Water Resources Management

Impact of Land Use Land Cover (LULC) Change on Surface Runoff in an Increasingly
Urbanized Tropical Watershed
--Manuscript Draft--

Manuscript Number: WARM-D-19-00147

Full Title: Impact of Land Use Land Cover (LULC) Change on Surface Runoff in an Increasingly
Urbanized Tropical Watershed

Article Type: General paper

Keywords: Hydrological modeling; Water balance; SWAT; land use land cover (LULC) change;
surface runoff

Corresponding Author: kamalakanta sahoo, Ph.D.

USDA Forest Products Laboratory
Madison, Wisconsin UNITED STATES

Corresponding Author Secondary

Information:

Corresponding Author's Institution: USDA Forest Products Laboratory

Corresponding Author's Secondary

Institution:

First Author: kamalakanta sahoo, Ph.D.

First Author Secondary Information:

Order of Authors: kamalakanta sahoo, Ph.D.

Ike Sari Astuti, PhD

Adam Milewski, PhD

Deepak R. Mishra, PhD

Order of Authors Secondary Information:

Funding Information:

Abstract: Understanding the influence of land use and land cover(LULC) changes on water
resources is essential for proper watershed management. Upper-Brantas watershed in
East Java, Indonesia, is a tropical watershed experiencing rapid landscape change, a
phenomenon typical to developing countries. Accelerated sedimentation, drying-up of
springs, and downstream eutrophication can be attributed to unrestrained urbanization
and intense resource demands due to the growing population. This study provides
baseline information on how LULC changes impact surface runoff in a tropical, data
scarce, urbanized watershed. We estimated the LULC changes between 1995 and
2015, and investigated its impact on the hydrological processes, particularly surface
runoff using the Soil and Water Assessment Tool(SWAT) model. In the past 21 years,
the watershed experienced an increase in settlement(7% of total area) and dryland
agriculture(3%) compared to a decrease in the forest(3%), rice field(3%) and
sugarcane plantation(4%). The SWAT model results during calibration(2003-2008) and
validation(2009-2013) periods matched observed values[R2>0.91 and NSE(Nash-
Sutcliffe Efficiency)>0.91]. The model illustrated changes in long-term average
runoff(+8%), water yield(+0.28%), groundwater(-1.8%), and evapotranspiration(-
1.15%) due to changes in LULC. The most significant factors affecting surface runoff
were changes in the forest, dry agriculture, and settlements. The changes in the LULC
showed a linear relationship with runoff generation in the studied watershed. With
increasing urbanization, industry and agricultural intensification, the increased runoff
will enhance the flow of nutrients and sediments into the water bodies. Future work is
recommended to quantify how the changes in runoff affect sediment and nutrient yields
that contribute to downstream sedimentation and eutrophication.

Suggested Reviewers: JC Seong, PhD

Professor, University of West Georgia

Powered by Editorial Manager® and ProduXion Manager® from Aries Systems Corporation
 [email protected]
He is one of the expert in this research area

Debabrata Sahoo, PhD

Hydrology Engineer, Woolpert
[email protected]
He is expert in this research area

Charles Gyamf, PhD

Lecturer, Technische Universitat Munchen Ingenieurfakultat Bau Geo Umwelt
[email protected]
He is an expert in this research areas

