The study area comprises a semi-enclosed groundwater basin (Figure 1), and has the characteristics of a typical alluvial fan, formed by the Tedori River. The research

area is bound by the Sai River on the north-eastern side, the Kakehashi River on the south-western side, mountainous areas on the south-eastern side and the Japan Sea on the north-western side. Both sides of the Tedori River contain developed farmland consisting mostly of rice paddies and both sides of the Sai River and National Route 8, the main highway in the study region, are de veloped residential areas. The total study area contains 17,682 ha of flat land, of which paddies occupy 42.6%, upland fields 2.8%, residential areas 34.2%, road areas 15.0%, and rivers (including canals) 5.3% [24].

The elevation of the top of the fan is about 80 m above sea level and the distance from the top to the end of thefan is 11 km. The mean gradient is 1/140, which is relatively steep. The underground portions of the study area show an excellent aquifer horizon consisting of sand and gravel, with partially contained a clay lens. The region has, therefore, experienced extensive groundwater exploitation.

Additionally, to confirm the reliability of the nitrogen balance analysis, the nitrogen content of two small rivers in the study area, the Fushimi and Kurahashi Rivers, was analyzed by collecting water quality data after 1993 from the prefectural office. The nitrogen content of the Fushimi River was investigated every month. The river has a 6200-ha catchment area, a predominantly residential area located in the northern part of the study area. The nitrogen content of the Kurabe River was investigated every two months. The river has a catchment area of 1700 ha and is representative of land use in the central part of the study area.

A two-horizon water balance model was applied in the study area to simulate the exchange of water between the channel/soil and aquifer horizon fractions and the Tedori River (Figure 2). The flow among the various components is represented by Equations (1)-(3) [26].

In the atmosphere:

In the channel-soil horizon:

In the aquifer horizon:

where Pr is precipitation, Prd is direct runoff, Prg is infiltration, Qin is the intake water to the study area from head works (irrigation canal off-take points), ET is evapotranspiration, Per is the percolation from the soil surface, Qout is the outflow discharge from the drainage canal, Pup is pumped water from the aquifer, Tin is recharge water from the Tedori River, Tout is effluent water into the Tedori River, and Gout is discharge from the aquifer to the lower reach of the drainage canal or the Sea of Japan directly.

The water balance model and its components are reported in detail elsewhere [25]; however, a brief summary is provided below.

2.2.2.1. Precipitation and Evapotranspiration (ET)

Precipitation rain data, measured by tipping type automatic gauge, used in this study covered 35 years (1976- 2010) and were collected at the Kanazawa Local Meteorological Observation Station (hereafter Kanazawa), which is located about 10 km from the center of the study area. The mean annual precipitation for the region over the 35 year study duration was 2425 mm, 877 mm of which fell during the irrigation period (13 April to 10 September) and the remaining 1548 mm fell outside of the irrigation period (non-irrigation period). The precipitation in the non-irrigation period includes snowfall. Daily evapotranspiration (ET) over the same period was estimated by the Penman equation using data for Kanazawa. However, the ET of roads (15.0%) and roofs of houses (about 11.5%) was assumed to be zero; thus, only 73.5% of the total ET calculated by the penman equation was applied to the entire study area.

2.2.2.2. Water Use and Percolation Water use (water requirement), investigated by paddy lot, for rice paddies is a very important factor in the analysis because this differs markedly before and after farm land consolidation. Consolidation resulted in soil compaction caused by heavy machinery, with consequent reduction in soil permeability.

Farm land consolidation took place over thirty years between 1965 and 1995 (Figure 3), after which, Murashima [27] investigated water use in 25 test lots of paddy fields in Shichika and 12 test lots in the Miyatake irrigation district. The average water use was 15.6 mm∙day⁻¹ (standard deviation = 5.5 mm∙day⁻¹) over the irrigation period. Prior to farm land consolidation, the average use in 16 paddies during July 1950 was 28 mm∙day⁻¹ [28]. During the land development works water use was assumed to be between 15.6 mm∙day⁻¹ and 28.0 mm∙day⁻¹ in proportion to the areas of developed and undeveloped land, respectively.

The mean percolation in paddies during the irrigation period was estimated as the water use minus Penman evapotranspiration; the evapotranspiration coefficient was assumed to be 0.85. The percolation rate used for the entire study area (17,682 ha) was based on that for the cropped rice area.

