Application of SWAT Model for Assessment of Surface Runoff in Flash Flood Areas

Abstract

In addition to changes in the amount of rain, changes in land use upstream are considered a factor that directly affects the maximum runoff flow in a basin, especially in areas that have experienced floods and flash floods. This research article presents the application of the SWAT model to assess runoff in areas that have experienced flash floods, in order to analyze the proportion of land use change to the maximum runoff. Study areas that experienced flash floods were in three basins in Thailand (Lam Saphung Basin, Phrom River Basin, and Chern River Basin Part 1, which is a sub-basin of the Nam Chi River Basin). This study analyzed two main factors that influenced runoff in the river basin by considering two simulation situations: (1) changes in land use affecting runoff assessed by considering land use maps in 2006, 2008, 2010, 2015, 2017, 2019, and 2021 when using rainfall data only in the year 2021 for all cases and (2) changes in the amount of rainfall influencing runoff by considering the rainfall records in 2006, 2008, 2010, 2015, 2017, 2019, and 2021 when using the land use data only in the year 2021 for all cases. The results of the study found that the SWAT model can be effectively applied to estimate annual runoff in areas that have experienced flash floods under eight parameters with R² values of 0.74, 0.82, and 0.74 for the Lam Saphung River Basin, Phrom River Basin, and Chern River Basin Part 1, respectively. In addition, it was found that the proportion of land use changes that involved changes from forested areas to residential areas was the greatest from 2008 to 2010 in the Phrom River Basin and Chern River Basin Part 1. This resulted in an increase in the maximum runoff amounts of 77.78% and 46.87%, respectively. When land use was constant, it was found that the rainfall in 2010, which was the highest, also had the greatest impact on the runoff in all three areas.

1. Introduction

The occurrence of floods and flash floods remains an ongoing problem. There has also been an increase in their frequency and severity. At the same time, water shortages in the dry season are a continuous and long-standing problem. The continued increase in the severity of floods and droughts in line with rising global temperatures is a clear sign of global climate change. This is caused by the rapid expansion and growth of the economy throughout every region. This increases the demand for land and water and has increased the severity of events. This is considered to be the source of problems involving both floods and droughts [1,2]. Changes in rainfall patterns caused by climate change and extreme weather events can produce rain that varies in appearance and in increased frequency [3,4,5]. This ongoing climate change is having a major impact on both floods and droughts, increasing their impact on life, social resources, and infrastructure in every region of the world [6,7].

Changes in rainfall patterns caused by climate change and land use changes have resulted in flooding events. They have three main characteristics: river floods, coastal floods, and flash floods [8,9,10]. According to the National Weather Service and USGS, flash floods account for the highest proportion of flood-related deaths and flash floods are becoming an increasingly common occurrence around the world [11,12]. Flash flooding is often associated with intense rainfall [13,14]. However, in some areas, such as in cities, drainage is poorly managed, and slope changes can present a risk of flooding. In extreme cases, flooding can pose a threat to public safety and may result in loss of life. The timescale of flash floods in which hydrological and meteorological conditions are involved is short with periods lasting from minutes to hours [15,16]. Flash floods often occur in areas where the soil does not absorb water readily (e.g., clay soils) and there is an increase in runoff at rivers and other water channels [17,18]. Flash floods can be very severe but there are few opportunities for protection. Flash floods also affect the well-being of people who are most affected by widespread flooding. The causes of the risk of flash floods can be both natural events, such as changes in rainfall patterns, but also human agency, such as changes in land use, which could potentially be prevented by effective planning.

At present, the study and analysis of factors affecting flood occurrence still faces some limitations. Estimating the amount of runoff in areas that have previously experienced flash floods has not been widely studied [19,20]. There has been little study of factors related to flash floods in both river basins and economic areas [21,22]. Data from pilot case studies are important in developing methods for estimating runoff from areas that have experienced recent and past flash floods. Improving watershed areas as part of flash flood management will be an important and useful way for local authorities to prevent and mitigate future floods [23,24]. In addition, such studies will provide important information for decision making in future land use planning, such as changing land use from forested areas to cities or from forested areas to agriculture, etc.

However, the complexity and characteristics of the studied basin (such as its topographic size, land use, and hydrological processes) can influence a model’s parameter range and sensitivity in capturing the dynamics of flood events and runoff assessments. The choice of model and underlying equations can also have an impact on the parameter range and sensitivity. Therefore, the SWAT model is suitable for describing flood events with accurate details and has been accepted for evaluating the impact of land and climate change on flooding and runoff in a basin [25,26,27,28,29]. There is also a Soil and Water Assessment Tool—Calibration and Uncertainty Procedures (SWAT-CUP) model that was developed from SWAT to help determine parameters appropriate to the topography and area of the studied basin [30,31,32]. Hydrological analysis helps to quickly assess runoff during flood events to achieve a comprehensive assessment [33,34,35]. Additionally, the use of advanced models allows for a thorough understanding of the complex processes that result in flooding [36,37,38]. This makes it possible to use the results of a study to plan for and prevent future events.

