Home

Final Report for KINEROS/SM-hsB COMET Cooperative Project

1. S Daily streamflows can be download from http waterdata usgs gov nwis rt Select the desired streamflow site and select Timeseries Daily Data for 24 hour streamflow data and USGS Instantaneous data archive Offsite for 15 minute streamflow data 29 E Acquiring Hydrometeorologic Data for Calibration of KINEROS SM hsB Acquisition of the Hydrometeorologic Data for past simulations of KINEROS SM hsB can be a time consuming process because the hydrometeorologic datasets are very large The model uses hourly NLDAS data daily MPE data which is downscaled to hourly data using the NLDAS data and it compares to daily SNODAS snapshots All of the data are national and so the model can be applied consistently anywhere in the U S but the downside is that the entire national grid must be downloaded since there is no way to download a subset of the data so the data files are very large and extraction is slow To speed up the extraction process if the data is to be extracted multiple times it can be optionally be saved in a faster binary file format see the KINEROS SM hsB documentation section 2 4 The downloaded data is then extracted by the GetData program DHR event data is also used but the process for acquiring this data is straight forward 1 Download the NLDAS MPE and SNODAS data from ftp servers This can be a time intensive process and the datasets are very large The data can reside anywhere includi
2. and P A Troch 2005 Improved understanding of soil moisture variability dynamics Geophys Res Letters 32 L05404 doi 10 1029 2004GL021935 Troch P A J A Smith E F Wood F P de Troch 1994 Hydrologic controls of large floods ina small basin central Appalachian case study Journal of Hydrology 156 285 309 Veatch W P Brooks J Gustafson N Molotch Quantifying the Effects of Forest Canopy Cover on Net Snow Accumulation at a Continental Mid Latitude Site Ecohydrology in review Whiteman C D J Allwine Extraterrestrial Solar Radiation on Inclined Surfaces Environmental Software 1986 1 3 Woolheiser D A R E Smith and D C Goodrich 1990 KINEROS A Kinematic Erosion and Runoff Model Documentation and User Manual U S Department of Agriculture Agricultural Research Service ARS 77 130 pp 25 Appendix B MODEL SET UP GUIDE FOR KINEROS SM hsB Patrick Broxton Peter Troch Mike Schaffner Carl Unkrich David Goodrich The University of Arizona Tucson AZ US National Weather Service Salt Lake City UT US 3USDA ARS Southwest Watershed Research Center Tucson AZ US The following is meant to be a simple guide for setting up KINEROS SM hsB and getting it to the point where it can be run in real time This document does not detail steps required to do all tasks e g for GIS processing though it should be understandable to a trained hydrometeorologist A 1 2
3. update_pars date2doy ge_groundoverlay ge_output mapshow_img Image Processing Toolbox Setup_CalibMainModel define_pars update_pars ge_groundoverlay ge_output mapshow_img OptDSMMainCalib Image Processing Toolbox RunLSM RunSMM Setup_CalibMainModel Setup_CalibKineros Setup_CalibSimpleRunoff RTM_RunModel Table 4 Functions related to the processing of spatial data located in the lt Model Root gt Program Files HSB_Kineros Spatial directory Name Functions Called From border_nans upslope_area Function Calls To Image Processing Toolbox 39 pixel_flow Finds NaNs connected to the DEM border calculate_flow extract_hillslopes Determines flow properties for a dem d8 process_dem Convert dinf flow directions to d8 flow directions dem_flow process_dem Downslope flow direction for a DEM downnbr process_hillslopes Creates map of 1D indexes of the D8 downstream neighbors flow_matrix upslope_area Image Processing Toolbox pixel_flow Image Processing Toolbox extract_hillslopes load_spatial_data Setup_Aquifers Gets hillslope information about Kineros planes facet_flow pixel_flow Facet flow direction fill_sinks process_dem Fills Interior sinks ina DEM flowdistance extract_hillslopes process_hillslopes Computes hillslope flow distance flow_matrix calculate_flow System of linear equations representing pixel flow Setup_CalibMainModel Setup_CalibKineros load_spatia
4. ProgramPars txt must refer to the location where NDFD data is downloaded At this point the model is set up to ingest real time Stage IIl radar data using a supplied rain rates program which works with radar data within the National Weather Service forecast offices but because all model development has taken place outside of the National Weather Service it has not been tested with KINEROS SM hsB Future testing and possible modification to the model code must be performed prior to using real time radar data SNODAS data is downloaded from the internet for real time modeling much in the same way that it is downloaded for past simulations except the process is automatic Each day one snapshot will be downloaded to be compared to the modeled value from KINEROS SM hsB 20 If the user selects No for the Load Realtime SNODAS Data field in ProgramPars txt this check will be disabled and SNODAS data will not be downloaded Finally streamflow data is downloaded automatically from the USGS real time waterdata site see section 2 3 1 The web URL for USGS streamflow data for each watershed must be specified in the Data URL field of Watersheds txt 2 6 Running the model for Calibration Evaluation Setting up the model processing the data calibrating and validating the model is a computer assisted process as some data files need to be downloaded and pre processed manually but a majority of the data hand
5. Troch D Goodrich H Gupta Modeling Flood Response in Fast Responding Catchments over Diverse Terrain and Climatic Regimes of the Desert Southwest and Northeast oral presentation presented by Michael Schaffner at the National Weather Association Meeting Tucson AZ October 4 M Schaffner P Broxton P A Troch D Goodrich H Gupta Modeling Flood Response in Fast Responding Catchments over Diverse Terrain and Climatic Regimes of the Desert Southwest and Northeast Modified oral presentation presented by Michael Schaffner at the National Flood Workshop Huston TX October 26 2010 M Schaffner Taught a lesson on distributed modeling at the COMET COMAP Virtual Symposium on QPF Rapid Onset Floods Course was held remotely the week of November 15h Talk title was Improving Flash Flood Prediction in Application of Distributed Models at a Weather Forecast Office Talk lasted for 30 minutes of instruction time Audience not only included WFOs in the USA but also international participation from the Canadian weather service Talk focused on KINEROS model work P Broxton P A Troch M Schaffner C Unkrich D Goodrich Improved Flash Flood Predictions Using KINEROS SM hsB article in August 2011 National Hydrologic Warning Council NHWC Transmission P Broxton P A Troch M Schaffner C Unkrich D Goodrich Development of a Distributed All Season Flash Flood Forecasting System oral pres
6. 3 4 5 6 7 Setting up the Directory Structure Create a base model directory somewhere on your hard disk such as My Documents KINEROS_SM_hsB If running KINEROS SM hsB from within Matlab you will also need the function files to be located in the sub directory Program Files HSB_KINEROS Create a subfolder with an identifier for the modeled area The directory structure should look something like My Documents KINEROS_SM_hsB Delaware Basin this directory will hereafter be referred to as SMODELHOME Next create subdirectories in SMODELHOME called Prior Data and Parameter Files yt In the Prior Data directory create subdirectories called Forcing Spatial and Streamflow The Forcing and Streamflow directories contain hydrometeorologic and streamflow data used for past simulations The Spatial directory contains spatial files required by KINEROS SM hsB If using KINEROS or testing SM hsB with past events then create a RADAR subdirectory in the Forcing directory In the Streamflow directory create subdirectories called 15 minute and 24 hour Put the required model executables in the SMODELHOME directory correct for the architecture operating system that these are to be run on These should be the 26 GetData CalibrateModel RunContinuousRTModel and RunEventRTModel programs If KINE
7. also provided so that the model can be modified Of the folders in the model directory only the folder called Parameter Files and the one entitled Prior Data need to have preprocessed files in them The necessary files in the Parameter Files directory are described in section 2 2 and the necessary files in the Prior Data directory are described in sections 2 3 and 2 4 1 In addition MPE NLDAS and SNODAS forcing files for past model simulations need to be placed in another local or remote directory as described in sections 2 3 2 and 2 3 3 and the Real Time NDFD forcing files must be downloaded with an automated script running every hour see section 2 4 Figure 6 shows a schematic of the model directory structure 2 2 Parameter Files and Model Options It is important to configure the model properly when setting it up These configuration files reside in the Parameter Files directory In this folder there must be four required files that relate to the general functioning of the model The required files are called CalibDates txt EventDates txt ProgramPars txt and Watersheds txt CalibDates txt contains a listing of dates that are used for the calibration of SM hsB In this file there are three sets of dates The first set indicates the periods that the snow model is to run during calibration so that it is not run needlessly during the summer when there is no snow The second set
8. conform to certain standards when using them with SM hsB and some of the maps need to be in Geographic Coordinates units of degrees and some of the maps need to be in a projected Coordinate System such as UTM where the units are in meters The third step is to acquire hydrometeorologic data for calibration Many forcing variables for the continuous model are from NLDAS MPE can also be used as a precipitation source In addition Stage Ill radar data can be downloaded e g using the NWS s Weather and Climate Toolkit for testing the model at the highest resolution during past flood or non flood events The fourth step is to perform model calibration of both the SM hsB components and of KINEROS This process can be done by either manually adjusting model parameters by using built in scripts to perform automatic calibration or by using a hybrid approach that involves both methods Finally the modeler needs to set up and run the model in real time For real time operation both the model and an auxiliary script to download NDFD data both need to be automated to run every hour 1 2 Description of SM hsB Components SM hsB is a full featured hydrological model that includes an energy balance snow model It describes radiation over complex terrain potential evaporation using the Penman Monteith equation canopy storage of rain and snow snow accumulation and melt and root zone processes During execution each module proceeds in sequence du
9. current usage and planned improvements oral presentation presented by Mike Schaffner at the Mid Atlantic Regional Forecast Center State College PA September 25 2009 e M Schaffner C Unkrich poster presentation A Distributed Rainfall Runoff Model Applied to Flash Flood Forecasting at National Weather Service Binghamton NY presented by Mike Schaffner at the National Weather Association Annual Meeting Norfolk Virginia October 19 2009 e P Broxton P A Troch M Schaffner C Unkrich D Goodrich H Gupta T Wagener S Yatheendradas Improving Flash Flood Prediction in Multiple Environments poster presentation presented by Patrick Broxton at the American Geophysical Union Fall Meeting San Francisco December 15 2009 e M Schaffner Real Time Flash Flood Forecasting in Small Fast Responding Watersheds Using a Distributed Rainfall Runoff Model oral presentation presented by Michael Schaffner at the NWS Eastern Region Flash Flood Conference Wilkes Barre PA June 3 2010 e P Broxton Using a Continuous Hydrologic Model in Support of Flash Flood Predictions oral presentation presented by Patrick Broxton at the NWS Eastern Region Flash Flood Conference Wilkes Barre PA June 3 2010 e M Schaffner KINEROS oral presentation presented by Michael Schaffner at the NWS Eastern Region Flash Flood Conference Wilkes Barre PA June 4 2010 M Schaffner P Broxton P A
