Impacts of land-use changes on watershed discharge and water quality in a large intensive agricultural area in Thailand

ABSTRACT

The objectives of this research were to evaluate the effect of land-use change on streamflow, sediment and water quality data along the Lower Yom River, Thailand, covering an intensively agricultural area of 14 613.6 km², and to assess the relative impact of point and non-point sources of pollution from multiple-land-use watersheds. Long-term calibration and validation of the SWAT (Soil and Water Assessment Tool) model was performed on data for 2000–2013. Land-use change led to a 13–49% increase in runoff in the basin and resulted in 37–427% increased sediment yield. The amount of NO₃-N load doubled in the upper and middle parts of the study area, while the increase in PO₄^3– ranged from 37 to 377%, reflecting the increase in agricultural lands and urban areas. It is concluded that the changed land use is closely associated with the quantity of runoff, sediment yield and the NO₃-N and PO₄^3– concentrations.

KEYWORDS:

• land-use change
• intensively agricultural areas
• nitrate
• phosphate
• SWAT
• Yom River Basin
• Thailand

Editor A. Castellarin; Associate editor X. Fang

1 Introduction

Land-use/land-cover (LULC) change is one of the main contributing factors, affecting discharge and water quality in watersheds, due to anthropogenic activities, for example, increasing population and urbanization, encroachment of forests by agriculture and urban areas, and the degradation of forest resources (Konkul et al. 2014, Waiyasusri et al. 2016). This can result in a decrease in infiltration and, in turn, an increase in the rate and volume of surface runoff (Chotpantarat and Chanyotha 2003, Fitzpatrick 2005, Klongvessa et al. 2017), as well as having a negative impact on water quality, which is a major concern in terms of drinkable water supplies (Buda and DeWalle 2009, Oeurng et al. 2010). Nitrogen (N) and phosphorus (P) are nutrients that are often the limiting factors in plant growth. Thus, they are heavily applied as fertilizer on agricultural land, but significant other sources include wastewater treatment effluent, animal waste, antiseptics and detergents. Agricultural activities are an important source of these nutrients (Yang et al. 2008). When nitrate (NO₃^–) and phosphate (PO₄^3–) flow into water sources, the elevated levels of these nutrients can cause accelerated eutrophication or enrichment of the waters that, in turn, stimulates the growth of algae and changes the ecology of the stream as the algal populations increase (Bhavya et al. 2016). This is a serious problem that can have many negative impacts on aquatic systems and the use of water for different purposes, such as fishing, industry and farming, and especially for drinking (Caldy 2004, Maier et al. 2009). Along with potassium, NO₃-N and PO₄^3– are the common components of multi-nutrient fertilizers, where, for example, a fertilizer grade of 16–16–16 contains (by weight) 16% nitrogen, 16% PO₄^3–, and 16% potassium. Contamination of groundwater and rivers with NO₃-N and PO₄^3– from applied fertilizers, or runoff from animal feed, is ubiquitous in agricultural regions (Chotpantarat et al. 2011, Liljeström et al. 2012, Boithias et al. 2014, Li and Bush 2015, Liu et al. 2017). Inappropriate management of sewage treatment plants and septic frameworks are other significant sources through which effluent is released into waterways. Furthermore, PO₄^3– comes from residential sewage (especially from clothing cleansers). The national drinking standard in the USA for NO₃-N is ≤ 10 ppm, but a NO₃-N level of only 1 ppm is sufficient to cause algal blooms (Boeykens et al. 2017). The desired maximum total phosphorus level of the US Environmental Protection Agency (EPA) for streams entering lakes and for other flowing water bodies was 0.05 and 0.10 ppm, respectively (Borah and Bera 2004). An amount of 0.02 ppm of phosphorus has appeared to affect algal growth in lakes (Sharpley et al. 1987).

The SWAT (Soil and Water Assessment Tool) model, developed by the Blackland Research Center in Texas, USA, is a river basin model that includes physical, biogeochemical and hydrochemical processes, and combines components of a physical and conceptual semi-empirical nature (Neitsch et al. 2005a). Therefore, it can be capable of quantifying and simulating the impact of land-use change on water flow, sediment flux and nutrient loadings in a river basin (Arnold et al. 1999, Krysanova et al. 2005). Worldwide studies have been done using the SWAT model, and it has been applied to watershed-scale projects in different parts of the world such as the USA, Europe (e.g. Arnold et al. 2012, Mbonimpa et al. 2012, Jha and Gassman 2014, Havrylenko et al. 2016), China (Hao et al. 2004), India (Kaur et al. 2004), Australia (Sun and Cornish 2006), Africa (Schuol and Abbaspour 2006) and Southeast Asia (Oeurng et al. 2016, Shrestha et al. 2016). Results have shown that the SWAT model is capable of assessing streamflow and sediment loads of a river basin with satisfactory accuracy. Several researchers (e.g. Cui et al. 2015, Andriamananjara et al. 2016) have performed hydrological SWAT modelling to evaluate the impact of land-use changes, including agricultural management, on the amount of NO₃-N that affects the water quality of rivers: for example, at the Jajrood River watershed, Iran (Jamshidi and Tajrishy 2010), the Hii River Basin, Japan (Takeda 2012), and in the northeast of China (Ouyang et al. 2017). Therefore, the SWAT model is considered to be a suitable tool for evaluating the impact of land-use change on nitrate and phosphate levels in surface waters. The results from the SWAT modelling can provide useful information for preventing non-point source (NPS) pollution in terms of land-use management within those regions. Moreover, in recent years, the effects of climate change have been taken into account together with the effects of land-use changes by using SWAT modelling to explain hydrological processes (Eugenio et al. 2014, El-Khoury et al. 2015, Krysanova and White 2015). However, as mentioned above, most of the research focused on the effects of streamflow and only some focused on nutrient loads, and these were carried out in small- to medium-scale watersheds. Only a few studies have applied the SWAT model to assess the impact of land-use changes on water flow, sediment load and water quality in a large, intensively agricultural river basin, especially in tropical areas. Therefore, in a large and intensively agricultural area, the assessment of such impacts needs to use the appropriate hydrological model to assess the gradual changes in terms of water quantity and quality and, moreover, it needs to use distributed monitoring stations with long-term recording data.

