Rice paddy water management was integrated into a distributed three-dimensional surface and subsurface coupling hydrological model of the Sakuragawa River watershed. This watershed is located in the Kanto Plain in Japan and includes the hillside of Mt. Tsukuba. Therefore, this watershed includes both steep mountainous areas and rice paddy-dominated flat land. Thus, water management of rice paddies is important and was calculated separately using a paddy model. The use of groundwater for rice paddy irrigation was considered as well as a water supply from outside of the watershed (Kasumigaura Lake). The model parameters were calibrated and validated with reference to the predictability of river water flow and the groundwater level. Using the calibrated model, three-dimensional streamlines, water travel time distributions, and water balance in some grids were clarified. The developed model will facilitate sustainable water resource management in the watershed.
INTRODUCTION
The importance of fresh water for the health of society cannot be overemphasized. The primary concern to preserve high-quality freshwater resources is to conserve a good water environment in the watershed. Various efforts related to watershed management, such as the Integrated River Basin Management and the Integrated Water Resources Management (Abell et al. 2002; Humphrey & David 2005; Cook & Spray 2012) have been implemented. The WHO's Water Safety Plans, which require a risk assessment encompassing all steps in water supply from catchment to consumer, is another example (Davison et al. 2005; Breach 2012). In watershed management, it is important to study rivers that flow in highly populated plain areas because human activities, such as urbanization and agricultural activities, alter the water flow from its natural state in the post-development watershed (Yates & Miller 2011). These human activities alter not only surface water flows but also subsurface flows, and also affect water movement between surface and subsurface zones (e.g. changes in groundwater recharge due to land use changes, uptake of groundwater, etc.). Furthermore, increased environmental loads of chemical and biological pollutants as a result of human activities are also important concerns. These effects result in increased risks of shortages and deterioration of freshwater resources.
Rational watershed management based on an understanding of hydrology is important to minimize these risks. One typical cause of such risk is excess nutrient loading in water bodies (Sagehashi et al. 2009a, b; Kawahara et al. 2011) from non-point sources. In particular, nitrogen plays an important role as fertilizer, including in rice cultivation (Ghosh & Bhat 1998). Penetration of nitrate into groundwater (O'Shea & Wade 2009; Buczko & Kuchenbuch 2010; Aquilina et al. 2012) is another risk.
On the other hand, as recognized in the renewed interest in the role of rice paddies as groundwater rechargers (Liu et al. 2001; Yu et al. 2006), the effects of rice paddies on the amount of groundwater are significant.
The investigation of surface water and groundwater was performed by monitoring and modeling approaches (Cho et al. 2009). However, as monitoring of the dynamics of groundwater is difficult and expensive, there is increasing interest in the modeling of groundwater dynamics. Progress in the modeling of soil-vegetation-atmosphere through pedotransfer functions improved the representation of soil-water dynamics (Vassena et al. 2012). Meanwhile, groundwater resources suffered from the impact of various human activities, such as the demands of various water (Green et al. 2011) and land use changes (Kimaro et al. 2003; Cho et al. 2009). Such impacts should be notable in the densely populated and urbanized flat land. Discharge of groundwater for irrigation, domestic, and industrial water supply is another impact. Paddy rice culture, which is an important agricultural practice in the Asian monsoon region (Jeon et al. 2007), has unique features of water management, such as keeping ponded water at a desirable depth (Li & Migita 1992) and temporary drainage (Yoshinaga et al. 2007). Including the role of rice paddies as groundwater rechargers (Liu et al. 2001; Imaizumi et al. 2004; Yu et al. 2006), the water budget of rice paddies should be considered in the analysis of surface water and groundwater dynamics. Therefore, the development of a three-dimensional, surface/subsurface coupling hydrological model considering rice paddy cultivation offers a key to understanding watershed management, especially in the Asian monsoon region.
In this study, we selected a watershed in Kanto Plain to model the hydrology. The development of Kanto Plain began in ancient times, meaning that the surface water and groundwater systems have been influenced by human activities for a long time. Therefore, the groundwater travel time analysis (Schilling & Wolter 2007) or analysis of land use legacy based on the travel time model (Ray et al. 2012) can provide important information for appropriate watershed management.