Percolation water from irrigation is assumed to travel directly to the aquifer, because the study area has a deep groundwater level and a gravel and sand horizon beneath the topsoil.

2.2.2.3 Discharge in the Tedori River The amount of inflow to, or discharge from, the Tedori River is a very important water balance component. Therefore, we conducted a survey of discharge in several sections of the Tedori River during the irrigation and non-irrigation periods and calculated the difference in discharge between the two sections. Discharge clearly declined in the river stretch between 2.2 and 16.4 km

from the river mouth, indicating that water from the river was recharging the aquifer underlying the area. The discharge between 1.1 and 2.2 km from the river mouth increased slightly, reflecting the inflow of effluent from the area into the river.

2.2.2.4. Pumping Groundwater from Wells Ishikawa Prefecture has reported on the amount of groundwater pumped, investigated by pump capacity multing operation time, from wells in the area [29]. We sorted these data to correspond with irrigation and non-irrigation periods.

2.2.2.5. Intake Water for Paddy Irrigation To assess paddy intake water quantities, we used observed data surveyed at the top of the Shichika irrigation canal from 1976 to 2010. No observations were available for the Miyatake District, thus paddy intake water was estimated based on water right discharges between two the Districts. Five yearly averages observed intake water quantities are shown in Figure 4.

2.2.2.6. Estimation of Direct Runoff and Infiltration from the Ground Surface To estimate percolation and direct runoff, runoff analysis was conducted, using a ratio of precipitation volume to direct runoff for various land uses (Figure 5) [30].

Land use areas were determined from statistical data for Ishikawa Prefecture. Direct runoff from rotational use of upland paddy areas was estimated using of those runoff ratio. Direct runoff from the paddy areas during the irrigation period was estimated using the runoff ratio for the non-irrigation period. The roof area percentage for residential areas (33.7%) was determined using GIS information in which the runoff ratio was assumed to be 100% [31].

Based on the above procedure, the amount of direct runoff and percolation from the ground surface for various land uses was estimated using daily precipitation and ET over 35 years divided into five-yearly intervals (1976- 2010).

Based on the water balance analysis, the nitrogen pollution load potential (NPLP) was estimated principally from the product of water flow volume and nitrogen concentrations (hereafter total nitrogen, TN). The nitrogen balance model was constructed from the water balance components adding nitrogen loads from farmland, effluxent from WWTWs and denitrification in paddies (Figure 6). Nitrogen concentration data were investigated over three years (2007-2009) from surface and groundwater.

Nitrogen deposited by precipitation was monitored at weekly intervals over an extended period by the Ishikawa Environment and Health Research Institute [32], located 10 km from the study area. In addition to wet deposition from precipitation, there is dry deposition of nitrogen from the atmosphere. Atmospheric deposition showed remarkable seasonal changes with high deposition during winter and low deposition during the summer season. This was due to the influence of seasonal winds from Siberia.

The observed volume of intake water (the water extracted from the river into irrigation canals, investigated H~Q curve) is shown in Figure 4 and the nitrogen concentration was observed at the Hirose sampling site near the Hakusan Head works. The NPLP in intake water is shown in Figure 7. The nitrogen concentration of intake water was low during the irrigation period and high during the non-irrigation period because of rainfall deposition [30,31].

The nitrogen concentration of percolation water, equals to surface water, was measured at 53 surface locations four times during the irrigation period and three times during the non-irrigation period [35]. The average concentration was 0.39 mg∙l⁻¹ and 0.57 mg∙l⁻¹ during the irrigation and non-irrigation periods, respectively.

The nitrogen concentration of recharge water from the river was calculated according to the observed data; 0.36 mg∙l⁻¹ during irrigation and 0.48 mg∙l⁻¹ during the nonirrigation period in the upand mid-stream portions of the river. The effluent (river outflow) water concentration was determined to be 1.18 mg∙l⁻¹ and 1.11 mg∙l⁻¹ during the irrigation and non-irrigation periods, respectively [35].

The nitrogen concentration of the pumped (aquifer) water was assumed to be the average nitrogen concentration of the groundwater observed from 53 points across in the entire study area [35]. The concentration was not markedly different between the irrigation and non-irrigation periods.