10. hour streamflow the file should contain five columns the first four columns represent the dates of the streamflows excel timestamp year month day and the fifth column corresponds to the streamflows themselves in CFS and each row represents a daily streamflow value For the 15 minute data there are six columns and each row corresponds to an instantaneous streamflow measurement The first five columns contain date information year month day hour minute and the sixth column contains the streamflow measurement in CFS For the daily data all days should be included and missing data should be 999 For 15 minute data the data that are missing are just omitted like when they are downloaded from the USGS site 2 4 2 Hydrometeorologic Data For past simulations forcing data come from NLDAS and MPE SM hsB only and from NLDAS MPE and Stage III radar data KINEROS SM hsB NLDAS data must be obtained for the entire calibration and validation periods specified in CalibDates txt Unfortunately there is not currently a way to download NLDAS data for a specific region from the web so the user must download data for the entire U S and use a supplied program built into KINEROS SM hsB to extract the NLDAS data and to put it in a format that the model expects NLDAS data are downloaded as Gridded Binary GRIB files that contain hourly NLDAS forcing data These files can be obtained from hydro1 sci gsfc nasa gov The model
11. latlon_all txt in the SMODELHOME Parameter Files directory Set up the main model SMODELHOME RunContinuousRTModel xxx to be updated hourly 6 16 minutes after the hour If the program is already compiled and all the model files are present then it can be automated to run using the task manager Windows or using Cron Unix It is suggested that a batch file Windows or a Shell Script Unix is used to make sure that model execution occurs in the correct directory This script can also be used to move model generated output files to a public directory so they can be displayed using a web page If the RunContinuousRTModel model script needs to be compiled for example the model code was modified it can be done so using Matlab s mcc command This compiled version should usually work on the computer on which it was compiled However if it is to run on a computer that it was not compiled on care has to be given to the fact that the architectures must match e g you need to use a 32 Matlab version to get a 32 bit executable It must also be distributed with the Matlab MCR libraries which are freely distributable and can be installed using a simple installer However the version of the MCR libraries and the version of Matlab that created the executable must match exactly e g a program made with Matlab 7 7 will only run with version 7 7 of the MCR libraries If running the model on a different computer
12. meaning and their expected values in ProgramPars txt Parameter Name Description Calibration Task For Snow Model Rain Snow Temperature C The typical maximum temperature that accumulating snows Snow occur Heat Transfer Coeff The efficiency that energy is transferred to the snowpack by Snow sensible heat Rain Freezing Efficiency The efficiency that energy is transferred to the snowpack by Snow falling rain For Subsurface Model Vegetation Height m The typical vegetation height as related to aerodynamic Potential ET roughness Vegetation Resistance Overall resistance of the vegetation to evaporation of water Potential ET CDensity Maximum Maximum canopy density relative to what is depicted on Manual the canopy map CDensity Minimum Minimum canopy density relative to what is depicted on the Manual canopy map Temp CDensity Maximum C Weekly average temperature at which canopy density is ata Manual maximum Temp CDensity Minimum C Weekly average temperature at which canopy density is ata Manual minimum Wilting Point Wilting point for vegetation Manual Critical Moisture Content Critical moisture content for vegetation Manual Root Zone Depth m Typical depth at which there not many roots Actual ET K_inf cm day Hydraulic Conductivity for Infiltration Infiltration MaxRes mm Maxim size of the subsurface reservoir
13. of presentations and other outreach educational materials Carl Unkrich Programming related to KINEROS preparation of presentations and other outreach educational materials Appendix A MODEL DOCUMENTATION FOR KINEROS SM hsB Patrick Broxton Peter Troch Mike Schaffner Carl Unkrich David Goodrich The University of Arizona Tucson AZ US National Weather Service Salt Lake City UT US 3USDA ARS Southwest Watershed Research Center Tucson AZ US 1 MODEL DISCRIPTION In the United States flash floods kill more people annually than any other type of natural disaster Modeling such floods is the primary motivation for coupling the Soil Moisture Hillslope Storage Boussinesq SM hsB model a continuous hydrologic model capable of keeping track of distributed soil moisture and snow states and the Kinematic Erosion and Runoff KINEROS model an overland flow and channel routing model which is designed to predict flash floods in the semiarid Southwest US KINEROS is an effective flash flood model that is designed for small urban and agricultural watersheds but it is event based and does not keep track of antecedent catchment wetness and does not model snow The combination of SM hsB and KINEROS overcomes these deficiencies by utilizing SM hsB s energy balance snow model and continuous soil moisture baseflow model The model has a variable spatial resolution limited by computational resources and programmable temporal reso
14. tells the model which time period to calibrate the subsurface model for 14 The third set tells the model which time period to evaluate the subsurface model for It is required that all dates be in a standard date format such as 01 03 2000 2000 03 01 or 01 Mar 2000 EventDates txt contains a list of dates that correspond to events that KINEROS is to be calibrated to The first column lists the starting date for each event and the second column lists the duration that the model is to be run in days A separate listing of dates must appear for each modeled watershed ProgramPars txt contains a majority of the programmatic parameters that control the functioning of KINEROS SM hsB Table 1 lists the parameters in ProgramPars txt their expected values and their meaning Parameter Name Expected Description Value West Number Western extent degrees of modeled area East Number Eastern extent degrees of modeled area North Number Northern extent degrees of modeled area South Number Southern extent degrees of modeled area TZCorrection Number Local time offset from UTC for local standard time Load Realtime SNODAS Data Yes No Load SNODAS data during real time model operation for data assimilation or comparison SM HSB Only Yes No Yes Use the simple routing model for real time operation No Use KINEROS for routing during real time simulation Model Resolution Numbe
15. 0 F G 1 2 Configuring KINEROS SM hsB parameter files The last manual step prior to running the model is to configure the KINEROS SM hsB parameter files In the SMODELHOME Parameter Files directory there must be required files for the model setup For a detailed description of the function and contents of each of these parameter files see the KINEROS SM hsB documentation section 2 4 Model Calibration General steps are outlined below but for more information see the KINEROS SM hsB documentation section 2 6 Processing the spatial hydrometeorologic and streamflow data is a fully automatic process though calibration can be fully automated partially automated or manual The following can be performed within Matlab or by using a compiled version of the code The advantage of doing this portion in Matlab is that it is faster and easier to diagnose a problem and perhaps fix a portion of the code if something goes wrong but the advantage of doing it with the compiled code is that the modeler does not need to have Matlab installed only the MCR libraries described below Run the GetData program or Matlab Script Perform each of the steps 1 Load NLDAS data 2 load SNODAS data 3 load MPE data 4 Load daily streamflow data 5 load high resolution streamflow data and 6 prepare forcing data for simulations during the calibration and validation periods in sequence These will not require any user i
16. 23 moisture state and running the model for multiple precipitation scenarios It is also set up to ingest the most recent Stage III radar rainfall estimates but this point this capability is untested and may require revisions to the code see section 2 5 If the computer that the model was set up on and the computer that it is run on in real time are different it is important to copy the entire directory or if desired follow the instructions at the end of the model set up guide References Brutsaert W 1973 Review of Green s functions for linear open channels J Eng Mech Div ASCE 99 EM12 pp 1247 1257 Brutsaert W 2005 Hydrology An Introduction Cambridge Cambridge University Press Campbell G 1974 A simple method for determining unsaturated conductivity from moisture retention data Soil Sci 117 6 311 314 Carrillo G P A Troch M Sivapalan T Wagener C Harman K Sawicz 2011 Catchment classification hydrological analysis of catchment behavior through process based modeling along a climate gradient Hydrol Earth Syst Sci 15 1 20 Deardoff J W 1978 Efficient prediction of groundwater surface temperature and moisture with inclusion of a layer of vegetation J Geophys Res 83 1889 1903 Famiglietti J S E F Wood M Sivapalan and D J Thongs 1992 A catchment scale water balance model for FIFE J Geophys Res 57 D17 pp 18997 19007 Famiglietti J S and E F Wo
17. D NAME gt _hsid_geo tif and lt WATERSHED NAME gt _hsid_proj tif Finally if radar data is to be used there must be a final file called Radar_geo tif 2 4 Model Data for Past Simulations The model system can be run with KINEROS and SM hsB working together Windows Only or with just the SM hsB modules alone Linux Windows This choice will determine how the model is calibrated and run operationally In either case successful calibration requires instantaneous USGS streamflow measurements NLDAS data spatial data and optionally SNODAS data If KINEROS is run alongside hsB then the model should be tested with individual events as well with 5 minute radar data and spatial files associated with KINEROS 18 2 4 1 USGS Streamflow USGS streamflow data are downloaded into the Prior Data Streamflow directory Twenty four hour streamflow data can be downloaded from the USGS s real time streamflow page http waterdata usgs gov nwis rt and archived real time data can be downloaded from the USGS s Instantaneous Data Archive http ida water usgs gov Fifteen minute streamflow measurements should be placed in a sub directory called 15 minute and daily streamflow measurements should be placed in a sub directory called 24 hour The model expects streamflow data to be contained in comma separated csv file format where the file names are the same as the watershed names in Watersheds txt For 24