Interestingly, during the past few decades, land use in the Yom River Basin, Thailand, covering an area of approx. 23 948 km², has been gradually changing from mainly forest land to mostly agricultural land (paddy fields, field crops and perennial crops), especially in the lower part of the basin (area: 1359 km²), followed by urban and developed land (approx. 310 km²) (Kitichuchairit 2013). To bridge the gap in the previous research, the current study applied a SWAT model to assess the effect of land-use change on the streamflow and sediment, including nitrate and phosphate, in the Lower Yom River, one of the largest drainage areas in Thailand with distributed and long-term monitoring data (13 years of recorded data). The main objective of this research was to assess the effects of land-use changes on streamflow, sediment transport, and the NO₃-N and PO₄^3– concentrations discharging along the Lower Yom River over a long period of time (2000–2013). The SWAT model was used to investigate the influence of spatial and temporal changes in the intensive agricultural catchment in terms of point and non-point sources, such as urbanization and agricultural activities.

2 Materials and methods

2.1 Study area

The study was focused on the lower part of the Yom River Basin (Fig. 1), which is located in the northern part of Thailand, covering the seven provinces of Phrae, Sukhothai, Uttaradit, Phitsanulok, Kamphaeng, Phichit and Nakhon Sawan. The lower part of the Yom River and its tributaries drain a total area of 14 613.7 km², about 60.8% of the Yom River Basin (Hydro and Agro Informatics Institute 2012). The river is approximately 420 km long and the people living along it are mainly farmers (RID [The Royal Irrigation Department] 2008). The increased population in the area has led to a continuous deterioration of the river water quality, especially in the Lower Yom River. Recently, the situation has become worse due to increased runoff from various agricultural activities and wastewater discharged from urban areas and industries.

Figure 1. Lower Yom River Basin showing the 14 water sampling points.

Moreover, in the period May–October, when crops are grown and fertilizer is applied, between three and five typhoons strike Thailand, influencing an episodic runoff in the Lower Yom River Basin, resulting in a large amount of erosion and nutrient load to the catchment system. Furthermore, in this area, agriculture-associated land use is limited by government policy, with nutrient export contingent on the runoff and land use.

2.2 Analysis of nitrate-N and phosphate levels in the Yom River

The water quality of the surface water was ascertained from samples collected at 14 stations, as detailed in Table 1 and Figure 1.

Table 1. Location of surface water stations.

Station	Location
YO.01	Phothale district, Phijit province
YO.02	Phopratubchang district, Phijit province
YO.03	Samngam district, Phijit province
YO.04	Bangrakam district, Phitsanulok province
YO.05	Mueang district, Sukhothai province
YO.06	Tumbon Pakquare Meuang district, Sukhothai province
YO.07	Sawankalok district, Sukhothai province
YO.08	Srisatchanalai district, Sukhothai province
YO.09	Wangching district, Phrae province
W01	Discharging waste water point of Bangrakam district, Phitsanulok province
W02	Discharging waste water point of Samngam district, Phijit province
W03	Discharging waste water point of Meuang district, Sukhothai province
W04	Discharging waste water point of Sawankalok district, Sukhothai province
W05	Discharging waste water point of Srisatchanalai district, Sukhothai province

The samples were collected from these 14 stations (YO.01–09 and W01–05) in May (summer) and August (rainy season) during 2012 and 2013. The water samples were vertically collected at mid-depth at the centre of the width of the river and transferred into separate 1-litre HDPE plastic bottles and adjusted to pH < 2 with sulphuric acid. Samples were then stored for subsequent analysis of the NO₃-N and PO₄^3– contents. All samples were labelled with the station name, packed in foam boxes, and kept at 4°C to reduce bacterial, physical and chemical processes. The samples were then taken to the laboratory at the Department of Environmental Science, Chulalongkorn University, for further analysis. The Brucine method was employed for analysis of the NO₃-N in the water samples according to the standards set out in the Standard Methods for the Examination of Water and Wastewater (APHA, AWWA, WEF 2005).

