Urbanization, agriculture, fertilization, livestock farming and unprecedented precipitations are presumed to cause augmented loadings of nitrogen (N) and phosphorus (P) to rivers and downstream reservoirs. At present, point source (PS) pollution in watersheds can easily be controlled, and it has been well-managed due to growing awareness and strict low enforcements. However, the control of pollutants from non-point sources (NPS) is still challenging and NPS have been identified as the main cause of water pollution and eutrophication in watersheds at present. Limitations in technical, human and financial resources impede efficient monitoring of those influents at watershed scale. At this end, process-based modelling approaches play an important role in analysis of nutrient loading effects quantitatively and qualitatively. Process-based water and energy processes (WEP) hydrological model with its updated version to couple nutrient loading through the implementation of N and P circulation processes was studied in this paper highlighting the application of the model to the Takasaki River, a tributary to the Inba-numa Lake basin, Chiba prefecture, Japan to understand the impacts of PS and NPS to the river water quality.
River water quality is a main factor in sustainable water environment since it has huge impacts on domestic, agricultural and industrial water usages. With the rapid increase of global population and urbanization, the intensified risk of the deterioration of river water quality has to be considered seriously in sustainable water resources management. Therefore, proper understanding of the sources of water pollution is essential in managing and restoration activities in river basins. The sources which cause river water pollution have been categorized as point sources (PS) and non-point sources (NPS). Although technical developments and rigid regulations have minimized the impact of PS, still the control of NPS is a challenging task. NPS such as agricultural, urban areas, forests and atmospheric depositions are major contributors which cause water quality degradation in rivers and other water bodies. Field monitoring and suppression of NPS are difficult task since their special distribution is vast (Hayakawa et al. 2006). One of a significant feature of NPS is their relationship with the hydrological cycle, especially; the rainfall events intensify the pollutant loadings from NPS to water bodies. Numerical model is a useful tool in understanding the transport of various nutrients from their sources to receiving ends. Physically based, statistical and empirical models are frequently used in nutrient loading studies in river basins. In complex environments, such as river basins with various land use types, PS and NPS, application of physically based models is more rational, since those complex features can be incorporated in physically based models. Application of such a numerical model to study the nutrient loadings of N and P from PS and NPS is discussed in this paper.
In this study, the Takasaki River which is a sub-catchment of the Lake Inba-numa is selected to analyse the influence of PS and NPS nutrient loadings. The Lake Inba-numa is located in the north-western part of Chiba Prefecture, Japan (Figure 1). It has a catchment area of 541 km2. Approximately 12% of the prefectural population, i.e. 750,000 live in the catchment of the lake. With the rapid urbanization and agricultural activities, the water quality of the lake has been deteriorating since 1970s. In the recent past, the Lake Inba-numa was ranked as the lake with the worst water quality among those being used as water supply source in Japan. As a result of several restoration activities, the water quality in the lake has been improved. In water quality restoration activities, control of NPS are still challenging while a significant improvement can be seen in the controlling of PS pollution in the Lake basin. According to Yuasa et al. (2010), there are two issues in improving the water quality of the lake. One is that the reduction of pollution loads within the catchment solely does not lead to accomplishment of drastic improvement of water quality of the Lake Inba-numa, and the other is that the problem of NPS, which are harder to be controlled, still remains and become grater in portion, making further implementation of measures difficult. The Takasaki River, which is the focus of the present study (Figures 1 and 2), has a basin area of 85.6 km2 and the population reside in the basin is about 94,000. The river drains into the Inba-numa Lake with average discharge at Ryutou Bridge observation point (Figures 1 and 2) is 3.23 m3/s for the past 10 years. According to the recent land use, approximately 65% of the basin area is un-urbanized areas such as forests, farmlands and paddy fields while the remaining 35% is urbanized (Figure 2). In the upstream of the river, the land use consists of agricultural lands and forests while the downstream urbanization related land uses are dominant. As a result of this land use distribution, nutrient loadings from upstream is significant than downstream especially for N due to fertilization. Both commercial and natural fertilizers such as manure are being used in the basin. Ongoing urbanization and infrastructure developments change the land use types and it affects the water quality of the river. There are four rainfall gauging stations and one water level measuring point in the basin which provides hourly data for the simulation and those stations are regulated by the Chiba Prefectural government, Japan. The river width varies between 1 and 13 m while the maximum depth of the river is 3.5 m.
