Modeling Rainfall-runoff using SWAT model in data scarce area: in - - PowerPoint PPT Presentation

Modeling Rainfall-runoff using SWAT model in data scarce area: in the Weybo River Catchment of southwestern Ethiopia

By Mathewos Muke (MSc : in Geo -Info & Earth Obs. Science)

Esri Eastern Africa Education GIS Conference Creating a World of Opportunities

Presentation outlines

1. Introduction 2. Methods and Materials 3. Results and Discussions 4. Conclusions

Introductions

 The main slogan of Ethiopia in the new millennium is

to pull the country out of pulverize poverty and leading the country to development.

 Several

indicators

• f

poverty with focus

• n

alleviations are related to the water sector (EAH, 2007 and Tesfamichael, 2009).

 But this sector is challenged or disturbed by climate

change (Bronstert et al., 2002; Kim et al., 2011 and Ayalew (2012)) and the impact is happening now and will be expected for the future.

Cont…

 Because of the global warming, the water cycle

will be

accelerated, shifting seasonal patterns and increasing

extreme events in different regions.

 Land management problem in addition to climate

change also brings significant effect (LI and Ishidaira, 2010 and Homdee et al., 2011)

Cont…

 Land use land cover changes (LULC) on the

• ther hand can bring variation on flow or

increase runoff because of the conversion of particular kind of fram land from one type to

• ther (Menzel et al., 2009; LI and ISHIDAIRA,

2010; Homde et al., 2011; Rientjes et al., 2011).

 Run off variability not only attributed to the

above case only but also associated with catchment level mainly catchment size.

Cont…

 Assessing the relationship of the spatial and

temporal distribution of rainfall with runoff and

• bserving its future trends are fundamental input

parameters for securing sustainable agricultural production as Ethiopia in general and the area under study in particular depends on rain-fed agriculture.

 To this end, therefore; the SWAT model was

applied to model the amount of water converted to runoff as a result of biophysical controlling factors dynamics in the catchment.

Cont…

 The overall objective of this study was to model spatial

and temporal variations of rainfall runoff relationship in data scarce area of Weybo River catchment. And specifically:

• to identify the catchment’s sensitive parameters
• to determine the ability of SWAT to closely

simulate daily surface flow from excess precipitation.

• to asses rainfall-runoff relationships in Weybo river

catchments

Location



The Figure 3-1 depicts the location of the basin in general and the Weybo catchment (538.3km2).

Figure 3-1 Location Map of the study area

 Physiographically, the study area is located within western rift margin at the NW

• f mount Damota with the highest and lowest elevation of 2946m and 845m.a.s.l

respectively.  In general, the average elevation of the area is 1879m and its standard deviation is +193 and so that, the area can be described as less ruggedness in its physiographic setting.

Figure –shows the relief of the area

Methods

Data collection

Both primary and secondary data were used for this research studies.

 In primary data, GCPs by using GPS were collected for

identification of outlet of the catchment, location and elevation

• f the respective meteorological stations.

Where as,

 Secondary data like:

• long period of recorded metrological and discharge data,
• soil and land use land cover data and
• digital elevation model were collected from the respective offices

and websites.

Data setting method

Figure : schematic diagram that show the

• verall

data sources of the model and its process

 All the above inputs (Raster, shapefiles and database file formats) for

model were prepared as SWAT dataset in a manner which is compatible to ArcGIS. NB:

• Spatial input data sets and
• ArcSWAT data bases or
• lookup tables for each inputs

 Unlike to other hydrological models, SWAT helps to estimate the

missing data.

 So that, a negative 99.0 (-99.0) value was inserted for missing

value since this value tells SWAT to generate meteorological data for that day.

were prepared based on the procedure recommended by Neitsch et al. (2002).

Sensitivity analysis

 For

the selection

• f

hydrologic parameters, the sensitivity analysis was carried out for the period of 11 years so as to identify the parameters that have the greatest influence on model results.

 The years include both:

• the warm-up period (Jan. 1st 1992 – Dec 31, 1994) and
• the calibration period (Jan 1st, 1995-Dec 31, 2002) with

keeping out validation year (4).