2.3 SWAT model description

The SWAT is a physical watershed model operating on a daily time step (Neitsch et al. 2005a). It depends on the topographical, geological, soil property, vegetative and input data of land management and uses both physical and empirical components to predict runoff, sediment and the chemical properties of agricultural yields from the impacts of land management practices in large watersheds over long periods of time (Neitsch et al. 2005b). The model inputs consist of the digital elevation model (DEM), soil type, land use, daily rainfall, maximum and minimum temperatures, relative humidity, radiation and wind speed, which allow users to monitor and assess the scenarios by using substitute input data such as weather, land-use activities and land cover. The SWAT model is separated into homogenous units referred to as hydrological response units (HRUs) according to the unique characteristics of the soil, land use and slope in each sub-watershed. The Soil Conservation Service curve number (CN) technique and daily rainfall are used for calculating the surface runoff volume. Moreover, the sediment yields from each sub-basin are simulated by using the Universal Soil Loss Equation (USLE; Williams and Berndt 1977). In addition, non-point source pollutants discharged into the streamflow are evaluated by calculating the surface flow, lateral flow and seeped water from 10 cm of the main layer of soil (Arnold et al. 1998).

2.4 Preparing input data for the SWAT model

The watershed was delineated into 29 HRUs, with the necessary data types and their data sources for the SWAT model set-up shown in Table 2; the input map is shown in Figure 2. The main database of the SWAT model was considered and applied mainly from the Mekong River Commission (MRC), which developed the soil database for SWAT modelling in the Mekong River Basin, and which could be applied as the Southeast Asian database, as mentioned in previous studies (Wangpimool et al. 2010, MRC [Mekong River Commission] 2010, Oeurng et al. 2016). Details of agricultural activities and the period of fertilizer application in the study area, as derived from the Office of Agricultural Economics, Thailand (2012), are given in Supplementary Material (Table S1).

Table 2. Data sources for the SWAT (Soil and Water Assessment Tool) model.

Main theme	Year	Sub-theme	Data sources
DEM		DEM	Derived from elevation data with GIS
Soil group	2001	Soil texture	Derived from 1:50 000 scale digital map of the Land Development Department (LDD)
Land use	2003, 2009	Land use	Derived from the LDD
Hydrology	2004	Sub-basin	Derived from digital map of the Water Resources Department
		Stream
		Hydro station	Royal Irrigation Department (RID)
Meteorology	2000–2013	Precipitation	Thai Meteorological Department (TMD)
		Temperature
		Relative humidity
		Solar radiation
		Wind speed
		Meteorology station
Water quality	2008–2013	NO3^–N	Region 2 Environment Office (EO), Lampang province and Region 3 EO, Phitsanulok province, Ministry of Natural Resources and Environment
		PO4^3−
Sediment	2006–2013	Sediment yields	RID
Reservoirs and diversion data	2013		Region 3 RID Office, Phitsanulok province
Management operations	2013	Rotation timings in planting and fertilizer application	Agricultural Land Reform Office, Ministry of Agriculture and Cooperatives

Figure 2. Spatial input maps for the SWAT model in the Lower Yom River showing (a) the digital elevation model (DEM), (b) soil groups, (c) land use in 2003, and (d) land use in 2009.

2.5 SWAT model calibration and validation

The SWAT display was adjusted by comparing the simulated water flow with the field-measured flow values from the observation stations of the watershed outlet. For the calibration process, the model was run with input parameters, and some parameters were gradually adjusted to get a good agreement. Afterwards, the model was validated without any adjusted parameters, which were set in the calibration process. All input data, such as land cover and soil type, were considered the same, except for the inputs of weather conditions (Heuvelmans et al. 2004).

The set of 13 years of monitoring data (2000–2013) for the observed flow obtained from the hydrological stations of the Royal Irrigation Department (RID), Thailand, was separated into three periods consisting of a calibration period and two validation periods. The SWAT model was first calibrated using the land-use map of 2009 with the observed flow data for the period 2006–2009. The first validation was carried out using the observed flow data for 2010–2013 (see Fig. 4(a)). Then, the land-use map of 2003 was used for validation of the observed flow data for the period 2000–2004.

Figure 3. Observation stations in the Lower Yom River Basin used for the SWAT model calibration and validation.

Figure 4. Calibration and validation periods of (a) streamflow, (b) sediment and (c) nitrate and phosphate.

For the sediment transport, the SWAT model was calibrated and validated using the land-use map of 2009 with 7 years of observed sediment data (2006–2013), which were obtained from the RID (Fig. 4(b)). The observed sediment data for the period 2006–2009 were used for the calibration, whilst those for the period 2010–2013 were used for validation.

After finishing the hydrological and sediment calibration and validation, the SWAT model was calibrated using the land-use map for 2009 with the observed NO₃-N and PO₄^3– levels for the 2008–2010 period, which were obtained from the local environmental offices in the Lampang and Phitsanulok provinces. This was validated with the observed data for the period 2011–2013, including the observed results from this current study (see Fig. 4(c)). The details and location of the observation stations used for the model calibration and validation are shown in Table 3 and Figures 3 and 4.