In the past, the SWAT model has been applied to assess runoff from changes in climate conditions and various land uses [39,40,41,42]. In Thailand, there have been various analyses of land use changes on runoff that also used SWAT models [43]. However, this study was limited to a single watershed and did not map past land use changes over a sufficiently long timescale. The results still cannot explain the details of land use factors and cannot truly be used in planning prevention as they lack details on the proportion of land use changes that affect flash floods and changes in maximum runoff volume. Moreover, the parameters that directly affect the runoff volume are not classified. This may not allow comprehensive tracking of long-term trends or extreme events. Moreover, nowadays, it is rare to find research that applies the SWAT model to assess runoff in areas that have experienced flash floods. There is a need to analyze the causes and factors that will enable planning for response and prevention in the future, including the expansion of use in other areas.

Therefore, this research applied the SWAT model to assess the amount of runoff in areas that experienced flash floods in three river basins in Thailand (the Lam Saphung River Basin, Lam Phrom River Basin, and Chern River Basin Part 1), by analyzing two main factors from a simulation scenario of land use changes from the past and a simulation scenario of changing rainfall in the year considered. Analysis included the investigation of the proportion of land use changes to the maximum runoff volume.

2. Materials and Methods

This research applied the SWAT model (The US Department of Agriculture-Agricultural Research Service (USDA-ARS) and Texas A&M University AgriLife Blackland Research Center, Temple, TX, USA) to assess runoff in areas that experienced flash floods and used SWAT-CUP to help determine parameters appropriate to the topography and representative area of the studied watershed [44,45]. Two main events were considered: (1) the simulation scenario of land use change from the past to the present, using land use maps in the years 2006, 2008, 2010, 2015, 2017, 2019, and 2021, by considering the least amount of rain in the year 2021 only and (2) the simulated situation of changing rainfall in the year under consideration, with land use maps only in the year 2021 analyzed. The steps of this study are shown in Figure 1.

From Figure 1, it can be seen that this study began by collecting various data that were then used in the SWAT-CUP model, such as maps showing the height of the area, land use maps, and soil type maps. The input data such as spatial data, climate, and meteorological data as well as data on the ground level (DEMs for 30 m and 5 m) were used in the SWAT model [46]. These data were then imported into the SWAT model to estimate runoff. The evaluation process consisted of a calibration process (training) and a validation process performed by searching for various parameters obtained from the SWAT-CUP model that were appropriate to the data and then using them. Then, the amount of runoff could be assessed from simulated situations representing all 3 areas. Each area was processed in the same way: (1) Evaluate runoff from changes in land use by specifying the use of rainfall in 2021 only. Then, change the land use according to different years. (2) Evaluate runoff from changes in the amount of rainfall by specifying the use of land use maps only for 2021 and changing the amount of rainfall according to different years.

2.1. Study Area

The study area selected for research comprised three basins that experienced flash floods, namely the Lam Saphung Basin, Phrom Basin, and Chern Basin Part 1. These three basins are sub-basins of the Chi River Basin, which is located in the northeastern region of Thailand, as shown in Figure 2. In addition, complete land use and rainfall map data are available for all 3 basin areas, which were used to evaluate the occurrence of runoff at different times.

From Figure 2, it can be seen that these 3 basins have different characteristics under watershed conditions, even though they are adjacent areas. Information on the general conditions and details of each basin will be presented in Section 2.2.1., Section 2.2.2. and Section 2.2.3..

2.2. Data Preparation

This study employed the Soil and Water Assessment Tool (SWAT) for the evaluation of each basin’s water balance, serving as a hydrological modeling instrument. Developed by the Agricultural Research Service of the United States Department of Agriculture [47], this model incorporates both process-based and semi-distributed parameters. These parameters are designed to anticipate the impact of alterations in land use, climate, and management practices within a basin. Specialized data pertaining to climatology, soil topography, vegetation, and land cover in a basin are integrated into the model. Each basin was compartmentalized into sub-basins interconnected by a stream network, further subdivided into hydrological response units (HRUs) to enhance calculation precision. HRUs denote combinations of soil, land-use/cover, and slope type within each sub-basin. Calibration and validation of the SWAT model in the study basin was imperative to ensure that the representation of model parameters aligned with the study area. The model operates by employing a hydrological balance equation to simulate a hydrological cycle.

$${SW}_t={SW}_0+\sum _{1=1}^t(R_{day}-Q_{surf}-E_a-W_{seep}-Q_{gw})$$

(1)

where SW_t = final soil water content (mm);

SW₀ = initial soil water content (mm);

R_day = amount of precipitation on i^th day (mm);

Q_surt = amount of surface runoff on i^th day (mm);

E_a = amount of evapotranspiration on i^th day (mm);

W_seep = amount of percolation on i^th day (mm);

Q_gw = amount of return flow on i^th day (mm);

t = time in days.

