Download as pdf or txt
Download as pdf or txt
You are on page 1of 22

Impact of land cover change on hydrological process and sediment yield, A

case of Jedeb and Chemoga watershed, upper Blue Nile basin, Ethiopia

Shegaw Y. Birhanu 1, Mamaru A. Moges4, Berhanu G. Sinshaw 2,3,*, Agumase K. Tefera


2,3 Haimanot B. Atinkut5, Habtamu M. Fenta5, Mulatu L. Berihun3

1,2Yilmana Densa Wereda Water, Irrigation and Energy office, Adet, Ethiopia P.O.Box 84 Adet,
Ethiopia
2Faculty of Civil and Water Resource Engineering, BiT, Bahir Dar University, 26 Bahir Dar,
Ethiopia
3School of Civil and Water Resource Engineering, University of Gondar, Ethiopia, Gondar,
Ethiopia
4Water, Irrigation, and Energy Bureau, Amhara Regional State, Bahir Dar, Ethiopia
5College of Agriculture and Environmental Sciences, University of Gondar, Gondar 196,
Ethiopia

*Corresponding Author

Berhanu G. Sinshaw

School of Civil and water Resource Engineering, University of Gondar, Ethiopia

P.O.Box 196 Gondar, Ethiopia

Tel : +251932957129

Email: berhanugeremew0@gmail.com

Electronic copy available at: https://ssrn.com/abstract=3981555


Impact of land cover change on hydrological process and sediment yield,
Case study in the upper Blue Nile basin, Ethiopia

Abstract

Land use and land cover (LULC) changes in the Jedeb and Chemoga watersheds have been detected

in the past 29 years. The study used ERDAS IMAGINE 2015 tools to classify landsat5TM, landsat7 ETM+

and landsat OLI / TIR satellite images for the preparation of 1990, 2000, 2009 and 2018 year LULC maps

using the highest likelihood of supervised classification techniques. The SWAT model has been used to

simulate sediment and stream flow to determine the seasonal variability in stream flow and sediment yield

due to land use and land cover changes in the watersheds. To calibrate and validate stream flow and sediment

yields and to test SWAT model performance, the SWAT-CUP SUFI 2 algorithm was used. The outcome

showed that 29 years generalized shift of 62.3 % increase in farm land, 57.8 % decrease in forest cover, 52.4

% decrease in grass land, 64.2 % decrease in shrub land, 145.2 % increase in bare land and 46.5 percent

increase in Jedeb watershed settlement. Similarly, the analysis showed that 47.3 % increase in farm land,

81.8 % decrease in forest cover, 148.9 % increase in grass land, 89 % decrease in shrub land, 10.7 % increase

in bare land and 96.9 % decrease in Chemoga watershed water body. The average monthly flow of Jedeb

watershed flows increased by 39.31 m3 / s from 1990 to 2009 and decreased by 17.19 m3 / s from 2009 to

2018. In Chemoga watershed average monthly stream flow increase between 1990 to 2009 in 36.73 m3 / s

but decrease between 2009 to 2018 in 37.77 m3 / s. The average monthly Jedeb watershed sediment decreased

by 204.59 ton / month from 1990 to 2000, increased by 15,962.76 ton / month from 2000 to 2009 and

ultimately decreased by 9,012.83 ton / month from 2009 to 2018. Average monthly sediment rises in

Chemoga watershed from 1990 to 2000 at 13,618.72 ton / month, decreased by 11,316.6 ton / month from

2000 to 2009 but increased by 11,192.93 ton / month from 2009 to 2018. The result indicates a strong

agreement between observed ad simulated stream flow and sediment yield during calibration and validation,

with a minimum result of R2=0.68 and NSE=0.54 and a maximum of R2=0.88 and NSE=0.84. The SWAT

model therefore performs well in estimating stream flow and sediment yield in the watersheds of Jedeb and

Chemoga.

Keywords: Chemoga and Jedeb watershed; LULC, sediment Yield; Stream flow; SWAT

Electronic copy available at: https://ssrn.com/abstract=3981555


1. Introduction

1.1. Background
Particularly in developing countries with an agriculture-based economy and increasingly growing

populations, LULCs are highly changed because the demand for land increases as the population increases.

Ethiopia's main industry is agriculture, which relies on the availability of seasonal rainfall and water supplies,

and 85 percent of the population is primarily involved in agriculture. In Ethiopia, the complex existence of

land use resulting from growing populations, agricultural sector expansion and climate change is worrying.

Humans have made large-scale changes to the terrestrial biosphere, mainly by modifying LULC, expanding

and intensifying agriculture, rising urban areas and extracting timber and other natural resources, which are

likely to increase over the coming decades to meet the demands of a growing population (Demisse, 2011).

The major problems and most important threats in the Ethiopian highlands are arable land depletion

and water supply depletion. The deterioration and loss of soil resulting from soil erosion, the transport of

sediments and the depletion of organic matter and agricultural nutrients is significantly exacerbated by poor

land use practices characterized by a general lack of conservation measures and population, resulting in

intensive fragmented land agriculture, deforestation and overgrazing (Hari, 2014). Changes in land use

undoubtedly have a direct effect on the availability of water and agricultural production within catchments,

thereby impacting global water yield. It directly affects the quantity of evaporation, infiltration of

groundwater and overland runoff that takes place during and after precipitation events. Land cover and usage

changes affect both the runoff behavior and the balance between evaporation, groundwater recharge and

discharge of streams in particular areas and in entire watersheds, with a major effect on the quantity of water

available for both ecosystem and human use (Lee, 2013).