Table 3. Location of the hydrological observation stations used for the calibration and validation of flow, sediment yields and NO₃⁻ and PO₄³⁻ levels.

Station	Location	Calibration	Validation
Y.6	Srisatchanalai, Sukhothai	Flow (2006–2009)	Flow (2000–2004, 2010–2013)
Y.4	Mueang, Sukhothai	Sediments (2006–2009)	Sediments (2010–2013)
Y.16	Bangrakam, Phitsanulok	NO3^–N (2008–2010)	NO3^–N (2011–2013)
Y.17	Samnagm, Phichit	PO4^3− (2008–2010)	PO4^3− (2011–2013)
Y.5	Phothale, Phichit

The predicted discharge, sediment, and NO₃-N and PO₄^3– levels were statistically evaluated using the coefficient of determination (R²), which is used as an indicator of model performance for the calibration and validation periods. Models with higher R² are presumed to perform better than those with a lower coefficient (Gikas et al. 2006). Additional parameters, such as those for the flow, sediment, and NO₃-N and PO₄^3– calibration and validation procedures are shown in Table 4.

3 Results and discussion

3.1 Water quality analysis

The water quality was determined from the surface water samples collected at 14 stations of the Lower Yom River over two years (2012 and 2013), as described in Section 2.2. The average water pH and temperature were 7.6 and 30.6°C, respectively, while the average dissolved oxygen (DO), biological oxygen demand (BOD), and NO₃-N and PO₄^3– concentrations were 5.10, 3.50, 1.02 and 0.74 mg L⁻¹, respectively. The water quality for the Sawankalok district, Sukhothai province (station YO.07), and Wangching district, Phrae province (YO.09), was defined as good for use in fisheries, while river quality measured in the Mueang district, Sukhothai province stations (YO.06 and W03), was lower than the Thailand Surface Water Quality Standard and Classification: the ambient water quality standards established in 1994 that serve as guidelines for water based on major beneficial uses (Pollution Control Department 2000). The results of the water quality analysis for six parameters are shown in Figure 5.

Figure 5. Analysis of water quality of the Lower Yom River Basin showing the average temperature, pH, dissolved oxygen (DO), biochemical oxygen demand (BOD), NO₃-N and PO₄³⁻ levels.

3.2 Parameter sensitivity analysis

Sensitivity analysis for the water flow was performed for nine hydrological parameters that have the potential to influence monthly water flow. The range of variation of these parameters was based on a listing provided in the SWAT manual (Neitsch and Sims 2002). Further analysis was then performed for the four most influential hydrological parameter – CN (curve number), SOL_AWC (available soil water capacity), GWQMN (threshold depth of the water in the shallow aquifer required for return flow), and GW_DELAY (groundwater delay time) – by trial and error to establish a dimensionless number related to the land use, soil type and weather conditions. The response of the surface runoff was evaluated when the parameters were changed from -60% to +60% by a sequential increment of 20%. The two sediment parameters – USLE equation support practice (P) factor, and SPEXP (exponential parameter for calculating sediment restrained in channel sediment routing) – were used to evaluate the sediment yields. The parameters that had a high impact on the NO₃-N and PO₄^3– concentrations were RSDCO (residue decomposition coefficient), NPERCO (nitrogen percolation coefficient), SHALLST_N (initial concentration of NO₃-N in the shallow aquifer), PPERCO (phosphorus percolation coefficient), and GWSOLP (concentration of soluble P in groundwater). Therefore, parameters from the sensitivity analysis of the model can help us understand the complex hydro-geological interactions of the watershed.

3.3 Calibration and validation of the SWAT model

The simulated flow results were compared with the observed data from the Y.6, Y.4, Y.16, Y.17 and Y.5 hydrological stations (see Fig. 3 for station locations). The model input parameters for the calibration and validation processes in the observed and simulated streamflow were adjusted by matching the guidelines according to the parameter sensitivity analysis (Section 3.2). The calibration period (2006–2009) and the validation periods (2000–2004 and 2010–2013) showed good relationships (see figure captions for R² values) between the monthly data from the observed and simulated flow (Fig. 6), monthly sediment yield (Fig. 7), daily NO₃-N concentrations (Fig. 8), and daily PO₄^3– levels (Fig. 9). Thus, there was a strong correlation between the observed and simulated results.

Figure 6. Comparison of the observed and simulated results for the monthly water flow (m³ s⁻¹; CMS) at, from top to bottom, hydrological stations Y.6 (R² = 0.85), Y.4 (R² = 0.77), Y.16 (R² = 0.74), Y.17 (R² = 0.71), and Y.5 (R² = 0.71).

Figure 7. Comparison of the observed and simulated results for monthly sediment yield at, from top to bottom, hydrological stations Y.6 (R² = 0.69), Y.4 (R² = 0.64), Y.16 (R² = 0.71), Y.17 (R² = 0.63), and Y.5 (R² = 0.68).

Figure 8. Comparison of the observed and simulated results for the river water NO₃-N concentrations at, from top to bottom, hydrological stations Y.6 (R² = 0.66), Y.4 (R² = 0.73), Y.16 (R² = 0.86), Y.17 (R² = 0.64), and Y.5 (R² = 0.70).