Powered by Editorial Manager® and ProduXion Manager® from Aries Systems Corporation
Manuscript Click here to access/download;Manuscript;Ike et
al_V2_WRM.docx
Click here to view linked References
1
2
3
4 1 Impact of Land Use Land Cover (LULC) Change on Surface Runoff
5
6 2 in an Increasingly Urbanized Tropical Watershed
7
8 3 Ike Sari Astuti1, 2, Kamalakanta Sahoo3, 4, Adam Milewski5, and Deepak R. Mishra2*
9 1
10 4 Department of Geography, University of Georgia, Athens, GA 30602, US
11 5 2
Department of Geography, Universitas Negeri Malang, Jawa Timur, 65145, Indonesia.
12
3
13 6 College of Engineering, University of Georgia, Athens, GA 30602, US.
14 4
15 7 Forest Products Laboratory, United States Forest Service, Madison, WI 53726, US.
16 8 5
Department of Geology, Water Resources & Remote Sensing Group (WRRS), University of Georgia,
17 9 Athens, GA 30602, US
18
19 10 * Author to whom correspondence should be addressed; E-Mail: [email protected]; Tel.: +1-706-351-1037.
20
21 11 Abstract
22
23 12 Understanding the influence of land use and land cover(LULC) changes on water resources
24
25
26 13 is essential for proper watershed management. Upper-Brantas watershed in East Java, Indonesia,
27
28 14 is a tropical watershed experiencing rapid landscape change, a phenomenon typical to developing
29
30
15 countries. Accelerated sedimentation, drying-up of springs, and downstream eutrophication can be
31
32
33 16 attributed to unrestrained urbanization and intense resource demands due to the growing
34
35 17 population. This study provides baseline information on how LULC changes impact surface
36
37
38 18 runoff in a tropical, data scarce, urbanized watershed. We estimated the LULC changes between
39
40 19 1995 and 2015, and investigated its impact on the hydrological processes, particularly surface
41
42
43 20 runoff using the Soil and Water Assessment Tool(SWAT) model. In the past 21 years, the
44
45 21 watershed experienced an increase in settlement(7% of total area) and dryland agriculture(3%)
46
47
48
22 compared to a decrease in the forest(3%), rice field(3%) and sugarcane plantation(4%). The
49
50 23 SWAT model results during calibration(2003-2008) and validation(2009-2013) periods matched
51
52 24 observed values[R2>0.91 and NSE(Nash-Sutcliffe Efficiency)>0.91]. The model illustrated
53
54
55 25 changes in long-term average runoff(+8%), water yield(+0.28%), groundwater(-1.8%), and
56
57 26 evapotranspiration(-1.15%) due to changes in LULC. The most significant factors affecting
58
59
60 27 surface runoff were changes in the forest, dry agriculture, and settlements. The changes in the
61
62 1
63
64
65
 1
2
3
4 28 LULC showed a linear relationship with runoff generation in the studied watershed. With
5
6
7 29 increasing urbanization, industry and agricultural intensification, the increased runoff will enhance
8
9 30 the flow of nutrients and sediments into the water bodies. Future work is recommended to quantify
10
11
12 31 how the changes in runoff affect sediment and nutrient yields that contribute to downstream
13
14 32 sedimentation and eutrophication.
15
16
17 33 Keywords: Hydrological modeling, Water balance, SWAT, land use land cover (LULC) change,
18
19 34 surface runoff
20
21
35 1. Introduction
22
23
24 36 Changes in land use and land cover(LULC) due to anthropogenic drivers are often
25
26 37 associated with alteration in various natural support systems such as water resources quantity and
27
28
29 38 quality(Giri_and_Qiu_2016), biodiversity(Solar_et_al.-2016), food security(Rutten_et_al.-2014),
30
31 39 and energy supply(Preston_and_Kim 2016). From a hydrological perspective, LULC change
32
33
34 40 within a watershed has been recognized as one of the critical factors influencing runoff
35
36 41 generation(Chang_2007), which became much stronger with climate change(Li et al. 2012;
37
38
42 Vörösmarty et al. 2000). In tropical regions, the impact of LULC changes and climate on
39
40
41 43 streamflow can be severe due to greater energy inputs and faster anthropogenic changes.
42
43 44 Southeast Asian countries have been experiencing the fastest LULC change over the last
44
45
46 45 few decades(Lambin-et-al. 2003). For example, the loss of tropical forest in Indonesia has been
47
48 46 extremely large, the second fastest after Brazil (Margono-et-al. 2014). Indonesia’s major source
49
50
51 47 of water resources relies on approximately 5,590 rivers, 521 lakes, and
52
53 48 100 reservoirs(Lehmusluoto-et-al. 1997). The increased water use and degrading watershed
54
55
56
49 conditions have intensified overall stress on already diminishing water resources(Pawitan-and-
57
58
59
60
61
62 2
63
64
65
 1
2
3
4 50 Haryani 2011) and thus, require scientific management of watershed to improve their current
5
6
7 51 condition(Fulazzaky-2014).
8
9 52 Numerous studies have been carried out in different environments on the impact of land
10
11
12 53 use changes on hydrological parameters such as runoff mechanism(Eshtawi-et-al. 2016;Ghaffari-
13
14 54 et-al. 2010;Wagner-et-al. 2013). Quantifying the consequences of land use change to hydrological
15
16
17 55 response is challenging due to the variability of hydrological systems, limited numbers of
18
19 56 controlled experiments, relatively short hydrological records, and difficulties in controlling land
20
21 57 use(DeFries-and-Eshleman-2004). Some studies have shown contradictory findings and lack of
22
23
24 58 consistency(Beck-et-al. 2013;Bruijnzeel 2004). For example, simulation on land use change might
25
26 59 result in different magnitudes of changes(Beck-et-al. 2013) − a small unnoticeable change of
27
28
29 60 around 5 mm in hydrological responses (i.e., surface runoff, ground water flow and stream flow)
30
31 61 for a 40-year period(Ghaffari-et-al. 2010) to more observable changes, such as a -38% in discharge
32
33
34 62 due to land use change in Uruguay (Silveira-and-Alonso 2009). A canceling-out effect can appear,
35
36 63 where influences from sub-basins can potentially mask out the impacts of LULC change in the
37
38
64 whole watershed(Becket-al.-2013;Wagner-et-al.-2013). The lack of understanding of this issue has
39
40
41 65 primarily been attributed to the heterogeneity of watershed characteristics, land use pattern and
42
43 66 configuration, threshold behavior in hydrological processes, and inference of climate, which are
44
45
46 67 difficult to quantify (Wang et al. 2014; Zehe et al. 2009). Therefore, a similar study from a
47
48 68 particular watershed bears some merits in enhancing our understanding of the impact of LULC
49
50
51 69 change on the ecosystem.
52
53 70 Assessing the LULC change impact on hydrological processes has been carried out using
54
55
56
71 pair catchment experiments to hydrological modeling. Previous paired catchment studies are often
57
58 72 impractical due to the need for a long-time window, but they are simpler and produce more rapid
59
60
61
62 3
63
64
65
 1
2
3
4 73 results to be fit into the local decision-making process(Ghaffari-et-al.-2010;Ochoa-Tocachi-et-al.-
5
6
7 74 2016). In hydrological modeling, the Soil Water Assessment Tool(SWAT) model is a robust model
8
9 75 to assess the impacts of land use change on hydrological processes and hydrologically influenced
10
11
12 76 ecosystem services(Douglas-Mankin-et-al. 2010;Francesconi-et-al. 2016;Gassman-et-al.
13
14 77 2007;Neitsch-et-al. 1999). Despite numerous studies worldwide, there have been limited SWAT-
15
16
17 78 based hydrological studies carried out in Indonesia(Barkey-et-al. 2017;Marhaento-et-al.
18
19 79 2017;Othman-and-Sholichin-2008; Rahayuningtyas-et-al. 2014;Setyorini-et-al. 2017). Setyorini-
20
21 80 et-al.(2017) have investigated the land use change in the Upper Brantas watershed, Indonesia
22
23
24 81 between 1989 to 2006 and provide its impacts on stream flow and water balances without
25
26 82 quantifying the relationship between land use change and watershed characteristics. The purpose
27
28
29 83 of this study within the Upper Brantas watershed in Indonesia was to further develop the LULC
30
31 84 versus runoff relationship over larger time domains, given the more recent rapid urbanization
32
33
34 85 (post-2006).
35
36 86 Java Island is home to around 60% of the country’s population. It has experienced
37
38
87 increasing pressures on land resources, and quickly exhibited a transition from a mainly rural to a
39
40
41 88 largely urbanized environment(Handayani-2013). Several studies highlighted the impacts of
42
43 89 increasing urbanization on Java, such as increasing farm loss(Partoyo-and-Shrestha-2013), water
44
45
46 90 pollution(Djuangsih-1993), and accelerated sedimentation and flood events(Valentin-et-al.-2008).
47
48 91 Since hydrology serves as the main driver governing sediment and nutrients, knowing the
49
50
51 92 hydrological response, especially the runoff generation in Javanese watersheds is of importance.
52
53 93 The objectives of this study are to: (i) simulate the impacts of LULC change over the past two
54
55
56
94 decades (1995 – 2015) on the hydrological response of catchment with a focus on surface runoff,
57
58
59
60
61
62 4
63
64
65
 1
2
3
4 95 and (ii) investigate the relationship between runoff generation and the pathways of the LULC
5
6
7 96 changes especially settlement or dryland agriculture.
8
9 97 2. Materials and methods
10
11
12 98 2.1 Study Area
13
14 99 The study site is the Upper Brantas watershed located in East Java, Indonesia(Figure 1).
15
16
17 100 Upper Brantas watershed is part of the Wilayah Sungai(WS) Brantas. Wilayah Sungai, or “River
18
19 101 Region”. Brantas is a delineated river basin management area in Indonesia that holds strategic
20
21 102 economic and ecological importance.