18. EROS SM hsB model Table 2 Functions that perform integrated tasks located in the lt Model Root gt Program Files HSB_Kineros Program directory Functions Called From Function Calls To Extract_DailyStreamflow GetData GetProgramPars Extracts daily streamflow data from file Extract_HiResStreamflow GetData Image Processing Toolbox 35 Extracts 15 minute streamflow data from file Extract_MPEData GetData gridfit netcdf Extracts MPE data to use with SM hsB Calibration Extract_NLDASData GetData date2doy rsm_extract_record Extracts NLDAS data to use with SM hsB Calibration Extract_SNODASData GetData Extracts SNODAS data to use with SM hsB Calibration GetProgramPars CalibrateModel RTM_DisplayMaps RunEventRTModel RunContinuousRTModel Gets model program parameters RTM_DisplayMaps RunContinuousRTModel createMap GetProgramPars mapshow imgblend Image Processing Toolbox Displays spatially distributed model states for the real time model RTM_DistributeForcing RunContinuousRTModel LoadMPEData_RT LoadNDFDData_RT LoadSNODASData_RT load_spatial_data local_time_to_utc Image Processing Toolbox Statistics Toolbox Prepares hydrometeorologic forcing data for use with real time SM hsB RTM_RunModel RunContinuousRTModel OptDsmMainRT RunEventRTModel ReadRTStreamflow RunLSM get_model_pars get_runoff load_spatial_data preload_forcing_data_LSM 36 Runs SM hsB real time preload_forcing_data
19. Final Report for KINEROS SM hsB COMET Cooperative Project University The University of Arizona Name of University Researchers Preparing Report Patrick Broxton Peter Troch NWS Office Weather Field Office WFO NWS Binghamton NY Weather Field Office WFO NWS Tucson AZ Name of NWS Researcher Preparing Report Mike Schaffner Type of Project Cooperative Project Title NWS Flash Flood Forecasting in Two Hydrologically Distinct Regions Using an Improved Distributed Hydrologic Model UCAR Award No S09 75794 Date December 2011 SECTION 1 PROJECT OBJECTIVES The overall goal of this multiyear project was to develop a more robust site specific flash flood forecasting system that employs an improved distributed hydrological model and assimilation of NWS products including DHR and MPE as well as other gridded datasets to calibrate the model Specifically we coupled a subsurface flow model SM hsB which includes an energy balance snow model with the KINEROS overland flow model to better predict flash floods This combination is beneficial because KINEROS by itself is an event based flash flood model and it needs to be started up with prescribed initial conditions it does not simulate rain on snow events and it does not simulate subsurface flow SM hsB helps to overcome these limitations because the model is continuous and therefore has the ability to keep track of the hydrologic state during inter storm periods it includes an
20. NDFDData_RT RTM_DistributeForcing Loads NDFD data for use with KINEROS SM hsB Real Time LoadNLDASData Setup_DistributeForcing Loads NLDAS data for use with KINEROS SM hsB Calibration xml_read 41 LoadSNODASData Setup_DistributeForcing Image Processing Toolbox Loads SNODAS data for use with KINEROS SM hsB Calibration LoadSNODASData_RT RTM_DistributeForcing Image Processing Toolbox Loads SNODAS data for use with KINEROS SM hsB Real Time ReadRTStreamflow RTM_RunModel Statistics Toolbox Reads Real Time Streamflow Data for KINEROS SM hsB Table 6 Functions related to optimization located in the lt Model Root gt Program Files HSB_Kineros Calib directory Functions Called From Function Calls To OptDsmMainRT RTM_RunModel get_runoff Execute a multi start of the downhill simplex method for SM hsB real_time OptDsmMainHSB Setup_Aquifers Execute a multi start of the downhill simplex method for hsB OptDsmMainCalib Setup_CalibMainModel RunLSM Setup_CalibSimpleRunoff RunSMM Setup_CalibKINEROS get_runoff Execute a multi start of the downhill simplex method for SM hsB calibration Table 7 Miscillaneous Functions located in the lt Model Root gt Program Files HSB_Kineros Misc directory Functions Called From Function Calls To arcascii load_spatial_data process_dem preload_forcing_data_LSM_hiRes Reads an Arc Info ascii raster map createMap mapshow_img hillshading RTM_DisplayMaps truecolors
21. P site is the same as what is expected by the model There can be missing files for the SNODAS files 2 5 Model Data for Real Time Simulations Unlike for calibration model data for real time simulations is downloaded automatically provided that the data paths are set up correctly The model requires MPE data Stage III radar data NDFD data and optionally SNODAS data All real time data is handled by a sub program that downloads extracts and processes all data so that it can be used with the model Paths to the locations of these data are found in the file ProgramPars txt Specifically the field labeled RFC Address must refer to the location on the internet where one can find MPE data from the appropriate River Forecast Center For NDFD data the data acquisition is a little more complicated as the data needs to be downloaded from the internet continuously For this there is an external script currently a UNIX shell script that makes a request to the NDFD Soap Service to download XML files that contain forecast data for particular locations on a grid These XML files are then placed in a folder on a network server It is necessary to request additional data every hour the script that downloads the data should be put on a scheduler whether or not the model is running because if the model is not running it can catch up but if the data is not downloaded it cannot be recovered The field labeled NDFD Address in the file
22. ROS is to be used then also place Kineros2_ AGWA exe into this folder and if the model is to ingest real time radar data then place dhr2rr exe into this folder In addition it is likely that the user will need the wgrib program to decode the NLDAS data Note that this program may need to be compiled Copy the compiled werib program to the SMODELHOME directory B Acquiring Spatial Data 1 Download a Digital Elevation Model a canopy coverage map and an impervious surface map These can be downloaded from sites such as the USGS s National Map Seamless Server or the USDA s Geospatial Data Gateway It is recommended to download a 1 3 arc second DEM and NLCD most recent impervious surface and forest canopy maps It is suggested that all maps be downloaded using a rectangular bounding box as this bounding box will ultimately define the modeled area 2 The canopy cover and impervious surface maps must be in geographic coordinates units of degrees latitude and longitude while there must be copies of the DEM in both geographic and projected coordinates such as UTM because the model grid is in geographic coordinates to conform with hydrometeorologic data but the terrain analysis needs to be performed on the projected DEM A GIS package such as ArcGIS GrassGIS or GDA command line tools can be used to perform these transformations Place the processed spatial files which can either be in ArcInfo ASCII Gri
23. Some of the maps such as slope and aspect also require GIS software to generate Some of the spatial data should be generated with the Automated Geospatial Watershed Assessment Tool AGWA If the modeler wishes to use KINEROS or if he she wants subsurface variability in SM hsB which must be specified by the options SM HSB Only and Distribute Aquifer Properties in the file ProgramPars txt then watersheds must be delineated and discretized into hillslopes using AGWA and KINEROS must be set up as normal to see examples of how this is done visit http www tucson ars ag gov agwa index ph The watershed delineation maps that are generated during the process of setting KINEROS up 17 must be converted to a raster using ArcGis s Shape2Raster tool and exported with the same geographic coordinate system and the same projected coordinate system as the other spatial data files The KINEROS parameter files that are set up during this process are important as well as KINEROS SM hsB will directly use this set up version of KINEROS to handle overland flow channel routing It also uses KINEROS s par file to determine subsurface properties which are important for the subsurface portion of SM hsB The path to the directory where the parameter files are generated by AGWA needs to be specified for each watershed in the file Watersheds txt under the KINEROS Location field If the modeler does not want to use KINEROS and
24. _LSM_hiRes update_pars Image Processing Toolbox Setup_Aquifers CalibrateModel Calibrates hsB responses on individual hillslopes Setup_BaseflowSeparation CalibrateModel Separates runoff and baseflow from a streamflow timeseries Setup_CalibKineros CalibrateModel Calibrates or runs KINEROS past simulation Setup_CalibMainModel CalibrateModel Setup_CalibKINEROS Setup_CalibSimpleRunoff Calibrates SM hsB using historical hydrometeorological and streamflow data Setup_CalibSimpleRunoff CalibrateModel extract_hillslopes OptDsmMainHSB RunHSB Curve Fitting Toolbox Image Processing Toolbox Statistics Toolbox Statistics Toolbox get_model_pars update_pars Setup_CalibMainModel load_spatial_data OptDsmMainCalib preload_forcing_data_LSM preload_forcing_data_LSM_hiRes RunLSM get_runoff Image Processing Toolbox get_model_pars update_pars load_spatial_data OptDsmMainCalib preload_forcing_data_LSM preload_forcing_data_SMM RunLSM RunSMM get_runoff Image Processing Toolbox get_model_pars update_pars 37 Setup_CalibMainModel load_spatial_data OptDsmMainCalib preload_forcing_data_LSM preload_forcing_data_LSM_hiRes RunLSM get_runoff Image Processing Toolbox Calibrates or runs the simple runoff routing model past simulation Setup_DistributeForcing GetData LoadMPEData LoadNLDASData LoadSNODASData load_spatial_data Image Processing Toolbox Statistics Toolbox Pre
25. alibration e We implemented the subsurface flow component of the SM hsB model in a way that is acceptable in the context of a distributed flash flood forecasting system Initially the model was designed to run at a fixed temporal resolution and there was no way to run it efficiently in a distributed fashion Therefore this portion of the model has been rewritten almost completely and it now runs in concert with the snow model e We implemented a data based calibration procedure to calibrate the SM hsB and KINEROS models The procedure involves a combination of inferring some model parameters from the base flow recession characteristics while others are determined by fitting the model to the data e We implemented the real time model system see above This involved considerable coding to ingest real time meteorologic data to run the model at a very high temporal and spatial resolution and to display model output in a way that is useful and familiar to NWS forecasters e Hydrologists at the NWS Binghamton Office have tested the performance of KINEROS in watersheds within their WFO including watersheds that are part of this test project SECTION 3 BENEFITS AND LESSONS LEARNED OPERATIONAL PARTNER PERSPECTIVE e Project webinars were open to WFO staff to enhance their understanding of data sources uncertainty in snow and flash flood forecasting etc e Collaboration with River Forecast Center RFC and New York City Department of Environme
26. ations over an entire area and they require much less computational effort for remarkably similar results many orders of magnitude faster than solving the hsB equations at each time step over a distributed area 1 3 Overland Flow and Channel Routing Water that emerges from hillslopes as baseflow is discharged directly into streams and water that cannot infiltrate into the subsurface either because it is saturated or the rain snowmelt intensity is too high runs off and flows over the land surface and then ends up in streams This stream water is then routed to the channel outlet The model has two ways to handle such flows During continuous past simulations those involving many years of simulation at one time the model uses a simple routing model that is based on the catchment geometry and the 12 Saint Venant equations for shallow water transport Mesa and Mifflin 1986 In this model the unit impulse response function is given by fit fy ane OH dx 15 where H x is referred to as the catchment width function and q x t is the solution to the linearized diffusion equation with a unit impulse input Brutsaert 1973 The basin response is then calculated by a convolution of responses from individual events See Troch et al 1994 for further details During real time operation and for simulation of individual past flood events KINEROS can be used as the runoff model KINEROS is a standalone model that was developed by the S