Data preparation for this research involved the selected river basin areas that had experienced flash floods in 2021 using flash flood data from the system, (https://data.dmr.go.th/dataset/debris_flood, accessed on 23 April 2023) [48,49], collected by the Department of Mineral Resources. Flash flood events were selected for study only in the Lam Saphung Basin, Phrom River Basin, and Chern River Basin Part 1. Each basin has different characteristics. The input data format was collected from relevant agencies as shown in

Table 1. Data used in simulating water scenarios.

Data Sources During the Years 2006–2021
DEM	Department of Land Development
Soil Type Map	Department of Land Development
Land Use Map	Department of Land Development
Meteorological Data	Meteorological Department
Runoff	Department of Irrigation

.

The data shown in Table 1 were used as the input into the SWAT model, representing all 3 areas and consisting of the following: (1) a numerical elevation model (DEM) using ASTER DEM (30 m.); (2) a soil series map and land use maps for 2006–2019 from the Land Development Department; (3) meteorological data from 2006 to 2019 from the Meteorological Department; and (4) runoff data from 2006 to 2019 from the Royal Irrigation Department. The topographic maps show the location of the rain gauge station, runoff measuring station, river lines, basin boundaries, level, soil group, and soil type, as shown in Figure 3, Figure 4, Figure 5, Figure 6, Figure 7 and Figure 8. The details of each basin are described below.

2.2.1. Information on the Lam Saphung River Basin

The Lam Saphung River Basin is a sub-basin of the Chi Basin, with an area of 743.74 km² covering 1.51% of the main basin. This area covers Kaset Sombun District, Nong Bua Daeng District, and Khon San District in Chaiyaphum Province. The average annual rainfall is about 1203.4 mm. The total runoff volume is 229.6 MCM (million cubic meters), with a runoff volume per unit area of 9.8 L/s·m² (liters/second/square meter).

This basin has 1 rain gauge station, at location 403010 (see Figure 4). Rain data were used to estimate runoff and rainfall from flooding by considering flash floods for the year 2021. The rainfall records in 2006–2021 are shown in Figure 3. It can be seen that the highest rainfall occurred in 2006 and the least amount of rain occurred in 2019. Figure 4 shows the location of runoff measurement at E.83. The topography is an area of high mountains and a plain in the middle around the elevated watershed. There are 9 groups of soil series, most of which are found in river basins of Soil-7. One group of soil series that was formed from soil parent material such as river sediments was in the area of the sedimentary plains which where water logged during the rainy season. The soil is very deep with relatively poor to poor drainage. The topsoil is clay and clay loam. The bottom soil is clay. The 5 main land use types of this area are shown in Figure 4; the most common type of land use is forested area. In addition, the land use of Lam Saphung River Basin is shown on the land use map from 2006 to 2019.

2.2.2. Information on the Phrom River Basin

The Phrom River Basin is a sub-basin of the Chi River Basin, with an area of 2264.66 km² covering 4.60% of the main river basin. This area covers Kaset Sombun District, Phu Khiao District, Ban Thaen District, and Khon San District in Chaiyaphum Province, Chum Phae District in Khon Kaen Province, and Mueang Phetchabun District, Lom Sak District, and Nam Nao District in Phetchabun Province. The average annual rainfall is about 1080.3 mm. The total runoff volume is 435.3 MCM with a runoff volume per unit area of 6.1 L/s·m².

This basin has four rain gauge stations: 403002, 403003, 905354, and 905824 (as shown in Figure 6). The rainfall data were used to estimate the impact of rainfall on flooding by considering flash flooding in the year 2021. The rainfall data in 2006–2021 are shown in Figure 5. It can be seen that the highest rainfall occurred in 2006 and the least amount of rain occurred in 2019. Figure 6 shows the location of 1 runoff measurement location, E.93, which has the topographical characteristics of the area adjacent to Lam Saphung, the upper Lam Chi River basins, the Lam Nam Chern branch basin, Lam Chi River Section 2, and the area of the upper Nam Phrom River above Chulabhorn Dam. There are 14 groups of soil series, most of which are found in the river basins of Soil-7 that formed from soil parent material such as river sediments. These soils are found in the area of the sedimentary plains that are waterlogged during the rainy season. It is very deep soil with relatively poor to poor drainage. The topsoil texture is clay and clay loam. The bottom soil texture is clay and constitutes the soil present in the 5 main types of land use areas. However, the most common type of land use is forested area.