Figure 9. Comparison of the observed and simulated results for the river water PO₄³⁻ concentrations at, from top to bottom, hydrological stations Y.6 (R² = 0.70), Y.4 (R² = 0.61), Y.16 (R² = 0.87), Y.17 (R² = 0.91), and Y.5 (R² = 0.65).

3.4 Impact of land-use changes in the Lower Yom River Basin

Most of the land use in the Lower Yom River Basin was represented by forest area followed by agricultural, urban and stream areas. The 2003 land-use map was overlaid with the land-use map for 2009 to identify the changes in land use between 2003 and 2009 (Fig. 10 and Table 5). According to the change detection results presented in Table 5 and Figure 10, much of the forest land has been transformed into agricultural land (paddy fields, field crops, and perennial crops), with an area of 603.35 km² or 4.29%; this was followed by water bodies (area: 48.07 km², 0.34%), and urban and built-up land (area: 23.46 km², 0.17%). In contrast, most of the urban and built-up land areas increased, to cover 251.69 km² (1.79%). Most of the urban and built-up land was transformed from field crops (area: 140.66 km², 1.00%), followed by paddy fields (area: 84.34 km², 0.60%) and some perennial crops (area: 3.23 km², 0.02%).

Table 4. Description and optimal values used in the SWAT model calibration/validation.

			Optimal value at sub-basins:
Parameter	Model process	Description	1–15	16–22	23–25	26–29
CN	Flow	SCS runoff curve number for moisture condition II	*0.65	*0.3	*0.4	*0.4
SOL_AWC	Flow	Available water capacity of the soil layer (mm H2O/mm soil)	*0.9	*0.9	*0.9	*0.98
GWQMN	Flow	Threshold depth of water in shallow aquifer for percolation to occur (mm)	23	15	15	15
GW_DELAY	Flow	Groundwater delay time (d)	50	100	100	70
ESCO	Flow	Soil evaporation compensation coefficient	0.05	0.05	0.05	0.05
RCHRG_DP	Flow	Deep aquifer percolation fraction	0.5	0.5	0.5	0.5
ALPHA_BF	Flow	Baseflow alpha factor	0.08	0.08	0.08	0.08
GW_REVAP	Flow	Groundwater “revap” coefficient	0.02	0.02	0.02	0.02
REVAPMN	Flow	Threshold water level in shallow aquifer for “revap”	0.0	0.0	0.0	0.0
USLE_P	Sediments	USLE equation support practice (P) factor	0.0005	0.0005	0.0005	0.0005
SPEXP	Sediments	Exponential parameter for calculating sediment re-entrained in channel sediment routing	1.5	1.5	1.5	1.5
SPCON	Sediments	Linear parameter for calculating the maximum amount of sediment	0.3	0.1	0.1	0.1
RSDCO	Nitrate-N	Residue decomposition coefficient	0.09	0.09	0.09	0.09
NPERCO	Nitrate-N	Nitrogen percolation coefficient	0.05	0.05	0.05	0.05
SHALLST_N	Nitrate-N	Initial concentration of nitrate in shallow aquifer	0.10	0.10	0.12	0.12
RSDCO	Phosphate	Residue decomposition coefficient	0.05	0.05	0.05	0.05
PPERCO	Phosphate	Phosphorus percolation coefficient	10	10	10	10
GWSOLP	Phosphate	Concentration of soluble P in groundwater	0.09	0.10	0.10	0.10

Note: * means the default values of the parameters are multiplied by the number following the *

Table 5. Change detection for the Lower Yom River, 2003–2009.

Change	Area (km^2)	Percentage area (%)
Forest to Fieldcrop	355.69	2.53
Forest to Paddy	147.98	1.05
Forest to Perennial	99.68	0.71
Forest to Water	48.07	0.34
Forest to Urban	23.46	0.17
Fieldcrop to Urban	140.66	1.00
Paddy to Urban	84.34	0.60
Perennial to Urban	3.23	0.02

Figure 10. Land-use changes in the Lower Yom River from 2003 to 2009.

As mentioned above, the largest changes in area were into paddy fields and urban areas due to commercial and industrial growth. In order to investigate the land-use changes in an intensive agricultural catchment, the impact of land-use changes on the runoff, sediment, NO₃-N and PO₄^3– levels was simulated by rerunning the SWAT model using the 2003 and 2009 land-use data with the same 2009 meteorological input from the simulation periods from the stations of the RID and the water quality monitoring stations of the Regional Environment Office in Thailand. The sub-basin area simulated by the SWAT model covered five stations (Y.6, Y.4, Y.16, Y.17 and Y.5) on the lower reaches of the Yom River (Fig. 10).

3.4.1 Runoff

The effect of the change in land use between 2003 and 2009 on the amount of runoff is shown in Figure 11. The simulations revealed that the annual runoff had increased by 34.0% (1620 × 10⁶ m³, MCM (million cubic meters)), 13.2% (889 × 10⁶ m³), 48.6% (3926 × 10⁶ m³), 33.2% (2357 × 10⁶ m³) and 28.8% (1274 × 10⁶ m³) at stations Y.6, Y.4, Y.16, Y.17 and Y.5, respectively. Furthermore, the monthly peak flow at station Y.6 in August 2009 was 680 m³ s⁻¹ (CMS), which was 35% higher than in September 2003. For the other stations (Y.4, Y.16, Y.17 and Y.5), the monthly peak flow in September 2009 was 675, 1126, 856 and 856 m³ s⁻¹, respectively, which was 15, 91, 25 and 25% higher, respectively, than in August 2003.

