MIKE SHE WM – User Manual
Contents
1. 6 MIKE SHE PP Calibration Statistics Utility DHI Water amp Environment DHI Institut for Vand amp Milj use If one file contains several time series the file name must be repeated for each record to be used Only three time series are visible but many more can be used in the calculation of statistics Use the Up and Down buttons to browse the list of specified time series Use the Insert and Delete buttons to make changes in the list For adding time series to the end of the list it is not necessary to use the Insert simply specify the new time series in the empty field at the end of the list Time series file names can either be typed using a path relative to the application directory or be selected using the TIME TO button A file selection box will be evoked by default showing the content of the time directory Selecting a TO file and pressing OK will close the file selection box and store the filename in a buffer Point to the field where you want the filename to be pasted and press the mouse select button R1 R2 R3 R4 criteria and tolerance The criterion is the fraction of observations in one time series that must satisfy the different formula R1 R4 The observations within the specified period are checked against the R1 R4 measures and the percentage of observation to fulfil the measures is calculated for each well These percentages are then checked again2. MIKE SHE PP User Manual Calibration Statistics Utility DHI Water amp Environment DHI Institut for Vand amp Milj CONTENTS DHI Software 2000 MIKE SHE PP Calibration Statistics Utility i DHI Water amp Environment DHI Institut for Vand amp Milj DHI Software 2000 ii MIKE SHE PP Calibration Statistics Utility DHI Water amp Environment DHI Institut for Vand amp Milj 1 GENERAL DESCRIPTION The calibration statistic utility is a tool tailored to produce performance statistics goodness or fit for a MIKE SHE simulation The main purpose is to calculate statistics on the residuals between observed and simulated data The current version enables the user to compare observed and simulated groundwater heads The present utility is developed under the project Developing a Small Scale Integrated Surface Water and Groundwater Model for the South Florida Hydrogeologic System The calibration utility was developed based on SFWMD traditions when calibrating a MODFLOW ground water model The main input is time series files TO files of observed groundwater heads These files must include information on the location for the observation A simulation result file including results from the saturated zone is required too The main output is statistical measures for each observation location and statistics on the number of locations to fulfil a number of criteria Output can opt
3. The basic input observed and simulated groundwater levels Observed head must be specified in time series files TO files and simulated heads are extracted from a simulation result file Including the information on the position of the individual observation locations together with the observed time series time series files type 3 see PP Manual enables the programme determine the corresponding location in the model thereby making comparison with simulated heads possible DHI Software 2000 MIKE SHE PP Calibration Statistics Utility 5 DHI Water amp Environment DHI Institut for Vand amp Milj DHI Software 2000 Catchment area T2 file Specify a T2 file delineating the area of interest Calculations are performed for observation wells placed within the area Grids included in the area should have a integer code and boundary grids should have a integer code 2 i e if the whole model area is of interest the catchment grid code file can be specified Select gives the opportunity to import a file from a list of files in the maps directory Use the file selection box to browse to other directories if needed Start and end date Specify here start and end date within which comparison between observed and simulated groundwater levels should be performed The period must be within the simulation period i e the start date should be equal or later than simulation start date and the end date must be equal to o
4. R4 Numerical criteria for goodness of fit Nash Sutcliffe coefficient Tolerance constants used for criteria R1 and R4 frac m Pri Pro Pr3 Pra Percentage of observation where R1 R4 is meet Cri Cro Cr3 Cra Success criteria for percentages Pr Prs Figure 1 Definitions DHI Software 2000 MIKE SHE PP Calibration Statistics Utility 3 DHI Water amp Environment DHI Institut for Vand amp Milj 3 USER S GUIDE 3 1 General DHI Software 2000 In every project the calibration and verification procedures are crucial to the accuracy and reliability of the model results The results of the calibration should be evaluated both qualitatively visual comparison and quantitatively mathematically In the beginning of a modelling study before any modelling is done it is useful and sound to assess the precision required by the model and to formulate calibration targets on that basis Calibration targets can seldom be considered strict success criteria but they should serve a quality measure In most model calibrations some observations will not meet the targets This is not necessary because the mathematical model is wrong It is however the modeller s responsibility to identify a justified explanation of the problem and to suggest means to overcome the problem and to assess the importance of this on the overall model performances The reason may be for instance errors in field data used for calibration or perhaps small scale