2.2.3. Information on the Chern River Basin Part 1

The Chern Basin Part 1 is a sub-basin of the Chi Basin, with an area of 1899.65 km² covering 3.86% of the main basin. This area covers Kaset Sombun District, Phu Khiao District, Ban Thaen District, and Khon San District in Chaiyaphum Province, Chum Phae District in Khon Kaen Province, and Lom Sak District and Nam Nao District in Phetchabun Province. The average annual rainfall is about 1104.4 mm. The total runoff volume is 363.8 MCM with a runoff volume per unit area of 6.1 L/s·m².

This basin has five rain gauge stations: 379028, 381003, 041302, 041302, and 041307 (as shown in Figure 8). The rainfall data were used to estimate the impact of rainfall on flooding by considering flash flooding in the year 2021 and rainfall in 2006–2021 as shown in Figure 7. It can be seen that the highest rainfall occurred in 2006 and the least amount of rain occurred in 2019. Figure 8 shows the location of 1 runoff measurement location (E.85), and more than 50% of the watershed’s topography comprises steep mountains and dense forests interspersed with low hills. There are 19 groups of soil series, most of which are of Soil-7. The most common type of land use is forested area.

The soil levels of each area are shown in Figure 4, Figure 6 and Figure 8 for the Lam Saphung River Basin, Phrom River Basin, and Chern Basin Part 1, respectively. These provide important information for input data affecting runoff data. In addition, the soil series maps and the land use maps are shown in terms of changes from past to recent years. A change in land use, such as changing from forestland to agriculture or building a city, affects the drainage process and flooding. The changes in land use and rainfall influence runoff data. The consideration of changing rainfall data will help to understand how rainfall distribution patterns affect stream flow in the basin.

The types of land use changes in the three river basins are summarized in

Table 2. Summary of the areas of each type of land use in each watershed.

No.	Land Use	Area (km2)
		2006	2008	2010	2015	2017	2019
1.	Lam Saphung
	Agricultural land	25.8	25.7	29.8	30.3	30.7	33.1
	Forested land	639.5	639.4	632.2	628.0	628.1	625.6
	Urban and built-up land	1.1	1.1	1.2	1.3	1.3	1.3
	Water body	1.1	1.2	3.8	3.8	3.8	3.7
	Other	75.9	75.9	76.4	79.9	79.4	79.7
	Total	743.4	743.3	743.3	743.3	743.3	743.4
2.	Nam Phrom
	Agricultural land	769.8	769.5	770.2	770.5	795.8	809.4
	Forested land	1287.1	1286.5	1267.5	1265.8	1257.8	1243.1
	Urban and built-up land	51.3	51.3	61.0	61.2	65.0	68.0
	Water body	36.7	36.7	56.6	56.6	56.1	58.5
	Other	119.1	119.6	108.3	109.6	89.1	84.8
	Total	2263.9	2263.7	2263.7	2263.7	2263.7	2263.7
3.	Chern Part 1
	Agricultural land	755.4	756.9	760.3	763.9	783.7	786.2
	Forested land	902.6	894.4	877.1	860.4	854.3	847.3
	Urban and built-up land	90.4	90.3	98.1	105.5	109.5	115.5
	Water body	19.2	18.7	29.1	32.3	33.2	36.1
	Other	131.2	138.2	134.0	136.4	117.9	113.4
	Total	1898.7	1898.5	1898.5	1898.5	1898.6	1898.5

. From the table, it can be seen that the Phrom River Basin has a total area of 2263.8 km² (square kilometers) over which increases were mainly in agricultural and other areas. Forestry decreased, accounting for the greatest proportion of land use changes from forested to residential areas from 2008 to 2010. The Chern River Basin Part 1 has a total area of 1898.5 km². The lowest proportion of land use change from forested area to residential area occurred in the Lam Saphung River Basin, with a total area of 743.4 km². Agricultural areas comprised the greatest increase in land use.

2.3. Estimating Runoff Scenarios

This research evaluated runoff from three simulated situations: (1) the influence of changes in land use to runoff; (2) the influence of changes in rainfall to runoff; and (3) the influence of changes in the proportion of land use to runoff. The procedure for each situation is described as follows.

2.3.1. Changes in Land Use to Runoff Volume

The effect of land use change on runoff was evaluated using the watershed areas selected as having been affected by flash floods in Thailand during 2021. This study considered only 3 areas, as described in the previous section. The selected land use maps for 2006, 2008, 2010, 2015, 2017, and 2019 were used for each run with the same rainfall data for the year 2021 in each case. Therefore, the 2019 land use map was used as the baseline data for comparing changes in land use affecting runoff data. In this study, data on land use changes in various years were used with the rainfall data from the year in which the flash flood occurred. When running the SWAT model according to various simulation situations, the amount of runoff in different years according to the selected land use map was obtained. The procedure is shown in Figure 9. The runoff data of each selected area were then recorded.