For such practices, we should combine the surface water and groundwater dynamics in some manner. Cho et al. (2009) concluded that there are three types of modeling approach, i.e. integrated surface-groundwater model development, groundwater model and surface model linking, and utilization of packages of existing groundwater models (Cho et al. 2009). Some previous studies have applied a surface and groundwater model to Kanto Plain. The National Institute for Environmental Studies Integrated Catchment-based Ecohydrology model, which is a process-based model, reproduced water cycle changes and drying phenomena in a watershed, including ground surface, unsaturated, and saturated layers in underground (Nakayama & Watanabe 2004), was applied to describe the groundwater dynamics related to Kasumigaura Lake, the second largest lake in Japan, which is located in Kanto Plain (Nakayama & Watanabe 2005, 2008; Nakayama et al. 2007). Some studies dealt with a combined model with irrigation (Liu et al. 2013), or models incorporating paddy water use (Taniguchi et al. 2009a, b, c). In addition, some studies applied a tank model to the groundwater model (Mekpruksawong et al. 2006; Takeuchi et al. 2009, 2010). To understand the spatial flow behavior of the water, a distributed model is required. However, there have been few reports of runoff simulation for rice paddies (Kang et al. 2006). In general, hydrological descriptions in mountainous areas and flat areas are different (Yu et al. 2006). Kubota et al. (2007) introduced a three-dimensional surface and subsurface coupling hydrological model with upland, rice paddy, and forest submodels to describe the water and nitrogen dynamics in the Koisegawa River watershed located in the Kanto Plain, and pointed out that it will be necessary to validate groundwater level and groundwater quality in future studies (Kubota et al. 2007). Sufficient reproducibility of surface water flow and groundwater level taking water management in rice paddies into consideration are keys to clarifying the water circulation in Asian flat lands by a distributed model.
The objectives of this study were to integrate the rice paddy water management into a three-dimensional hydrological model coupling surface and subsurface water flows in the watershed of Sakuragawa River in the Kanto Plain, and to clarify the water circulation characteristics that are essential for watershed management, i.e. streamlines, travel time to the mouth, and water balance, using the developed model. Especially, the groundwater level is strongly affected by the rice paddy, and we tried to reproduce this behavior with considering the separately developed paddy model.
MATERIALS AND METHODS
Site description
Hydrogeological model
In this study, the detailed boring data could not be obtained. Therefore, the watershed was divided roughly into four geologically different zones, namely the right bank far from the Sakuragawa (A), the right bank near Sakuragawa (B), the left bank (C) and the mountainous zone in the left bank (D), based on the land use, location, and literature information (Yoshitani et al. 2001; Geosphere Environmental Technology 2006) (Figure 2). A 2 m of top soil layer was assumed for each zone. Under the topsoil, 10 m of loam and clay layers were assumed to be located in zones A and B, whereas 10 m of alluvium layer was assumed to be located at the same position of the zone C. Under these layer(s), two layers (∼100 m in depth and 100 m in depth under elevation level) were assumed to exist. This assumption is the presumption based on the reports by Sekimoto et al. (2009), and the bed lock is assumed to be located at 150 m in depth under sea level for zones A, B, and C, whereas, it was assumed to be located under the 2 m of topsoil for zone D.
One of the most important goals of this modeling approach is to replicate the artificial water flows in the watershed. Actually, drinking water and industrial water are used in the watershed as well as agricultural water. At the Tonegawa River system in Ibaraki Prefecture, which includes the Sakuragawa River watershed, the amounts of supply (and demand) of drinking water, industrial water, and agricultural water were 7,423 (7.363) (m3/s), 16,693 (10.226) (m3/s), and 55.31 (63.86) (m3/s), respectively, in 2004 (Ibaraki Prefecture 2007). Considering this ratio and their circulation property, only the artificial water flow of agricultural water was considered in this model. Actually, irrigation is introduced for agricultural fields other than rice paddies. However, considering the dominance of the irrigation water used in paddy fields in Japan (MAFF Home Page), the irrigation was assumed to take place only in paddy fields.