Figure 11. Effects of land-use change between 2003 and 2009 on the runoff (m³ s⁻¹; CMS).

Thus, the change in land use between 2003 and 2009 led to a 13–49% increase in runoff in the Lower River Yom Basin, ranging from 900 × 10⁶ m³ (station Y.4) to 3920 × 10⁶ m³ (station Y.16) (Fig. 12(a)). Moreover, the maximum flow velocity increased by 15–90%, ranging from about 15 m³ s⁻¹ (Y.4) to 380 m³ s⁻¹ (Y.16). This is likely to be because most of the forest area above station Y.16 was converted to agricultural areas (rice fields, field crops and perennial crops), with an area of 526.2 km². In addition, the runoff chart at each station revealed that the maximum runoff in 2009 occurred one month sooner than in 2003. At the upstream station Y.16, the highest monthly runoff was in September 2003 and in August 2009, while for the middle and lower reaches of the Yom River, the highest monthly runoff occurred in October 2003 and September 2009, respectively. Thus, changes in land use are closely associated with the quantity of runoff, the ratio of highest runoff, and flow characteristics.

Figure 12. Effect of land-use change between 2003 and 2009 on (a) annual runoff, (b) sediment yield, (c) NO₃⁻, and (d) PO₄³⁻ levels in the Lower River Yom Basin.

Several other studies have modelled the impact of land-cover and land-use changes on hydrological responses. For example, Githui et al. (2009) used the SWAT model in the Nzoia River watershed in Kenya to examine the impacts on baseflow and streamflow under the current land-use changes. They found that flood risk was exacerbated as the current land-use gradually changed to more smallholder agricultural areas, and the change in surface runoff increased by 55–68%.

3.4.2 Sediment yield

The effect of the change in land use between 2003 and 2009 on the sediment yield is shown in Figure 13. The simulations revealed that the annual sediment yield increased by 37.0% (8.64 × 10⁵ t), 315% (2.79 × 10⁶ t), 44.2% (1.79 × 10⁶ t), 427% (3.47 × 10⁶ t) and 76.9% (3.30 × 10⁵ t) at stations Y.6, Y.4, Y.16, Y.17 and Y.5, respectively. Furthermore, the highest sediment yields in 2009 and 2003 for stations Y.6, Y.4, Y.16, Y.17, and Y.5, respectively, were 1.73 and 1.30 × 10⁶ t, 1.69 and 6.94 × 10⁵ t, 2.25 and 1.48 × 10⁶ t, 1.80 × 10⁶ t and 3.27 × 10⁵ t, and 2.37 and 1.80 × 10⁵ t.

Figure 13. Effects of land-use change (2003–2009) on sediment yield.

Thus, the change in land use from 2003 to 2009 resulted in a 37–427% increase in the annual sediment yield or an increase of approx. 8.6 × 10⁵ t at station Y.6 and an approx. 3.47 × 10⁶ t at station Y.17 (Fig. 12(b)). Moreover, the maximum sediment yield was generally higher in September and October. For the upstream region, the highest sediment yield was in October in both 2003 and 2009, which corresponded to the time of maximum monthly runoff. The major change of land use in the upper and middle part of the Lower Yom River Basin was from forest to agricultural land (~5.67%) and from agriculture to urban areas (~2.26%). As a result, a reduction in sediment, erosion and sediment deposition occurred at the lower elevations, especially downstream (station Y.16).

Similarly, rapid changes in the upland areas (Mau Forest) of the River Njoro watershed, Kenya, caused an increase in the streamflow and flashier flows, which are associated with higher erosion and sediment delivery. The Kenya Wildlife Service identified sediment transport as a significant issue in Lake Nakuru National Park (Baker and Miller 2013).

3.4.3 Nitrate-N levels

The effect of the change in land use between 2003 and 2009 on the river NO₃-N concentration is shown in Figure 14. The simulations reveal that the annual NO₃-N load at stations Y.6, Y.4, Y.16, Y.17 and Y.5 had increased by 93% (2004 t year⁻¹), 141% (2288 t year⁻¹), 127% (12 126 t year⁻¹), 41% (3025 t year⁻¹), and 47% (101 t year⁻¹), respectively. Furthermore, the monthly peak of NO₃-N load was increased by 156%, 1265%, 112%, 57% and 84% to 2581, 209, 3474, 1487 and 99 t, respectively.

Figure 14. Effects of land-use change (2003–2009) on NO₃-N load in the Lower Yom River Basin.