2.3.2. Changes in Rainfall in Relation to Runoff

The changes in rainfall data to runoff volume were investigated similarly to Section 2.3.1, by using land use maps only in the year 2021 with the various rainfall data, as shown in Figure 10. The selected land use areas of this study were treated similarly using rainfall changes for 2006, 2008, 2010, 2015, 2017, and 2019. The procedure of each selected area was performed, and the runoff output was obtained and recorded.

2.3.3. Changes in the Proportion of Land Use in Relation to Maximum Runoff

Changes in the proportion of land use in relation to maximum runoff volume were investigated by using the procedure described in Section 2.3.1. The constant rainfall data were used only in the year with the least rainfall (2019), whereas land use maps were used from various years. The runoff volume was obtained from each run of SWAT. The trends of land use change and maximum runoff volume were analyzed.

3. Results and Discussions

The results of this study are presented in four parts: (1) the efficiency of the SWAT model; (2) the effect of changes in land use on the amount of runoff; (3) the effect of changes in the amount of rainfall on the amount of runoff; and (4) the change in the proportion of land use to the maximum runoff. The details of each part are presented below.

3.1. SWAT Model Performance Results

Adjusting the parameters in a SWAT model is an essential procedure for evaluating the accuracy and reliability of a model before using it to simulate any scenarios. This step helps the model correspond to the real situation, and increase efficiency and confidence in the runoff results.

3.1.1. Appropriate Parameters

This study used SWAT-CUP to analyze the sensitivity of eight important parameters during flooding. The results from the SWAT-CUP of over 500 simulation rounds are shown in

Table 3. Optimal results obtained from the calculations of the model using 8 parameters over 500 simulation rounds.

No.	Parameter	Description	File Type	Ranking Range	Optimal Value
					E.83	E.93	E.85
1.	Cn2	Initial curve number (II) value	.Mgt	0.1–0.1	0.045766	0.0458	0.0093825
2.	Sol_Awc	Available water capacity [(mm water) (mm soil)−1]	.Soil	0–0.7	0.7476	0.7476	0.6941
3.	Esco	Soil evaporation compensation factor	.hru	0–0.2	−5.67	−5.67	0.19005
4.	Gwqmn	Threshold water depth in the shallow aquifer for flow [mm]	.gw	0–500	500	500	500
5.	Gw_Revap	Groundwater ‘revap’ coefficient	.gw	0.6–0.95	0.191	0.191	0.7614
6.	CH_N2	Manning’s N value for the main channel	.rte	0–0.3	0.6	0.6	0.0999
7.	Gw_Delay	Groundwater delay [days]	.gw	−15–15	0.6941	0.6941	0.57
8.	Alpha_Bf	Baseflow alpha factor [days]	.gw	0–1	0.0555	0.0555	0.7476

. It was found that the results were close to the actual data from water-measuring stations E.83, E.93, and E.85, in 2021. The SWAT model was used to analyze runoff data both from the dry season to the rainy season and from the rainy season to the dry season, with acceptable criteria for all three selected areas. As shown in Table 3 (which reports the statistical indices of the inspection for the period of October–December 2021), the SWAT-CUP made it possible to identify appropriate parameters. It had good performance in predicting runoff volume in all three river basins.

In addition, the various parameters that affected the realism of the model were studied. It was found that the parameter with the greatest effect on the model was Gwqmn, which was the variable that had a notable impact on the drainage process in the watershed, which was part of the objective of this study. The adjustment of the Gwqmn value can have a significant impact on the simulated water behavior in a basin. The parameters Gw_Revap and Alpha_Bf were also important in communicating information related to drainage processes and rainfall distribution in the basin. The Gw_Revap watershed had a low p-value and a high t-statistic. This shows a relationship that is statistically significant and had an effect on the model. Similarly, Alpha_Bf is a parameter that had a significant impact on the model due to its low p-value and high t-statistic. Therefore, adjusting Gwqmn, Gw_Revap, and Alpha_Bf are important steps in model adjustments, as reported in previous studies [50,51,52]. The SWAT-CUP model adjusted in this study achieved simulation results that were most consistent with realistic water behavior in the three representative basins.

3.1.2. Calibration and Validation

The calibration results of the model to check its accuracy by considering statistical values, including the coefficient of determination (R²), root mean square error (RMSE), percent bias of estimation (PBIAS), and Nash–Sutcliffe coefficient (Nash–Sutcliffe efficiency, NSE), according to Equations (2)–(5) [53,54] and the results, are shown in

Table 4. Performance indices of the SWAT model.

