SWAT-CUP亚小时率定SWAT-CUP-Hourly_par

共7个文件

txt：2个

pdf：1个

def：1个

版权申诉

SWAT

水文模型

5星 · 超过95%的资源 46 浏览量 2024-07-22 09:36:59 上传评论收藏 1.12MB 7Z 举报

资源推荐

资源详情

资源评论

收起资源包目录

SWAT_CUP_Hourly.7z （7个子文件）

SWAT_CUP_Hourly

SUFI2_extract_rch_hrly.exe 531KB

Read Me.docx 221KB

SUFI2_extract_rch_hrly.def 1KB

SUFI2.OUT

FLOW_250.txt 18KB

SUFI2.IN

observed_hrly_rch.txt 26KB

fig.fig 153KB

swes-v09-i1-pp-21-27.pdf 853KB

Int. J. of Sustainable Water and Environmental Systems

Volume 9, No. 1 (2017) pp. 21-27

Calibration and Validation Of Swat For Sub-Hourly Time Steps Using

Swat-Cup

Sa’d Shannak*

Postdoctoral Researcher, Biological and Agricultural Engineering, Texas A&M AgriLife, Dallas, Texas

Abstract

SWAT is a semi-distributed, lumped parameter, continuous time model that simulates hydrology and water quality in

watersheds. Traditionally, the model operated at a daily time step and it estimated the influence of landuse and

management practices on water and agricultural chemical yields in a watershed. The daily time step format may not be

sufficient to capture the impact of flashy storms where peak flows last for minutes only and are not reflected in daily

average flows. A sub-hourly SWAT model for urban applications was developed but is not widely used. The main goal of

this study was to present a basic methodology to calibrate sub-hourly SWAT models using SWAT-CUP. SWAT was

tested using data from the Blunn Creek Watershed in Austin, Texas. The model was calibrated and evaluated using two

separate representative 2-year periods bracketing hydrologic conditions experienced in the watershed. Results show that

the sub-hourly SWAT provides reasonable estimates of stream flow for multiple storm events.

Keywords: Sub-hourly, SWAT, SWAT-CUP, SUFI-2, Uncertainty, Hydrological Modelling

1. Introduction

SWAT (Soil & Water Assessment Tool) is a river basin

scale model developed to quantify the impact of land

management practices in large, complex watersheds [1]. Since

SWAT is a distributed hydrological model, there are potentially

many parameters that can affect the stream flow assessment.

Investigating the potential impact of all these parameters can

be a very difficult task due to the high number of input

parameters at one of several different levels of detail:

watershed, subbasin, or HRU and which also highly interlinked

and interdependent on each other. Most of the available SWAT

studies use a daily time step format to assess hydrological

changes. [2,3,4]. A sub-hourly time step was developed and

released in 2010 [5]. The intention was to increase accuracy in

modeling single storm events, peak flows and provide essential

hydrologic metrics that maybe important predictors of stream

health and in studying the impact of low impact development.

A sub-hourly time step format accounts for high temporal

resolution that is needed in urban scenario analysis including

controlling for stormwater runoff, reducing potential flooding

and providing healthy environment for aquatic life. On the

other hand, sub-hourly time step model results in a more

complex model since it accounts for more details than daily or

monthly time steps, thus making the calibration-validation

process more complicated. The complexity of modern

hydrologic models requires narrowing model parameters down

to just those that have the greatest influence on the processes

being modeled. Sensitivity analysis is one method used before

the calibration process in order to study the variability in model

outputs with respect to changes in individual model

parameters. This type of analysis is considered essential in

order to determine the parameters that should be included in

the calibration process [6]. These methods also assist in

facilitating model evaluation in terms of the accuracy of the fit

of simulated data to measured or historical data using several

combinations of input parameters.

Most hydrological models must be calibrated so their

predictions can be used for tasks ranging from regulation to

research [7]. Distributed hydrological models often incorporate

inputs from numerous sources including weather, soils, land

use, surface water, groundwater and management practices.

Manual calibration depends heavily on the modeler adjusting

model parameters until the output match closely the measure

data. This can be difficult and time-consuming process due to

the complexity of some large scale models with many

objectives and the numerous interactions between these

objectives [8]

The process of calibration entails testing a model’s ability to

accurately simulate the behavior of the system of interest using

known inputs and uncertain parameters, and comparing the

outputs to observed data [7]. These parameters can be

Corresponding author.