Bare land expansion, increased development of surface runoff and soil degradation are major

environmental harms to local agricultural efficiency, the availability of water supplies and community food

security. (Berhanu et al., 2017). Land depletion has been one of the big problems impacting land production

in the highlands. Intensive steep slope agriculture without appropriate soil conservation measures has resulted

in soil impoverishment due to soil erosion and, in some cases, the complete loss of agricultural land due to

the development of gulls. Investment in soil conservation has been further discouraged by land tenure

regimes, inadequate extension methods, and declining farm incomes, whereas crop pests and diseases have

Electronic copy available at: https://ssrn.com/abstract=3981555


increased incidences of crop intensification. The spiral of land destruction, reduced production, reduced farm

incomes, and mining of the land resources has reduced the once prosperous highland communities to poverty

and food insecurity (Amede et al., 2006).

Watershed features may lead to hydrological response differences. Certain features affecting the

hydrological regime include unique characteristics such as soil properties, geology, watershed scale, local

climate , topography, anthropogenic activity and vegetation cover (Sivapalan, 2005). Whether the type of

land use is positively or negatively related to a particular hydrological attribute, such as total flow, high flow

and low flow or runoff coefficients, helps to describe the identification of watershed characteristics and their

effect on the hydrological regime. In places like the Blue Nile, where the effect of land use on various flow

regimes is uncertain, the characterization of the hydrological response of watersheds is critical. In the Blue

Nile basin, however, the spatial variations in land use and other watershed characteristics related to water

scarcity are not well defined (Gebrehiwot, 2012). The understanding of how LULC changes impact

watershed hydrology will enable local governments and policymakers to devise and enforce efficient and

acceptable response plans to reduce the adverse effects of potential LULC changes or improvements. (FAO,

2002).

Hydrological modeling and management studies of water supplies are closely linked to the

hydrological cycle 's spatial processes. The relationship between LULC and the hydrological cycle should

therefore be well known. Priority areas of research should be given to having a scientific understanding of

the land use and land cover change process, the impacts of various land use decisions, and how decisions are

influenced by the hydrological cycle and increasing variability (Abraha et al., 2011). On this basis, the main

objective of the study is to determine the effects of changes in land cover on the stream flow and sediment

yield of the watersheds of Jedeb and Chemoga.

Electronic copy available at: https://ssrn.com/abstract=3981555


2. Material and Methods
2.1. Study Site
The research in Chemoga covering a watershed area of 346.7 km2 and Jedeb covering a watershed

area of 296 km2 in the Upper Blue Nile was carried out. Both watersheds source at an elevation of around

4000 m a.s.l. from Choke Mountain and flow to the basin of the Abbay / Blue Nile river. Administratively,

five districts / weredas, such as Machakel, Gozamn, Debay Telatign, Bibugn and Awabel, constitute the

watersheds located in the East Gojjam administration zone of Amhara National Regional State , North

Western Ethiopia. They are found about almost 20 km from Debremarkos and around 300 to 320 km from

North West of Addis Ababa. Jedeb and Chemoga are located between 37°42'0" to 37°52'30"E longitude and

10°10'0” to 10° 40'0"N latitude.

Figure 1 Here

Soil type: Eutric Leptosols, Haplic Alisols and Haplic Luvisols were the main types of soil groups in the

study field (Tekleab et al., 2015). Haplic Alisols, which are a deep soil with silty-clay texture, are also known

the soil types in these watershed; Haplic Luvisols, well-drained soil with clay to silty clay texture; and Eutric

Leptosols (moderately deep soil), with clay loam to clay texture. But, due to erosion issues, much of the soil

is becoming thin under different land use.

2.2. Materials
The materials used were listed below in (Tables 2 and 3) .

Table 1. The different sources of data used for this analysis

No Types of Data Source Purpose/usage


1 Stream flow and sediment Abbay basin authority Calibration and validation

2 Satellite image https://earthexplorer.usgs.gov LULC map


3 Soil MoWRIE Soil map
4 STRM DEM https://earthexplorer.usgs.gov/ Watershed delineation
5 Meteorological data meteorological stations Inputs for SWAT model
simulation

Electronic copy available at: https://ssrn.com/abstract=3981555


Table 2. Various decision support methods used in this analysis and their uses

No Materials Version Purpose


1 ARC GIS 10.1 data processing and map preparation
2 SWAT 12 simulation
3 SWAT-CUP 5.1.6 Sensitivity analysis, calibration and validation

4 ERDAS IMAGINE 2015 LULC classification and accuracy assessment.


5 GPS GARMMIN Reading outlet and ground truth point
6 EndNote X7 Citation and Reference
7 MS EXCEL 2007 Statistical analysis, Chart and graphs
8 XLSTAT 2014 Data quality test
9 Weather database. v01803 weather generator preparation
10 Surfer 16 coordinate system conversion
11 R studio 3.4.3 Purpose for Plotting of graph
12 Google earth pro 7.2.3 Investigation of study area
13 Minitab 16 Data analysis /staking of data in one column)

2.3. Image classification and accuracy assessment


ERDAS IMAGINE 2015 applications for land cover classification and accuracy evaluation using

distinct landsat images was used in the study. Next, the selection of the training sites that are usually

representative of the classes for land cover. The training sites were collected based on personal experience,

expertise and existing research of the physiographical of the region. Furthermore, to help distinguish the land

cover groups, image enhancement and composition were applied. Using these approaches around 150 and

above training sites were collected from each year land use and land cover data (1990, 2000, 2009 and 2018)

with a proportion of above 30 samples for each land use types. Using ERDAS IMAGINE 15, an accuracy

assessment was done for the classified land use land cover picture. 150 points for each categorized land use

land cover image were created randomly from the classifier. The user defined all the randomly generated

points and allocated them to various classes. For the 4 supervised classification photos (i.e. 1990, 2000, 2009

and 2018) of each watershed, this process was carried out. An Error matrix and Kappa statistics were also

created from the ERDAS Imagine 15 software report section 's reference and classified data.