Watershed	Interval	Year (B.E.)	Average Runoff (m3)		Index
			Observed	SWAT	R2	PBIAS	RMSE	NSE
Lam Saphung	Calibration	January–September 2021	2977.83	1762.78	0.84	40.80%	15.61	0.654
	Validation	October–December 2021	2118.65	2389.61	0.60	12.81%	21.57	0.608
	Total	January–December 2021	5096.48	4152.39	0.65	17.31%	17.31	0.637
Nam Phrom	Calibration	January–September 2021	2478.59	2347.074	0.88	5.31%	7.95	0.874
	Validation	October–December 2021	4856.31	3733.487	0.82	57.02%	23.20	0.737
	Total	January–December 2021	2377.72	6080.561	0.81	25.25%	13.53	0.795
Chern Part 1	Calibration	January–September 2021	3646.84	4034.21	0.79	10.62%	15.37	0.783
	Validation	October–December 2021	2316.92	3357.38	0.74	44.63%	33.91	0.205
	Total	January–December 2021	5963.76	7391.60	0.72	23.83%	21.60	0.604

.

$$R^2=\frac{{\left[\left(Q_{obs}-Q_{avr}\right)\times (S_{sim}-S_{avr})\right]}^2}{\sum {(Q_{obs}-Q_{avr})}^2\times \sum {(S_{sim}-s_{avr})}^2}$$

(2)

$$RMSE=\sqrt{{\sum }_{i=1}^N\frac{{\left(S_{sim}-Q_{obs}\right)}^2}{N}}$$

(3)

$$PBIAS=\frac{\sum _{i=1}^n{(Q_{obs}-S_{sim})}^2}{Q_{obs}}\times 100$$

(4)

$$NSE=1-\frac{\sum _{i=1}^n{(Q_{obs}-S_{sim})}^2}{\sum _{i=1}^n{(Q_{obs}-Q_{avr})}^2}$$

(5)

From Table 4, it was found that the Lam Saphung River Basin’s total annual runoff volume was 4152.89 m³ (cubic meters), 22.7% lower than station E83, with the SWAT-CUP giving good results (R² = 0.84, PBIAS = 40.80%, RMSE = 15.61, and NSE = 0.654). For the Nam Prom Basin, the total annual runoff was 6080.56 m³, 20.1% higher than station E93, also with good results of an R² = 0.88, PBIAS = 5.31%, RMSE = 7.95, and NSE = 0.874. For the Chern River Basin Part 1, the total annual runoff volume was 7391.60 m³, 23.9% higher than station E85, with the good results of an R² = 0.79, PBIAS = 10.62%, RMSE = 15.37, and NSE = 0.783. The relationships between the stream from the record and the runoff from the simulation model are shown in Figure 11.

Figure 11 allows for a comparison of the results between the runoff estimation and the flow record in 2021 at measurement stations E.83, E.93, and E.85. It also presents the values for the R² and PBIAS, indicating that the calibration of the data over a period of approximately 1 year between the simulation and real data revealed a consistent correlation of 74.9% between the two datasets, while the validation provided an additional 24.1%. The accuracy and reliability of simulation results are acceptable for both flood and drought events in 2021. In addition, the selection of the time period during calibration and verification was very important in order to extract standardized data from the combined stream flow records and adapt them to the simulation results. This information provided a useful database to evaluate the reliability of the simulations and the validity of the results of the flow correlation results from the two data sources according to previous studies [48,49,53,54].

3.2. Effects of Land Use Changes on Runoff Volume

Table 5. Maximum runoff volumes of the selected study areas when using rainfall data in 2021 with the various land use maps.

Watershed	Rainfall (Year)	Land Use Map (Year)	Period of Occurrence	Maximum Runoff (m3)
Lam Saphung	2021	2019	10 October	169.3
		2017		169.1
		2015		169.1
		2010		168.7
		2008		168.0
		2006		168.0
Nam Phrom	2021	2019	10 October	157.0
		2017		156.6
		2015		156.1
		2010		156.0
		2008		154.8
		2006		154.8
Chern Part 1	2021	2019	10 October	299.8
		2017		299.0
		2015		298.3
		2010		296.3
		2008		294.3
		2006		293.3

shows the maximum runoff volumes of the selected study areas when considering rainfall data in 2021 only with the various land use maps in 2006, 2008, 2010, 2015, 2017, and 2019. The results indicate that the highest runoff event was on 10 October for all selected areas. There was a maximum runoff of 169.3, 169.1, 169.1, 168.7, 168.0, and 168.0 m³ for land use years 2019, 2017, 2015, 2010, 2008, and 2006, respectively, for the Phrom River Basin. The maximum runoff volumes of the Nam Prom Basin were 157.5, 156.6, 156.1, 156.0, 154.8, and 154.8 for land use years 2019, 2017, 2015, 2010, 2008, and 2006, respectively. For the Chern River Basin Part 1, the highest value was in the year 2019 of 299.0 m³ whereas the lowest was in 2006 of 293.3 m³. These results are plotted in Figure 12.