27. c imgblend Statistics Toolbox Displays and returns a matrix as a map date2doy Extract_NLDASData RunLSM Converts the date to a decimal day of the year 42 gridfit LoadMPEData_RT Extract_MPEData Estimates a surface on a 2 d grid based on scattered data hillshading createMap create a hillshading image from a digital terrain model createMap RTM_DisplayMaps Blend two images local_time_to_utc RTM_DistributeForcing RunContinuousRT Model Convert local time date information to UTC mapshow RTM_DisplayMaps createMap imblend Visualize the spatial output from KINEROS SM hsB mapshow_img RunLSM createMap RunSMM Image Processing Toolbox Creates an RGB image from a matrix netcdf Extract_MPEData Function to read NetCDF files rsm_extract_record Extract_NLDASData Extract a grib record truecolorsc createMap make a truecolor bitmap from a matrix xml_read LoadNDFDData_RT Reads XML files and converts them into a Matlab structure tree Table 8 Google Earth toolbox functions located in the lt Model Root gt Program Files HSB_Kineros ge_tools directory Functions Called From Function Calls To authoptions ge_groundoverlay ge_output ge_groundoverlay RunLSM authoptions RunSMM parsepairs 43 Generate a KML ground overlay of an image ge_output RunLSM RunSMM Write KML object to file parsepairs ge_groundoverlay ge_output authoptions parsepairs Data Processing Extract_DailyStreamfl
28. categories as well as the general flow of the program 44
29. contributing to Infiltration baseflow Time Infiltration Reset day Time with no rain that the time compression approximation Manual for infiltration resets For Simple Runoff Average Velocity m s Average water velocity in the channel Simple Runoff Froude Number Froude number Simple Runoff For KINEROS KINEROS Manning n Multiplier KINEROS Manning n Multiplier KINEROS Table 2 Parameter values that can be modified during model calibration and their descriptions 16 Watersheds txt contains information specific to each watershed Under each watershed heading there is a field called Area m2 which is the area in square meters that is upstream from the USGS gauging station there is a field called Data URL which lists the internet address of the USGS streamflow data and there are fields that specify the location of the KINEROS files for Windows and Linux which are the paths to the KINEROS parameter files that were generated by AGWA In addition to the four files listed above the Parameter Files directory must contain for each modeled watershed a file that lists model parameters for that watershed These can be varied during the calibration processes either manually or automatically see section 2 6 These parameters and their meanings are listed in Table 2 Note that most entries contain three numbers representing respectively the actual values the lower limit and the upper limit fo
30. d or Geotiff formats called Canopy_geo xxx DEM_geo xxx DEM_proj xxx and ImpGround_geo xxx into the SMODELHOME Prior Data Spatial directory 3 Using GIS software create slope and aspect maps making sure that the final processed maps for the model are in geographic coordinates These files called Aspect_geo xxx and Slope_geo xxx should be placed in the SMODELHOME Prior Data Spatial directory 1 KINEROS SM hsB uses a wgrib interface for Matlab by Emanuele Di Lorenzo More details can be found at http www o3d org wegrib http seamless usgs gov website seamless viewer htm gt http datagateway nrcs usda gov Note that the maps in geographic coordinates need to be clipped out such that the edges correspond to the edges of the modeled area but the maps in projected coordinates do not necessarily need to be clipped like this Also all maps in a given coordinate system need to have the same resolutions bounding boxes etc 27 4 If Stage III Radar Data is to be used with KINEROS SM hsB then the model needs to know were in space the radar grid cells are It does this by automatically creating a weighting scheme to convert radar coordinates to its coordinates but it needs a map showing the locations of radar grid cells To do this take a shapefile of a polar radar grid center it on the coordinates of the desired Doppler radar site and clip it to the modeled area From thi
31. ece of the program while the supplementary Model Set Up Guide Appendix B How it Works Past Simulations for Calibration Evaluation Continuous Model Snow Baseflow Simple Routing Fully Distributed Hourly Timesteps Real Time Continuous Modeling Event Model Snow Baseflow KINEROS Fully Distributed 5 min Timesteps Calibration Validation Fully Distributed Continuous Model Snow Baseflow KINEROS Hourly Timesteps ee mma Workflow Past Simulations for Process Spatial Data Calibration Evaluation 1 Assimilate Reanalysis Data NLDAS SNODAS Real Time Continuous Modeling 1 Assimilate Real Time Data ais 2 Run Real Time Model High resolution Provides state for event model if needed Project 24 hours into the future 2 Calibrate Model Baseflow Separation Analysis Calibrate hsB response Calibrate Snow Potential ET Actual ET Infiltration Calibrate KINEROS Run Model Code Figure 2 Representation of a functioning of KINEROS SM hsB and b the workflow related to running the model details the steps required for setting up the model A general overview of these steps is given below First the modeler needs to acquire the spatial data which can be downloaded for free over the internet Next the modeler needs to process the spatial data and set up KINEROS using the AGWA extension for ESRI s ArcGIS The maps that are downloaded and are produced by AGWA must
32. ected from Rain On R H AE Ap G AQ G Ground Heat Flux Melt Q Change in internal Energy Precipitation which is given as a forcing variable is Figure 3 R presentation of the processes that are important in SM hsB s partitioned into rainfall and snowmelt model snowfall depending on whether the temperature is above or below a critical value usually 0 C Though this threshold usually does not make much of a difference in mountainous areas in the southwestern USA it can considerably influence snowpack accounting and snowmelt in areas where many storms occur with rain falling at or below freezing For incoming precipitation canopy storage is taken into account using the method of Deardoff 1978 When there is snow it is allowed to accumulate on the land surface until it receives enough energy to melt The overall energy balance see figure 3 for the snowpack is calculated as Qm Rn H AE A G XQ 2 where Qn is the energy available for melt R is the net radiation computed from above H is the sensible heat exchange AE is the latent heat exchange Ap is the heat advected from precipitation and G is the ground heat exchange The sensible heat exchange H depends on both the temperature gradient between the snow surface and the air and the wind speed and the latent heat exchange is computed as the vapor pressure gradient between the snow surface and the air modified by wind speed Heat can also be advected
33. en area which are specified by a text file that is generated during the data processing step called latlon_all txt and saved in the Parameter Files directory This script will then place the XML files in a specified public directory so they can be retrieved by KINEROS SM hsB The XML files are saved for seven days so that if the model is offline for a few days it can be restarted and catch up Currently this script is scheduled using UNIX s Cron Scheduler The other component that must be automated is the model itself After correctly setting up the fields called RFC Address which is the web URL where real time MPE files can be found and NDFD Address which is the URL of where the NDFD SOAP script puts the downloaded XML files RunContinuousModel exe can be put on a task scheduler and left to run RunContinuousModel exe gets and processes the real time forcing data and runs the model using default options the model should be updated every hour By default the model ingests streamflow data and if specified by Load Realtime SNODAS Data in ProgramPars txt SNODAS data and adjusts current streamflows to match In addition there is another program called RunEventModel exe which can be run in between continuous model runs for example during a flash flood event This program also provides options for changing the model snow state to a specified value or to the SNODAS value adjusting the soil
34. energy balance snowmelt module and it is a physically based subsurface flow model The models are coupled such that there is a continuous portion that updates hourly and an event portion which is initialized during events using information about model states using the continuous portion SECTION 2 PROJECT ACCOMPLISHMENTS AND FINDINGS The project started August 1 2009 At the start of the project KINEROS was already used operationally as a flash flood model but SM hsB along with the energy balance snow model was not yet ready for operations The bulk of our efforts have been to revamp and revise the SM hsB code to make it compatible with real time operations We have written algorithms to acquire data both for real time modeling and for calibration validation we incorporated the snow energy balance model into the model system we have developed and tested calibration strategies for the combined snow SM hsB system we have introduced formulations that vastly improve model performance by several orders of magnitude and we have given it the ability to run at a variety of temporal and spatial resolutions Currently SM hsB runs at hourly and 5 minute temporal resolutions at spatial resolutions down to 1 by 1 km and it is coupled with the KINEROS overland flow channel routing model The results of these efforts have yielded a fully operational model that accounts for snow variable infiltration evapotranspiration canopy interception subsu
35. entation Sixth Southwest Hydrometeorology Symposium Tempe AZ September 28 2011 P Broxton P A Troch M Schaffner C Unkrich D Goodrich Improving Flash Flood Predictions in Moderate Sized Watersheds by Adding Snowmelt and Baseflow to KINEROS poster presentation presented by Peter Troch at the American Geophysical Union Fall Meeting San Francisco December 6 2010 These presentations highlight the goals and accomplishments of the project in regards to the KINEROS and SM HSB models and the assimilation of various national gridded datasets which are used to run the models SECTIOM 6 SUMMARY OF UNIVERSITY OPERATIONAL PARTNER INTERRACTIONS AND ROLES Patrick Broxton Programming related to SM hsB model calibration assimilation of forcing datasets and coupling with KINEROS preparation of presentations and project documents Peter Troch SM hsB model development development of methodologies related to calibration and the coupling of KINEROS and SM hsB Mike Schaffner Test KINEROS in New York watersheds facilitate communication between the University of Arizona and National Weather Service as well as other operational partners such as the New York City Department of Environmental Protection NYCDEP preparation of presentations and other outreach educational materials Dave Goodrich KINEROS model development facilitate communication with outside entities such as River Forecast Centers RFCs preparation
36. er must use NLDAS2 forcing files in a directory called s4pa NLDAS NLDAS_FORA0125_H 002 The directory structure on the FTP site inside the NLDAS_FORA0125_H 002 folder is the same as what is expected by the data extraction program Copy all the folders with the desired data to a local or network disk in a subfolder called Raw NLDAS Data this location must be referenced in the file ProgramPars txt under the field NLDAS Data Location This data need not reside in the same location as the rest of the model files as these data sets are large and it is likely that they will be placed in a remote location that is designed to store large datasets There cannot be missing files in the forcing period 19 2 4 3 SNODAS Data This data is not required but is necessary if the modeler desires to calibrate the energy balance snow model with SNODAS data SNODAS data like NLDAS data is downloaded as national gridded datasets and is then extracted for use with the model The SNODAS data represents daily snapshots at 06 00 UTC representing Snow Water Equivalent SWE snow temperature and accumulated rain snow and sublimation SNODAS data can be downloaded from ftp sidads colorado edu DATASETS NOAA G02158 to a local or network disk in a subfolder called Raw SNODAS Data this location must be referenced in the file ProgramPars txt under the field SNODAS Data Location Again the directory structure on the FT