Irrigation water from agricultural canals, Sakuragawa River, and repetition irrigation were also assumed, as well as drawing of groundwater. Uptake of water from the river was assumed to take place from the river as a whole (the river was treated as one unit). The groundwater uptake is assumed to be held only at the southern part of the watershed, and, due to the limitation of groundwater well locations and its depth data, the simulation model postulates the groundwater sucking grids with the depth 100–150 m as indicated in Figure 2(d). In addition, the drawn water is assumed to be supplied equally to paddies in the watershed. Actually, groundwater is drawn up in many places for irrigation. It was reported that 13,559 facilities pumped 22,5071 × 103 m3 water annually for the irrigation of 11,232 ha of paddy fields in Ibaraki Prefecture (MAFF 2003). However, detailed information of the pump locations is not open to the public (and that is why we employed such a rough setting for the suction of the groundwater as described above). Meanwhile, the local government makes it mandatory to report the amount of groundwater uptake by pumps with discharge ports more than a certain size. Considering the aggregate of the amount of groundwater uptake in each year from 1997 to 2001 (Department of Planning, Ibaraki Prefecture, personal communication 2014), and the area of paddy fields in each corresponding year (MAFF), the relative usage of the aggregated groundwater uptake to the paddy fields was considerably lower in the northern part of the watershed compared with the southern part. The Kasumigaura Canal supplied from Kasumigaura Lake is well developed in the northern part (i.e. Sakuragawa City and Chikusei City). In Tsukuba City located at the southern side of the watershed, large amounts of groundwater were used for the irrigation of rice paddy fields (Tobita et al. 2004). These conditions support the assumptions regarding groundwater uptake in the model to a certain extent. However, information about the groundwater uptake obtained in this study is limited, and the resultant uncertainty of the model cannot be denied. It would be possible to reduce the uncertainty in the model if we could obtain more detailed information regarding groundwater uptake.
Paddy model
There have been few studies of distributed, physically based, three-dimensional surface water and groundwater models of a watershed ranging from steep mountainous areas to rice paddy-dominated flat land. One possible reason is that the distributed model employed simplified equations, such as Manning's equation, while the water level in rice paddies is controlled artificially, meaning that it is difficult to describe the water budget in rice paddies by such simplified equations.
In the paddy rice non-cultivation period, no irrigation water was supplied and the surface runoff and penetration were calculated using the Soil Conservation Service Curve Number (SCS-CN) method (Mockus et al. 1972; McCuen 1982). The parameter for the SCS-CN, i.e. CN, was set as 67 based on a previous study by Im et al. (2007).
In the paddy rice cultivation period, irrigation was supplied to maintain the desired water level (Figure 3). The desired water level was set according to the conventional pattern based on the Fertilization Standard for Ibaraki Prefecture (provided by MAFF). Irrigation water from Sakuragawa, Lake Kasumigaura, and groundwater were considered. The infiltration rate of water differs depending on the situation (Hitomi et al. 2010). In this study, a constant rate (1.1 cm/day) (Nakamura et al. 2004) was set for penetration. Surface runoff occurred when the water depth on a day passed the desired water level. Resultant calculations of water balance, the amounts of surface runoff, and penetration were input into GETFLOWS as daily discharge to the river (surface runoff) and penetration into the subsurface zone (penetration). The lateral flows into the paddy field grids were assumed to occur the same as other grids, and the calculations of paddy model were considered as additional flows.
Watershed initialization
Prior to the hydrological calculation, the initial conditions of water distribution in the watershed were estimated by steady-state calculation under a constant condition. First, all of the grid points except the atmosphere and surface were assumed to be filled with water. Then, calculations for 10,000 days under the assumption of uniform precipitation and evapotranspiration (3.26 mm/day in precipitation and 2.18 mm/day in evapotranspiration) were used considering the annual average precipitation, temperature, day length in the area, and Thornthwaite's equation. Running the model for 10,000 days resulted in a steady state that reflected the proper starting conditions for the predictive simulation. In this initialization, the water budget in paddy fields was calculated by the SCS-CN method. The initial parameters shown in Tables 1 and 2 were used in this initialization.