From Figure 12, it can be seen that the highest runoff in all three selected river basins was also caused by land use in 2019, which is the most recent data for land use. In all three selected river basins, it was found that the maximum runoff volume tended to steadily increase from 2006 to 2019, with all values being similar. There was little difference in peak runoff volumes from 2006 to 2019. The highest maximum runoff value was from the Chern River Basin Part 1. In addition, it was found that the highest occurrence of runoff happened in two periods, September (dry season—flood season) and October (flood season—dry season) of every land use year in all three river basins, as previously reported [43,55,56,57]. Therefore, it can be concluded that recent land use changes under the highest annual rainfall will cause the basin’s highest runoff.

3.3. Effects of Changes in Rainfall on Runoff Volume

Table 6. The maximum runoff volumes of the Lam Saphung Basin, Phrom Basin, and Chern River Basin Part 1 when using land use maps for only 2021 with the rainfall data from various years.

Watershed	Land Use Map (Year)	Rainfall Data (Year)	Maximum Runoff (m³)	Period of Occurrence
Lam Saphung	2021	2019	9.8	10 October 2019
		2017	81.6	4 October 2017
		2015	88.3	19 September 2015
		2010	316.1	18 October 2010
		2008	162.8	20 September 2008
		2006	258.6	21 October 2006
Nam Phrom	2021	2019	9.25	11 October 2019
		2017	115.2	5 October 2017
		2015	88.3	19 September 2015
		2010	222.6	20 October 2010
		2008	164.7	21 September 2008
		2006	162.0	4 October 2006
Chern Part 1	2021	2019	63.4	26 September 2019
		2017	215.7	5 September 2017
		2015	140.6	19 September 2015
		2010	498.2	18 October 2010
		2008	273.3	19 September 2008
		2006	387.5	4 October 2006

shows the highest runoff volumes of the Lam Saphung Basin, Phrom Basin, and Chern River Basin Part 1 when only using land use maps for 2021 with the rainfall data from various years. It was found that the maximum runoff volumes of all selected study areas occurred in 2010. The highest maximum runoff value (498.2 m³) was from the Chern River Basin Part 1. Figure 13 shows the daily runoff of the Lam Saphung Basin, Phrom Basin, and Chern Basin Part 1 when using land use maps for 2021 with the rainfall data of the years 2006, 2008, 2010, 2015, 2017, and 2019. It can be seen that when using the latest year’s land use maps (2021) and rainfall in 2010, the highest runoff volumes were obtained for all selected areas. Therefore, it can be concluded that the current land use with more rainfall data can provide the highest maximum runoff volumes under the same situation reported for most river basins according to previous studies [58,59,60].

3.4. Results of Changes in the Proportion of Land Use Affecting Maximum Runoff

Table 7. The area sizes of residential land use types for the Lam Saphung Basin, Phrom Basin, and Chern Basin Part 1.

Watershed	Land Use Area (km2)
	2006	2008	2010	2015	2017	2019
Lam Saphung	1.1	1.1	1.3	1.3	1.3	1.3
Nam Phrom	51.3	51.33	61.0	61.2	65.0	68.0
Chern Part 1	90.4	90.3	98.1	105.5	109.5	115.5

shows the residential land use sizes from the years 2006, 2008, 2010, 2015, 2017, and 2019 of the Lam Saphung Basin, Phrom Basin, and Chern Basin Part 1. Figure 14 also shows the percent change rates for residential land use types as compared with the baseline year of 2006. The results show that the highest rate was from 2008 to 2010 for the Lam Saphung Basin and Nam Prom Basin, whereas the highest rate for Chern Part 1 occurred from 2008 to 2015 as compared with the baseline of 2006. These land use types were changed from forested areas to residential areas, especially in two river basins (the Phrom River Basin, increased from 51.3 to 61.0 km² (18.9%) and the Lam Saphung Basin, increased from 1.1 to 1.3 km² (18.2%) as compared to the baseline year of 2006). The results also indicate that the proportion of changes in land use types from forested areas to residential areas clearly increased in some years, especially in 2008 and 2010. These changes will have directly affected the maximum runoff volume. Therefore, it can be concluded that a higher proportion of land use change will result in higher runoff.