Electronic copy available at: https://ssrn.com/abstract=3981555


2.4. Soil and Water Assessment Tool (SWAT)models
It is a semi-distributed physical basin scale model that uses various data for hydrological and

climatological modeling on a daily or monthly basis, such as DEM, soil, land use, and climate data (Arnold

et al., 1998). Climate, hydrology, soil properties, and plant growth, nutrients, and practices of land

management are included in the model. The first modeling stage includes the delineation of watersheds. In

addition, the delineated watershed is subdivided into hydrologic response units (HRUs), which are a special

combination of land use, soil and slope management. For precipitation and temperature input, each HRU in

the model acts differently. (Yacoub and Foguet, 2012). By adjusting the threshold limit in the model, the sub-

basin is transformed to a single basin. Soil moisture (m3m-3) will be calculated by using equation 1 after

simulating SWAT production.


t

∆SMC = ∑(Rday ‒ Qsurf ‒ Ea ‒ Wseep ‒ Qgw)


i=1
( 1)

Where; SWt is the final soil water content (mm), Rday is the simulation time (days), Rday is the amount of

daily precipitation (mm), Qsurf is the amount of daily surface runoff (mm), Ea is the amount of daily

evapotranspiration (mm), wseep is the amount of water entering into the vadose zone from the soil profile on

a given day (mm), and Qgw is the amount of return flow on a given day (mm).

2.4.1. SWAT model calibration and validation

An automatic calibration procedure was implemented in SWAT - CUP. SWAT-CUP is a computer

program for calibration and validation of SWAT model. This program links GLUE, Parasol, SUFI 2, and

MCMC and PSO procedures to SWAT. It enables sensitivity analysis, calibration and validation of SWAT

model parameters. Model validation is comparison of the model outputs with an independent data set

without made further adjustment of the parameter values adjusted during calibration process.

Stream flow and sediment yield of the model simulations was calibrated at Jedeb and Chemoga

gauging station . The calibration was done on monthly time steps using the average measured stream

flow and sediment data of the Jedeb and Chemoga watershed covering from January 1990 to December

2002. After the model was calibrated and got acceptable results the model validated using measured data of

average monthly stream flow and sediment data of 7 years from January 2003 to December 2009.

Electronic copy available at: https://ssrn.com/abstract=3981555


2.4.2. Model performance evaluation
Based on (McFeeters, 1996) ,two quantitative statistics were used to assess model performance in

watersheds simulation .The Nash Sutcliffe Coefficient (NSE) and the goodness-of-fit (R2) those objective

function were widely used for model performance analysis (Tian et al., 2019). The determination coefficient

defines the proportion of variance according to the model in the calculated data. It is the linear magnitude

relationship between values observed and simulated and ranges from 0 (the model is poor) to 1 (the model is

good) with a higher value indicating less variances in error. The Nash-Sutcliffe simulation efficiency (NSE)

shows that 1:1 line suits well in the plots of observed versus simulated data importance. Using the following

equation it is calculated. The NSE value varies from negative infinity to 1. The NSE value below zero means

that the mean value observed is a better indicator than the simulated value, which also implies that the output

is unacceptable as illustrated in Table 4.

Table 3. Efficiency range for stream flow and sediment in the SWAT model

Rank NSE R2
Very good 1≥NSE≥0.75 0.8≤R2≤1.0
Good 0.75 ≥NSE ≥0.65 0.7≤R2≤0.75
satisfactory 0.65 ≥NSE≥0.5 0.5≤R2≤0.6

Electronic copy available at: https://ssrn.com/abstract=3981555


3. Result and Discussion
3.1. Land cover classification

The following Figures 2 show the land use and land cover maps created from images TM, TM,

ETM+ and OLS during 1990, 2000 2009 and 2018 of Jedeb and Chemoga. Agricultural land has increased

and decreased trees, grasslands and shrubs / bushes have been observed over the past 29 years, according to

land use land cover research.

Figure 2 Here

Table 4. Performance evaluation of satellite image for land LULC classification

Jedeb (Year) Classification Accuracy Overall Kappa Statistics


Producer’s User’s Overall
1990 90.3% 94.2% 90.8 0.88
2000 83.5% 87.9% 85.4% 0.80
2009 94.8% 94.7% 96% 0.95
2018 93.3% 94.1% 94% 0.92

Chemoga(Year) Classification Accuracy Overall Kappa Statistics


Producer’s User’s Overall
1990 91.9 % 95 % 92.3 % 0.89
2000 88.1 % 89.9 % 87.8 % 0.86
2009 95.5 % 95.5% 96.5% 0.93

2018 94.3% 94.9% 94.8% 0.92

Table 5 indicates the accuracy report of image classification in Jedeb and Chemoga

watershed is acceptable because; overall accuracy and kappa coefficient is greater than 85%

and 0.8 respectively based on Anderson (1976), (Sinshaw et al., 2019). But based on the

accuracy report the accuracy in Chemoga watershed was better than Jedeb watershed.