. | Manning's roughness coefficient [–] . | ||
---|---|---|---|
Land use . | Used for the initialization . | Used for the calibration and validation . | Literature valuesa . |
River and lake | 0.035 | 0.035 | 0.035 |
Waste land | 0.1 | 0.1 | 0.1 |
Paddy | 0.6 | – | 0.6 |
Other crop field | 0.2 | 0.2 | 0.2 |
Forest | 0.4 | 0.4 | 0.4 |
Golf course | 0.4 | 0.4 | – |
Transport | 0.05 | 0.05 | 0.05 |
Other | 0.1 | 0.1 | 0.1 |
Urban development | 0.05 | 0.05 | 0.05 |
. | Manning's roughness coefficient [–] . | ||
---|---|---|---|
Land use . | Used for the initialization . | Used for the calibration and validation . | Literature valuesa . |
River and lake | 0.035 | 0.035 | 0.035 |
Waste land | 0.1 | 0.1 | 0.1 |
Paddy | 0.6 | – | 0.6 |
Other crop field | 0.2 | 0.2 | 0.2 |
Forest | 0.4 | 0.4 | 0.4 |
Golf course | 0.4 | 0.4 | – |
Transport | 0.05 | 0.05 | 0.05 |
Other | 0.1 | 0.1 | 0.1 |
Urban development | 0.05 | 0.05 | 0.05 |
. | Permeability (mDarcy) . | Porosity (m3/m3) . | ||||
---|---|---|---|---|---|---|
Layer . | Initial . | Calibrated . | Literature valuesa . | Initial . | Calibrated . | Literature values . |
Topsoil (II) | 10,000 | 5,000 | 10,000b/5,000c,i | 0.4 | 0.4 | 0.4b |
Alluvium (IV-1) | 1,000 | 10,000 | 100–1,000/10,000b–90,000d | 0.2 | 0.17 | 0.2b |
Loam (IV-2) | 1,000 | 1,000 | 100–1,000b,j | 0.2 | 0.2 | 0.2b,j |
Clay (IV-3) | 100 | 0.287 | 1–10b,k/0.01–0.1e,k/–100f,l | 0.2 | 0.2 | 0.2b,k |
Layer A (V-1) | 10,000 | 25,000 | 10–10,000b,m/100–100–100,000f,h | 0.2 | 0.2 | 0.1–0.2b,m |
Layer B (V-1) | 100 | 200 | 100b,o/20–10,000g,o | 0.2 | 0.05 | 0.2b,n/0.15h,o/0.05–0.10h,p |
Layer C (VII) | 1 | 1 | 0.1–1b,q | 0.5 | 0.05 | 0.05b,q |
. | Permeability (mDarcy) . | Porosity (m3/m3) . | ||||
---|---|---|---|---|---|---|
Layer . | Initial . | Calibrated . | Literature valuesa . | Initial . | Calibrated . | Literature values . |
Topsoil (II) | 10,000 | 5,000 | 10,000b/5,000c,i | 0.4 | 0.4 | 0.4b |
Alluvium (IV-1) | 1,000 | 10,000 | 100–1,000/10,000b–90,000d | 0.2 | 0.17 | 0.2b |
Loam (IV-2) | 1,000 | 1,000 | 100–1,000b,j | 0.2 | 0.2 | 0.2b,j |
Clay (IV-3) | 100 | 0.287 | 1–10b,k/0.01–0.1e,k/–100f,l | 0.2 | 0.2 | 0.2b,k |
Layer A (V-1) | 10,000 | 25,000 | 10–10,000b,m/100–100–100,000f,h | 0.2 | 0.2 | 0.1–0.2b,m |
Layer B (V-1) | 100 | 200 | 100b,o/20–10,000g,o | 0.2 | 0.05 | 0.2b,n/0.15h,o/0.05–0.10h,p |
Layer C (VII) | 1 | 1 | 0.1–1b,q | 0.5 | 0.05 | 0.05b,q |
aLiterature values are described in the dimensions of [L · T–1], and converted to [mDarcy] under the assumption of 1 Darcy = 0.01 cm/s.
iValues of sandy topsoil.
jValues of the Kanto Loam Layer.
kValues of the Jysoso Clay Layer.
lExtrapolated range of value for clay by a regression formula based on various sizes of soils.
mValues of the Shimousa Layer.
nExtrapolated range of value for sand, fine sand, very fine sand and silt by a regression formula based on various sizes of soils.
oValues of the Kazusa Layer.
pValues of mud layer and clay layer.
qValues of the Bedlock.