5. features that are not accounted for by the model A programme for calculating statistical measures is provided and the use of it is described in the following The utility is aimed for projects where time series of observed potential head constitutes a major calibration reference As described in the technical documentation the goodness of fit of simulated potential heads are checked for a number of criteria The criteria provides help when considering the dynamics of the ground water table with respect to both correct mean level and short and long term temporal variations In order to run the calibration statistics programme the menu U 10 Calibration Statistics must be accessed see It is a sub menu to the general utility menu menu U accessible from the main menu 4 MIKE SHE PP Calibration Statistics Utility DHI Water amp Environment DHI Institut for Vand amp Milj 3 2 Input menu U 10 Calibration Statistics Calibstat Catchment Area MAPStcatch TZ Select year month day hour min start date 1981 or oi od 1 end date 1995 01 or od og Flow Result File aarhus500 fr4 Select Read Result File Observations Timeseries TIME dgu_79568 t0 l Up Timef TO TIME dgu_88173 t0 lt gt Delete Down Criteria Tolerance Output type View Message file View Output file Execute Calibration Statistics Figure 2 Menu U 10 Calibration Statistics Utility
6. assess the spatial distribution of the various calibration criteria tmp R2 dig iof x FILETYPE DATATYPE VERNO 31 0 544 TEXTLINE Calibration utility output R2 UTM XYUNIT 0 i 3 580 754 6237 891 343 3 563 743 6220 882 750 3 575 346 6200 391 702 Figure 4 Example of dig output DHI Software 2000 MIKE SHE PP Calibration Statistics Utility 9 DHI Water amp Environment DHI Institut for Vand amp Milj The last output option is the generation of a plot macro which can be used in the Graphical Presentation tool The plot macro will be constructed by placing 6 figures on each page Due to limitations in the plotting programme this option only allows up to 50 time series If the plotting is activated through the Graphical Presentation menu the number of time series should be restricted to 15 only im C SHE540 BIN MSHE_GD EXE 124 116 2 TIME dgu_7 40 00 Figure 5 Plot made by using auto generated the plot macro DHI Software 2000 10 MIKE SHE PP Calibration Statistics Utility DHI Water amp Environment DHI Institut for Vand amp Milj
7. ionally be chosen as table scattered data dig or a plot macro plt 2 TECHNICAL DOCUMENTATION The calibration utility calculates the following basic statistical measures Symbols and notations can be found in the list in the end of this section The residual or difference between observed and simulated values for each observation RES FE Hsm is For each location time series the average residual and the average of the absolute residuals DHI Software 2000 MIKE SHE PP Calibration Statistics Utility 1 DHI Water amp Environment DHI Institut for Vand amp Milj DHI Software 2000 ARS RES n DIRES n RES Root mean square of the residuals JY RES RMS _ _ n Standard deviation on the residuals and on the observed data so EET n D H ossi Hogs j STDoss The Nash Sutcliffe coefficient is calculated as RS Louis Homa gt Hons H ops These statistical measures are checked against different criteria Four criteria have been implemented The R1 criterion ensures that the difference between residuals and the standard deviation on the residuals is kept within limits relative to the range of observed values Tr determines the acceptable range of deviation Ri RES lt Ta H H ogs min j STD OBS max j The R2 criterion ensures that the acceptable difference between observed and simulated values
8. is less than the standard deviation of observations R2 H STD ons Homar EE OBS i j OBS j SIM ji j OBS i j The R3 criterion is applied to avoid that simulated values drop below observed minimum or exceed maximum values 2 MIKE SHE PP Calibration Statistics Utility DHI Water amp Environment DHI Institut for Vand amp Milj R3 H ogs min j lt H sm lt H ops max j The R4 criterion provides an absolute measure of the maximum allowable difference between observed and simulated values The allowable difference Tra is specified in meters R4 H lt H T SIM i j OBS i j R4 OBS Ji j T pa lt H The R1 R4 criteria are checked at discrete observations within the simulation period Often field observations and simulation results are not available at exactly the same time and normally simulation results are stored with a higher frequency Simulated values are interpolated linear in time to calculate the residuals for a specific observation At each well location percentage of residuals to fulfil the four criteria is calculated These percentages are checked against specified global fit criteria to determine the number of observed time series that satisfies it Statistics on the number of observation time series satisfying the global fit criteria are determined Pr1 ra 1 Index for time i STD Standard deviation of RES Standard deviation of Hops j at location j R1 R2 R3