Electronic copy available at: https://ssrn.com/abstract=3981555


3.2. Stream flow modeling

3.2.1. Sensitivity analysis, calibration and validation

The sensitivity analysis was performed by considering 25 flow related parameters in order to select

the top 11 sensitive parameters that have a significant impact on the management of stream flow in the Jedeb

and Chemoga watersheds. SOL K, SOL Z, CN2, CANMX, HRU SLP, GW REVAP, GWQMN, ALPHA BF,

RCHRG DP, SOL AWC and SURLAG provide certain parameters as shown in Table 6 .

Table 5. Sensitive parameters and its rank in A) Jedeb and B) Chemoga watersheds

A. Jedeb watersheds

SN Parameter Name t-Stat P-Value Rank


1 RCHRG_DP.gw 3.37 0.00 1
2 CANMX.hru 1.76 0.09 2
3 SOL_K (...).sol 1.25 0.23 3
4 GWQMN.gw 1.21 0.24 4
5 HRU_SLP.hru 1.21 0.24 5
6 GW_REVAP.gw 0.58 0.57 6
7 ALPHA_BF.gw 0.49 0.63 7
8 SOL_Z (...).sol 0.43 0.67 8

B. Chemoga watersheds

SN Parameter Name t-Stat P-Value Rank


1 SOL_Z (...).sol 3.48 0.00 1
2 CN2.mgt 1.29 0.20 2
3 ALPHA_BF.gw 0.73 0.47 3
4 CANMX.hru 0.57 0.57 4
5 HRU_SLP.hru 0.56 0.58 5
6 SURLAG.bsn 0.30 0.77 6
7 SOL_AWC (...).sol 0.16 0.88 7

As shown in Table 6 Land use cover for 1990 , 2000, 2009 and 2018 was adjusted for 13 years from

January 1990 to December 2002 and validated for 7 years from January 2003 to December 2009. A good

agreement between the observed and simulated stream flow with the following determination coefficient (

R2) and Nash cliff coefficient (NSE) for the land use land cover plot for 1990 , 2000, 2009 and 2018 was

shown in the calibrated and validated value monthly stream flow result.

Electronic copy available at: https://ssrn.com/abstract=3981555


Table 6. Value of objective function during stream flow modeling

Land use land Jedeb watershed Chemoga watershed


cover Calibration Validation Calibration Validation
R2 NSE R2 NSE R2 NSE R2 NSE
1990 0.83 0.6 0.8 0.78 0.84 0.76 0.83 0.73
2000 0.85 0.84 0.82 0.62 0.84 0.75 0.82 0.65
2009 0.84 0.83 0.82 0.82 0.78 0.6 0.88 0.72
2018 0.81 0.75 0.8 0.8 0.74 0.54 0.82 0.64

During the calibration and validation of land use land cover maps for 1990, 2000, 2009 and 2018,

the following line graph shows the relationship between observed and simulated stream flow as well as

rainfall in the Jedeb and Chemoga watershed. There was a great relationship between stream flows observed

and simulated, as we can see from the graph. During Jedeb and Chemoga watershed modelling, there was

also a good relationship between stream flow and rainfall as the stream flow through depend on the amount

of rainfall as shown in Table 7. It corresponds to peak flow when high amounts of rainfall have occurred.

Figure 3 Here

3.3. Sediment yield modeling

3.3.1. Sensitivity analysis, calibration and validation

In 1000 simulations, 11 sediment parameters were tested, and only 4 parameters were defined as

having a significant impact on sediment inflow in the watersheds. These are USLE-P (USLE help practice

factor), erodibility factor (USLE-K), SPCON.bsn (Channel sediment routing linear factor), LAT-SED

(Lateral and ground water flow) and SPEXP (Channel sediment routing exponential factor).

Table 7. sensitivity parameters during modeling of sediment in A) Jedeb and B) Chemoga Watersheds.

A) Jedeb watershed

Parameter Name t-Stat P-Value Rank


SPEXP.bsn 1.76 0.09 1
LAT_SED.hru 1.55 0.13 2
SPCON.bsn 0.49 0.63 3
USLE_K(..).sol 0.39 0.70 4

Electronic copy available at: https://ssrn.com/abstract=3981555


B) Chemoga watershed

Parameter Name t-Stat P-Value Rank


SPCON.bsn 9.48 0.00 1
SPEXP.bsn 0.49 0.63 2
USLE_P.mgt 0.45 0.66 3
USLE_K(..).sol 0.36 0.72 4
LAT_SED.hru 0.20 0.84 5

Table 8 referes to SPEXP.bsn was the most sensitive sediment parameter in the Jedeb watershed

and SPCON.bsn was the most sensitive parameter in the Chemoga watershed, based on the above tables. The

sediment yield was adjusted for the 13-year period from January 1990 to December 2002 and the 7-year

model validation from January 2003 to December 2009 showed a strong agreement between the average

monthly sediment yield observed and simulated with the following coefficient of determination ( R2) and

(NSE) for the years 1990, 2000, 2009 and 2018.