Thus, the change in the land use from 2003 to 2009 increased the NO₃-N load by about 40–140%, or up to 21 644 t year⁻¹ or 1880 kg km⁻² year⁻¹ (station Y.16). The amount of NO₃-N load doubled in the upper and middle reaches of the Lower Yom (stations Y.6, Y.4 and Y.16), as shown in Figure 12(c); this is because most of the forest area above station Y.16 was transformed into agricultural land (area: 526.2 km²) and urban areas (area: 251.7 km²) (see Table 5 and Fig. 10), especially along the main river. This increase in nitrate load corresponds to the 315% increase in annual sediment yield (from 0.89 to 3.69 × 10⁶ t year⁻¹) at station Y.16.

Nitrates on the soil surface may leach into rivers and groundwater, together with the NO₃-N that is discharged into rivers from wastewater, and the resultant high NO₃-N concentrations can be harmful to human health through water consumption. As mentioned above, since the southern parts of the Yom River are in agricultural areas, people grow rice and other crops, and because of the number of houses and communities, there is a growing tendency for increased NO₃-N concentrations in the river, which may lead to eutrophication in the foreseeable future (Ratkrow 2004).

Martínková et al. (2011), similarly, found that future trends in nitrate loads are not likely to be driven only by environmental change, but mainly by nitrogen sources such as fertilizer addition in cultivated areas. As can be seen from Table 6, there is a corresponding relation between the proportion of anthropogenic areas and the nitrate yield in various countries throughout the world. In Europe, studies in both Spain and Germany (Languisch and Matzner 2002, Merchan et al. 2013) reported that the nitrate yield from pristine watersheds was in the range 336–561 kg N km⁻² year⁻¹. However, in areas under anthropogenic change, with > 48% agricultural land, Müller et al. (2018), Rode et al. (2009), and Merchan et al. (2013) found a remarkable concentration of nitrate yield in the range of 1850–4120 kg N km⁻² year⁻¹. As for the USA, pristine areas yielded 58 kg N km⁻² year⁻¹, while nitrate contribution in intensively agricultural areas yielded as much as ~4000 kg N km⁻² year⁻¹ (David et al. 1997, Kalita et al. 2007). In Taiwan, as a representative subtropical area, Kao et al. (2004) and Huang et al. (2012) found that the nitrate yield was approximately 620 kg N km⁻² year⁻¹ from less than -1% “anthropogenic” land, yet moderately anthropogenic land resulted in a nitrate yield as high as 2550 kg N km⁻² year⁻¹. According to a study by Takamatsu et al. (2014) in the tropical area of the Sesan, Srepok and Sekong (3S) rivers, the largest tributaries of the Mekong River, major future land-use changes in the 3S basins will be driven by deforestation of native rainforest, expansion of agricultural and urban areas, and expansion of commercial plantations such as rubber trees, and this will cause the greatest change in future nitrate loads. Comparing the 3S basins of Cambodia, Vietnam and Lao PDR, where high levels of fertilizer application are not widespread among farmers, some pristine basins have a nitrate load of 135 kg N km⁻² year⁻¹, while the sub-basins with intensively agricultural areas have a nitrate load as high as 1619 kg N km⁻² year⁻¹.

Table 6. Nitrate load observed in this study compared to previous studies around the world (modified from Huang et al. 2012, Oeurng et al. 2016).

Region	Anthropogenic area* (%)	Nitrate yield (kg N km^−2 year^−1)	Reference
Illinois, USA	91	~4000	David et al. (1997)
Northeastern USA	9.81	132.9	Mayer et al. (2002)
Northeastern USA	32.85	582.8	Mayer et al. (2002)
California, USA	3.2	58.0	Sobota et al. (2009)
California, USA	24.4	170.9	Sobota et al. (2009)
Central Germany	66.2	1850–4120	Rode et al. (2009)
Central Germany	61.6	2391	Müller et al. (2018)
Germany	< 1.0	336–493	Langusch and Matzner (2002)
Spain	< 0.1	561	Merchan et al. (2013)
Spain	48	2702	Merchan et al. (2013)
Northern Taiwan Northern Taiwan	< 0.1 5	660 2550	Kao et al. (2004) Kao et al. (2004)
Central Taiwan	< 0.1	585	Huang et al. (2012)
Central Taiwan	0.6–8.9	2327	Huang et al. (2012)
Central Taiwan	24.7	40 827	Huang et al. (2012)
3S Basin, Cambodia, Lao PDR and Vietnam	~12–44	135–1619	Oeurng et al. (2016)
Lower Yom River, Thailand	70.45	2770	This study

* Refers to areas subject to anthropogenic activities, such as increased population and urbanization, encroachment of forests by agriculture and urban areas, and degradation of forest resources.

The results of this study thus clearly illustrate that the large nitrate yield of 2770 kg N km⁻² year^–1 in the Lower Yom River area in 2009 (Table 6) was caused by uncontrolled nitrate usage serving the quantity-based agricultural production, as well as wastewater and household sewage water. In this area, corn and rice are the major agricultural crops, covering areas of > 8000 km². Following the recommendations of the government agency guidebook (see Table S1 in the Supplementary material), the fertilizer application level for these crops is approx. 150–160 kg N ha⁻¹ batch⁻¹ or 300–320 kg N ha⁻¹ year⁻¹ (generally, farmers plant two rounds of crops during the year). However, farmers might apply fertilizer at levels higher than the recommended level in order to achieve greater agricultural yields. Consequently, some residual fertilizer-N in the soil after harvesting is released into the surface water and, finally, into the subsurface environment.