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55 103
56 104 Figure 1. Map of Indonesia, East Java and Upper Brantas watershed with corresponding
57
58
105 elevation and location of gaging stations.
59
60
61
62 5
63
64
65
 1
2
3
4 106 The WS river basin covers multiple watersheds(called Daerah Aliran Sungai,-DAS). It is
5
6
7 107 managed under a concept known as “one river basin one management”(Bappenas-
8
9 108 2012;Kementrian-2011). The WS Brantas contains the longest river network of East Java, and
10
11
12 109 provides 73% of the total water demand of the province. It also produces around 63% of the
13
14 110 electricity that is consumed by the province, and plays a critical role in water conservation, water
15
16
17 111 resource utilization and acts as a water hazards controller(Adi-et-al.-2013;Kementrian-2011). WS
18
19 112 Brantas river management encompasses three river sub-basin management areas; Lower, Middle
20
21 113 and Upper Brantas river basins. Over the years, the Upper Brantas has attracted scrutiny due to
22
23
24 114 increasing reports of environmental problems such as sedimentation(Adi-et-al.-2013), reservoir
25
26 115 eutrophication(Sulastri-and-Suryono-2004) and discharge of pollutants(Fulazzaky-2009). Upper
27
28
29 116 Brantas watershed is surrounded by mountains creating a circularly-shaped drainage area of almost
30
31 117 2,000 km2. The elevation ranges from 223 m to 3673 m above sea level, with slope ranging from
32
33
34 118 0° to 78°. Upper Brantas experiences abundant rainfall with a mean total annual rainfall of 2063
35
36 119 mm between 1991 and 2015. Similar to other monsoon-influenced regions, Upper Brantas exhibits
37
38
120 distinct rainfall distribution with intense precipitation during the wet season(October–March),
39
40
41 121 peaking in December-January, and low precipitation during April – September(Dry Season). The
42
43 122 watershed experiences slight temperature variation ranging from 19°C to 30°C.
44
45
46 123 2.2 SWAT model
47
48 124 SWAT is a continuous, semi-distributed, watershed-based hydrological model developed
49
50
51 125 to predict the impact of land management practices on water, sediment and chemicals from
52
53 126 agriculture within a watershed(Neitsch-et-al.-2011). SWAT was designed to assess the impact of
54
55
56
127 different management practices on hydrology and water balances in a watershed. It has been
57
58 128 widely applied to study hydrological responses and pollutant loads at varying spatial and temporal
59
60
61
62 6
63
64
65
 1
2
3
4 129 scales with adequate accuracy and reliability(Becker-et-al.-2012;Francesconi-et-al.-2016;
5
6
7 130 Gassman-et-al.-2007;Milewski-et-al.-2009,2014). SWAT estimates the runoff using biophysical
8
9 131 data such as precipitation, soil properties, topography, land use and land cover, and SCS curve
10
11
12 132 number equation(Abbaspour-2015).
13
14 133 In SWAT, the watershed is divided into multiple sub-watershed and hydrologic response
15
16
17 134 units(HRUs). Each HRU is a designated area having a unique combination of land use, soil
18
19 135 characteristics, and slope. It uses a water balance principle to simulate the hydrological process
20
21 136 within a watershed and all flow parameters are routed to a sub-basin level, and eventually to a
22
23
24 137 basin level as described in(Neitsch-et-al.-2011) as:
25
26 𝑡
27
28
SWt = SWt−1 + ∑( 𝑅𝑖 − 𝑄𝑖 − 𝐸𝑇𝑖 − 𝑃𝑖 − 𝑄𝑅𝑖 ) (1)
29 𝑖=1
30 138 Where SWt , SWt−1, 𝑅𝑖 are the final soil water content, initial soil water content, the amount
31
32
33 139 of precipitation respectively on day 𝑖 in mm. 𝑄𝑖 , 𝐸𝑇𝑖 , 𝑃𝑖 , and 𝑄𝑅𝑖 are the amount of surface runoff,
34
35 140 evapotranspiration, the amount of water percolating through the soil profile, and amount of return
36
37
38 141 flow on day i respectively. Despite increasing urbanization, rural areas still dominated the land
39
40 142 cover in Java. Therefore, the influence of artificial drainage networks can be considered minimum.
41
42
43 143 2.3 Input data
44
45 144 SWAT input data includes topography, climate, land use and land cover, and soil.
46
47
48
145 Topography data was obtained from a 30m ASTER Digital Elevation Model(ASTER GDEM V2,
49
50 146 obtained from(USGS-2017). The land use maps(Figure 2) were obtained from a previous study
51
52 147 consisting of land use in 1995 and 2015, which derived from an Object-Based Image
53
54
55 148 Analysis(OBIA) classification technique using Landsat images from the previous study(Astuti -
56
57 149 2018). The land use maps(adopted from the Indonesian Ministry of Forestry’s land use
58
59
60 150 classification scheme) were reclassified to meet the nomenclature used in the SWAT land cover
61
62 7
63
64
65
 1
2
3
4 151 database namely Agricultural Land Row Crop(AGRL), Forest Evergreen(FRSE), Rice(RICE),
5
6
7 152 Sugarcane(SUGC), Urban Residential Medium Density(URMD), Range Shrubland(RNGB),
8
9 153 Range Grassland(RNGE) and Water(WATR).
10
11
12 A B
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30 Agricultural land row crop (AGRL)
31 Forest evergreen (FRSE)
32 Sugarcane (SUGC)
33
34
Rice field (RICE)
35 Shrubland/Bushland (RNGB)
36 Grassland/(RNGE)
37 Urban Residential Medium Density (URML)
38 Rivers/Reservoir (WATR)
39
40
41 154 Figure 2. Land Use/Land Cover(LULC) in Upper Brantas Watershed for(A) 1995 and (B)2015.
42
43 155 URMD was selected instead of Urban Residential High Density(URHD) or Urban
44
45
46 156 Residential Low Density(URLD) with an assumption that settlement areas in Java, Indonesia are
47
48 157 not heavily covered with impervious surface. URMD class is assumed to have around 38% of the
49
50
51 158 impervious surface in the settlement areas(Neitsch-et-al.-2011), which is relatively applicable to
52
53 159 most urban areas in Java. Slope classes were used in this study are in accordance with the Guideline
54
55
56
160 of Land Rehabilitation from Indonesia’s Ministry of Forestry(MOF-1987). The land use maps for
57
58 161 the site were produced with an overall accuracy of 85% (Kappa Coefficient: 81%). SWAT data
59
60
61
62 8
63
64
65
 1
2
3
4 162 inputs are summarized in Table-1. Soil data were obtained from the Harmonized Soil World
5
6
7 163 Database(Figure-3). Due to data unavailability, global solar radiation and wind that have been used
8
9 164 for hydrological simulation in Indonesia(Barkey-et-al.-2017) were used.
10
11
12 165 Table 1. SWAT model data input and sources
13
14 Data type Source Remarks
15 Land Use Map Previous study(Astuti-2018), Derived from LANDSAT 30m
16
17 Streamflow Perum Jasa Tirta, Malang Daily
18
19 Soil Harmonized World Soil Database(FAO ISRIC)
20 DEM ASTER DEM V2(USGS) 30m
21
22 Slope classes Indonesia’s Ministry of Forestry
23
24 Rainfall Weather Agency(BMKG Malang) Daily
25 Relative humidity, Solar National Centers for Environmental Prediction Climate Forecast
26 Daily
radiation, and Wind System Reanalysis(http://globalweather.tamu.edu/)
27
28 166
29
30
31
32
33
34
35
36
37 FAO 4509 - Lithosols
38 FAO 4518 - Eutric Cambisols
39
FAO 4538 - Vertic Luvisols
40
41 FAO 4570 - Eutric Regosols
42 FAO 4573 - Mollic Andosols
43 FAO 4575 - Ochric Andosols
44
45 FAO 4580 - Vitric Andosols
46
47
48
49
50 167 Figure 3. Soil Map for Upper Brantas(Source: FAO, Harmonized World Soil Database).
51
52
53 168 2.4 Model evaluation
54
55
56 169 The model evaluation covers calibration, sensitivity analysis and validation. Sensitivity
57
58 170 analysis represents the model response to changes made in the model parameters. This helps
59
60
61
62 9
63
64
65
 1
2
3
4 171 identify important input parameters. The initial parameters were selected based on watershed
5
6
7 172 characteristics and previously published literature(Ghaffari-et-al.-2010;Seyoum-et-al.-2015;van-
8
9 173 Griensven-et-al.-2006). Due to the variability of input data, model design, and parameters, the
10
11
12 174 model requires uncertainty analysis. SWAT was calibrated(2000-2008) and validated(2009-2013)
13
14 175 at a monthly time step with three years of warm-up.
15
16
17 176 The SWAT-Calibration Uncertainty Program(SWAT-CUP) was used to perform
18
19 177 calibration, validation and sensitivity analysis using the Sequential Uncertainty Fitting
20
21 178 version 2(SUFI-2) routine. Several statistical measures were developed and are used to quantify
22
23
24 179 the degree of uncertainty by the P-factor(percentage of the simulated values fall within the 95%
25
26 180 probability band), r-factor(the width of the 95% probability band). Ideally, a value of 1 for P-factor
27
28
29 181 and 0 for r-factor indicate that the simulated data perfectly match with the observed
30
31 182 data(Abbaspour-2015).The relative sensitivity of each parameter is measured by t-test and p-
32
33
34 183 values. The higher the absolute value of t-test, the more sensitive the parameter. p-denotes the
35
36 184 statistical significance of sensitivity of each parameter. Overall, the model performance is assessed
37
38
39
185 using R2, NSE, and PBIAS. The coefficient of determination(R2) measures the proportional
40
41 186 variation in the simulated variable explained by the measured variable and indicates the linear
42
43 187 relationship between the estimated and measured variables. Nash-Sutcliffe Efficiency(NSE)
44
45
46 188 determines the relative magnitude of the residual variance compared to the observed data(Nash-
47
48 189 and-Sutcliffe-1970). It ranges from -∞ to 1 with 1 representing a perfect agreement between the
49
50
51 190 simulated and measured values. Percent BIAS(PBIAS) measures the percentage of the
52
53 191 overestimation or underestimation of the simulated variables.
54
∑𝑛𝑖=1(𝑂𝑖 − 𝑆𝑖 )(𝑆𝑖 − 𝑆̅)
55
2
(2)
56 𝑅 = [ 𝑛 ]2
57 (∑𝑖=1(𝑂𝑖 − 𝑂̅)2 )0.5 (∑𝑛𝑖=1(𝑂𝑖 − 𝑂̅)2 )0.5
58
59
60
61
62 10
63
64
65
 1
2
3
4 (3)