E-mail: sshannok@tamu.edu

Shannak / Int. J. of Sustainable Water and Environmental Systems, 9 (2017) 21-27

quantified through direct measurement or in some cases, these

input parameters are unknown and must be estimated using the

current literature or through the calibration process itself. It is

important that every possible effort be made to minimize the

difference between simulated and measured data. This would

be essential to estimate high flows and potential flooding and

erosion in addition to assessing the impact of Best

Management Practices (BMPs) scenarios. Additionally, to

make informed decisions concerning remedial action or

environmental compliance, there should be a clear

demonstration that model simulations are reasonably

representative of the site being studied [9].

Therefore, utilizing an automated model such as SWAT-CUP, a

program designed to integrate various calibration/uncertainty

analysis algorithms into the SWAT model, rather than manual

calibration becomes very important to save time and achieve

higher accuracy. SWAT-CUP provides good visualization of

calibration parameters and allow the user to perform several

sensitivity, calibration and validation analyses of SWAT.

Model validation follows model calibration. The validation

process involves using observed or measured data of a different

time period than the calibration period and analyzing the

goodness-of-fit (estimation responses) and checking whether

the calibrated model’s predictive performance is in accordance

with observed/ measured data. The definition of sufficient

accuracy of the validation process can vary based on the use

and model’s goals [10].

The wider use of SWAT using sub-hourly data for more

accurate hydrologic modeling depends on a demonstrated

successful calibration and validation of SWAT and uncertainty

analysis. This research offers a framework for evaluating sub-

hourly SWAT models using data from the Blunn Creek

Watershed in Austin, Texas. The selection of the watershed was

based on various criteria including; climatic conditions and

potential flashy storms, current and future urban development

and finally availability of measured data.

Model and Calibration and Sensitivity Program

Description

SWAT

The hydrologic model used for this study was the SWAT 2012

model, a semi-distributed, lumped parameter, river basin scale,

continuous time step model developed to assess and predict

hydrological processes and changes in large basins [1]. This

model was developed to operate on a daily time step and it is

often used to estimate the impact of land use and management

on water and agricultural chemical yields in a watershed. The

major components of the model include hydrology, soil, land

management, plant growth, pesticides, nutrients, weather,

reservoir routing, and erosion.

A sub-daily time step SWAT model was initially developed and

added to SWAT 2005 in order to simulate rainfall-runoff

processes and realistically capture the long-term flow and

water quality trends in watersheds that are experiencing

urbanization [5]. It is worth mentioning at this point that

buildup/washoff cannot be done on daily basis to estimate first

flush water quality while it can be easily done using subhourly

time step models. Additional efforts are needed to improve low

flow predictions since low flows are dominated by base flow

and SWAT still utilizes soil water and ET estimation routines at

daily time steps [5].

SWAT-CUP

SWAT Calibration and Uncertainty Procedures (SWAT-CUP) is

a program designed to integrate various calibration/uncertainty

analysis algorithms into the SWAT model. These algorithms

include SUFI-2 [11], GLUE [12], and ParaSol [13]. SWAT-

CUP allows the user to select one of these algorithms and run

the procedure several times until convergence between

simulated and observed objectives is reached [14].

SUFI-2 is a multi-site semi-automated inverse modeling

routine [11]. Calibration in SUFI-2 is achieved when two rules,

termed the P-factor and the R-factor, are satisfied. The P-factor,

which ranges between 0 percent and 100 percent is the

percentage of measured data bracketed by the 95 percent

Prediction Uncertainty (95PPU). The R-factor, which ranges

between 0 and infinity, is the average width of the 95PPU band

divided by the standard deviation of the measured data. A Latin

Hypercube is used to sample from the distributions of each

uncertain parameter to create n parameter sets and n values for

each output. Then an objective function is calculated using the

simulated output and measured values for that same output. A

matrix that uses a relationship between a change in the

objective function that corresponds to a change in each

parameter value is generated to assess the simulations.