3.4.4 Phosphate levels

The effects of the changes in land use between 2003 and 2009 on the river PO₄³⁻ concentration are shown in Figure 15. The simulations revealed that the annual PO₄³⁻ load at stations Y.6, Y.4, Y.16, Y.17 and Y.5 increased by 65% (868 t year⁻¹), 43% (737 t year⁻¹), 377% (18 015 t year⁻¹), 83% (9383 t year⁻¹) and 101% (763 t year⁻¹), respectively. Furthermore, the monthly peak PO₄³⁻ load in 2009 increased by 104, 47, 793, 41 and 177% to 1098, 1028, 12 793, 7359 and 1098 t, respectively.

Thus, the change in land use from 2003 to 2009 led to a 37–377% increase in PO₄³⁻ levels, increasing to 22 795 t year⁻¹ (station Y.16), and 20 657 t year⁻¹ at station Y.17 (Fig. 12(d)). This reflects both the change from the forest area above station Y.16 to agricultural land (526.2 km²) and the increased urban and built-up land along the main river in the lower part of the study area (251.7 km²) (see Supplementary material, Fig. S1). The highest monthly PO₄³⁻ level increased by 41–793%, which followed the increased maximum PO₄³⁻ level at station Y.16 (an 11 000-t increase) that was associated with the increased runoff and sediment yield (see Fig. 15). This can be explained by the fact that most of the area between stations Y.16 and Y.5 had been changed to urban and built-up areas, which mainly release wastewater and household sewage effluent containing detergents and other cleaning products (Fig. 10 and Fig. S1).

Figure 15. Effects of land-use change (2003–2009) on PO₄³⁻ load in the Lower River Yom Basin.

As revealed by Dabrowski (2014), who applied the SWAT model to predict ortho-phosphate loads, four reservoirs in the Upper Olifants River in South Africa experienced eutrophication and the presence of algae blooms due to high phosphate loading. The sources of these phosphate loadings were mainly the poorly maintained sewage treatment plants, contributing approximately 7% of the mean annual runoff volume to this basin. Moreover, according to the study of Baldyga et al. (2007), the presence of algal blooms appearing in Lake Nakuru in more recent years is considered to be a result of higher nutrient inputs (for example, phosphate, nitrate and potassium) from agricultural runoff from the middle and uplands portions into the downstream River Njoro watershed. Pisinaras et al. (2009) studied the amount of NO₃-N and PO₄³⁻ in the approx. 52-km-long Kosynthos River in northern Greece, with a catchment area of about 440 km², and found that eutrophication tended to occur due to agricultural activity, housing and wastewater industries that released NO₃-N and PO₄^3– into the water sources. The overwhelming plant and algal growth from eutrophication may impede infrastructure systems, such as municipal water, and transportation on the water surface. It may also overexploit dissolved oxygen and adversely affect the aquatic ecosystem. For better long-term water quality there must be effective regulations for wastewater management in both the agricultural and household sectors.

4 Summary and conclusions

The SWAT model was proposed to evaluate the streamflow, sediment yield and chemical pollutants (NO₃-N and PO₄³⁻) at the watershed outlet from a large watershed situated in a tropical zone under a sub-humid climate in Thailand. The average monthly flow calibration and validation processes used measured data for the period 2000–2013 and yielded a good correlation (R² = 0.68–0.86) between the measured and simulated flow volumes. Model calibration and validation was performed for the average monthly sediment for the period 2006–2013, giving R² values of 0.54–0.82. The measured and simulated average daily NO₃-N and PO₄³⁻ levels for the period 2008–2013 were 0.55–0.90 and 0.52–0.55, respectively. Furthermore, the simulated results from two land-use change scenarios exceeded the water quality standards. After both calibration and validation processes, statistical assessment of the model output was accepted following the criteria suggested by Moriasi et al. (2007).

The results reveal that the SWAT model can be used effectively in describing the streamflow, sediment yield, and NO₃-N and PO₄³⁻ levels of the surface water, and so it could be used as an administration instrument of choice in solving many water resources and water pollution issues. The outcomes of the study could be used as a decision-making tool by policy makers in creating a suitable strategy for the management of the runoff and sediment yields from the watershed, and for fertilizer management in the agricultural area to mitigate the point and non-point sources of pollution, with improved efficiency of nutrient use for rainfed crops.

Acknowledgements

The authors gratefully acknowledge the Center of Excellence on Hazardous Substance Management (HSM) and the International Postgraduate Programs in Environmental Management, Graduate School, Chulalongkorn University, Bangkok, for their invaluable support in terms of facilities and scientific equipment.

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplementary material

Supplementary material can be accessed on this article here.

Additional information

Funding

The support of the Thailand Research Fund through the Royal Golden Jubilee Ph.D. (RGJ-PHD) Program [Grant no. PHD/0137/2558] is gratefully acknowledged. We would also like to express our sincere thanks to the Ratchadaphiseksomphot Endowment Fund 2014 of Chulalongkorn University [CU-57-061-CC] for financial support.