5 ∑𝑛𝑖=1(𝑂𝑖 − 𝑆𝑖 )2
6 𝑁𝑆𝐸 = 1 − 2
7 ∑𝑛𝑖=1(𝑂𝑖 − 𝑂̅)
8 (4)
9 ∑𝑛𝑖=1(𝑂𝑖 − 𝑆𝑖 )2
𝑃𝐵𝐼𝐴𝑆 =
10 ∑𝑛𝑖=1 𝑂𝑖
11
12 192 Where 𝑂𝑖 and 𝑆𝑖 are the observed value and simulated value respectively. The mean of the
13
14
15 193 observed and simulated values for ‘n’ number of observations are represented as 𝑂̅ and 𝑆̅ .
16
17 194 2.5 Model implementation
18
19
20 195 The calibrated SWAT model was used to simulate the impact of LULC change over the
21
22 196 past two decades on hydrological responses. The “impact” was determined as the quantified
23
24
25
197 change, which was calculated by following the method used in similar studies(Ghaffari-et-al.-
26
27 198 2010;Wagner-et-al.-2013;Wang-et-al.-2014). This was implemented by running the model using
28
29 199 different land use data, while keeping the other data inputs the same. In this approach, all the
30
31
32 200 changes to hydrological variables were assumed to be the result of the changes in land use inputs.
33
34 201 The validated parameters from the LULC 1995 were then applied to the 2015 LULC period while
35
36
37 202 keeping other parameters such as climate, soil, and topography same as before. The climate during
38
39 203 2000 – 2013 was assumed to have no significant changes and was confirmed using a Mann-Kendall
40
41
42 204 (MK) trend test. The MK test was used to detect the presence of a rainfall trend, testing whether
43
44 205 the rainfall trends show significant variability. The null hypothesis for the MK test revealed no
45
46
206 trend in the data, which was tested at a significance level of 0.05. This test is often used in
47
48
49 207 hydrological applications(Liu-et-al.-2009;Pingale-et-al.-2014). The test results showed that the
50
51 208 Thiessen weighted rainfall within the study area did not show any statistically significant trends
52
53
54 209 over the last 20 years as shown in Figure-4.
55
56 210
57
58
59
60
61
62 11
63
64
65
 1
2
3
4
5 A
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20 B
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35 211 Figure 4. (A) Rainfall trend over 25 years(1990 – 2015) and (B) Mean monthly rainfall and
36 212 discharge data for the period 1998 – 2014 in Upper Brantas.
37
38 213 2.6 Influence of LULC change on hydrological processes
39
40
214 Varying methods have been employed to quantify the impacts of LULC change on
41
42
43 215 hydrological variables. Several studies applied simple comparison of the variables of interest
44
45 216 between different years(Ghaffari-et-al.-2010;Khoi-and-Suetsugi-2014;Sajikumar-and-Remya-
46
47
48 217 2015). Others have applied statistical analyses using either correlation or regression between
49
50 218 hydrological parameters and LULC(Gyamfi-et-al.-2016;Wagner-et-al.-2013;Wang-et-al.-2014).
51
52
53 219 A comparison approach was performed here to provide an overview of changes in surface runoff
54
55 220 due to LULC change in 1995 and 2015. The correlation was used to identify the relative
56
57
58 221 importance of each major land cover type. To better understand the hydrologic response to
59
60 222 particular land use classes, a gradual land use change was applied by running SWAT on one land
61
62 12
63
64
65
 1
2
3
4 223 use map using a specified scenario(Eshtawi-et-al.-2016;Ghaffari-et-al.-2010). For this purpose, a
5
6
7 224 gradual change(0% to 100%) scenario on LULC 1995 was simulated to examine the relative
8
9 225 impact of urbanization and agricultural expansion on surface runoff. For these, the scenario was
10
11
12 226 applied to the selected sub-basin(sub-basin-6). This sub-basin was selected because it represents
13
14 227 one of the major sub-basins having all LULC types, allowing simulation of important land use
15
16
17 228 change scenarios. This sub-basin has been subject to forest and farm loss for settlement
18
19 229 development to support trade and tourism activities(Widianto-and-Lestariningsih-2001).
20
21
230 3. Results
22
23
24 231 3.1 Main parameters and model performance
25
26 232 The calibration and validation of the watershed model/SWAT model were initiated with 27
27
28
29 233 parameters using SWAT-CUP. The model was simulated/run/iterated four times before reaching
30
31 234 the final six sensitive parameters.
32
33
34 235 Table 2. List of 12 most sensitive model parameters with their relative sensitivity
35 Default Fitted
Parameter Description t-stat p-value
36 range value
37
38 v__GW_DELAY.gw Groundwater delay time(d) 0-500 203.13 19.38 0.000
39 v__LAT_TTIME.hru Lateral flow travel time 0-180 4.89 -15.16 0.000
40 r__SOL_K.sol Saturated hydraulic conductivity -0.9-1 -0.54 4.43 0.000
41 v__CH_K2.rte Effective hydraulic conductivity -0.01-500 21.57 2.85 0.005
42
43 Threshold depth of water in the
44 v__GWQMN.gw shallow aquifer 0-5000 649.65 2.57 0.010
45 v__GW_REVAP.gw Groundwater revap coefficient 0.02-0.2 0.10 2.57 0.011
46 r__SOL_BD.sol Moist bulk density 0.9-2.5 0.32 1.57 0.118
47
48 v__ALPHA_BNK.rte Baseflow alpha factor 0-1 0.72 1.12 0.263
49 r__CN2.mgt SCS runoff curve number 68-89 -0.14 0.93 0.355
50 v__CH_W2.rte Average width of main channel 0-1000 549.57 0.78 0.435
51
52
v__SURLAG.bsn Surface runoff lag time 0.0-24 0.86 -0.62 0.534
53 v__CH_D.rte Average depth of main channel 0-30 13.52 0.55 0.583
54 236
55
56
57
58
59
60
61
62 13
63
64
65
 1
2
3
4
5
6 A
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24 B C
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41 237 Figure 5. (A) Plot of observed and simulated flow used for calibration and validation.
42 238 Observed versus simulated flow for (B) calibration, and (C) validation.
43
44 239 Table 3. SWAT model performance for calibration and validation
45
46 Value of statistical fits
47 Simulation 2
48
R NSE p-factor PBIAS(%)
49 Calibration(January 2003−December 2008) 0.94 0.94 0.85 -0.1
50
Validation(January 2009−December 2013) 0.91 0.91 0.75 1.4
51
52
53 240 Table 2 summarizes the 12 most sensitive parameters with six of which being statistically
54
55 241 significant (absolute t-stat values ≥2 and p-value ≤0.05)(Moriasi-et-al.-2007). Most of these
56
57
58 242 parameters are related to soil and groundwater characteristics, which are similar to the findings
59
60 243 from the study of(Rahayuningtyas-et-al.-2014) in Lesti catchment, East Java. The statistical fits
61
62 14
63
64
65
 1
2
3
4 244 produced for streamflow on a monthly time step according to Moriasi et-al.(2007) were “very
5
6
7 245 good”, indicating that the SWAT model set up was reliable for the further application(Figure 5).
8
9 246 The high R2 and NSE values in the calibration and validation simulations (Table 3) suggest that
10
11
12 247 the calibrated model can reasonably describe the variability of streamflow within the basin. Hence,
13
14 248 one can assume that the calibrated model with the optimized parameters can be further applied to
15
16
17 249 assess the impacts of LULC change in Upper Brantas on hydrological responses.
18
19 250 3.2 LULC change over 20 years in the Upper Brantas watershed
20
21 251 Comparison between LULC 1995 and 2015 revealed that the Upper Brantas watershed has
22
23
24 252 experienced noticeable LULC changes. In the last 20 years, there has been a reduction in forested
25
26 253 areas, plantation and rice fields, ranging from 2% to - 4% of the total watershed size. On the other
27
28
29 254 hand, mixed-dryland agricultural areas within Upper Brantas increased by 62 km2, adding around
30
31 255 3% of the total area. Over the past two decades, the changes in each LULC class were relatively
32
33
34 256 small, ranging from 0.06% to 6.83% of the total basin area.
35
36 257 Table 4. LULC change over the past 20 years(1995–2015) in Upper Brantas watershed
37
38 Change % Change of % Change of
LULC 1995 2015
39 (km2) 1995 Total Area
40
Agriculture(AGRL) 684.69 746.61 61.9 9.0 3.1
41
42 Forest evergreen(FRSE) 364.12 302.68 -61.4 -16.9 -3.1
43 Sugarcane Plantation(SUGC) 263.73 175.17 -88.6 -33.6 -4.5
44 Rice field(RICE) 415.46 371.16 -44.3 -10.7 -2.3
45 49.57 49.15 -0.4 -0.8 0.0
Rangeland Bushland(RNGB)
46
47 Rangeland Grassland(RNGE) 3.29 0.19 -3.1 -94.4 -0.2
48 Urban Residential Medium 170.23 304.92 134.7 79.1 6.9
49 Density(URMD)
50 Water(WATR) 15.04 16.24 1.2 8.0 0.1
51 Total (km2) 1966.12 1966.12
52
53
54 258 The settlement became the LULC class that experienced the biggest shift with an increase
55
56 259 of around 135 km2 by 2015 (Table 4). This represents almost 80% change from its urban size in
57
58
59 260 1995, accounting for 6.9% of the total catchment area. From post-classification change detection
60
61
62 15
63
64
65
 1
2
3
4 261 in Table 5, it was found that forest land area reduction was mainly caused by the conversion of
5
6
7 262 forest land to mixed-dryland agriculture. Meanwhile, rice and mixed-dryland were converted to
8
9 263 the developed area to accommodate the increase in the urban population. Larger changes involved
10
11
12 264 changes in dryland agriculture and urban development. Approximately, 9% and 14% of mixed
13
14 265 dryland agriculture in 1995 was converted to rice fields and urban areas respectively by 2015. In
15
16
17 266 this period, a small reduction of forest land (3% of total watershed) occurred. From this converted
18
19 267 forest, approximately 20% was replaced by mixed dryland agriculture. Relatively larger shift in
20
21 268 the landscape was the reduction of rice fields. Around 26% of the rice fields in 1995 were
22
23
24 269 converted to settlement in 2015.
25
26 270 Table 5. Transition Matrix of LULC from 1995 to 2015
27
28 Land use 2015
29
30 AGRL FRSE SUGC RICE RNGB RNGE URMD WATR Total(Km2)
31 AGRL 482.7 14.5 34.8 92.6 1.2 0.0 57.6 1.3 684.7
32
33 FRSE 70.8 268.5 0.0 1.7 20.7 0.1 2.4 0.0 364.1
34 SUGC 96.3 0.0 79.3 60.5 0.0 0.0 27.1 0.5 263.7
Land use 1995