Parameter calibration and validation
The values of the Manning's roughness coefficient used in this model are shown in Table 1. All parameter values except for the golf course are from NILIM (2006). The value for the golf course was set the same as that of the forest. The calibrated values for permeability and porosity are shown in Table 2. The clay layer represents an aquiclude and the following layers are confining aquifers. Layer C, considered as Bedlock, can be classified as an aquitard. The ratio of irrigation from Kasumigaura Canal, Sakuragawa River, and the repeated use of irrigation water was based on personal communications with Ibaraki Prefecture (Department of Public Works) (2010) and Japan Water Agency (Kasumigaura Canal O&M Office) (2010), and actual inspection (2010), whereas the ratio of groundwater to total irrigation was calibrated.
With reference to the ranges, a parameter set that gives good reproducibility for the river flow rate at Fujisawa Shinden and the groundwater level at Toride (Figure 1) in 1997 (provided by the Ministry of Land, Infrastructure, Transport, and Tourism) was estimated. Then, the predictive capability of the parameter set was confirmed by comparing the calculations and observations of the flow rate and groundwater level in 1998–2001.
Hydrological analyses
Using the finally calibrated model, the movements of water particles precipitated on the surface were calculated to understand the water flow behavior in the watershed. Two days, i.e. January 2, 1997 (dry case) and September 16, 1998 (wet case) were chosen as typical examples of dry and wet conditions, respectively. The former was just after the watershed initialization, and the latter was a period of heavy rain and resultant high river flow. Especially, the total amount of rainfall during the period September 15–16 in 1998 was 148 mm, corresponding to 10.3% of the total precipitation in 1998, and the average river flow rate observed on September 16 and 17 was 149.4 m3/s, which is 14.4 times larger than the average river flow in 1998. In this calculation, the water particles were assumed to be dropped at the center of each surface grid point, and the trajectory of each particle from its original position (surface) to the river mouth was calculated based on the water velocity distribution within the grid system (Figure 2) of the certain time. That is, the calculated trajectory and resultant travel time were predictions if the vector field at each time was kept constant.
RESULTS AND DISCUSSION
Paddy model
Parameter calibration
Based on the preliminary sensitivity analysis, three geological parameters (Manning's roughness coefficient, permeability, and porosity) of the watershed, excluding the surface of the paddy fields as well as the ratio of groundwater usage for the total irrigation of paddy fields, were calibrated to give good prediction of the river water flow and ground water level in 1997. The finally calibrated ratio of groundwater for the total irrigation of paddy fields was 23.6%. The groundwater dependence ratios of agricultural irrigation water in Ibaraki and Kanto were 70% and 15%, respectively (Isozaki 1974). In addition, the specific groundwater supply to rice paddies (=total of groundwater drawn for rice paddy irrigation/total area of paddy fields that used the groundwater) for Ibaraki Prefecture was 2,003 mm, whereas a value of 1,476 mm was estimated for the Hitachi Terrace, in which the southern part of the Sakuragawa River watershed is included (Ooi et al. 2013). As mentioned in the previous section, the Kasumigaura canal supplies irrigation water mainly at the northern part of the watershed. It was concluded that such conditions reduce the watershed dependence on groundwater. Meanwhile, the irrigation water from rivers, lakes, and repeated use are 48.7%, 6.6%, and 21.0%, respectively.