37. es out a suitable model of the form q aX to represent the characteristic baseflow response for each hillslope Models of this form are used in the main time loop in KINEROS SM hsB because they can be vectorized performed as matrix operations over an entire area and they require much less computational effort for remarkably similar results they are many orders of magnitude faster than solving the hsB equations at each time step over a distributed area This allows the model to be calibrated and run on standard commercial computers The Calibrate Individual SM HSB Modules and Batch Calibrate SM HSB Modules tasks are nearly identical except the first allows for calibration of one module at a time while the second allows for nearly automatic calibration of different modules in sequence Each module is calibrated with just a few parameters so that during the calibration process the model is tuned a few parameters at a time similar to the method in Carrillo et al 2011 This way each section of the model can be calibrated independently in sequence or iteratively The automatic programs employ the downhill simplex method Nelder and Mead 1965 to find optimized parameters but the parameters can be manually calibrated as well by changing their values in the lt WATERSHED NAME gt _Pars txt file With the automatic programs the modeler can 1 Calibrate the Snow Module 2 Calibrate the Potential Evapotranspiration Module 3 Ca
38. ial streamflow forcing and calibration data needed for setting up the model see sections 2 2 2 4 In addition there should be four executables GetData exe CalibrateModel exe RunContinuousRTModel exe and RunEventRTModel exe These programs are at the core of the model s operation see sections 2 6 and 3 1 GetData exe handles the acquisition and processing of past streamflow and forcing data CalibrateModel exe allows the user to pick various options about how to calibrate the model RunContinuousRTModel exe should be run with a scheduler to produce model output at hourly increments and RunEventRTModel exe allows a modeler to manually run the model between hourly runs and provides advanced 13 GetData exe Data extraction and processing program Calibrate Model exe Program to perform the model calibration RunContinuousRTModel exe Program to run the Continuous Model RunOneTimeRTModel exe Program to manually run high resolution model between hourly runs Model Root Directory Display Z These folders will be created and filled automatically as the model runs Model States CalibDates txt File with the dates for the calibration of SM hsB Parameter Files EventDates txt File with the dates for the calibration of KINEROS Prior Data ProgramPars txt File with model options Watersheds txt File with watershed information lt WATERSHED NAME gt Pars txt File s with model parameters for each wate
39. ing the need for the toolbox Also while many of the following functions were written for this model not all of them were For example many of the functions in the Spatial sections either derive from Patrick Bogaart s HSB Toolbox or Steve Eddins Upslope 34 Area Toolbox In addition many of the Miscellaneous functions are from a variety of sources on the Matlab File Exchange or from the HSB Toolbox Authorship or a record of modification is retained in the header of all the function files There is also a chart at the end of this document depicting graphically the flow of the model functions Table 1 Top level Programs located in the lt Model Root gt lt Region Name gt directory Functions Called From Function Calls To CalibrateModel GetProgramPars Setup_Aquifers Setup_BaseflowSeparation Setup_CalibKINEROS Setup_CalibMainModel Setup_CalibSimpleRunoff Performs all the tasks that are necissary for calibrating KINEROS SM hsB GetData Extract_DailyStreamflow Extract_HiResStreamflow Extract_NLDASData Extract_MPEData Extract_SNODASData Setup_DistributeForcing Prepares the hydrometeorologic and streamflow data that are used by KINEROS SM hsB RunContinuousRTModel GetProgramPars RTM_DisplayMaps RTM_DistributeForcing RTM_RunModel local_time_to_utc Executes the real time continuous KINEROS SM hsB model RunEventRTModel GetProgramPars RTM_RunModel Executes the real time event KIN
40. iven area can be modeled in KINEROS SM hsB These need to be set up separately in AGWA repeat steps 1 3 above Acquiring Streamflow Data for Calibration of KINEROS SM hsB Streamflow at the outlet of the modeled watersheds need to be downloaded from the USGS Both 24 hour and 15 minute streamflows are required 24 hourly data is used when calibrating hsB to a continuous multi year long time series and 15 minute data is used when using KINEROS SM hsB with past events The data can be downloaded as excel spreadsheets and the data range must include both the calibration and validation periods for the model as well as any event that the model is to be run for The streamflow data needs to be formatted so that it can be ingested into KINEROS SM hsB All streamflow files must be in comma separated csv format and be called SWATERSHEDNAME csv For daily streamflows the files need to have five columns excel timestamp year month day streamflow value and they must be placed in the SMODELHOME Prior Data Streamflow 24 hour directory For 15 minute streamflows the files need to have six columns year month day hour minute streamflow value and they must be placed in the SMODELHOME Prior Data Streamflow 15 minute directory Note that it is standard convention that streamflows are in local standard time and this offset from UTC must be noted in a model options file see the KINEROS SM hsB documentation section 2 4
41. l_data Setup_CalibSimpleRunoff Setup_DistributeForcing RTM_DistributeForcing RTM_RunModel Loads spatial data layers for KINEROS SM hsB nbrtable process_hillslopes Creates a neighborhood index table calculate_flow flowdistance process_hillslopes Mapping Toolbox Image Processing Toolbox extract_hillslopes arcascii process_dem Image Processing Toolbox 40 pixel_flow dem_flow border_nans facet_flow Downslope flow direction for DEM pixels process_dem load_spatial_data d8 dem_flow fill_sinks arcascii Get slope and aspect from a DEM process_hillslopes extract_hillslopes Computes hillslope and catchment characteristics ProcessWFuncs process_hillslopes Computes width functions for all hillslopes upslope_area calculate_flow Upslope area measurements for a DEM widthfunction process_hillslopes Compute hillslope width functions Mapping Toolbox downnbr flowdistance nbrtable processWFuncs widthfunction flowdistance widthfunction border_nans Table 5 Functions related to the acquisition and processing of hydrometeorologic data located in the lt Model Root gt Program Files HSB_Kineros Data directory Functions Called From loadMPEData Setup_DistributeForcing Loads MPE data for use with KINEROS SM hsB Calibration LoadMPEData_RT RTM_DistributeForcing Loads MPE data for use with KINEROS SM hsB Real Time Function Calls To Image Processing Toolbox gridfit Load
42. librate the Infiltration Module 4 Calibrate the Actual Evapotranspiration Module 5 Calibrate the Runoff Module using a continuous series and 6 Calibrate the Runoff Module using individual events The modeled snow water equivalent is calibrated to SNODAS data the modeled potential evapotranspiration PET is calibrated to NLDAS estimates of PET and the changes in Infiltration Actual Evapotranspiration and Runoff are tuned based on their effects on modeled streamflow Since changing parameters related to Infiltration Actual Evapotranspiration and Runoff not only affect streamflow but involve components that are coupled they should be calibrated iteratively multiple times Calibration is only performed ona period specified by the Total Calibration Period field in CalibDates txt and the model is checked using an evaluation period specified by the Total Evaluation Period field in CalibDates txt Table 2 lists the parameters that are utilized in each calibration step The Calibrate Kineros task lets the user select the watershed to calibrate and it then calibrates the KINEROS portion of the model by using hillslope output from SM hsB Calibration is performed similarly to the SM hsB modules using the Downhill Simplex Method but the temporal and spatial resolutions of the model are higher and modeled streamflows are 22 calibrated to observed streamflow for multiple events See Table 2 for parameters that are
43. ling and model operation is accomplished by four programs The two programs that relate to the Calibration Evaluation phase of the model are called GetData exe and CalibrateModel exe To get started open the GetData exe Program This will allow the modeler to 1 Load NLDAS data 2 load SNODAS data 3 load MPE data 4 Load daily streamflow data 5 load high resolution streamflow data and 6 prepare forcing data for simulations during the calibration and validation periods The first task Load NLDAS Data extracts the desired NLDAS variables from each GRIB file and then cuts out the data using the bounds that are specified in ProgramPars txt The second task Load SNODAS Data performs essentially the same tasks but for the SNODAS snapshots The third task Load MPE Data extracts the MPE precipitation data for the desired area The fourth task Load Daily Streamflow Data extracts the daily streamflow data from the data files and formats them for use with the model The fifth task Load High Resolution Streamflow Data formats the hourly streamflow data for use with the model Finally the sixth task Prepare Forcing Data resizes and distributes the already extracted NLDAS and SNODAS data preparing it for the model runs Each one of these tasks is fully automatic and requires no further user input The actual calibration evaluation is performed by another set of tasks that are accessed u
44. lutions currently set to 1 hour and 5 minutes The combined model system which will be referred to as KINEROS SM hsB has the following features e Fully Distributed i e divides watersheds up gag Vv P into smaller elements 4 e D e Continuous i e keeps track of catchment 5 Aa Sao n 9t Evapotranspirati wetness and snow water equivalent ae a ee wpe Canopy 1 z Accounts for snow variable infiltration e isi on evapotranspiration canopy interception ian Sv Tepe Precipitation subsurface flow overland flow and channel Radiation 4 routing see Figure 1 Flexible spatial and temporal resolution Includes a framework to easily calibrate the models using historical data from the North American Land Data Assimilation System Figure 1 Model schematic for the KINEROS SM hsB system that shows important modeled physical processes NLDAS and Multisensor Precipitation Estimator MPE e Real time operation using hourly precipitation estimates from the Multisensor Precipitation Estimator MPE forecasts from the National Digital Forecast Database NDFD and optionally Stage IIl Radar Data 5 minute 1 degree by 1 km We believe that this system will make a meaningful contribution to the flash flood forecasting community especially because it has very high spatial and temporal resolutions and its ability to run in continuous mode allows for improved consistency of flash flood predictions The m