35
36 RICE 77.0 0.1 54.9 177.5 0.0 0.0 103.4 2.6 415.5
37 RNGB 4.6 17.7 0.0 0.1 26.9 0.1 0.2 0.0 49.6
38
39 RNGE 1.3 1.4 0.0 0.0 0.3 0.0 0.1 0.0 3.3
40 URMD 13.1 0.5 5.9 36.9 0.1 0.0 113.2 0.7 170.2
41
42 WATR 0.8 0.0 0.3 1.8 0.0 0.0 0.9 11.2 15.0
2
43 Total(km ) 746.6 302.7 175.2 371.2 49.2 0.2 304.9 16.2 1966.1
44
45 271 The loss of rice fields has become an increasing concern due to its significant role as the
46
47
48 272 staple food source for Indonesians. From 1995 to 2015, the population in Malang city and Malang
49
50 273 municipality increased from 2,869,596 to 4,556,648(BPS 2016). This increase accounted for
51
52
53 274 almost 60% increase in two decades with an average change rate of 2.78% per year. This rapid
54
55 275 increase in the urban areas was experienced due to the rapid economic development within the
56
57
58 276 watershed for the last few decades.
59
60
61
62 16
63
64
65
 1
2
3
4 277 3.3 Changes in overall water balance due to LULC change
5
6
7 278 The resulting simulation using two land use maps from 1995 and 2015 revealed changes
8
9 279 in all water balance components. The average annual basin surface runoff in 2015 increased from
10
11
12 280 460 mm to 496 mm, an 8% increase compared to 1995.In contrast, the annual evaporation average
13
14 281 decreased from 481 mm to 476 mm. Conversely, an increase in water yield from 1628 mm to 1633
15
16
17 282 mm and a decrease in annual groundwater recharge from 893 to 877 mm were simulated during
18
19 283 the two decades. One limitation in SWAT is that SWAT does not rigorously estimate the
20
21 284 groundwater process(Rostamian-et-al.-2008). Thus, groundwater-related estimates should be
22
23
24 285 considered relative, not absolute.The increase in water yield can be attributed to the increase in
25
26 286 surface runoff. This was due to the decrease in groundwater percolating into the soil. Figure 6
27
28
29 287 summarizes the changes in the annual average of water budget components for the whole
30
31 288 watershed. The changes for an annual average of surface runoff, evapotranspiration groundwater,
32
33
34 289 and water yield were relatively small(less than 40 mm). Compared to others, changes in surface
35
36 290 runoff was the most evident. This observation was also reflected in the long-term simulation result.
37
38
291 Changes in surface runoff from LULC 1995 to LULC 2015 was more pronounced than changes in
39
40
41 292 other hydrologic facets.
42
43 293
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 17
63
64
65
 1
2
3
4
5 A
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20 B
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35 C
36
37
38
39
40
41
42
43
44
45
46
47
48 Figure 6. LULC change impacts on watershed’s water balance between 1995 and 2015: (A) % change;
49
50
(B) absolute change; and (C) annual trends of changes.
51 At the sub-basin level, 25 of 27 sub-basins experienced an increase in annual average runoff, with
52
53
54 changes varying from 1% to 50% (Figure 7). Some sub-basins underwent an increase, while some others
55
56 exhibited a decrease, creating a canceling-out effect for the entire watershed. Variation of LULC change
57
58
59 was observed among sub-basins. The mixed dryland agriculture varied from -17% to 29% of the sub-
60
61
62 18
63
64
65
 1
2
3
4 basin area size. Variations in the changes in urban areas among sub-basins ranged from 0.6% to 20%.
5
6
7 The canceling-out effect observed in this study has been reported by many prior studies(Eshtawi-et-al.-
8
9 2016;Ghaffari-et-al.-2010;Wagner-et-al.-2013).
10
11
12 A
13
14
15
16
17
18
19
20
21
22
23
24
B
25
26
27
28
29
30
31
32
33
34
35 C
36
37
38
39
40
41
42
43
44
45
46
47 D
48
49
50
51
52
53
54
55
56
57
58 Figure 7. Changes (in %) at sub-basin level for (A) water yield, (B) groundwater, (C)
59 evapotranspiration and (D) surface runoff.
60
61
62 19
63
64
65
 1
2
3
4 294 3.4 Relative impact of land use change on surface runoff
5
6
7 295 The 21-year LULC change in Upper Brantas appeared to cause an increase in surface runoff
8
9 296 at varying degrees at the sub-basin level. This suggests that each land use might have a different
10
11
12 297 relative sensitivity to runoff generation and other hydrologic components. However, it should be
13
14 298 noted that in actual cases, the sensitivity of a particular watershed to runoff generation is also
15
16
17 299 impacted by other factors such as magnitudes of land use change(Marhaento-et-al.-
18
19 300 2017;Sajikumar-and-Remya-2015), soil conditions(Milewski-et-al.-2014), and different climatic
20
21 301 factors(Pingale-et-al.-2014).
22
23
24 302 Table 6. Correlation between changes in hydrologic components and four major LULC types
25 303 from 1995 to 2015*
26
27 ET SURQ GW WYLD LATQ FLOW ARGL FRSE RICE SUGC URML
28 ET
29
30 SURQ 0.56
31 GW -0.48 -0.45
32 WYLD -0.97 0.30 0.34
33
LATQ 0.48 -0.25 -0.60 -0.23
34
35 FLOW -0.78 0.01 0.22 0.61 0.02
36 ARGL 0.22 0.63 -0.57 -0.11 0.23 0.13
37 FRSE 0.19 -0.70 0.19 -0.17 0.03 0.39 -0.59
38
39 RICE -0.28 -0.37 0.44 0.25 -0.36 0.12 -0.28 -0.17
40 SUGC -0.06 -0.45 0.46 0.04 -0.07 -0.08 -0.56 0.05 -0.07
41 URML 0.34 0.53 -0.60 -0.55 0.43 -0.48 0.26 -0.06 -0.58 -0.23
42 304 *) ET: Evapotranspiration, SURQ: Surface Runoff; WYLD: Water Yield; LATQ: Lateral Flow; GW: Groundwater
43 305 flow, FLOW: Streamflow, AGRL: Mixed-Dryland Agriculture, FRSE: Dryland Forest, RICE: Rice field, SUGC:
44 306 Sugar cane/Napier grass association plantation. URML: Settlement. Bold numbers are for p < 0.05
45
46
47 307 In order to relate the land use change to water budget component, the relationship of
48
49 308 changes in each water budget component to changes in each land use in each sub-basin was
50
51
52
309 computed using the non-parametric Spearman rho-correlation(Hughes-et-al.-2012;Zemke-2016).
53
54 310 As indicated in post-classification change detection matrix, some of the major changes in LULC
55
56 311 in 21 years involved forest(FRSE) conversion to dryland agriculture, loss of rice field(RICE) and
57
58
59 312 mixed dryland agriculture(AGRL) to settlement development(URMD). These major LULC
60
61
62 20
63
64
65
 1
2
3
4 313 trajectories were confirmed with relatively high negative correlation(Table-6). Correlation
5
6
7 314 between changes in runoff and changes in LULC types suggest that runoff generation was
8
9 315 associated mainly with the increase in dryland agriculture and settlement, and reduction in the
10
11
12 316 forest area.
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28 A B
29
30
31
32
33 C D
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48 317 Figure 8. Relative impact of LULC changes on runoff due to changes in (A) FRSE, (B) AGRL,
49 318 (C) URMD, and (D) impact on surface runoff for four different LULC change trajectories.
50
51
52
319 Relationship between changes in runoff and these three LULC types(AGRL, FRSE, and
53
54 320 URMD) can be seen in Figure 8A, B and C. The changes to FRSE accounted for 52% (R2 = 0.52)
55
56 321 of runoff variations, as opposed to AGRL and URMD which showed an R2 of 0.35 and 0.11
57
58
59 322 respectively. All three LULC types’ impact on runoff were significant(p <0.1 at level 90% for
60
61
62 21
63
64
65
 1
2
3
4 323 URMD and p <0. for FRSE and AGRL). The lower R2 values are attributed to the presence of
5
6
7 324 several LULC types in each sub-basin with varying degrees of change, the change in runoff might
8
9 325 be a resultant impact of changes in these LULC types. Thus, it is difficult to discern the relative
10
11
12 326 impact of a particular LULC type. The magnitude of increase in runoff can result from an increase
13
14 327 in AGRL, URML or FRSE together in a particular sub-basin. Results from gradual land use change
15
16
17 328 simulations using four major land use change trajectories was depicted in Figure 8D. A linear
18
19 329 relationship between runoff and forest/farm loss and urban development was observed in all types
20
21 330 of simulations. This linear relationship was similar to the simulation result of urban sprawl in
22
23
24 331 Gaza(Eshtawi-et-al.-2016). However, somewhat contrasting findings were reported by Ghaffari et
25
26 332 al.(2010), who found a non-linear relationship between runoff change after a 70% removal of
27
28
29 333 rangeland. Different gradients for different LULC change paths demonstrate the sensitivity of each
30
31 334 LULC type removal to runoff generation. The result confirmed that forest conversion to other
32
33
34 335 LULC types(to AGRL or to URMD) has a pronounced effect on surface runoff(Figure 8D).
35
36 336 4. Discussion
37
38
337 Simulating a 21-year LULC change in Upper Brantas watershed revealed the changes in
39
40
41 338 hydrological processes occurring in this watershed. At the whole basin level, the 21-years of
42
43 339 changes in LULC resulted in a slight increase in long-term annual average runoff and water yield,
44
45
46 340 and a decrease in evapotranspiration and groundwater flow. The overall magnitudes of changes at
47
48 341 basin or catchment level were relatively small. However, at the sub-basin level, the percentages of
49
50
51 342 changes of these elements were higher. Variation among sub-basins was noticeable, especially for
52
53 343 water yield and evapotranspiration. Wagner et al.(2013) highlighted the influence of LULC change