9. nd plot macros it is recommended to use suffixes dig and plt respectively to comply with other parts of the MIKE SHE system Execute Calibration Statistics Press this button to make the calculation of the calibration statistics View Message file Press this button to view the message file The message file will contain error messages if any View Output file Press this button to view the result file This is most relevant is output type is chosen to Table Output in a table will result in generation of a text file The file consists of header specifying information on which flow result file was used and some headlines Then follows a line for each time series specifying information about observation location and the statistical data about how well the simulation performs compared to the observations The statistical data listed are number of observation min and max observation mean residual RMS STD and the percentages of observations fitted by simulated values in accordance with the R1 R4 measures and in the end is found the R5 value Nash Sutcliffe Finally statistics on how many wells fulfilled the different criteria is written An example is found in 8 MIKE SHE PP Calibration Statistics Utility DHI Water amp Environment DHI Institut for Vand amp Milj If the statistics are not calculated a message is found No criteria calculated check input data Typically there are no observations in
10. r earlier than the simulation end date Flow Result File frf file Specify here the name of the flow result file The flow simulation must include storing of the potential head in the saturated zone datatype 15 Select gives the opportunity to choose a flow result file from a list of files in the res directory You can only select files within the res directory even though using the file selection box to browse to other directories seems to work the calculation of the calibration statistics will not work Read Result File Activating Read Result File will result in retrieving simulated potential head for grid cells corresponding to where observations are measured Retrieved data is placed in the tmp directory There will be placed one file for each observation location in the tmp directory containing the simulated time series of the potential head The names of these files will use the syntax Hi j k TO Where i j k refers to the calculation cell in the saturated zone corresponding to the location of the observation well Retrieving data can be time consuming and if different output type is chosen just after one another retrieving are only needed the first time Without Read Result File switched on information in the tmp directory is used Observation time series and record no Specify here filenames of the time series of observed groundwater levels Indicate by the record number what time series in the file to
11. side the specified period cstat tat WordPad Oy x File Edit View Insert Format Help oela SI al Eee Macro file calibstat txt Flow Result file aarhus500 frf mea Record Coordinates Celle Observations Residual dbs wrr x Yy ix iy iz No Min Max Mean RMS Mean STD RL RZ R3 R4 RS km km im im m im im im 4 4 4 i Il TINEVaga 7986800 1 580 754 6237 891 124 116 2 35 40 20 44 60 1 40 30 1 45 1 10 100 0 34 3 57 1 71 4 2 77 2 TIMEVdgu 88173 t0 1 563 743 6220 882 90 82 4 56 7 50 2 26 38 41 2 33 3 07 100 0 75 0 100 0 62 5 21 3 TIME dgu_108152 t0 1 575 346 6200 391 113 41 4 336 9 38 11 24 33 03 41 43 100 0 70 2 100 0 100 0 16 Ho of walisidnciided dacalsuacion R TT No of wells satisfying Rl criteria 3 No of wells satisfying R2 criteria 2 No of wells satisfying R3 criteria 3 No of wells satisfying R4 criteria 3 For Help press F1 INOM 4 Figure 3 Example of table output A number of dig files can be produced containing the location of the observation wells and a statistical number for each well The selection can be made between 9 number the percentage of observation to fulfilled by simulations according to R1 R2 R3 R4 criteria and residual mean and standard deviation of the residuals and root mean square of the residuals and the average of the absolute residuals and finally the Nash Sutcliffe number The digfiles can be overlayed on a grid plot or imported to ArcView to
12. st the criteria to summarise the number of observation wells where the criteria are meet The formulations of the different measures can be found in the technical documentation section For the R1 criterion a tolerance constant must be specified it is merely a multiplication number used to multiply on the difference between the maximum and the minimum head observation For the R4 criterion a tolerance constant must be specified The values is used directly in the R4 measure as an absolute measure of the maximum allowable difference between observed and simulated values Unit of the tolerance is m DHI Software 2000 MIKE SHE PP Calibration Statistics Utility 7 DHI Water amp Environment DHI Institut for Vand amp Milj 3 3 Output DHI Software 2000 Output type Select here the desired output type Table gives a text file containing calculated criteria for each observation well Results about R1 R4 are listed along with data of the residuals and R5 Nash Sutcliffe coefficient The different calculated data can be stored in dig files These files contain the x and y location of the observation wells and the calculated statistical data for each well An option for generating a plot macro for the graphical presentation programme is also available Due to limitations in the plot programme the number of time series should not exceed 50 Output file Enter here a file name of the output data file For dig files a
Download Pdf Manuals
Related Search