Estimated daily loadings of TN and TP for 2008 based on the basic unit load method by the Chiba Prefectural government for four municipalities in the Takasaki River Basin, Sakura city, Shisui city, Tomisato city and Yachimata city are shown in Figure 3. The figure illustrates the contribution of nutrient loadings from different land use types including both PS and NPS. Among the different sources, farmlands contribute highest TN loading rates while livestock is responsible for the highest discharge of TP according to the data shown in Figure 3. Individual residences and urban areas have discharged second and third largest amounts of TN and TP to the river in 2008. The Tomisato city, which is located in the upstream agricultural area of the basin, has contributed the highest loading of TP compared to other three cities and that loading was mainly from livestock. The Yachimata city, which is located in the upstream residential area has high rate of TP loadings and its contribution has mainly come from residential and urban sources. As overall estimated loading rates are concerned, it is obvious that the Tomisato city and Yachimata city had high loading rates for both TN and TP in 2008. Moreover, the types of sources show a significant impact to the different loading rates as they are illustrated in Figure 3.
The annual rainfall of the selected basin is shown in Figure 4. It can be noticed that the maximum and minimum of annual rainfalls have been occurred in 2006 and 2007. In this study, initially the hydrological model was calibrated for 2005 discharge and then water quality model was calibrated for the two rainfall events in 2005. The calibrated model was applied to study the nutrient loading of 2006 and 2007, where the Takasaki River received its maximum and minimum rainfalls in the recent past. The average discharges for those consecutive years were 3.61 m3/s and 2.24 m3/s respectively. The annual volume of water passed though Ryutou Bridge measuring station for 2006 and 2007 were 114 × 106 and 71 × 106 m3. According to those discharge figures, those 2 years show a significant difference in rainfall and river discharge. In this study, we try to analyse the impact of this rainfall variation on the nutrient loading to the river.
Development of the material loading model
The water and energy processes (WEP) model is a grid-based distributed parameter hydrological model developed by Jia et al. (2001, 2005). The initial hydrological model is capable of simultaneous simulation of hydrological phenomena of evapotranspiration, infiltration, subsurface and groundwater flow, overland flow/runoff and river flow with the energy transfer processes (radiation and heat fluxes). The model is based on the Green-Ampt model for infiltration simulation during heavy rain events to optimize computation time and capable of modelling multi-layered aquifers with heterogenic land use patterns by using a mosaic method. Penman − Monteith equation is adopted in the model for the evapotranspiration calculation. River flow routing is conducted for each tributary and a main river by using kinematic wave method. Overland flow (1-D) is simplified as lateral inflow to rivers following slope direction. In addition, a two-dimensional simulation of multi-layered aquifers, i.e. quasi-3D simulation, is performed for groundwater flow (Jia et al. 2005). The inputs to the hydrological model can be categorized into three groups as meteorological data, geographical/hydrological data, and anthropogenic activities’ data. Mosaic method is used in the model for the re-classification of land use within the grids which allows simulating more realistic land use distribution in the model.
Later the model was updated to accommodate N and P circulation in surface and subsurface environments. The movements of dissolved N and P (DN, DP) were assumed to follow the same flow paths in overland and channel (1-D advection diffusion) and groundwater (2-D multi-layer advection diffusion) flows in the updated model, while particulate N and P (PN, PP) were modelled similarly in overland and channel flows. The coupled biogeochemical model is a process-based nutrient dynamics model at watershed level (Kinouchi et al. 2005) and includes detailed flow modules to simulate overland and stream pathways of dissolved and particulate matters of N and P (DN, PN, DP and PP), percolation and subsurface flows of dissolved matter. It generates overland, groundwater and stream concentrations of nutrients in time and space, based on holistic simulation of all involved processes like fertilizer input, atmospheric deposition, turnover through litter mineralization, nitrification/depuration in river beds, and outflows due to immobilization, crop uptake and harvest, surface runoff and leaching, and loading to downstream lakes via river network (Rajapakse et al. 2009).
Both NPS (from farmlands, forested and urbanized areas) and PS (from household, industrial, livestock farming and recreation areas) pollution were considered in the present model. The initial values of parameter were presumed based on published literature, and subsequently adjusted during calibration. The historical data on annual fertilizer input to crop and harvest from paddies, farmlands and orchards were collected by analysing the GIS/digitized land use map data and published data in each municipalities which are located in the basin.