Goodness- of- fit is assessed and the process is done again,

narrowing the uncertainty bands on each parameter. A perfect

match between observed and simulated flows is indicated by

Nash-Sutcliff (NS) and coefficient of determination, R2, values

of 1. This would reflect a simulation where, based on

parameter uncertainty, 100 percent of the observed data fell

within the 95PPU; however, due to measurement errors and

conceptual model uncertainty, this is a rare occurrence. SUFI-2

starts by assuming a large parameter uncertainty that is within a

physically meaningful range, to ensure the measured data fall

within the 95PPU [11]. The model decreases this uncertainty

range gradually while monitoring the values of goal factors

(e.g. Objective functions such as NS) between the measured

and simulated data. Generally, model calibration can be

considered satisfactory if Nash Sutcliffe (NS) value > 0.50

[15]. The NS coefficient can be expressed as follows [16]:

)(

QoQo

QmQo





(Equation 2)

Where: Q

is observed values

is modeled values

is average observed values

SUFI-2 allows several iterations, and in each iteration previous

parameter ranges are updated by calculating the sensitivity

matrix and the equivalent of a Hessian matrix [17], followed by

the calculation of a covariance matrix, 95 percent confidence

intervals of the parameters, and a correlation matrix.

Parameters are updated so the new ranges are always smaller

than the previous iteration and continue until centered around

the best simulation [11]. In addition to being widely used in the

literature, SUFI-2 procedures were applied due to its easy

implementation in comparison to other procedures and the low

number of model runs needed to reach good prediction [18].

Methodology

SWAT Setup

The case study is based on research conducted in Austin, Texas

and specifically the Blunn Creek Watershed (Figure 1). The

watershed was estimated to have 34.8 percent impervious cover

in 2003 and the creek has a length of 4.82 km. The study area

is a rapidly urbanizing 2.58 km2 watershed with a total

population of 6,000 as of 2013 and is expected to increase by

12 percent in less than 10 years [19]. Part of the selection of

this model is driven by plans by the City of Austin to evaluate

the impact of low impact development structures on stream

health in all watersheds within city limits. Very few studies

were done in urban environments using SWAT although current

Shannak / Int. J. of Sustainable Water and Environmental Systems, 9 (2017) 21-27

effort in SWAT development is targeting urban applications

[20].

Figure 1. Blunn Creek Watershed in Austin, Texas

SWAT was run using 15-minute rainfall data from the Flood

Early Warning System (FEWS) and Water Quality Monitoring

(WQM) sections at COA

(http://www.austintexas.gov/department/flood-early-warning-

system), sub-hourly temperature data from the Austin and

Austin-Bergstrom NOAA weather stations

(WGEN_US_COOP_1960_2010), a 3-meter integer Digital

Elevation Model (DEM) developed by COA based on 2003

LIDAR data and SSURGO soils data from NRCS. A landuse

raster layer develop by COA was used. Geometry of the

channels for each sub-basin was modified after conducting a

cross section analysis for the DEM layer. The cross section

analysis was done by converting the DEM layer using an

interpolation line tool under 3D Analyst menu in ArcGIS,

creating a profile graph (Figure 2) and calculating the

dimensions of an equivalent trapezoidal cross-section [21]. The

data was then included in SWAT’s main channel input file

(.rte).

Figure 2. Screenshot of Cross section analysis for the DEM

layer using 3D Analyst tool in ArcGIS that can be directly

used in ArcSWAT.

SWAT-CUP Setup

A sub-hourly 15minute time step SWAT model was run through

the ArcGIS environment. Because the default output is in a

daily time step, the output file was modified to include sub-

hourly output. It is worth noting that SWAT-CUP version

5.1.4.2 used in this study is designed for daily, monthly and

yearly time steps and while it was assumed that sub-hourly

input data can be read and processed as well, minor

modification to the fig.fig file located in the TXINOUT folder

in a SWAT project by including a new saveconc command for

sub-hourly output. This command saves flow, sediment and

water quality data from a specified point to a file in SWAT that

is directly used by SWAT-CUP. A case study was designed and

SWAT-CUP was applied using SUFI-2 calibration and

uncertainty parameters.

Calibration Parameter Selection

The primary parameters responsible for streamflow simulation

for the Blunn Creek Watershed are shown in Table 1.