 Summary statistics are usually applied as a measure of

how well the model hydrograph fits, or agrees with, the

• bserved hydrograph.

 A wide range of statistics like Coefficient of determination

(R2), the Nash-Sutcliffe model efficiency (NSE) coefficient and the Relative Volume error (RVe) has been used to evaluate SWAT hydrologic predictions.

 

1 . 3 , , , ,

2 1 1 2 2 1 2

                                                 

  

       n i n i n i

S i Si O i O S i Si O i O R

 

 

2 . 3 , , ,

1 2 1 2

1          

  

         

n i n i

O i O i O i Si

NSE

 

3 . 3 . 100 , ,

1 1 1

                           

  

  

x

• i

i

• i

si RVe

n i n i n i

Cont…

• 3. RESULTS AND DISCUSSION

Rainfall

 As indicated in Figure 4-1, the graph reveals that the area receives a bi-modal rainfall

distribution which extends over the period of April to September with its peak in April and August.

 Total amount of annual rainfall and distribution influence the nature of surface runoff and

volume of its flow.

 The highest rainfall of the area is recorded in April, July, August and September whereas

the minimum rainfall is exhibited in November, December, January and February.

1 2 3 4 5 6 7 8 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Rainfall (mm) Months

Figure 4-1 Graphical representation of the mean monthly rainfall from 1992-2006

5 10 15 20 25 30 35 1/1/1992 5/1/1992 9/1/1992 1/1/1993 5/1/1993 9/1/1993 1/1/1994 5/1/1994 9/1/1994 1/1/1995 5/1/1995 9/1/1995 1/1/1996 5/1/1996 9/1/1996 1/1/1997 5/1/1997 9/1/1997 1/1/1998 5/1/1998 9/1/1998 1/1/1999 5/1/1999 9/1/1999 1/1/2000 5/1/2000 9/1/2000 1/1/2001 5/1/2001 9/1/2001 1/1/2002 5/1/2002 9/1/2002 1/1/2003 5/1/2003 9/1/2003 1/1/2004 5/1/2004 9/1/2004 1/1/2005 5/1/2005 9/1/2005 1/1/2006 5/1/2006 9/1/2006 Discharge (m-3/s)

Years

Figure 4-2 Daily time series graphical representation of observed discharge value (1992-2006)

The figure here reveals, its natural trends and its hydrological pattern in the catchment by making use of daily based discharge values for its plotting.

 The hydrograph below shows a fluctuating stream pattern in response to the

rainfall .

 They start to rise from April to reach the maximum in August (rainy season)

and shows a decreasing trend in flow during dry season, which continuous to achieve its minimum in December.

Figure 4-3 Daily time series comparison of observed discharge and Rainfall (1992-2006)

Years

20 40 60 80 10 20 30 40 1/1/1992 7/1/1992 1/1/1993 7/1/1993 1/1/1994 7/1/1994 1/1/1995 7/1/1995 1/1/1996 7/1/1996 1/1/1997 7/1/1997 1/1/1998 7/1/1998 1/1/1999 7/1/1999 1/1/2000 7/1/2000 1/1/2001 7/1/2001 1/1/2002 7/1/2002 1/1/2003 7/1/2003 1/1/2004 7/1/2004 1/1/2005 7/1/2005 1/1/2006 7/1/2006 Discharge (m3/s) Rainfall (mm/day) Discharge (m-3/s) Rainfall (mm)

Sensitivity Analysis

From the total stream flow parameters, the table below

shows the top five ranked parameters based on their degree of sensitivity.

Parameters Definitions Rank Mean Surlag Surface runoff lag coefficient 1 6.94E-01 Alpha_Bf Baseflow alpha factor (days) 2 4.86E-02 Cn2 Initial SCS runoff Curve Number for moisture condition 2 3 2.75E-02 Ch_K2 channel effective hydraulic conductivity (mm/hr) 4 1.17E-02 Ch_N2 Manning coefficient for channel 5 1.07E-02 Table 4-3 Top 5 sensitive parameters applied for optimization

 To calibrate the model, nine parameters were varied individually

within the SWAT model (Table 4-4).

