Đang tải... (xem toàn văn)
A distributed water quality tank model was developed to evaluate nitrogen load reduction by artificial wetlands in the A10 area, located in the Lake Kasumigaura watershed in Japan. In order to collect data for model calibration and verification, a year-long field investigation was conducted. The study area was divided into a 50 m × 50 m grid, and attributes and the model are added to each cell. The model simulates the runoff and the total nitrogen (TN) concentration with average relative errors of 10% and 2%, respectively, except for no-rain periods. Scenario analyses were conducted considering artificial wetlands at different locations. The three scenarios are as follows: no artificial wetland, cell located at the head of the valley and cell located on a flow path through animal waste cell. Compared with case 1, the annual TN concentration was lower by 4% and 11% in cases 2 and 3, respectively. The results suggest that artificial wetlands on water flow paths with a large nitrogen load effectively reduce these loads and that the model quantitatively evaluates the difference in the nitrogen concentration across locations.
Journal of Water and Environment Technology, Vol.3, No.2, 2005 - 235 - A DISTRIBUTED WATER QUALITY TANK MODEL FOR NITROGEN LOAD REDUCTION BY ARTIFICIAL WETLANDS T. Kato* 1 , H. Kuroda * 2 and H. Nakasone * 3 * Faculty of Agriculture, Ibaraki University, 3-21-1 Chuoh, Ami, Inashiki, Ibaraki 300-0393, Japan 1. tkato@mx.ibaraki.ac.jp, 2. kuroda@mx.ibaraki.ac.jp, and 3. nakasone@mx.ibaraki.ac.jp ABSTRACT A distributed water quality tank model was developed to evaluate nitrogen load reduction by artificial wetlands in the A10 area, located in the Lake Kasumigaura watershed in Japan. In order to collect data for model calibration and verification, a year-long field investigation was conducted. The study area was divided into a 50 m × 50 m grid, and attributes and the model are added to each cell. The model simulates the runoff and the total nitrogen (TN) concentration with average relative errors of 10% and 2%, respectively, except for no-rain periods. Scenario analyses were conducted considering artificial wetlands at different locations. The three scenarios are as follows: no artificial wetland, cell located at the head of the valley and cell located on a flow path through animal waste cell. Compared with case 1, the annual TN concentration was lower by 4% and 11% in cases 2 and 3, respectively. The results suggest that artificial wetlands on water flow paths with a large nitrogen load effectively reduce these loads and that the model quantitatively evaluates the difference in the nitrogen concentration across locations. KEYWORDS: Diffuse pollutant source; Purification; Scenario analysis; Total nitrogen INTRODUCTION Nutrients are often found in high concentrations in agricultural watersheds due to excess fertilizers and animal wastes. In general, it is difficult to effectively reduce the amount of diffuse pollutants; however, it is known that paddy fields or wetlands have a purification function, especially for nitrogen pollutants. In a small valley, spring water contaminated by a diffuse pollutant load from farmlands located on the surrounding uplands is naturally purified in the paddy fields located on the lowlands; this characteristic is called the ‘topographic chain’ or ‘topographic sequence’ effect (Tabuchi, 1986; Tabuchi et al., 1991; Nakasone et al., 1996; Kuroda, 1998) and is observed in many farmlands in Japan. In this study, a distributed water quality tank model was developed, and the difference in nitrogen concentration was simulated considering artificial wetlands at different locations. An ‘artificial wetland’ is land formerly used as a paddy or bare field that is now filled with water contaminated with diffuse pollutants. A distributed model is generally used to simulate location information (Lu et al., 1989; Goto et al., 1993; Abbott et al., 1996; Vieux, 2001; NIES, 2002); therefore, we modified a ‘water quality tank model classified by land use’ (Kato et al., 2003) into a distributed model based on detailed data obtained from an approximately year-long investigation and simulated the nitrogen load under three scenarios considering artificial wetlands at different locations. STUDY AREA AND FIELD INVESTIGATION Study area Journal of Water and Environment Technology, Vol.3, No.2, 2005 - 236 - The study area considered was the A10 area, located in the Kasumigaura watershed in Japan (Fig. 1). Lake