Table 1. Initial calibration parameters given by SWAT-

CUP

Parameter name Description

r__CN2.mgt Initial SCS runoff curve number for

moisture condition II

v__ALPHA_BF.gw Baseflow alpha factor (1/day)

v__GW_DELAY.gw Groundwater delay (days)

v__GWQMN.gw Threshold depth of water in the

shallow aquifer required for return

flow to occur (mm)

v__GW_REVAP.gw Threshold depth of water in the

shallow aquifer for "revap" to occur

v__ESCO.hru Soil evaporation compensation factor

v__CH_N2.rte Manning roughness for main channel

v__CH_K2.rte Effective hydraulic conductivity

v__ALPHA_BNK.rte Baseflow alpha factor for bank storage

r__SOL_AWC(1).sol Available water capacity of the soil

layer (mm H2O/ mm soil)

r__SOL_K(1).sol Saturated hydraulic conductivity

(mm/hr)

r__SOL_BD(1).sol Moist bulk density (g/cm

)

v__SFTMP.bsn Snowfall temperature (

Observed stream discharges at USGS Station 08157700 were

retrieved in 15-minute format from USGS (USGS, 2014). The

stream data was only available for the two-year period of 1998-

1999 which was used for calibration and for the 2-year period

of 2001-2002 which was used for validation purposes. A 24-

year simulation from 1986 to 2010, using the weather data

from COA, was executed for the sensitivity analysis, excluding

outputs from the first two years that were used as a warmup

period for antecedent conditions. The minimum and maximum

ranges suggested by the SWAT-CUP analysis were used to

define the parameters. The calibration and validation periods

were representative typical flow regime for this creek

according to COA staff.

Uncertainty Analysis

In this study, an uncertainty analysis was performed after each

calibrating procedures based on the ranking results provided by

statistical measurements such as; P-factor, R2 and NS. Default

values suggested by van Griensven et al. [22] were chosen as

minimum and maximum ranges for the respective model

parameters (Table 1). The objective function type was selected

to be NS coefficient with 0.5 minimum value of objective

function threshold to separate the behavioral solutions (with

NS better than 0.5) from the non-behavioral ones (with NS

lower than 0.5). The effect of using behavioral solutions is to

评论收藏

内容反馈

版权申诉

2301_77140588

2024-08-27

总算找到了想要的资源，搞定遇到的大问题，赞赞赞！

KmBase

粉丝: 1w+
资源: 38

SWAT-CUP亚小时率定 SWAT-CUP-Hourly

最新资源

SWAT-CUP亚小时率定 SWAT-CUP-Hourly

Swatcup简单使用说明_SWAT-cup_swatcup说明书_SWAT_swatcuppso_SWAT-CUP调参

SWAT-CUP2012手册简化版

SWAT-CUP-Hourly 示例TxtInOut文件及cup配置，可以成功运行swat及率定

SWAT-cup的使用说明

任务63：基于SWAT-CUP的模型率定.mp4

Swatcup简单使用说明,swatcup参数怎么带回swat,Fortran

SWAT-CUP参数辅助修改小程序

SWAT-CUP校准软件及其说明

SWAT-CUP-setup-Dec2016-5.1.6.2

Usermanual_Swat_Cup

SWAT-ArcGIS开发与应用-(1).docx

SWAT-MODFLOW.rar

Swatcup简单使用说明.pdf

SWAT-CUP-setup-Dec2016-5.1.6.2.zip

Swat_Cup 说明

SWAT_CUP使用说明

PyPI 官网下载 | swat-em-0.1.3.tar.gz

SWAT-EM:电机专用绕组分析工具-开源

期刊论文《SWAT模型参数敏感性分析及应用》

samba-swat-3.0.33-3.39.el5_8.i386.rpm

SWAT软件安装程序

Python库 | swat-1.9.2-0-cp27-cp27m-manylinux2014_ppc64le.whl

PyPI 官网下载 | swat-1.7.1-0-cp38-cp38-manylinux2014_ppc64le.whl

PyPI 官网下载 | swat-1.8.0-0-cp37-cp37m-manylinux1_x86_64.whl

论文研究 - 在SWAT模型中使用三个N循环选项模拟的小麦产量的比较

最新资源