|Mean Absolute Error (MAE)||0.59|
|Relative Root Mean Square Error (RRMSE)||0.55|
|N–S Coefficient of Efficiency (EF)||0.73|
|Coefficient of Determination (CD)||1.04|
|Mean Absolute Error (MAE)||0.59|
|Relative Root Mean Square Error (RRMSE)||0.55|
|N–S Coefficient of Efficiency (EF)||0.73|
|Coefficient of Determination (CD)||1.04|
RESULTS AND DISCUSSION
Firstly, the hydrological model of the WEP was calibrated for the 2005 data set. The obtained results are shown in Figure 5 for the calibration period. The simulated discharge time series correlated reasonably well with the observed hourly discharge data at the Ryutou Bridge observation point. The calibrated hydrological model shows reasonable performance indicators as shown in Table 1 with N-S coefficient 0.73, RRMSE 0.55 and MAE 0.59, which implies that the model is satisfactory to be coupled with water quality model. From June to September of 2005, there were several rainfall events which exceeded 10 mm/hr. The simulated discharges corresponded well with the rainfall for those events and the low flow conditions during less or zero rainfall periods were also simulated satisfactorily by the model as shown in Figure 5. Although the river discharge data was available continuously for 2005, the measured river water quality data was available only for two rainfall events. The model results for TN and TP are illustrated in Figures 6 and 7 for the two rainfall events in 2005.
Two measured data sets for discharge, TN and TP were available for the two rainfall events occurred in 2005, where the rainfalls were above 15 mm/hr as shown in Figure 5. The first event's duration was 65 hours spreading 3 days from July 25th and the second rainfall event lasted 91 hours starting from August 24th. During the first event average measured river discharge, TN loading and TP loading were 21.51 m3/s, 108.04 g/s and 11.73 g/s, while the simulated averages were 17.67 m3/s, 86.24 g/s and 7.52 g/s, respectively. In the case of second event, average measured river discharge, TN loading and TP loading were 22.52 m3/s, 88.74 g/s and 8.73 g/s while the simulated averages were 17.72 m3/s, 80.84 g/s and 7.65 g/s, respectively. According to the obtained results, the simulated values show under-estimation compared to the measured values for certain extent. As shown in Figures 6 and 7, even the rainfall magnitudes of two events were almost same, the loadings from the first event is significantly higher than the second event. The main reason for such a phenomenon is the accumulation of nutrients during the dry season before the first rainfall event. As shown in Figure 5, there was a dry period before the first event which caused the accumulation of nutrients on the ground surface. Although there were several small rainfalls between two events it was not enough to flush significant nutrient loading to the river until the second large rainfall event occurred which had maximum rainfall about 17 mm/hr. However, those small rainfalls have hindered the accumulation of nutrients on the surface and it was the reason for less loading occurrence in the second rainfall event than the first. As far as the total loadings were concerned, in the first rainfall event, the total TN and TP loadings were 25.28 ton and 2.74 ton based on measured values while 20.18 ton and 1.76 ton for simulated values. For the second event, loadings were 29.07 ton and 2.86 ton based on the measured values and 26.48 ton and 2.51 ton based on the simulated results for TN and TP, respectively. The relationships developed between river discharge and nutrient loadings are shown in Figure 8. For both TN and TP, linear regression lines for measured discharges and loadings shows high R2 values than that of simulated loadings. The shown relationship for the nutrient loadings can be used as a prediction tool for future events if the discharge can be measured during rainfall events.
The nutrient loading simulations for the maximum and minimum rainfall years were conducted to analysis the loadings in 2006 and 2007. Figures 9−11 illustrated the daily river discharges, TN and TP loading, respectively, for 2006 where maximum annual rainfall has been occurred in recent past while Figures 12−14 depicts similar components for 2007 where minimum annual rainfall has been occurred recently. In 2006, there were two rainfall events which had daily rainfalls exceeding 100 mm. The daily averaged discharges of those two events were above 40 m3/s. Contrary to 2006, in 2007 the maximum daily rainfall occurred was below 100 mm and the peak discharges were below 40 m3/s. Similarly, the TN loadings which are co-related to the rainfall show high values for 2006 and low values for 2007. The WEP model's performances for the NPS during rainfalls events were reasonable and it was able to simulate the TN and TP loadings from PS and NPS for expectable levels according to the obtained results.
The application of improved WEP model to the water quality assessment during rainfall events is discussed in this paper. The calibrated model results show its capability in simulating the river discharge and water quality components successfully. The event based N and P loadings corresponded to the measured values with reasonable accuracy in 2005 simulation. The model was applied to 2006 to check its performance for high rainfall events. In 2007, where low rainfalls were recorded, the model was tested to compare the impact of rainfall in generating NPS loadings. Although in real practice, identification of PS and NPS nutrient loadings is a difficult task, with available land use, hydrological and other detailed nutrient discharge data this simulation approach is able to produce valuable information for the river water quality leading better watershed management where river water quality issues are prevailing. Further, relationships between river discharges and nutrient loadings can be established. Those can be supportive in predicting nutrient loadings in ungauged basins.