54
55
56
344 on hydrological responses can be masked by complexities found in a large catchment. It can be
57
58 345 argued that in 21 years, the changes in LULC with the biggest change of only 7% of the total area
59
60
61
62 22
63
64
65
 1
2
3
4 346 might not be significant enough to create a distinct change in the long-term annual average. Yet,
5
6
7 347 the canceling-out effect was also found in other studies with more pronounced LULC changes
8
9 348 such as higher than 10% of the total watershed(Gyamfi et al.-2016;Marhaento et al.-2017;Wagner
10
11
12 349 et al.-2013). It is also important to note that the sub-basin areas in Upper Brantas delineated by
13
14 350 SWAT vary greatly from 130 to 24,000 hectares. In addition, Upper Brantas is a catchment with
15
16
17 351 large variation in topography and LULC types. AGRL class represents generic dryland agriculture
18
19 352 in East Java, which can vary in terms of plant types. Farmland classified as mixed-dryland can
20
21 353 have a composition of crops, fruits, vegetables and shrubs or bushes, or even hard trees such as
22
23
24 354 coconuts and mangoes, whose density and species diversity can vary from one piece of land to
25
26 355 another. The variation in magnitudes of changes in four hydrological components supports the
27
28
29 356 study from Bruijnzeel (2004) that variability in topography, soil types, and geological settings can
30
31 357 all play a role in affecting the water balance. Ongley(1996) highlight that smaller basins (< several
32
33
34 358 hundred km2) for more accurately discerning the impact of various bio-physical factors in the
35
36 359 watershed.
37
38
360 Quantitative analysis showed that changes in runoff can mainly be attributed to the loss of
39
40
41 361 forest and an increase in mixed-dryland agriculture and settlement. Sub-basin level analysis for
42
43 362 the Upper Brantas watershed revealed that distinct runoff increases were observed mostly in sub-
44
45
46 363 basins located either in north-east upper parts, where forest encroachment and agricultural
47
48 364 expansion occurred, or in the central part where urbanization is more concentrated. Another factor
49
50
51 365 that can possibly contribute to this is the role of lithology. Groundwater contribution to the stream
52
53 366 flow has been dominant in Brantas Watershed(ratio of GWQ/P and LATQ/P around 40% and
54
55
56
367 20%). The geological formations in these sub-basins are mixed between impervious and pervious
57
58 368 rocks that restrict the water movement to the groundwater. In addition, the relatively dense
59
60
61
62 23
63
64
65
 1
2
3
4 369 dominating parallel – dendritic river patterns in the watershed support close association of rainfall-
5
6
7 370 runoff, which suggest that excessive rainfall, can barely be absorbed. While most sub-basins
8
9 371 exhibited runoff increase, sub-basins 12 and 26 were the only ones experiencing slightly reduced
10
11
12 372 runoff. LULC change in these sub-basins showed a noticeable reduction in mixed dryland
13
14 373 agriculture and smallest forest loss that might have contributed to lessening the runoff magnitudes.
15
16
17 374 However, considering sub-basin 12 is only 139 hectares in area, being the smallest sub-basin, the
18
19 375 pronounced change in groundwater might not be a representative hydrological process.
20
21 376 Results from gradual LULC change simulations using four different LULC change
22
23
24 377 trajectories confirmed that forest conversion to mixed-dryland agriculture is most detrimental. The
25
26 378 changes in runoff were greater than the changes resulted from conversion of forest to settlement,
27
28
29 379 rice to settlement, and mixed-dryland agriculture to settlement. The gradient of runoff changes
30
31 380 from forest to dryland agriculture is steeper than from forest to urban, suggesting that the increase
32
33
34 381 in agriculture might have caused higher runoff than the increase in settlement. The plausible
35
36 382 explanation for this is that the combined effect of topography and LULC characteristics. For
37
38
383 instance, around 43% of total AGRL areas were located on steep terrains with slopes between 25
39
40
41 384 – 45%, while settlement areas were mainly distributed on gentler slopes(below 15%). With intense
42
43 385 rainfall and mild temperature at a higher elevation, mixed-dryland agricultural types in the upper
44
45
46 386 part of the basin are dominated by vegetable farming. The land management practices for growing
47
48 387 vegetables in this region involve traditional terracing against the contour lines that allow water
49
50
51 388 running down the slope directly for better aeration, a favorable condition for vegetable plots. This
52
53 389 condition can amplify the runoff generation within the landscape. On the other hand, settlement
54
55
56
390 areas on gentler slopes, except in the city, are relatively sparse and the presences of trees with
57
58
59
60
61
62 24
63
64
65
 1
2
3
4 391 denser canopy might allow a higher amount of water to be retained from soil and thus produces
5
6
7 392 smaller runoff.
8
9 393 With the apparent impact of surface runoff increases due to urbanization and forest
10
11
12 394 conversion to agriculture, proper management for mitigating the impact of LULC is of importance.
13
14 395 Despite relatively small changes in runoff generation due to the 21-year LULC change, the
15
16
17 396 findings suggest that increasing urbanization and forest conversion to dryland agriculture in Upper
18
19 397 Brantas can impose serious threats in the long run. This will be more complicated with the on-
20
21 398 going rapid economic development within the watershed. Threats from increasing point-source
22
23
24 399 pollutions as a result of increasing industries can exacerbate the watershed condition. Findings
25
26 400 from several studies suggested that increasing runoff leads to increasing sediments and nutrients
27
28
29 401 brought across the landscape through the runoff. Polyakov et al. (2010) found a non-linear increase
30
31 402 in sediment yield with an increase in runoff events in semi-arid watersheds of Arizona
32
33
34 403 while, Michaelides et al.(2012) found a linear increase in nutrient yield associated with the runoff.
35
36 404 Elevated nutrients runoff has been regarded to increase the risks of eutrophication in the receiving
37
38
405 water bodies(Smith et al.-1999). Sutami reservoir within Upper Brantas watershed has been
39
40
41 406 reported to undergo increasing frequency of occurrence of eutrophication. Considering this,
42
43 407 remaining challenges are: how to establish agricultural land use practices that reduce the surface
44
45
46 408 runoff, how to strengthen forest protection to reduce encroachment for agricultural expansion, and
47
48 409 how to establish a settlement that accommodates more green spaces for runoff reduction in the
49
50
51 410 watershed.
52
53 411 This study showed the potential of using a hydrologic modeling framework for a data scare
54
55
56
412 region such as in Upper Brantas, Indonesia. Setyorini et al.(2017) also studied both LULC
57
58 413 change(1987-2006) and climate impacts in the Brantas basin, Indonesia. The results from that
59
60
61
62 25
63
64
65
 1
2
3
4 414 study showed that the significant changes in LULC(from forest to others land use types) and
5
6
7 415 climate, especially temperature impacted the surface runoff and other hydrologic characteristics of
8
9 416 the watershed. However, this study provided the most recent changes in LULC(1995-2015) and
10
11
12 417 its impacts on the watershed characteristics. The increase in surface runoff was predicted by both
13
14 418 studies at the watershed level but this study showed that the effect of LULC change on surface
15
16
17 419 runoff has been was lower in the recent past(this study). The use of Object-Based Image
18
19 420 Analysis(OBIA) using Landsat images that produce very accurate LULC type in this study and
20
21 421 demonstrated that the LULC change in the basin is multidirectional(Table 5) compared to
22
23
24 422 unidirectional, i.e., from forest to another LULC type, as shown in Setyorini et al.(2017).
25
26 423 Moreover, this study provided more in-depth results at the sub-watershed scale and quantified the
27
28
29 424 impact of each LULC change type to the surface runoff (Figure 8). Results can be used to identify
30
31 425 and prioritize the implementation of best management practices in the most vulnerable areas to
32
33
34 426 prevent the deterioration of current situations. Given the escalating issues of sedimentation and
35
36 427 eutrophication in Upper Brantas, future work should focus on expanding SWAT modeling to
37
38
428 investigate sediment and nutrient management in the rapidly urbanizing tropical watershed with a
39
40
41 429 growing population.
42
43 430 5. Conclusions
44
45
46 431 Upper Brantas represents a typical tropical watershed in a developing country undergoing
47
48 432 rapid LULC change. As a trend in most developing countries, the LULC change is expected to be
49
50
51 433 more severe due to increasing globalization, leading to forest conversion for agriculture, and
52