Table 8. Value of objective function during sediment modeling


Land use land Jedeb watershed Chemoga watershed
cover Calibration Validation Calibration Validation
R2 NSE R2 NSE R2 NSE R2 NSE
1990 0.83 0.8 0.88 0.62 0.78 0.71 0.79 0.79
2000 0.81 0.76 0.85 0.59 0.88 0.72 0.68 0.76
2009 0.8 0.79 0.81 0.73 0.77 0.77 0.7 0.73
2018 0.79 0.79 0.77 0.77 0.88 0.73 0.72 0.61

The relationship between observed sediment, simulated sediment and stream flow in the Jedeb and

Chemoga watersheds during the calibration and validation of land use land cover maps from 1990 , 2000,

2009 and 2018 were clearly explaned in Table 9. According to Figure 4 there was a great relationship

between sediment being observed and simulated. The sediment yield of the watersheds also had a strong

relationship with the observed stream flow, based on the graph sketched below.

Figure 4 Here

Electronic copy available at: https://ssrn.com/abstract=3981555


3.4. Impact of land use and land cover change
Four separate SWAT model simulations were constructed using the various classes and data from

the four-land use / cover maps generated from the four separate Landsat imagery. The simulated stream flow

and sediment production of 1990 , 2000, 2009 and 2018 were then isolated from the entire result of the stream

flow and sediment and compared to their respective stream flow and sediment yield observed.

3.4.1. Evaluation of steam flow due to land use and land cover change
3.4.1.1. Change on Monthly stream flow
The average monthly stream flows of Jedeb watershed increase from year 1990 to 2000 in 35.58

m3/s, year 2000 to 2009 increased in 3.73 m3/s and finally decrease from year 2009 to 2018 in 17.19 m3/s.

The increase of flow may be due to slowly change of land use land covers especially the incremental of

agricultural in 17.6% (between 1990 to 2000) and 11.2% (between 2000 to 2009) and bare land in 51.7%

(between 1990 to 2000) and 8.8 (between 2000 to 2009) whereas the decrease of grass land in 8.2% (between

1990 to 2000) and 11.6% (between 2000 to 2009), forest in 16.1% (between 1990 to 2000) and 74.4%

(between 2000 to 2009) and shrubs land in 26.7% between 1990 to 2000. The reduction of stream flow

between year 2009 and 2018 was due to high percentage decrease of shrubs land in 64.2%, Forest in 57.8%

and grass land in 52.4% and high percentage increase of agricultural land in 62.3%, bare land in 145.2% and

establishment of settlement in 46.5 % in the watershed.

In Chemoga watershed average monthly stream flow increase between year 1990 to 2000 in 1.51

m3/s , between year 2000 to 2009 in 35.22 m3/s but decrease between the year 2009 to 2018 in 37.77 m3/s.

The increase of stream flow in Chemoga watershed may due to the increase of grass land between the year

1990 and 2000 in 2.4% and between the years 2000 to 2009 in 14.5%. In Table 10 the decrease of monthly

stream flow between the years 2009 to 2018 may be caused by the sum effects of the reduction water bodies

in 96.9% , shrubs in 89% and forest land in 81.8% and the increment of agricultural land in 47.3%, bare land

in 10.7% and establishment of settlement with in 420.7%. The following table shows the mean monthly

stream flow due to land use land cover of 1990 to 2018.

Electronic copy available at: https://ssrn.com/abstract=3981555


Table 9. Average monthly flow in Jedeb and Chemoga watershed for each land use land cover

Land use land cover map (year) Stream flow (m3/s)


Jedeb watershed Chemoga watershed
1990 89.1 58.14
2000 124.68 59.65
2009 128.41 94.87
2018 111.22 57.1

3.4.1.2. Change on Seasonal Stream flow


The seasonal stream flow of the Jedeb watershed increases more in the rainy season than in the dry

season because the rise in agricultural land and other land use changes make more flow for the wet season,

but no more for the variation of the dry season. Based on the result, the stream flow in the dry season was

increased in 21.4 m3 / s, while wet season in Jedeb watershed was increased in 71.4 m3 / s between 1990 and

2018 years. Seasonal stream flow in Chemoga watershed in dry season was increased in 2.92 m3 / s, but

between 1990 and 2018 the wet season stream flow in Chemoga watershed was decreased in 21.26 m3 / s.

Table 10. Land use land cover change impact on Seasonal variability of stream flow.

Land use land cover Seasonal mean monthly flow(m3/s)


Jedeb Chemoga
Dry season Wet season Dry season Wet season
1990 77.4 541.6 41.6 385.2
2000 181.85 636.11 45.6 382.2
2009 120.5 751.2 84 588.8
2018 98.8 613 44.52 363.94

3.4.1.3. Change on surface run off and ground water flow

To determine the change in the contribution of the components of stream flow due to change in

LULC of Jedeb and Chemoga watershed, study was made on the surface run off (SURQ) and ground water

flow (GWQ). Table 4 describes the simulated SURQ and GWQ of the stream flow based on the LULC map

of 1990 , 2000, 2009 and 2018 for a similar period of time.

Electronic copy available at: https://ssrn.com/abstract=3981555


Table 11. SURQ and GWQ of Jedeb and Chemoga watershed during different years LULC maps
simulation.