45. ndling of the forcing data and running the model subroutines Functions in the lt Model Root gt Program Files HSB_Kineros Spatial directory perform spatial processing required for KINEROS SM hsB They include functions to perform generalized terrain analysis as well as functions to perform specific terrain analysis that is required by HSB Functions in the lt Model Root gt Program Files HSB_Kineros Data directory are helper functions for processing hydrometeorologic data both past and real time Functions in the lt Model Root gt Program Files HSB_Kineros Calib directory perform calibration for various model components Finally functions in the lt Model Root gt Program Files HSB_Kineros Misc are miscellaneous functions such as for creating map figures reading various file formats and handling timestamp information These functions along with the top level programs make up KINEROS SM hsB Only the top level programs are compiled yielding four executables the functions are automatically compiled with the top level programs The tables below give the name and location of each of these functions as well as a very brief description what the function does In it there is a reference to all of the functions that are in separate files not including any sub functions Requirements for Matlab toolboxes are also noted though in some cases the required toolbox functions can be rewritten in the future thus eliminat
46. ng a network file system but the directory structure inside of a Raw Data subdirectory for each data set must match the original directory structure For example for NLDAS data there must be an SNLDASROOT folder with a Raw Data subfolder which in turn contains an NLDAS_FORA0125_H OO2 subfolder and so on The model must also know where to look for these directories see the KINEROS SM hsB documentation section 2 4 2 If past events are to be simulated with Doppler radar precipitation estimates then radar data must also be downloaded This can be done using NOAA s Weather and Climate Toolkit along with the NCDC NEXRAD Data Inventory Search Just fill out the order form note the HAS job number and retrieve the data using the Weather and Climate Toolkit Radar grids can be exported for the desired area same as the modeled area as ArcInfo ASCII Grid or Geotiff formats These exported files must be placed in the SMODELHOME Prior Data Forcing RADAR directory where each day s radar data is in a subdirectory that identifies the date yyyymmdd MPE http directory http water weather gov precip p_download_new 1 NLDAS ftp directory hydro1 sci gsfc nasa gov data s4pa NLDAS NLDAS_FORA0Q125_H 002 11 SNODAS ftp directory ftp sidads colorado edu DATASETS NOAA G02158 The NOAA weather and climate toolkit can be obtained at http www ncdc noaa gov oa wct http www ncdc noaa gov nexradinv 3
47. nning a range of topographies and climates five in the West Branch Delaware River basin in New York and three near Tucson AZ e We have analyzed the hydrology of the New York and Arizona basins Of note we have detected a potentially significant radar bias that favors the southwestern portion of the New York study area Also initial model runs with the SM hsB system indicate that it will be much harder to successfully implement the system for the Arizona basins e We have developed algorithms to acquire meteorologic data both for real time modeling and for calibration validation for past flows KINEROS already had built in the capability to assimilate real time radar data but SM hsB only had the ability to operate on data that were manually prepared The ability to automatically acquire and process meteorologic data for SM hsB is new e We implemented the snow energy balance module for inclusion into the model system This has involved a considerable amount of work to run it both in hind cast mode as well as in forecasting mode One challenge has been to give the module the ability to be run at multiple temporal and spatial resolutions This is essential because we run the model at a lower resolution during inter storm periods and at a higher resolution during events Also it is impossible to calibrate the model at extremely high resolutions because past data is only available at hourly timesteps and efficiency becomes a major concern during c
48. nput Run the CalibrateModel program or Matlab Script Perform each of the steps 1 Perform Baseflow Separation 2 Find Distributed Drainage Characteristics in sequence These will not require any user input The next steps depend on the method of calibration whether KINEROS is used or if the model is just to be run with SM hsB and whether past events are to be simulated Details of the steps 3 Calibrate individual SM HSB Modules 4 Batch calibrate SM HSB Modules 5 Calibrate KINEROS using selected events 6 Run SM HSB for the calibration and evaluation periods fully coupled 7 Run SM HSB for the calibration and evaluation periods fast model and 8 Run model for selected events can be found in section 2 6 of the KINEROS SM hsB documentation 31 H 1 2 3 4 Setting up KINEROS SM hsB to run in real time To run KINEROS SM hsB in real time both the model and a separate script to download NDFD forecasts must be automated Set up the NDFD script to be executed hourly 5 15 minutes after the hour Right now this script only exists as a shell script on a Linux server It can be automated in UNIX with CRON The NDFD script also needs be modified so that it puts the downloaded NDFD files in the correct directory modify the script called DownloadNDFD_all sh and it must be provided with a list of latitude longitude values which are produced during the model set up process look for the file called
49. ntal Protection NYCDEP staff e Potential to add value and increase lead times for flash flood warnings Will be able to issue flash flood warnings by basins and estimate the timing and magnitude of flash flood events e Potential to forecast flash floods for rain on snow events e Staff at numerous WFOs have had exposure to KINEROS and there is an increased understanding of the data that are necessary desirable for flash flood forecasting SECTION 4 BENEFITS AND LESSONS LEARNED UNIVERSITY PARTNER PERSPECTIVE e UA researchers have developed an increased understanding and appreciation for the challenges of implementing operational flash flood models including o Ingesting various forcing inputs o Addressing performance issues o Understanding uncertainties in model inputs and model structure e UA researchers have participated in numerous conference calls involving other organizations USDA NWS and NYCDEP RFCs e Access to NWS data sources e Research experience and professional development for the graduate student Patrick Broxton involved with the project including three professional presentations and extensive coding experience SECTION 5 PUBLICATIONS AND PRESENTATIONS There have been twelve external presentations as well as numerous scientific exchanges between the various parties in the form of teleconferences and webinars field visits and at scientific conferences e M Schaffner KINEROS2 Site Specific Model Background
50. od 1994 Multiscale modeling of spatially variable water and energy balance processes Water Resour Res 30 11 3061 3078 doi 10 1029 94WR01498 Khumalo G J Holechek M Thomas F Molinar 2008 Soil Depth and Climatic Effects on Desert Vegetation Dynamics Rangeland Eco Manage 61 269 274 Lyne V and Hollick M 1979 Stochastic time variable rainfall runoff modelling Institute of Engineers Australia National Conference Publ 79 10 89 93 McCumber M C and R A Pielke 1981 Simulations of the effects of surface fluxes of heat 24 and moisture in a mesoscale numerical model 1 Soil layer J Geophys Res 86 9929 9938 Mesa O J and Mifflin E R 1986 On the relative role of hillslope and network geometry in hydrologic response In Gupta V K Rodriguez Iturbe and Wood E F Editors 1986 Scale Problems in Hydrology D Reidel Dordrecht pp 1 17 Milly P C D 1986 An event based simulation model of moisture and energy fluxes at a bare soil surface Water Resour Res 22 1680 1692 Nelder John A R Mead 1965 A simplex method for function minimization Computer Journal 7 308 313 doi 10 1093 comjnl 7 4 308 Shuttleworth W J 2012 Terrestrial Hydrometeorology Wiley Blackwell ISBN 978 0 470 65938 0 Tarboton D G 1997 Anew method for the determination of flow directions and upslope areas in grid digital elevation models Water Resources Research 33 2 309 319 Teuling A J
51. odels are coupled in such a way that the watershed can be modeled with SM hsB only or KINEROS SM hsB SM hsB has its own simpler routing procedure that is used during calibration whereas KINEROS can be used when running the model in real time and during individual past events if specified KINEROS is run with hillslope output from SM hsB runoff but baseflow is always routed using the simple runoff procedure Both models are distributed SM hsB runs on a regular grid and KINEROS runs on a series of delineated overland flow planes so the model system can diagnose if areas where there is heavy rain are wet dry or snow covered snow free SM hsB is written using Mathworks Matlab while KINEROS is written in FORTRAN though the entire suite of programs is run as a series of stand alone executables that require an additional installation of the provided Matlab Compiler Runtime MCR libraries The functioning of the entire program is very fast for being a distributed model and it can be run on most commercial Linux SM hsB only and Windows KINEROS SM hsB machines 1 1 Workflow The process of setting up and running the model is complicated by the fact that there are many input datasets both spatial and hydrometeorologic The model requires calibration though much of the model programming is designed to aid or automate this process The supplementary Function Reference Appendix C gives an overview of each individual pi
52. orage Boussinesq equation Os kcosa S OS S w eae l on k cosa 4 ax wax Here S is the hillslope storage is the slope w is the width of the hillslope x is a coordinate Ww along a sloping bed and k and f are parameters representing horizontal hydraulic conductivity and drainable porosity Different versions of SM hsB can be run on daily or finer time steps in distributed mode separately on each individual hillslope or in semi distributed mode where hillslopes are lumped into a single catchment hillslope but such that their collective geometries are still preserved in a hillslope width function It is however fairly slow to run because it numerically solves the Boussinesq equation For KINEROS SM hsB to satisfy the requirement of high temporal and spatial resolutions as well as numerical speed the subsurface flow part of SM hsB has been simplified and optimized to effectively simulate the rapid response of floods and to apply on a distributed model grid in a computationally efficient manner Characteristic hsB responses are computed for each hillslope prior to the main model run to preserve within catchment subsurface variability and to take advantage of hsB s ability to simulate different behaviors on different hillslopes These responses are represented with storage discharge models of the form q aX for each hillslope see figure 4 Models of this form can be vectorized performed as matrix oper
53. outhwest Watershed Research Center SWRC of the US department of Agriculture USDA It is an event oriented physically based model describing the processes of interception infiltration surface runoff and erosion from small agricultural and urban watersheds Woolhiser et al 1990 In KINEROS overland flow and channel routing are simulated using a kinematic wave approach over a network of planes and channels It also simulates runoff either infiltration excess or saturation excess In KINEROS SM hsB the KINEROS model is used to simulate overland flow and channel routing on runoff from SM hsB though the routing of baseflow is always performed by the simple routing model More information about KINEROS can be found at http www tucson ars ag gov kineros 2 SETTING UP THE MODEL 2 1 Directory Structure and Model Basics Files for a particular modeled area with the exception of some data files AGWA parameterization files which can reside on external hard disks network drives etc can be located anywhere provided they are in a single directory with a prescribed directory structure The first level directories are called Display which contains graphical model output files Model States which contains state files and other files that are needed during and between model runs Parameter Files which contains files defining model parameters and program options see section 2 5 and Prior Data which contains spat