 Of course, variations of other parameters were also attempted but

variations in them showed no significant effect on daily streamflow simulations for this watershed.

Parameters Minimum Maximum Optimized Value Process Surlag 150 1 Surface runoff Alpha_Bf 1 0.05 Groundwater Cn2 35 98 35 Surface runoff Ch_K2

• 0.01

500 Pertain to peak flow Ch_N2 0.01 0.3 0.01 Pertain to peak flow Additional parameters OV_N 0.01 30 30 Surface runoff Sol_K 2000 30 Groundwater Slsubbsn 10 150 100 Surface runoff HRU_SLP 0.6 0.1 Surface runoff

Table 4-4 Default values, optimized parameter values and process

So that, the model efficiency after calibration for daily

runoff simulation of the Weybo catchment shows:

• R2=0.68 and
• NSE= 0.54
• RVe 3.2%

for the optimization period (1995-2002)

 After the transfer of the parameter estimates from calibration to

validation, there was a variation in model efficiency i.e.

• 0.68

0.672 for R2

• 0.54

0.53 for NSE

• 3.2%

4.5% for Rve

 Moreover, the visual inspection reveals that the simulated runoff

not well fitted to observed flow in some situations as indicated in circle and there are also some outliers that not equally competed to each other as shown by arrow.

Figure 4-4 Daily time series comparison of simulated Vs observed flow data for calibration and validation period (m3/s)

Taken as a whole, simulated peak flow fall within the magnitude of the measured peak flows.

5 10 15 20 25 30 35 1/1/1995 5/1/1995 9/1/1995 1/1/1996 5/1/1996 9/1/1996 1/1/1997 5/1/1997 9/1/1997 1/1/1998 5/1/1998 9/1/1998 1/1/1999 5/1/1999 9/1/1999 1/1/2000 5/1/2000 9/1/2000 1/1/2001 5/1/2001 9/1/2001 1/1/2002 5/1/2002 9/1/2002 1/1/2003 5/1/2003 9/1/2003 1/1/2004 5/1/2004 9/1/2004 1/1/2005 5/1/2005 9/1/2005 1/1/2006 5/1/2006 9/1/2006 Discharge m3/s

Years

Sim Observ

Validation Calibration

 The simulation of runoff and comparing it alongside to rainfall in a watershed

provides insight to the processes that affect runoff.

 The standard linear regression statistics value shows that,

• its significance P value is 0.00 and
• the R squared value is 0.07 and

20 40 60 80 10 20 30 40 1/1/1995 5/1/1995 9/1/1995 1/1/1996 5/1/1996 9/1/1996 1/1/1997 5/1/1997 9/1/1997 1/1/1998 5/1/1998 9/1/1998 1/1/1999 5/1/1999 9/1/1999 1/1/2000 5/1/2000 9/1/2000 1/1/2001 5/1/2001 9/1/2001 1/1/2002 5/1/2002 9/1/2002 1/1/2003 5/1/2003 9/1/2003 1/1/2004 5/1/2004 9/1/2004 1/1/2005 5/1/2005 9/1/2005 1/1/2006 5/1/2006 9/1/2006 Discharge (m3/s) Years

Rainfall (mm)

Simulation (m3/s) Rainfall (mm)

Figure 4-5 Graphical representation of rainfall and runoff simulation relationship

A storm hydrograph as a form of peak flow in some instances reveals how a simulated runoff responds following a period of heavy rainfall.

this indicate that there is a significant correlation and have similar seasonal pattern in most cases between the two.

Conclusions

 The successful calibration of the model indicates that

with similar parameters, SWAT2005 model can be calibrated on different river basins in sub-humid to semi-arid climatic zones and the result also can be used for the parameterization and hydrological analysis of ungaged catchments in the other rivers.

 The statistical model efficiency assessment for both

• ptimization and validation period verifies that there is

a prospect for its wide applications to compensate the lack of measurements in the region.

Questions and comments??