LULC map Chemoga flow (mm) Flow difference in mm


2000 - 1990 2009 - 2000 2018 – 2009

SURQ GWQ SURQ GWQ SURQ GWQ SURQ GWQ


1990 167.89 24.29
2000 177.65 24.35 9.76 0.06 2.59 -2.66 11.09 -0.62
2009 180.24 21.69
2018 191.33 21.07
LULC map Chemoga flow(mm) Flow difference in mm

2000 - 1990 2009 - 2000 2018 – 2009


SURQ GWQ SURQ GWQ SURQ GWQ SURQ GWQ
1990 139.95 29.78
2000 139.95 29.78 0 0 18.15 -1.83 18.06-2.79
2009 158.1 27.95
2018 176.16 25.16

As shown in Table 12, SURQ and GWQ stream flow components were 139.95 mm and 29.78 mm

during simulation of the 1990 and 2000 LULC maps, 158.1 mm and 27.95 mm during simulation of the 2009

LULC map, 176.16 mm and 25.16 mm during simulation of the 2018 LULC map for the period 1990 to 2018

in the Jedeb watershed. In the Chemoga watershed, the SURQ and GWQ were 167.89 mm and 24.29 mm

during the 1990 LULC map simulation, 177.65 mm and 24.35 mm during the 2000 LULC map simulation,

180.24 mm and 21.69 mm during the 2009 LULC map simulation, and 191.33 mm and 21.07 mm during the

same simulation time during the 2018 LULC map simulation.

The surface runoff contribution increased from 139.95 mm to 176.16 mm, where ground water flow

decreased from 29.78 mm to 25.16 mm due to the generalized change in land use cover between 1990 and

2018. This is due to the expansion of agricultural land over forest spending, which results in a rise in SURQ

after rainfall events. In terms of crop soil moisture requirements and rest time for infiltration, we will explain

this. In the first example, crops need less soil moisture than forests, so rain satisfies the soil moisture deficit

more quickly in agricultural lands than in forest land by producing more surface runoff where there is more

area under cultivated land. And this allows the soil moisture and the storage of ground water to differ. The

Electronic copy available at: https://ssrn.com/abstract=3981555


second cause of forest retardation was the time for surface flow of rainfall events and gave more time for

infiltration where water infiltration rate was reduced as agricultural land expansion.

3.4.1.4. SWAT Water Balance estimation of watersheds

The study uses 29 years of climate data and the study uses similar climate data from 1988 to 2018

only by adjusting the land use cover data in order to understand the change in stream flow and sediment

yield due to land use and land cover change. The research simulated and obtained the following water

balance components on the basis of this approach indicated in Table 13.

Table 12. Water balance components of Jedeb and Chemoga watersheds

A) Jedeb watersheds

Year Water balance components in mm

PRECIP SURQ GW_Q ET PET SW WYLD PERC LAT_Q


1990 93.8 13.2 23.7 51.5 89.2 118.6 40.8 26.7 2.5
2000 93.8 13.2 23.7 51.5 89.2 118.6 40.8 26.7 2.5
2009 93.8 14.8 22.5 51.4 89.2 118.8 40.9 25.4 2.4
2018 93.8 16.4 21.7 50.7 89.2 122.0 41.5 24.6 2.2

B) Chemoga watershed

Year Water balance components in mm

PRECIP SURQ GW_Q ET PET SW WYLD PERC LAT_Q


1990 92.8 14.3 24 49.9 87.5 123.3 41.4 27 1.7
2000 92.8 1.7 22.9 50.4 87.5 121.7 40.9 25.9 1.7
2009 92.8 15 21.4 52.1 87.5 115.6 39.2 24.3 1.5
2018 92.8 16.2 20.5 52 87.5 116.6 39.3 23.3 1.5

3.4.2. Impact of land use and land cover change on sediment yield

3.4.2.1. Change on Monthly sediment yield

The Jedeb watershed's average monthly sediment yield decreased by 204.59 ton / month from 1990

to 2000; from 2000 to 2009 it increased by 15,962.76 ton / month and finally decreased by 9,012.83 ton /

month from 2009 to 2018. Table 14 indicates that average monthly stream flow in Chemoga watershed

increased by 13,618.72 tons / month between 1990 and 2000, decreased by 11,316.6 tons / month between

2000 and 2009, then decreased by 11,192.93 tons / month between 2009 and 2018. As discussed above, the

rise in sediment yield in the watershed was due to a change in stream flow.

Electronic copy available at: https://ssrn.com/abstract=3981555


Table 13. Average monthly sediment yield in Jedeb and Chemoga watershed

Land use land cover Sediment yield (ton/month)


Jedeb watershed Chemoga watershed
1990 87,780.75 12,993.69
2000 87,576.16 26,612.41
2009 71,613.4 14,895.81
2018 62,600.57 26,088.74

3.4.2.2. Change on Seasonal sediment


According to Table 15 seasonal sediment yield in the Jedeb watershed decreased during the dry

season at 2,201.12 ton / month and decreased between 1990 and 2018 at 155,192.92 ton / month. Whereas

sediment yield in the Chemoga watershed decreased during the dry season at 1,826.39 ton / month and

decreased between 1990 and 2018 at 79,218.9 ton / month.

Table 14. Land use land cover change impact on Seasonal variability of sediment yield in the watersheds.

Land use land Seasonal mean monthly sediment (ton/month)


cover Jedeb watershed Chemoga watershed
Dry season Wet season Dry season Wet season
1990 26,674.42 567,596 14,114.11 76,824.5
2000 26,561.2 571,072 15,920.3 155,577.2
2009 24,923.2 458,716.6 11,860.1 91,268
2018 24,473.3 411,909.8 15,940.5 156,043.4

3.4.2.3. Identification of Sediment hotspot areas

Sediment sources from 17 sub basins of the Jedeb watershed were identified in Table 16 . Based on

the results of sub-basins 7,1,12, 17 and 15 of the 17 sub-basins of the Jedeb watersheds, sediment yield were

the highest contributor. IThe sediment yield of each sub-basins have the following sediment yield in Table

16 .