Predictability of hydrological behavior
Hydrology analyses
Table 3 shows the results. The ratios of runoff and penetration were about 70–80% and 20–30%, respectively, in the grids analyzed here, indicating that the surface runoff is two to three times larger than the penetration within the grids of 70,000–80,000 in the area located at Mt. Tsukuba under the wet conditions assumed in this study. On the other hand, there are about 10 grids from the top of Mt. Tsukuba to the river (Figure 16). It is a very simple calculation under the assumption that only a straight line flow from the top of the mountain to the river occurred, but 3% (i.e. 0.710) of water precipitated at the top of Mt. Tsukuba is estimated to reach the river directly by surface runoff. Conversely, 97% of water precipitated at the top of Mt. Tsukuba penetrates into the subsurface zone at least once.
Grid . | Area (m2) . | (1) Precipitation (m3/day) . | (2) Runoff (inflow) (m3/day) . | (3) Runoff (outflow) (m3/day) . | (4) Penetration (m3/day) . | Runoff (outflow)/inflow ratio (3)/(1) + (2) . | Penetration/inflow ratio (4)/((1) + (2)) . |
---|---|---|---|---|---|---|---|
A | 77,122 | 5,385 | 59 | 3,800 | 1,624 | 0.70 | 0.30 |
B | 72,177 | 5,197 | 2,334 | 5,579 | 1,826 | 0.74 | 0.24 |
C | 71,676 | 5,161 | 3,616 | 6,650 | 2,001 | 0.76 | 0.23 |
D | 82,984 | 5,794 | 6,408 | 8,850 | 3,420 | 0.73 | 0.28 |
Grid . | Area (m2) . | (1) Precipitation (m3/day) . | (2) Runoff (inflow) (m3/day) . | (3) Runoff (outflow) (m3/day) . | (4) Penetration (m3/day) . | Runoff (outflow)/inflow ratio (3)/(1) + (2) . | Penetration/inflow ratio (4)/((1) + (2)) . |
---|---|---|---|---|---|---|---|
A | 77,122 | 5,385 | 59 | 3,800 | 1,624 | 0.70 | 0.30 |
B | 72,177 | 5,197 | 2,334 | 5,579 | 1,826 | 0.74 | 0.24 |
C | 71,676 | 5,161 | 3,616 | 6,650 | 2,001 | 0.76 | 0.23 |
D | 82,984 | 5,794 | 6,408 | 8,850 | 3,420 | 0.73 | 0.28 |
FUTURE PERSPECTIVES
There are some perspectives to utilize the model developed here for future watershed management in the Sakuragawa River. Simulation of the effects of land use changes on water resources is one example. Actually, fallow and abandoned rice paddies have increased in Japan (Kusumoto et al. 2005). From the viewpoint of regional vitalization, the effective utilization of these unused paddy fields (Sagehashi et al. 2008; Yang et al. 2012) is essential. Meanwhile, there has been renewed interest in the role of paddy fields as groundwater rechargers (Liu et al. 2001; Yu et al. 2006), and they are considered an effective means of securing freshwater resources. Furthermore, rice paddies also have the potential to serve a role in flood mitigation (Yu et al. 2006). Therefore, it is expected that a freshwater resource management system will be realized based on the appropriate utilization of unused rice paddies in Japan as well as other Asian countries facing a lack of freshwater resources. The model developed in this study will become a powerful tool to simulate such management practices. The increase in extremely hot days in summer observed in this plain (Fujibe 1998; Watarai et al. 2009) may also lead to increased water usage and evaporation, and may result in serious water shortages. The model developed in this study will play an important role in estimating such changes in climate and in discussing rational countermeasures.
CONCLUSIONS
The rice paddy water management was integrated into a distributed three-dimensional surface and subsurface coupling hydrological model of the Sakuragawa River watershed. The predictive capability of this model for river water flow and groundwater level was judged to be sufficient. Especially, the changes in groundwater level in the irrigation periods were predicted successfully by considering the paddy model. Three-dimensional streamlines of the water, the distribution of water travel time, and the water balance in some grids, which represent important information for watershed management, were clarified using the developed model.
ACKNOWLEDGEMENTS
This research was supported in part by the River Fund in charge of the Foundation of River and Watershed Environment Management (FOREM). We thank Dr Yuso Kobara for providing the information about the agricultural activities in the Sakurawgawa watershed. We are also grateful to the staff of the Geosphere Environmental Technology Co. for their assistance in the operation of GETFLOWS.