• Michaelis, C. and Ames D., 2012. Considerations for Implementing OGC WMS and WFS Specifications in a Desktop GIS. Journal of Geographic Information System 4(2):161–167, DOI: x.4236/jgis.2012.42021. [CrossRef] [Google Scholar]
• NOAA , 2015. Advanced Hydrologic Prediction Service. National Oceanic and Atmospheric Assistants'south (NOAA) National Atmospheric condition Service (NWS). http://water.weather.gov/ahps/most/about.php, accessed February, 2015.
• Open Geospatial Consortium , 2015a. Web Map Service. http://www.opengeospatial.org/standards/wms, accessed June 2015.
• Open up Geospatial Consortium , 2015b. Web Feature Service. http://world wide web.opengeospatial.org/standards/wfs, accessed June 2015.
• Open Knowledge Foundation , 2013. DataStore Extension. CKAN v2.2. http://docs.ckan.org/en/ckan-two.2/datastore.html, accessed June 2015.
• Pappenberger, F. and Beven K.J., 2006. Ignorance Is Bliss: Or 7 Reasons Not to Use Doubt Analysis. Water Resources Research 42, W05302, DOI: x.1029/2005WR004820. [CrossRef] [Google Scholar]
• Pappenberger, F. , Cloke H.Fifty., Parker D.J., Wetterhall F., Richardson D.S., and Thielen J., 2015. The Monetary Benefit of Early Flood Warnings in Europe. Environmental Science & Policy 51:278–291, DOI: 10.1016/j.envsci.2015.04.016. [CrossRef] [Google Scholar]
• Pappenberger, F. , Stephens E., Thielen J., Salamon P., Demeritt D., Andel Southward.J., Wetterhall F., and Alfieri L., 2013. Visualizing Probabilistic Alluvion Forecast Information: Expert Preferences and Perceptions of Best Exercise in Uncertainty Communication. Hydrological Processes 27(1):132–146. [Google Scholar]
• Salas, F. , Somos G., Gochis D., Maidment D., David C., Tolle K., Navarro C., and Corby R., 2014. Edifice Cyberinfrastructure to Support a Real‐Time National Flood Model In: AGU Fall Coming together Abstracts, San Francisco, December 15–19, Vol. 1, p. 1207 https://agu.confex.com/agu/fm14/webprogram/Paper25860.html, accessed May2016; https://www.cuahsi.org/Files/Pages/documents/13623/poster_salasetal_buildingcyberinf.pdf, accessed May 2016.. [Google Scholar]
• Slingo, J. and Palmer T., 2011. Uncertainty in Weather condition and Climate Prediction. Philosophical Transactions of the Royal Guild A: Mathematical, Physical and Engineering science Sciences 369(1956):4751–4767. [PMC gratis article] [PubMed] [Google Scholar]
• Swain, N.R. , 2015. Tethys Platform: A Development and Hosting Platform for H2o Resources Web Apps . All Theses and Dissertations, Paper 5832. http://scholarsarchive.byu.edu/etd/5832, accessed May 2016.

How To Add Nws River Gauges Web Services To Openlayers,

Source: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6686479/

Posted by: tuckerforepland.blogspot.com

Share this post