Electronic copy available at: https://ssrn.com/abstract=3981555


Table 15. Sediment sources of each sub basin in Jedeb watershed

Sub-basins Sediment (tons) Area (km2) Rank


1 34130 16.7 2
2 5821 5.8 10
3 1076 8.4 17
4 7083 6.5 7
5 5007 6.6 11
6 2483 33.9 15
7 54660 66.5 1
8 3535 44.4 14
9 4971 10.3 12
10 5971 3.1 9
11 7599 12.8 6
12 31340 40.7 3
13 4845 10.5 13
14 7042 11.1 8
15 8087 1.1 5
16 1975 12.1 16
17 10360 5.5 4

Sediment sources from 25 sub basins of the chemoga watershed were identified in Table 17 . Based

on the results of sub-basins 7,9,13, 1 and 2 of the 25 sub-basins of the Jedeb watersheds, sediment yield were

the highest contributor.

Table 16. Sediment sources of each sub basin in Chemoga watershed.

Sub-basins Sediment (tons/year) Area (km2) Rank


1 12350 16.2 4
2 10120 7.6 5
3 3982 9.1 11
4 1650 15.9 25
5 8888 13.4 7
6 2436 25.1 22
7 35890 39.2 1
8 9079 11.4 6
9 28860 31.6 2
10 2058 13.5 24
11 2980 6.2 16
12 7661 9.8 8
13 13350 18.8 3
14 2992 1 15
15 2931 26.3 17
16 7152 16.7 9
17 2879 21.5 18

Electronic copy available at: https://ssrn.com/abstract=3981555


18 3608 9 12
19 2486 6.8 21
20 3514 10.1 13
21 2205 1.2 23
22 2560 8.1 19
23 2540 17.5 20
24 3073 9.9 14
25 5694 0.6 10

4. Conclusion and Recommendation

4.1. Conclusion

Both Jedeb and Chemoga watersheds, land use and land cover change have been recorded in the

past 29 years from 1990 to 2018. The massive portion of the watersheds are agricultural lands which have

risen from year to year due to growing land use usage which cover due to population growth. For well

classification and accuracy outcome, pre-processing and processing was performed before landsat image

classification. In the picture of landsat7 ETM +, there were black stripes, these stripes were removed using a

tool called landsat tool box. The result of image classification showed 29 years generalized change of 62.3%

increase in agricultural land, 57.8 reductions in forest cover, 52.4% decrease in grass land, 64.2% reduction

in shrub land, 145.2% increase in bare land and 46.5% increase settlement in Jedeb watershed. Similarly the

study revealed that 47.3% increase in Agricultural land, 81.8% reduction in forest cover, 148.9% increase in

Grass land, 89% decrease in shrub land, 10.7% increase in bare land and 96.9% decrease in water body in

Chemoga watershed. Stream flow and sediment yield from the watershed were determined from SWAT

model. The model evaluation statistics of stream flow and sediment yield gave a good result of NSE and R2

both for calibration and validation. The sediment yield from watershed was high due to high population

growth, overgrazing and deforestation practice in the watersheds. The Sediment yield from the watershed

was high in wet season which have direct relationship with the watershed discharge. The average monthly

stream flows of Jedeb watershed increase from year 1990 to 2000 in 35.58 m3/s, year 2000 to 2009 increased

in 3.73 m3/s and finally decrease from year 2009 to 2018 in 17.19 m3/s. The increase of flow may be due to

slowly change of land use land covers especially the incremental of agricultural and bare land, the decrease

of grass land and shrubs land between the year 1990 and 2009. The reduction of stream flow between year

Electronic copy available at: https://ssrn.com/abstract=3981555


2009 and 2018 was may be due to high percentage increase of bare land, high percentage decrease of shrubs

and grass land, establishment of settlement in the watershed. In Chemoga watershed average monthly stream

flow increase between year 1990 to 2000 in 1.51 m3/s , between year 2000 to 2009 in 35.22 m3/s but decrease

between the year 2009 to 2018 in 37.77 m3/s. The increase of stream flow in Chemoga watershed may due

to the increase of grass land between the year 1990 and 2000 in 2.4% and between the years 2000 to 2009 in

14.5%. The decrease of monthly stream flow between the years 2009 to 2018 may be caused by the reduction

water bodies, shrubs, forest land and grass land and the increment of agricultural land, bare land and

establishment of settlement. The average monthly sediment yield of Jedeb watershed decreased from year

1990 to 2000 in 204.59 ton/month; between years 2000 to 2009 increased in 15,962.76 ton /month and finally

decrease from year 2009 to 2018 in 9,012.83 ton/month. In Chemoga watershed average monthly sediment

increase between years 1990 to 2000 in 13,618.72 ton/month, between years 2000 to 2009 decreased in

11,316.6 ton /month but decrease between the years 2009 to 2018 in 11,192.93 ton/month. Seasonal stream

flow of Jedeb watershed increase in wet season more than in dry season because the increase in agricultural

land and other land use change makes more flow for wet season but no more for the variation of dry season.