The NPLP in surface drainage water was based on measured nitrogen concentrations and estimated surface drainage flows. The concentration was measured near the Sea of Japan, four times at 5 - 6 sites during the irrigation period (n = 22, σ = 0.19), and four times at 6 - 8 sites during the non-irrigation period (N = 28, σ = 0.17) [35]. The nitrogen concentration was measured by appliance for continuous micro flow (OVATTRO 2ch-TNTR mulch).

Similarly, the NPLP in groundwater discharge was estimated from measured nitrogen concentrations and estimated groundwater discharge. The nitrogen concentration was investigated at 63 locations during the irrigation period (σ = 0.60) and at 55 locations (σ = 0.70) during the non-irrigation periods [35].

NPLP from farmland was assessed in previous studies [37,38]; NPLP was estimated as the difference in nitrogen between fertilizer application rate and the yields of various crops. The fertilizer input was determined from standard application rates for various crops, as recommended by the local government office in the study area. The cultivated crops were rice, soybean, barley, vegeta-

Table 1. Temporal change of water balance components in every five years over 35 years (mm∙day⁻¹).

bles, and orchard fruits. The nitrogen content in the unit yields for crops was determined from standard tables of food composition. The yield of crops was determined from statistical data collected by the prefectural govern ment for crops every five years. NPLP shows remarkable temporal changes (Figure 8).

The peak of NPLP occurred in 1990 at around 1000 t∙yr⁻¹, decreasing gradually thereafter to about 400 t∙yr⁻¹ during the period 2005-2010. The largest NPLP source was rice cultivation, because of the large total cropping area. The second was horticultural crops; vegetables and orchards, due to heavy fertilizer per unit area application rates.

Nitrogen in effluent from WWTWs is a very important factor for NPLP, and existing estimates were available for the study [41]. In the study area WWTWs only provide secondary stage treatment, which does not remove nitrogen (whereas tertiary treatment removes nitrogen completely); consequently, large quantities of nitrogen flow into the environment.

NPLP from WWTW was estimated from population equivalent nitrogen loads for various WWTWs; These were 4.8 g∙person⁻¹∙d⁻¹ for activated sludge, 1.6 g∙person⁻¹∙d⁻¹ for the oxidation ditch method (a kind of public sewage treatment), 6.1 g∙person⁻¹∙d⁻¹ for the rural sewage method, 6.5 g∙person⁻¹∙d⁻¹ for combined septic tanks, and 7.6 g∙person⁻¹∙d⁻¹ for night soil septic tanks [42]. Data from the Prefectural Government show how both the number and types of WWTWs gradually changed with time.

The peak WWTW NPLP was about 800 t∙yr⁻¹ in 1994 (Figure 9) [42]. With the application of public WWTWs, which discharge directly into the Japan Sea, the NPLP within the study area had gradually decreased from approx. 600 t∙yr⁻¹ in 1984 to approx. 100 t∙yr⁻¹ in 2010.

The nitrogen balance analysis results for the irrigation period and non-irrigation periods are presented in Tables 2 and 3 respectively; the NPLP was obtained from the product of the volume of each component and its nitrogen concentration (acronyms refer to Table 1).

Tables 4 shows the input and Table 5 shows output of the NPLP which obtain by adding irrigation and nonirrigation periods (Tables 2 and 3). In the two tables, the first term of the acronyms indicate the water balance components and second term shows nitrogen.

The difference between the input and output is the pollution amount, i.e. the mass of pollutant stored in each horizon. In addition, denitrification (De-N) from planted paddy was estimated as 17.1 kg∙ha⁻¹ from the test paddy [24] (Table 5).

The total input of nitrogen peaked 2155 t∙yr⁻¹ at period

(3), after which rapidly decreased 1035 t∙yr⁻¹ at period (7). The dominant input components were Agr-N and SwN mentioned detail in following section. The other components keep relatively constant (Table 4).

The total output of nitrogen peaked 1141 t∙yr⁻¹ at period (1) after which rapidly decreased. The dominant decreased component is Gout-N, while Qout-N is gradually increased. The De-N gradually decreased because of paddy area decreasing. The other components maintained relatively constant (Table 5).

The difference of input and output, index of pollution of the study area peaked 1129 t∙yr⁻¹ at period (3), after which markedly decreases 174 t∙yr⁻¹ at period (7). The main reason of the phenomena was remarkable decreasing of Sw-N and Agr-N in the input component.