53 434 eventually farm loss for urban development. Lack of regulations and poor watershed management
54
55
56
435 practices have remained as troubling issues in Upper Brantas. With this condition, the LULC
57
58
59
60
61
62 26
63
64
65
 1
2
3
4 436 change can impose more serious threats for the watershed, especially in regions where data is very
5
6
7 437 limited for monitoring changes.
8
9 438 The land use change in Upper Brantas for the past 21 years was relatively similar to those
10
11
12 439 typically observed in developing countries in Southeast Asia, where LULC changes mainly
13
14 440 involve forest loss to compensate for the agricultural expansion and increasing urbanization. Over
15
16
17 441 the past two decades, the overall changes in LULC in Upper Brantas was relatively small but
18
19 442 urban/settlement experienced the largest growth, accounting for an increase around 7% of the total
20
21 443 watershed, followed by an increase in dryland agriculture and forest loss, denoting 3% - 5% of
22
23
24 444 total area. Simulation using two different land use maps revealed slight changes in the long-term
25
26 445 average of water balance components, leading to increased runoff and water yield, and decreased
27
28
29 446 groundwater flow and evapotranspiration. Among all, surface runoff became the component with
30
31 447 the most pronounced increase, denoting to 8% change in long-term average runoff depth, while
32
33
34 448 other variables were less noticeable. The effect of land use change at sub-basin level was more
35
36 449 pronounced, showing high variability in changes of hydrological variables. Surface runoff was
37
38
450 found to be associated with the conversion of forest to dryland agriculture and rice field and
39
40
41 451 dryland agriculture to urban areas/settlement. Significant runoff increases were mostly restricted
42
43 452 to sub-basins experiencing higher loss in forest and an increase in urban expansion. The simulation
44
45
46 453 results indicate the importance of watershed management for a sustainable future. The impact of
47
48 454 increased runoff and water yield is expected to be more evident with continuing urbanization,
49
50
51 455 especially in the upper part of the watershed. Considering the ability of SWAT model for
52
53 456 simulating the hydrological responses in Upper Brantas watershed, future studies should
54
55
56
457 investigate the impact of LULC change on water quality, which will ultimately help to develop a
57
58 458 plan for sustainable use of the watershed.
59
60
61
62 27
63
64
65
 1
2
3
4 459 Acknowledgements
5
6
7
8
460 We gratefully acknowledge the Fulbright Exchange Program for Indonesia because of which the
9
10 461 first author was able to pursue a PhD program at the Department of Geography, University of
11
12 462 Georgia, USA.
13
14
15
16
463 Conflict of Interest:
17
18
19 464 The authors declare that they have no conflict of interest.
20
21
22 465 References
23
24
25 466 Abbaspour KC(2015) SWAT‐ CUP: SWAT Calibration and Uncertainty Programs ‐ A User
26 467 Manual. Swiss Federal Institute Aquatic Science and Technology.
27 468 http://www.neprashtechnology.ca/wp-content/uploads/2015/06/Usermanual_SwatCup.pdf
28
29 469 Adi S, Jänen I, Jennerjahn TC(2013) History of development and attendant environmental changes
30 470 in the Brantas River Basin, Java, Indonesia, since 1970 Asian Journal of Water, Environment and
31
32 471 Pollution 10:5-15
33 472 Astuti, IS (2018) Drivers and Implications Multi-Decadal Land Use Land Cover Change on the
34
35
473 Function of Indonesian Watersheds, PhD thesis, University of Georgia, Athens, GA, USA.
36 474 Bappenas(2012) Analisis Perubahan Penggunaan Lahan di Ekosistem DAS dalam Menunjang
37
38
475 Ketahanan Air dan Ketahanan Pangan; Studi Kasus DAS Brantas. Direktorat Kehutanan dan
39 476 Konservasi Sumber Daya Air. Bappenas.
40
477 Barkey RA, Mappiasse MF, Nursaputra M(2017) Model Of Climate And Land-Use Changes
41
42 478 Impact On Water Security In Ambon City, Indonesia Geoplanning: Journal of Geomatics and
43 479 Planning 4:97-108 doi:10.14710/geoplanning.4.1.97-108
44
45 480 Beck HE, Bruijnzeel LA, van Dijk AIJM, McVicar TR, Scatena FN, Schellekens J(2013) The
46 481 impact of forest regeneration on streamflow in 12 mesoscale humid tropical catchments Hydrol
47 482 Earth Syst Sci 17:2613-2635 doi:10.5194/hess-17-2613-2013.
48
49 483 Becker D, Sultan M, Milewski A et al.(2012) Integrated solutions for hydrologic investigations in
50 484 arid lands Geosphere 8:1588-1605 doi:10.1130/GES00777.1
51
52 485 Badan Pusat Statistik (BPS)(2016) Kota Malang dalam Angka, Indonesia.
53 486 https://malangkota.bps.go.id/
54
55 487 Bruijnzeel LA(2004) Hydrological functions of tropical forests: not seeing the soil for the trees?
56 488 Agriculture, Ecosystems & Environment 104:185-228 doi:10.1016/j.agee.2004.01.015.
57
58 489 Chang H(2007) Comparative streamflow characteristics in urbanizing basins in the Portland
59 490 Metropolitan Area, Oregon, USA Hydrological Processes 21:211-222 doi:doi:10.1002/hyp.6233
60
61
62 28
63
64
65
 1
2
3
4 491 DeFries R, Eshleman KN(2004) Land-use change and hydrologic processes: a major focus for the
5
6 492 future Hydrological Processes 18:2183-2186 doi:10.1002/hyp.5584
7 493 Djuangsih N(1993) Understanding the state of river basin management from an environmental
8
9 494 toxicology perspective: an example from water pollution at Citarum river basin, West Java,
10 495 Indonesia Science of The Total Environment 134:283-292 doi:10.1016/S0048-9697(05)80029-4.
11
12
496 Douglas-Mankin KR, Srinivasan R, Arnold JG(2010) Soil and Water Assessment Tool (SWAT)
13 497 Model: Current Developments and Applications Transactions of the ASABE 53:1423-1431
14 498 doi:10.13031/2013.34915
15
16 499 Eshtawi T, Evers M, Tischbein B(2016) Quantifying the impact of urban area expansion on
17 500 groundwater recharge and surface runoff Hydrological Sciences Journal 61:826-843
18 501 doi:10.1080/02626667.2014.1000916
19
20 502 Francesconi W, Srinivasan R, Pérez-Miñana E, Willcock SP, Quintero M(2016) Using the Soil
21 503 and Water Assessment Tool (SWAT) to model ecosystem services: A systematic review Journal
22 504 of Hydrology 535:625-636 doi:10.1016/j.jhydrol.2016.01.034.
23
24 505 Fulazzaky MA(2014) Challenges of Integrated Water Resources Management in Indonesia Water
25 506 6: 2000-2020 doi:10.3390/w6072000
26
27 507 Fulazzaky MA (2009) Water Quality Evaluation System to Assess the Brantas River Water Water
28 508 Resources Management 23:3019–3033 doi:10.1007/s11269-009-9421-6
29
30 509 Gassman PW, Reyes MR, Green CH, Arnold JG(2007) The Soil and Water Assessment Tool:
31 510 Historical Development, Applications, and Future Research Directions Transactions of the
32
511 ASABE 50:1211-1250 doi: 10.13031/2013.23637
33
34 512 Ghaffari G, Keesstra S, Ghodousi J, Ahmadi H(2010) SWAT-simulated hydrological impact of
35 513 land-use change in the Zanjanrood basin, Northwest Iran Hydrological Processes 24:892-903 doi:
36
37 514 10.1002/hyp.7530.
38 515 Giri S, Qiu Z(2016) Understanding the relationship of land uses and water quality in Twenty First
39
40 516 Century: A review Journal of Environmental Management 173:41-48 doi:
41 517 10.1016/j.jenvman.2016.02.029
42
43 518 Gyamfi C, Ndambuki JM, Salim RW(2016) Hydrological Responses to Land Use/Cover Changes
44 519 in the Olifants Basin, South Africa Water 8:588
45
46
520 Handayani W(2013) Rural-Urban Transition in Central Java: Population and Economic Structural
47 521 Changes Based on Cluster Analysis Land 2:419-436
48
522 Hughes JD, Petrone KC, Silberstein RP(2012) Drought, groundwater storage and stream flow
49
50 523 decline in southwestern Australia Geophysical Research Letters 39 doi:10.1029/2011GL050797
51
524 Kementrian PU(2011) BBWS Brantas. Kementrian Pekerjaan Umum. http://bbwsbrantas.org/.
52
53 525 Accessed 11/11 2018
54 526 Khoi DN, Suetsugi T(2014) The responses of hydrological processes and sediment yield to land-
55
56 527 use and climate change in the Be River Catchment, Vietnam Hydrological Processes 28:640-652
57 528 doi:10.1002/hyp.9620
58
59 529 Lambin EF, Geist HJ(2003) Regional differences in tropical deforestation Environment: Science
60 530 and Policy for Sustainable Development 45:22-36
61
62 29
63
64
65
 1
2
3
4 531 Lehmusluoto P, Machbub B, Terangna N et al.(1997) National inventory of the major lakes and
5
6 532 reservoirs in Indonesia. Research Institute for Water Resource Development, Indonesia.
7 533 Li H, Zhang Y, Vaze J, Wang B(2012) Separating effects of vegetation change and climate
8
9 534 variability using hydrological modelling and sensitivity-based approaches Journal of Hydrology
10 535 420-421:403-418 doi: 10.1016/j.jhydrol.2011.12.033
11
12
536 Liu Q, Yang Z, Cui B, Sun T(2009) Temporal trends of hydro-climatic variables and runoff
13 537 response to climatic variability and vegetation changes in the Yiluo River basin, China
14 538 Hydrological Processes 23:3030-3039 doi: 10.1002/hyp.7414
15
16 539 Margono BA, Potapov PV, Turubanova S, Stolle F, Hansen MC(2014) Primary forest cover loss
17 540 in Indonesia over 2000–2012 Nature Climate Change 4:730 doi:10.1038/nclimate2277
18
19 541 Marhaento H, Booij MJ, Rientjes THM, Hoekstra AY(2017) Attribution of changes in the water