Based on the result the stream flow in dry season was increased in 21.4 m3/s, whereas wet season was

increased in 71.4 m3/s in Jedeb watershed between 1990 and 2018 years. In Chemoga watershed seasonal

stream flow in dry season was increased in 2.92 m3/s, but the wet season stream flow was decreased in 21.26

m3/s in Chemoga watershed between year 1990 and 2018. Generally seasonal sediment yield in Jedeb

watershed decreased in 2,201.12 ton/month during dry season and decreased in 155,192.92 ton/month

between years 1990 to 2018. Whereas sediment yield in Chemoga watershed decreased in 1,826.39 ton/month

during dry season and decreased in 79,218.9 ton/month between years 1990 to 2018.

4.2. Recommendation

The SWAT model is a very powerful tool in the watersheds of Jedeb and Chemoga to model flow

and sediment and to fill the gap we have in the area of scarce sediment and stream flow data. It is advisable

to increase the number of hydrological and meteorological stations and to improve their efficiency in order

to obtain better modeling results and increase the performance of the SWAT model. Bare land and steep

slopes should be recommended in order to increase water supplies in the dry season and to minimize

sedimentation in wet season soil and water storage systems in particular biological structures and

Electronic copy available at: https://ssrn.com/abstract=3981555


reforestation of deforested areas. Special systemic treatment for soil and water protection has been suggested

for sub-basins 7,1,12, 17 and 15 of the Jedeb Watershed and sub-basins 7, 9, 13, 1, 2 , 8 and 5 of the Chemoga

Watershed, as sedimentation has seriously affected them. By ensuring different data quality, more studies

with more on this land use survey could be easier. In Jedeb watershed, the monthly sediment increase was

higher than in Chemoga, so special attention was needed for Jedeb watershed to manage watersheds.

Acknowledgement

The authors gratefully acknowledge Bahir Dar University, united states geological survey (USGS),

National Meteorology Agency (NMA) of Ethiopia, Yilmana Densa Woreda Water and Irrigation and Energy

office for their contribution.

Reference

ABRAHA, G. L., RAO, K. & KASSAWMAR, N. T. 2011. Analyzing the Impact of Land Use and Land

Cover Change on Ground Water Recharge Using RS and GIS Approach: The Case of Awassa

Catchment, Ethiopia. IUP Journal of Earth Sciences, 5.

AMEDE, T., GERMAN, L. A., RAO, S., OPONDO, C. & STROUD, A. 2006. Integrated natural resource

management in practice: enabling communities to improve mountain livelihoods and landscapes:

Proceedings of the African highlands initiative conference, 12-15 October, 2004, Nairobi, Kenya.

ANDERSON, J. R. 1976. A land use and land cover classification system for use with remote sensor data,

US Government Printing Office.

ARNOLD, J. G., SRINIVASAN, R., MUTTIAH, R. S. & WILLIAMS, J. R. 1998. Large area hydrologic

modeling and assessment part I: model development 1. JAWRA Journal of the American Water

Resources Association, 34, 73-89.

BERHANU, M., GOSHU, A., QURAISHI, D. & SHOEB, E. 2017. Identification of Soil Conservation

Priority Area in Jedeb River Watershed, East Gojjam Zone, Amhara Region, Ethiopia. Haramaya

University.

DEMISSE, B. A. 2011. Discharge and sediment yield modeling in Enkulal watershed, Lake Tana region,

Ethiopia. Cornell University.

FAO 2002. Land water linkages in rural watersheds. 9.

Electronic copy available at: https://ssrn.com/abstract=3981555


GEBREHIWOT, S. G. 2012. Hydrology and forests in the Blue Nile Basin.

HARI, K. 2014. A Review-Impact of land use land cover change and best management practices in a

watershed by using swat model. International Journal of Pure & Applied Bioscience.

LEE, H. 2013. Intergovernmental Panel on Climate Change.

MCFEETERS, S. K. 1996. The use of the Normalized Difference Water Index (NDWI) in the delineation of

open water features. International journal of remote sensing, 17, 1425-1432.

SINSHAW, B. G., MOGES, M. A., TILAHUN, S. A., DOKOU, Z., MOGES, S., ANAGNOSTOU, E.,

ESHETE, D. G., KINDIE, A. T., BEKELE, E. & ASESE, M. Integration of SWAT and Remote

Sensing Techniques to Simulate Soil Moisture in Data Scarce Micro-watersheds: A Case of

Awramba Micro-watershed in the Upper Blue Nile Basin, Ethiopia. International Conference on

Advances of Science and Technology, 2019. Springer, 294-314.

SIVAPALAN, M. 2005. Pattern, process and function: Elements of a new unified hydrologic theory at the

catchment scale. Contribution to: Encyclopaedia of Hydrologic Sciences, MG Anderson (Managing

Editor), Chapter 13 (Vol. 1, Part 1). John Wiley & Sons.

TEKLEAB, S., UHLENBROOK, S., SAVENIJE, H. H., MOHAMED, Y. & WENNINGER, J. 2015.

Modelling rainfall–runoff processes of the Chemoga and Jedeb meso-scale catchments in the

Abay/Upper Blue Nile basin, Ethiopia. Hydrological Sciences Journal, 60, 2029-2046.

TIAN, F., HU, H., SUN, Y., LI, H. & LU, H. 2019. Searching for an Optimized Single-objective Function

Matching Multiple Objectives with Automatic Calibration of Hydrological Models. Chinese

Geographical Science, 29, 934-948.

YACOUB, C. & FOGUET, A. P. 2012. Slope effects on SWAT modeling in a mountainous basin. Journal

of Hydrologic Engineering, 18, 1663-1673.

Electronic copy available at: https://ssrn.com/abstract=3981555

You might also like