54. ow Extract_HiResStreamflow Extract_MPEData GetDate Extract_NLDASDate Extract_SNODASData Setup_DistributeForcing Program Information GetProgramPars Real Time Modeling RTM_DisplayMaps RTM_DistributeFordng RunContinuousRTModel rTM RunModel RunEventRTModel Model Calibration Validation Setup_Aquifers Setup_BaseflowSeparation Setup_CalibMainModel Setup_CalibKineros Setup_CalibSimpleRunoff CalibrateModel Past Data Load_MPEData Load_NLDASData Load_SNODASData Real Time Data Load_MPEData_RT Load_NDFDDsta_RT Load_SNODASData_RT ReadRTStramflow arcascii create_map date2doy gridfit hillshading imgblend local_time_to_ute mapshow mapshow_img netcdf rsm_extract_record truecolorsc xmi_read OptDsmMainCalib OptDsmMainHSB OptDsmMainRT Spatial Processing calculate_flow d8 extract_hillslopes load_spatial_data process_dem process_hillslopes ProcessWFuncs HSB Toolbox define_constants downnbr flowdistance nbrtable widthfunction Upslope Area Toolbox border_nans dem_fiow facet_flow fill_sinks flow_matrix pixel_flow upslope_area Parameters define_pars get_model_pars update_pars Forcing Data preload_forcing_data_LSM preload_forcing_data_LSM_hiRes preload_forcing_data_SMM Main Model get_runoff RunHSsB RunLSM RunSMM authoptions ge_sround_overlay ge_output parsepairs Figure 1 Graphical structure of KINEROS SM hsB showing all of the functions grouped into
55. pares hydrometeorologic forcing data for use with SM hsB calibration Table 3 Functions that perform model specific tasks located in the lt Model Root gt Program Files HSB_Kineros Model directory define pars Defines constant model parameters for SM hsB get_model_pars Reads model parameters for SM hsB from file get_runoff preload_forcing data_LSM Preloads forcing data files for the land surface model Functions Called From Function Calls To RunLSM RunSMM Setup_CalibMainModel Setup_CalibSimpleRunoff Setup_CalibKineros RTM_RunModel Setup_CalibSimpleRunoff Setup_CalibMainModel Setup_CalibKineros RTM_RunModel OptDSMMainRT OptDSMMainCalib Setup_CalibMainModel Setup_CalibSimpleRunoff RTM_RunModel Setup_CalibKineros Curve Fitting Toolbox preload_forcing data_LSM_hiRes Preloads forcing data files for the land surface model high resolution preload_forcing data_SMM Preloads forcing data files for the sub surface model RunHSB Executes the HSB Aquifer Model Runs the Land Surface Model RunSMM Runs the Sub Surface Model update_pars Updates model parameters for SM hsB Setup_CalibSimpleRunoff arcascii Setup_CalibKineros RTM_RunModel preload_forcing_data_LSM_hiR es RTM_RunModel Image Processing Toolbox Setup_CalibMainModel Setup_Aquifers OptDSMMainHSB Setup_CalibMainModel OptDsmMainCalib Setup_CalibSimpleRunoff RTM_RunModel Setup_CalibKineros define_pars
56. platforms Copy Model Root lt Model Site Name gt Parameters RunContinuousRTModel_win32 exe Copy Model Root lt Model Site Name gt Parameters RunEventRTModel_win32 exe 33 Appendix C FUNCTION REFERENCE FOR KINEROS SM hsB Patrick Broxton Peter Troch Mike Schaffner Carl Unkrich David Goodrich The University of Arizona Tucson AZ US National Weather Service Salt Lake City UT US 3USDA ARS Southwest Watershed Research Center Tucson AZ US KINEROS SM hsB is made up of four main programs and sixty nine functions that perform tasks related to data processing both hydrometeorologic forcing data and terrain data model calibration and real time data acquisition model operation There are six main groupings of functions which are located in separate sub directories in the lt Model Root gt Program Files HSB_Kineros directory Functions in the lt Model Root gt Program Files HSB_Kineros Program directory perform a variety of tasks similar in nature to stand alone programs except these are tied together by the four top level programs They handle tasks related to processing of hydrometeorologic data setting up and running the model both for the calibration validation periods and in real time and they keep track of program information Functions in the lt Model Root gt Program Files HSB_Kineros Model directory relate to the execution of KINEROS SM hsB including handling of the parameters ha
57. r The spatial resolution in degrees of the SM hsB model grid during real time operation simulation of past events Forcing Resolution Number The spatial resolution in degrees of the processed forcing data for use with past simulations of SM hsB Distribute Aquifer Properties Yes No Yes Use soil characteristics and hillslope delineations to calculate variable subsurface hsB response No treat subsurface properties as uniform Correct Past SWE Yes No Use SNODAS to correct the snow model during the calibration period to match real time operation where SNODAS data can be used to correct real time snow estimates Use MPE for Calibration Yes No Use MPE data instead of NDFD data for precipitation estimates during the calibration validation periods Use Rain Rates Program Yes No Use the provided rain rates program to ingest real time radar data untested NLDAS Data Location String Path to NLDAS data for past simulation SNODAS Data Location String Path to SNODAS data for past simulation Linux NLDAS Data Location String Location of the calibration NLDAS files Linux Linux SNODAS Data Location String Location of the calibration SNODAS files Linux Linux MPE Data Location String Location of the calibration MPE files Linux Linux DHR Data Location String Location of the real time DHR files Linux Linux NLDAS Data Location String Location of the calibration NLDAS files Windows Linu
58. r each parameter The lower and upper limits are used by the automatic calibration subroutines they place limits on possible values for each parameters Not all parameters in the files need to have upper and lower limits as these are not changed during automatic calibration Any numbers can be changed but it is suggested that reasonable bounds be placed on the parameters to ensure a reasonable representation of the watershed physics Additional model parameters are not shown in these files and should not be adjusted 2 3 Spatial Data Because KINEROS SM hsB is distributed it requires a large amount of spatial data Required maps include a digital elevation model DEM a slope map an aspect map a canopy coverage map and an impervious surface map these data can be downloaded from the USGS s Seamless Map Server http seamless usgs gov website seamless viewer htm The modeler must define coordinates of the corners of the desired map data these corners are then specified in the parameter file ProgramPars txt under the fields North South East and West All data must be in geographic coordinates as the modeling takes place on a regular latitude longitude grid to conform to the forcing data However there must also be a version of the DEM in projected coordinates because some of the terrain processing requires pixels that represent actual distances GIS Software must be used to perform these transformations
59. rface flow overland flow and channel routing The combined model system is spatially distributed and has a temporal resolution of 1 hour five minutes during events It uses readily available gridded data sources for past and present forcing data including North American Land Data Assimilation System NLDAS Multisensor Precipitation Estimator MPE National Digital Forecast Database NDFD and Stage III radar data The model is also validated with USGS streamflow data and model output can be compared with the Snow Data Assimilation System SNODAS dataset Calibration of the combined model system is computer assisted and involves inferring catchment characteristics using baseflow analysis and subsequent calibration of parameters related to snow evapotranspiration infiltration and runoff Calibration steps are optionally fully automatic manual or a combination of both and involve changing parameters related to specific processes a few at a time and iterating Once calibrated the model is set up to run in real time every hour using newly downloaded hydrometeorologic data It also ingests real time streamflow data to correct predictions and it has the capability of running at 5 minute intervals during events We have also prepared extensive documentation see Appendices A C regarding the functioning setup and execution of the combined model system Specific accomplishments are as follows e We selected eight test basins to be modeled spa
60. ring every model time step Below is a synopsis of the important processes simulated by SM hsB 1 2 1 Radiation The first process that is computed in the model is the radiative balance incoming outgoing radiation over complex terrain The overall radiation balance equation is Ry R 1 a sRig Riu 1 The model keeps track of net radiation at the ground and at the top of the canopy separately so that different processes can use either canopy radiation estimates e g for potential evapotranspiration or ground radiation estimates e g for snowmelt The radiative balance is computed as in Shuttleworth 2012 The model first computes extraterrestrial incoming solar radiation and then applies atmospheric corrections involving cloud cover Terrain influences are taken into account for direct shortwave radiation by comparing flat surface incoming solar radiation with that hitting an inclined plane Whiteman and Allwine 1986 Incoming longwave radiation is computed as a function of atmospheric temperature and cloud cover and outgoing longwave radiation is computed as a function of the temperature and emissivity of the surface For computational efficiency surface temperature is assumed to be the same as that of the atmosphere for snow free surfaces R wo Snowfall Rainfall Q Energy Available For Melt JE H R Net Radiation Kz H Sensible Heat Exchange 1 2 2 Snow Model gt AE Latent Heat Exchange Ap Heat Adv
61. rshed Forcing This folder will be created and filled automatically by the data extraction program Spatial Aspect_geo tif Aspect Map geographic coordinates Canopy_geo tif Canopy Map geographic coordinates DEM_geo tif Digital Elevation Model geographic coordinates DEM_proj tif Digital Elevation Model projected coordinates Streamflow ImpGround_geo tif Impervious Ground Map geographic coordinates Radar_geo tif Radar Grid Map geographic coordinates Slope_geo tif Slope Map geographic coordinates lt WATERSHED NAME gt _hsid_projtxt Watershed Discretization geographic coordinates lt WATERSHED NAME gt _hsid_geo txt Watershed Discretization projected coordinates 15 Minute gt lt WATERSHED NAME gt txt File s with 15 minute streamflows Daily gt _ lt WATERSHED NAME gt txt File s with daily streamflows Processed This folder will be created and filled automatically by the data extraction program Figure 6 Directory structure of root model directory Bold text represent directories while plain text represent files Text in italics are descriptions Note that in addition to the main model directory NLDAS data and SNODAS data for past simulations as well as NDFD data for real time modeling must reside in separate directories options about running ensembles and updating the model states Matlab scripts from which the exe files are compiled as well as a large number of helper functions are