Table 8. The maximum runoff volumes when using the year with the least rainfall (2019) with the various land use maps of 2006, 2008, 2010, 2015, 2017, and 2019.

Watershed	The Maximum Runoff Volumes (m3/s)
	2006	2008	2010	2015	2017	2019
Lam Saphung	4.5	4.5	8.0	9.7	9.8	9.8
Nam Phrom	7.2	6.4	9.4	9.8	9.5	9.6
Chern Part 1	59.3	59.2	63.6	63.5	66.4	60.5

shows the maximum runoff volumes when using the year with the least rainfall (2019) with the various land use maps for the years 2006, 2008, 2010, 2015, 2017, and 2019. The results indicate that the high changes in maximum runoff occurred from year 2008 to 2010 for all selected study areas. Figure 15 show the percent of maximum runoff changes as compared with the baseline year of 2006 when using the year with the least rainfall (2019) with the various land use maps of 2006, 2008, 2010, 2015, 2017, and 2019. The data indicate that high percent changes of maximum runoff occurred from year 2008 to 2010 for all selected study areas. In addition, the highest percent change was 77.8% for the Lam Saphung River Basin, whereas the lowest percent change was 7.3% for the Chern Basin Part 1. Therefore, it can be concluded that the proportion of the increase in runoff volume was a result of changes in land use from forested areas to residential areas, and this result agrees with previous studies [61,62,63,64].

4. Conclusions

This research article presents the application of the SWAT model to assess runoff in areas with a history of flash floods. Three areas that previously experienced flash floods were selected as the case study, namely the Lam Saphung Basin, Phrom River Basin, and Chern River Basin Part 1, which is a sub-basin of the Chi River in Thailand. There were three objectives of this study. The first was to assess the effectiveness of the SWAT model in assessing runoff in areas that experienced flash floods. The second was to analyze the factors that influence the amount of runoff especially with regard to changes in rainfall and land use. Finally, the proportion of land use change affecting the maximum runoff volume was analyzed. Three simulation scenarios were employed as the following: (1) a scenario of land use change affecting runoff volume by using constant rainfall data only in the year 2021 and considering land use changes from land use maps in 2006, 2008, 2010, 2015, 2017, 2019, and 2021; (2) a scenario of rainfall change affecting runoff using constant land use data only in 2021 and considering rainfall in 2006, 2008, 2010, 2015, 2017, 2019, and 2021; and (3) a scenario of changes in the proportion of land use to runoff using constant rainfall data in the year with the least rainfall of 2019 and considering land use changes from land use maps in 2006, 2008, 2010, 2015, 2017, 2019, and 2021.

It was found that the application of the SWAT model produced highly reliable results that could be applied to estimate annual runoff in areas that have experienced flash floods. It was also shown that eight parameters were effective conditions, with R² values of 0.74, 0.82, and 0.74 for the Lam Saphung Basin, Phrom River Basin, and Chern River Basin Part 1, respectively. In addition, the analysis found that land use changes in 2019 (which was the latest year studied), influenced the maximum runoff volume in all three areas which was 169.3, 157.5, and 299.8 m³/s (cubic meters/second), respectively. When considering changes in rainfall volume and constant land use, it was found that the highest rainfall in 2010 also had the greatest impact on the runoff volume in all three areas. In addition, it was found that the proportion of land use changes involving changes from forested areas to residential areas was the greatest from 2008 to 2010 in the Phrom River Basin and Chern River Basin Part 1, which would have resulted in the maximum runoff volume increasing by 77.78% and 46.87%, respectively, as compared to the year 2006. It can be concluded that the maximum amount of rain in each area affected the maximum amount of runoff. Also, a high proportion of changes in land use from forested to residential areas affected the maximum runoff volume in the basin. For improved accuracy and a more detailed characterization of runoff, future studies should consider hourly rainfall data. However, despite these valuable insights, it is essential to acknowledge the limitations of past research in the field. Previous applications of the SWAT model in Thailand have predominantly focused on single watersheds and lacked comprehensive mapping of historical land use changes over extended periods. The resulting data limitations hindered detailed analysis of land use factors affecting flash floods and changes in maximum runoff volume. Additionally, parameters directly influencing runoff volume were not adequately classified, impeding comprehensive tracking of long-term trends and extreme events.

Though the SWAT model effectively estimates annual runoff in flash flood-prone areas, its application is limited. Being focused on specific Thai basins, its results may not be generalizable to other regions. Evaluation at an annual scale suggests potential for predicting flash floods at shorter intervals, yet uncertainties persist. As a recommendation for future research, investigators are encouraged to conduct thorough calibration and validation for site-specific applications, considering local hydrological characteristics. Integrating hourly rainfall data in subsequent studies could enhance the model’s accuracy in predicting flash floods.