20 542 balance of a tropical catchment to land use change using the SWAT model Hydrological Processes
21 543 31:2029-2040 doi:10.1002/hyp.11167
22
23 544 Michaelides K, Lister D, Wainwright J, Parsons AJ(2012) Linking runoff and erosion dynamics
24 545 to nutrient fluxes in a degrading dryland landscape Journal Of Geophysical Research 117:G00N15
25 546 doi:10.1029/2012JG002071
26
27 547 Milewski A, Sultan M, Al-Dousari A, Yan E(2014) Geologic and hydrologic settings for
28 548 development of freshwater lenses in arid lands 28:3185-3194 doi:10.1002/hyp.9823
29
30 549 Milewski A, Sultan M, Yan E, Becker R, Abdeldayem A, Soliman F, Gelil KA(2009) A remote
31 550 sensing solution for estimating runoff and recharge in arid environments Journal of Hydrology
32
551 373:1-14 doi: 10.1016/j.jhydrol.2009.04.002.
33
34 552 Misnistry of Forestry (MOF)(1987) A Guideline of Land Rehabilitation and Soil Conservation.
35 553 Ministry of Forestry, Jakarta, Indonesia.
36
37 554 Moriasi DN, Arnold JG, Van Liew MW et al.(2007) Model Evaluation Guidelines for Systematic
38 555 Quantification of Accuracy in Watershed Simulations Transactions of the ASABE 50:885-900 doi:
39
40 556 10.13031/2013.23153
41 557 Nash JE, Sutcliffe JV(1970) River flow forecasting through conceptual models part I — A
42
43 558 discussion of principles Journal of Hydrology 10:282-290 doi: 10.1016/0022-1694(70)90255-6.
44 559 Neitsch SL, Arnold JG, Williams JR(1999) Soil and water assessment tool user’s manual version
45
46
560 2000. Texas Water Resources Institute, College Station, Texas, US.
47 561 Neitsch SL, Arnold JG, Kiniry JR, Williams JR(2011) Soil and water assessment tool theoretical
48
562 documentation version 2009. Texas Water Resources Institute, College Station, Texas, US.
49
50 563 Ochoa-Tocachi BF, Buytaert W, De Bièvre B(2016) Regionalization of land-use impacts on
51
564 streamflow using a network of paired catchments Water Resources Research 52:6710-6729
52
53 565 doi:10.1002/2016WR018596.
54 566 Ongley ED (1996) Control of water pollution from agriculture. Food and Agriculture Organization
55
56 567 of the United Nations.
57 568 Othman F, Sholichin M(2008) Predicting the Impacts of Fertilizer Usage on Water Quality,
58
59 569 Ssutami Reservoir, Indonesia Journal of Environmental Hydrology 16
60
61
62 30
63
64
65
 1
2
3
4 570 Partoyo, Shrestha RP(2013) Monitoring farmland loss and projecting the future land use of an
5
6 571 urbanized watershed in Yogyakarta, Indonesia Journal of Land Use Science 8:59-84
7 572 doi:10.1080/1747423X.2011.620993
8
9 573 Pawitan H, Haryani GS(2011) Water resources, sustainability and societal livelihoods in Indonesia
10 574 Ecohydrology & Hydrobiology 11:231-243 doi: 10.2478/v10104-011-0050-3
11
12
575 Pingale SM, Khare D, Jat MK, Adamowski J(2014) Spatial and temporal trends of mean and
13 576 extreme rainfall and temperature for the 33 urban centers of the arid and semi-arid state of
14 577 Rajasthan, India Atmospheric Research 138:73-90 doi: 10.1016/j.atmosres.2013.10.024
15
16 578 Polyakov VO, Nearing MA, Nichols MH, Scott RL, Stone JJ, McClaran MP(2010) Long-term
17 579 runoff and sediment yields from small semiarid watersheds in southern Arizona Water Resources
18 580 Research 46:W09512 doi: 10.1029/2009WR009001
19
20 581 Preston TM, Kim K(2016) Land cover changes associated with recent energy development in the
21 582 Williston Basin; Northern Great Plains, USA Science of The Total Environment 566-567:1511-
22 583 1518 doi: 10.1016/j.scitotenv.2016.06.038
23
24 584 Rahayuningtyas C, Wu R-S, Anwar R, Chiang L-CJT, Atmospheric, Sciences O(2014) Improving
25 585 AVSWAT Stream Flow Simulation by Incorporating Groundwater Recharge Prediction in the
26
27
586 Upstream Lesti Watershed, East Java, Indonesia Terrestrial, Atmospheric & Oceanic Sciences
28 587 25:881-892.
29
588 Rostamian R, Jaleh A, Afyuni M, Mousavi SF, Heidarpour M, Jalalian A, Abbaspour KC(2008)
30
31 589 Application of a SWAT model for estimating runoff and sediment in two mountainous basins in
32 590 central Iran Hydrological Sciences Journal 53:977-988 doi:10.1623/hysj.53.5.977
33
34 591 Rutten M, van Dijk M, van Rooij W, Hilderink H(2014) Land Use Dynamics, Climate Change,
35 592 and Food Security in Vietnam: A Global-to-local Modeling Approach World Development 59:29-
36 593 46 doi: 10.1016/j.worlddev.2014.01.020
37
38 594 Sajikumar N, Remya RS(2015) Impact of land cover and land use change on runoff characteristics
39 595 Journal of Environmental Management 161:460-468 doi: 10.1016/j.jenvman.2014.12.041
40
41 596 Setyorini A, Khare D, Pingale SMJAG(2017) Simulating the impact of land use/land cover change
42 597 and climate variability on watershed hydrology in the Upper Brantas basin, Indonesia Applied
43 598 Geomatics 9:191-204 doi:10.1007/s12518-017-0193-z
44
45 599 Seyoum WM, Milewski AM, Durham MC(2015) Understanding the relative impacts of natural
46 600 processes and human activities on the hydrology of the Central Rift Valley lakes, East Africa
47
601 Hydrological Processes 29:4312-4324 doi:10.1002/hyp.10490.
48
49 602 Silveira L, Alonso J(2009) Runoff modifications due to the conversion of natural grasslands to
50 603 forests in a large basin in Uruguay Hydrological Processes 23:320-329 doi:10.1002/hyp.7156
51
52 604 Smith VH, Tilman GD, Nekola JC(1999) Eutrophication: impacts of excess nutrient inputs on
53 605 freshwater, marine, and terrestrial ecosystems Environmental Pollution 100:179-196 doi:
54
55 606 10.1016/S0269-7491(99)00091-3.
56 607 Solar RRDC, Barlow J, Andersen AN, Schoereder JH, Berenguer E, Ferreira JN, Gardner
57
58 608 TA(2016) Biodiversity consequences of land-use change and forest disturbance in the Amazon: A
59 609 multi-scale assessment using ant communities Biological Conservation 197:98-107 doi:
60 610 10.1016/j.biocon.2016.03.005
61
62 31
63
64
65
 1
2
3
4 611 Sulastri. AAM, Suryono T(2004) Blooming Alga Dinoflagelata Ceratium hirudinella di Waduk
5
6 612 Karangkates , Malang, Jawa Timur. Oseanologi dan Limnologi Indonesia 36:51-67
7 613 United States Geological Survey (USGS)(2017) EarthExplorer - Home.
8
9 614 https://earthexplorer.usgs.gov/
10 615 Valentin C, Agus F, Alamban R et al.(2008) Runoff and sediment losses from 27 upland
11
12
616 catchments in Southeast Asia: Impact of rapid land use changes and conservation practices
13 617 Agriculture, Ecosystems & Environment 128:225-238 doi:10.1016/j.agee.2008.06.004
14
618 van Griensven A, Meixner T, Grunwald S, Bishop T, Diluzio M, Srinivasan R(2006) A global
15
16 619 sensitivity analysis tool for the parameters of multi-variable catchment models Journal of
17 620 Hydrology 324:10-23 doi: 10.1016/j.jhydrol.2005.09.008
18
19 621 Vörösmarty CJ, Green P, Salisbury J, Lammers RB(2000) Global Water Resources: Vulnerability
20 622 from Climate Change and Population Growth Science 289:284 doi:10.1126/science.289.5477.284
21
22 623 Wagner PD, Kumar S, Schneider K(2013) An assessment of land use change impacts on the water
23 624 resources of the Mula and Mutha Rivers catchment upstream of Pune, India Hydrol Earth Syst Sci
24 625 17:2233-2246 doi:10.5194/hess-17-2233-2013
25
26 626 Wang G, Yang H, Wang L, Xu Z, Xue B(2014) Using the SWAT model to assess impacts of land
27 627 use changes on runoff generation in headwaters Hydrological Processes 28:1032-1042
28 628 doi:10.1002/hyp.9645
29
30 629 Widianto D, Lestariningsih ID(2001) Implementasi Kaji Cepat Hidrologi (RHA) di Hulu DAS
31 630 Brantas. Jawa Timur, Working paper.
32
33 631 Zehe E, Sivapalan MJH, Sciences ES(2009) Threshold behaviour in hydrological systems as
34 632 (human) geo-ecosystems: manifestations, controls, implications Hydrology and Earth System
35 633 Sciences 13:1273-1297
36
37 634 Zemke JJ(2016) Runoff and Soil Erosion Assessment on Forest Roads Using a Small Scale
38 635 Rainfall Simulator Hydrology 3:25
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 32
63
64
65
Map of Indonesia, East Java and Upper Brantas watershed with corresponding Click here to access/download;colour figure;Figure 1.JPG
elevation and location of gaging stations.
Land Use/Land Cover (LULC) in Upper Brantas Watershed for (A) 1995 and (B) 2015. Click here to access/download;colour figure;Figure 2.JPG
Soil Map for Upper Brantas (Source: FAO, Harmonized World Soil Database). Click here to access/download;colour figure;Figure 3.JPG
(A) Rainfall trend over 25 years (1990 – 2015) and (B) Mean monthly rainfall and Click here to access/download;colour figure;Figure 4.JPG
discharge data for the period 1998 – 2014 in Upper Brantas.
(A) Plot of observed and simulated flow used for calibration and validation. Observed Click here to access/download;colour figure;Figure 5.JPG
versus simulated flow for (B) calibration, and (C) validation.
LULC change impacts on watershed’s water balance between 1995 and 2015: (A) % Click here to access/download;colour figure;Figure 6.JPG
change; (B) absolute change; and (C) annual trends of changes.
Changes (in %) at sub-basin level for (A) water yield, (B) groundwater, (C) Click here to access/download;colour figure;Figure 7.JPG
evapotranspiration and (D) surface runoff.
Relative impact of LULC changes on runoff due to changes in (A) FRSE, (B) AGRL, (C) Click here to access/download;colour figure;Figure 8.JPG
URMD, and (D) impact on surface runoff for four different LULC change trajectories.