62. s clipped shapefile create a grid with the same resolution as the above maps in geographic coordinates where the values of each cell correspond to the radar bin C Setting up KINEROS and processing spatial data using AGWA 1 Instructions for installing AGWA and setting up KINEROS using AGWA can be found at the AGWA home page It is recommended that the modeler if he or she is not already familiar with AGWA upon downloading AGWA download one of the tutorials for delineating discretizing and parameterizing an example watershed Note that AGWA2 for ArcGIS 9 3 was used with the initial version of KINEROS SM hsB The DEM that is downloaded for KINEROS SM hsB is adequate to use as a starting point for the AGWA set up The modeler will also need to download a NLCD land cover map and a soils map either STATSGO or SSURGO for the AGWA set up When performing the AGWA steps it is necessary that care be given to the fact that spatial data is properly projected in a consistent spatial reference system as this can be the source of many errors especially because ArcGIS will display maps with multiple projections concurrently When defining a precipitation scenario it is OK just to use dummy values because these will be updated by KINEROS SM hsB as needed 2 Note the location of the par file that is generated or copy it to a new location as KINEROS SM hsB will need to know where to look for this file see the Kineros SM hsB documen
63. sing the CalibrateModel exe program When this program is first opened the modeler is prompted to select the watershed that is to be evaluated within a region one or many watersheds can be modeled independently Available options are to 1 Perform Baseflow Separation 2 Find Distributed Drainage Characteristics 3 Calibrate individual SM HSB Modules 4 Batch calibrate SM HSB Modules 5 Calibrate KINEROS using selected events 6 Run SM HSB for the calibration and evaluation periods fully coupled 7 Run SM HSB for the calibration and evaluation periods fast model and 8 Run model for selected events The Perform Baseflow Separation and Find Distributed Drainage Characteristics programs are important for calibrating the HSB aquifer The hsB response is calibrated to the baseflow recession characteristics of the watershed These characteristics are determined by separating baseflow from quickflow using a lowpass filter Lyne and Hollick 1979 finding a master 21 recession curve from the individual baseflow recessions and then determining a suitable model of the form q aX to match this curve The second program actually runs hsB on individual hillslopes determined from AGWA or from a hillslope representing the entire catchment calibrates its response to match the master recession curve while preserving differences resulting from variations in slope and soil type specified in the KINEROS par files and figur
64. tation section 2 4 3 AGWA hillslope discretization files need to be made available to KINEROS SM hsB to be able to correctly map the forcing data and the model data to the KINEROS hillslopes This is accomplished by a weighting procedure similar to what is used for the radar data Just like the radar grid needs to be converted to a raster map for the model to be able to do this the KINEROS hillslopes which are represented by a shapefile in AGWA ArcGIS s Shape2Raster conversion tool will perform the conversion from shapefile to raster at a given resolution 6 7 2 r http www tucson ars ag gov agwa index php documentation mainmenu 41 7 An example location is C Working AGWA2 workspace delawarebasin walton d1 simulations kineros kineros par 28 D 4 1 2 need to be converted to a raster map where the values of each cell correspond to the KINEROS hillslope id There must be copies of this map in both geographic and projected coordinates because they need to be used with both the DEM processing steps as well as with the regular model grid They need to be put into the SMODELHOME Prior Data Spatial directory and called SWATERSHEDNAME_hsid_geo xxx and SWATERSHEDNAME_hsid_proj xxx where SWATERSHEDNAME is an identifier given to a specific watershed see the KINEROS SM hsB documentation section 2 4 and again the files can be in either ArcInfo ASCII Grid or Geotiff formats Multiple watersheds in a g
65. than it was set on it is convenient to only copy the necessary files over to save space and to make copying faster Section lists the files that need to be copied over to the new computer Real time model output is saved to image and text files and placed in the SMODELHOME Display Real Time directory It is recommended that these results be linked to a web page for transmission over the internet 32 5 The program called SMODELHOME RunEventModel xxx can be used to run the model in between hourly runs It also provides additional options such as the ability to alter the model state e g add more Snow Water Equivalent and to run the model multiple times using various precipitation scenarios l Running the model on another machine than where it was set up Copy the directory Model Root RealTime Data Copy portions of the lt Model Site Name gt directory Copy the directory Model Root lt Model Site Name gt Parameters Copy the following files and folders from Model Root lt Model Site Name gt Model States Real Time Runoff SM_hsB Calibration Spatial Forcing Data_av mat Note the subdirectory called Calibration is quite large and not needed for real time model functioning so it should not be copied over Copy the following folder from Model Root lt Model Site Name gt Prior Data Spatial Copy the appropriate platform specific executables for the real time and event models For example for win32
66. to the snowpack by precipitation and through the subsurface as a ground heat flux These energy inputs can cause sublimation heating of the snowpack or melt 10 The snow model is also able to capture behavior related to variability in snow depth and snow albedo Snow depth is computed from the snow density and snow water equivalent The snow density is allowed to change through time depending on the age of the snow in the snowpack Also as the snow surface ages its albedo can decrease dramatically These considerations are also taken into account through the use of parameterizations Discharge NLin Mod Soil Mod 2Lin Mod hsB Mod Actual Q mm day 5 10 15 20 25 30 Time day General model Poweri f x a x b Coefficients wi 95 confidence bounds a 0 001432 0 001413 0 00145 Q BF b 2 05 2 046 2 053 Goodness of fit SSE 76 54 R square 0 9928 Figure 4a Watershed delineation for KINEROS SM hsB b model schematic showing the important mass balance F3 a t M Q Gi 1 u 2 w 4 0 o O RMSE 0 08358 considerations for the subsurface portion of SM hsB e 10 20 30 40 50 60 Storage mm Discharge mm day a 1 2 3 Baseflow and Soil Moisture Figure 5a Shape of typical baseflow recessions and the ability of various models to reproduce such Snowmelt from the above formulation behavior b storage discharge behavior predicted combined with rainfall input minus canop
67. tuned during the KINEROS calibration In addition there are several options to run and display results for calibration and validation periods The option called Run SM HSB fully coupled runs the SM hsB snow model included for the entire calibration and validation periods The option called Run SM HSB fast model runs just the subsurface model no snow model as it is much faster so it is used extensively for calibration To run the fast model the full model must first be run to get outputs from the snow model which are saved as hourly inputs to disk This splitting which is used for calibration because it helps significantly with speed is made possible by the fact that the subsurface parameterizations do not affect the snow model The final option Run model for selected events runs KINEROS SM hsB for past streamflow events the same that are used to calibrate KINEROS KINEROS or the simple runoff module can be used in this step This option is useful if the modeler wants to manually calibrate KINEROS or the simple runoff module 3 RUNNING THE REAL TIME MODEL Once the model is calibrated it is relatively straight forward to run the model in real time There are two components which must be automated run every hour The first is a script that queries the NDFD SOAP XML service to download NDFD data every hour This script currently a UNIX shell script retrieves an XML file containing the NDFD data for all points in a giv
68. wants to disregard subsurface variability in SM hsB then the hillslope discretization map can be replaced with a simple watershed delineation map Finally if radar data is to be used with KINEROS SM hsB then a raster map showing the locations of the radar bins for the local Doppler Radar must also be created again using the same geographic coordinate system as the other spatial data files This map can be generated by projecting a generic radar grid at coordinates specified by the radar and then clipped to the modeled area and converted into a raster Given this raster KINEROS SM hsB will automatically generate weight files to map the radar data to the latitude longitude grid The model expects all spatial data to be in geotiff tif or in ArcInfo ASCII Grid asc format These files must be placed in the Spatial Data subdirectory in the Prior Data directory and be named the following Aspect_geo tif geographic version of the aspect map Canopy_geo tif geographic version of the canopy map DEM_geo tif geographic version of the digital elevation model DEM_proj tif projected version of the digital elevation model ImpGround_geo tif geographic version of the impervious surface map and Slope_geo tif geographic version of the slope map In addition there must be watershed maps for each of the modeled watersheds in both geographic and projected coordinates named lt WATERSHE
69. x SNODAS Data Location String Location of the calibration SNODAS files Windows Linux MPE Data Location String Location of the calibration MPE files Windows Linux DHR Data Location String Location of the real time DHR files Windows RFC Region String River Forecast Center Region used for the naming convention of real time MPE data RFC Address String Internet address of MPE data from the local River Forecast Center NDFD Address String Address of NDFD data that is downloaded using the SOAP XML server see section 2 4 NDFD ID String Identifier used for the naming convention of real time NDFD data 15 NDFD Precip Duration hours Number Duration of forecasted QPF since NDFD only outputs QPF once every six hours Calibration NStarts Number Number of repetitions for optimization algorithm used for calibrating SM hsB Calibration Niters Number Number of starts for optimization algorithm used for calibrating SM hsB Nudging NStarts Number Number of repetitions for optimization algorithm used during real time data assimilation Nudging Niters Number Number of starts for optimization algorithm used during real time data assimilation HSB NStarts Number Number of repetitions for optimization algorithm used for calibrating the HSB response on each hillslope NSB Niters Number Number of starts for optimization algorithm used for calibrating the HSB response on each hillslopes Table 1 Program options their
70. y ye e E E EE nee storage give the total water input into the soil The infiltration of this precipitation is computed as in Troch et al 1994 The infiltration equation is based on the time compression approximation to Philip s equation Famiglietti et al 1992 It depends on the accumulated rainfall and the soil volumetric water content at saturation Once infiltrated it adds to the soil water which is either lost to evapotranspiration or is subject to drainage Evapotranspiration is computed using the Penman Monteith model for the whole canopy A Rn G 22D Qr rr 3 A y ra Here rr the bulk stomatal resistance of the whole canopy is treated as a user defined parameter while A y 2 D r are computed using the standard methodology Shuttleworth 11 in press Ra Gis the available energy to be partitioned into sensible and latent heat G is calculated as in Famiglietti and Wood 1994 Vegetation characteristics are allowed to vary both spatially using a vegetation cover map and temporally as a function of weekly temperature when the average temperature is above below a certain threshold the canopy density and LAI are maximized minimized with a smooth change in between Once water has recharged the confined aquifer it is handled by the hsB routing portion of the model see figure 3 hsB accounts for complex hillslope geometries and subsurface properties when predicting baseflow according to the hillslope st

Final Report for KINEROS/SM-hsB COMET Cooperative Project

Contents

Download Pdf Manuals

Related Search

Related Contents