README.md

In mtalebizadeh/swatsens:

swatsens quick start manual

Introduction

swatsens is a package in R for performing sensitivity and model performance analysis for the SWAT model.The package is part of land and water resources management studies at USDA-ARS Grazinglands Research Laboratory.

How it works

1) swatsens automatically extracts SWAT model information from TextInOut folder. To achieve this, users are required to make a backup of TextInOut project (see "backUpTextInOutPath" in SA section) and provide its path along with other set of user-supplied inputs.

2) After setting the inputs, "genParmRange" function with "backUpTextInOutPath" as its argument is called to generate a list containing empty ranges of SWAT model parameters. Using this generated list, user can set uncertainty range for different SWAT model parameters.

3) Once, the range of parameter uncertainties are set, "createGsaObject" function is called to create a GSA list containing samples from the ranges of uncertain parameters and other sensitivity analysis (SA) settings.

4) "runMonteCarlo" function is called to run Monte Carlo simulation using the generated samples and save the timeseries of different SWAT outputs, located inside SWATS's "output.sub" file, as well as a performance measure (current version of swatsens only supports "mean" as performance measure).

At last, the program simply loops through performance measure files and SA coeeficients are calculated and save inside "gsa_output" folder.

Steps for performing SA

# 1) Setting required inputs:
     # original TextInout (used for reading SWAT parameters)
     backUpTextInOutPath <- "D:/backUpTxtInOut"

     # TextInOut Folder (used for performing MC simulations)
     textInOutPath <- "D:/TxtInOut"

     # SWAT executable file inside name TextInOut folder
     exe_fileName = "rev670_64rel.exe"

     # Path to output.sub file inside TextInOut folder
     outputSubFile= "D:/TxtInOut/output.sub"

     # Output folders where performance metric (i.e. mean), SA coeffs, and timeseries
     # are stored inside pmfs, gsa_outputs, and subs are stored, respectively
     pmfFolderPath = "D:/outputs/pmfs"
     folder_path_GSA_Outputs <- "D:/outputs/gsa_output"
     subbasinFolderPath = "D:/outputs/subs"

     # Beginning year of simulation as it appears inside "file.cio"
     # Please make sure "IDAF" and "IDAL" values are set to 0 inside "file.cio"  
     startDate="2003-01-01"
     outputVars = c("PET", "PRECIP", "ORGN", "SYLD")
     gsaType <- "src" # Available methods are: "src", "srrc", "morris", "fast"
     n = 1000         # Number of generated samples

# 2) Generating range of parametrs and setting relative uncertainty ranges
     parmRange <- genParmRange(backUpTextInOutPath)

     parmRange$gw$GW_DELAY$Idx_1_1$low = -0.02 # i.e., lower uncertainty: default*(1 - 0.02)
     parmRange$gw$GW_DELAY$Idx_1_1$up = 0.02 # i.e., upper uncertainty: default*(1 + 0.02)

     parmRange$hru$SLSUBBSN$Idx_1_1$low = 0.1
     parmRange$hru$SLSUBBSN$Idx_1_1$up = 0.3

     parmRange$bsn$P_UPDIS$low = -0.02
     parmRange$bsn$P_UPDIS$up = 0.02

     parmRange$wwq$RHOQ$low = -0.2
     parmRange$wwq$RHOQ$up = 0.2

     parmRange$plant$FRST$BIO_E$low = -0.2
     parmRange$plant$FRST$BIO_E$up = 0.2

# 3) Create GSA object
     GSA <- swatsens::createGsaObject(parmRange = parmRange,
                       gsaType = gsaType,
                       rangeType = "rel", n = n)

# 4) Run Monte Carlo simulations and save model outputs and performance measure
     swatsens::runMonteCarlo(parmDataFrame = GSA$X,
              projectPath = textInOutPath,
              outputSubFile= outputSubFile,
              startDate=startDate,
              skip=8,
              outputVars = outputVars,
              subbasinFolderPath = subbasinFolderPath,
              pmfFolderPath = pmfFolderPath,
              exe_fileName = exe_fileName)

# 5) Calculating SA coefficients for different variables inside "output.sub" file  
  for (f in list.files(pmfFolderPath) ) {

  y = read.table(paste(pmfFolderPath, "/", f, sep=""), header = TRUE)
  SACoeffs <- swatsens::caclulateSACoeffs(GSA_Object = GSA,
                                Y = y,
                                GSA_Type = gsaType)
  print(f)
  write.table(SACoeffs,
              paste(folder_path_GSA_Outputs,
                    "/", "sub_",swatsens::getSubbasinNumberFromFileName(f),
                    "_", names(y),
                    ".txt", sep = ""),
              sep = "\t")
}

mtalebizadeh/swatsens documentation built on Nov. 13, 2020, 12:18 p.m.