Kasumigaura is the second largest lake in Japan. Its water resources are used for municipal, domestic, industrial and irrigational purposes. Despite the countermeasures for improving water quality, the annual average total nitrogen (TN) concentration did not improve; rather, it remained at 0.93 mg·L –1 in 2002 (Ibaraki Prefecture environment white paper, 2003). The area of the A10 is 55.4 ha; in 2001, the human population in the area was 118 and the livestock population was 2,400. Animal wastes, in particular, have accumulated in unlined storage ponds and have affected the water quality (Shimura et al., 1996). Though the acreage is small, the land use is diverse—paddy fields, upland vegetable fields, forests and livestock farms—and the TN concentration in the drainage water varies across the points. Therefore, we believe that the area is suitable for comparing the effect of artificial wetlands using the distributed model. Field investigation A field investigation was conducted at four points to measure the nitrogen concentrations and collect runoff data for the period from August 2001 to July 2002. Sampling point A is the mouth of the river; point B, the drainage in the paddy fields surrounded by forests; point C, the drainage of the paddy fields surrounded by the upland fields; and point D, the spring water near the livestock farms (Fig. 1). The measurement parameters are the TN, NO 3 -N, NO 2 -N and NH 4 -N concentrations (measurement instruments: YANACO TN-301P; DIONEX DX-100). The water was analyzed by our laboratory staff at Ibaraki University. At point A, a weir was built in the drainage and the runoff was measured using a pressure-type water level gauge (HI-TECHS, HIT-M8) every 10 min for the period from 10 August 2001 to 24 July 2002; in addition, the rating curve (water level discharge curve) was graphed based on the data from an electromagnetic current meter (Tokyo Keisoku SF-5001) every week. The water was sampled using an automatic sampling machine (ISCO, ISCO6400) everyday. At points B, C and D, the water was manually sampled every week for the period from 1 April to 24 July 2002. Results of the investigation The average measured data are shown in Table 1. The NO 3 -N concentrations for points A and D are considerably higher than the maximum legal level of 10 mg·L –1 . The TN concentration is lowest at point B because the main land use in this area is forests. The TN concentration at point C is higher than that at point B because upland fields are located behind it. The TN concentration at point D is 46.7 mg·L –1 because the water is spring water contaminated by fertilizers and animal wastes. The data measured at point A are illustrated in Fig. 2. Data for October 2001 and April 2002 are unavailable due to machine trouble. The TN and NO 3 -N concentrations were diluted by rainfall. TN mainly consists of NO 3 -N; these results support those of the previous study, which revealed that nitrogen contained in fertilizers and animal wastes accumulates in soil (Kuroda, 1996). The NH 4 -N concentration is typically high when N 0 200 m (ha) Paddy Forest Upland Residence 9.3 23.1 21.5 1.5 Sampling point A D C B River Fig. 1 Land use and sampling points in the A10 ・ Ibaraki Tokyo Lake Kasumigaura Drainage Unlined storage pond Journal of Water and Environment Technology, Vol.3, No.2, 2005 - 237 - rainfall occurs, indicating that animal wastes directly flow in the drainage along with rainwater. We confirmed that a large difference exists in the nitrogen concentrations across locations and that fertilizers and animal wastes are the main sources of pollutants. DISTRIBUTED WATER QUALITY TANK MODEL Structure of the model The study area was divided into a 50 m × 50 m grid with a total of 350 cells. Attributes such as elevation, land use and drainage data were added to each cell and were determined by means of a field investigation by using a land-use map and a 1:5,000 map (Fig. 3). The Geographical Survey Institute in Japan publishes elevation data and the land-use map. A channel network (flow directions) is determined based on the elevation data, and the water and nitrogen load move to the lowest elevation cell among the four adjacent directions. A tank model, comprising a set of two tanks, is applied to each cell (Fig. 4), and a tank is also applied to a drainage cell. Water and the nitrogen load arrive at the drainage cell and then flow into a drainage tank. The time step is set to 10 min. The equations used in the model are as follows. Symbols used in the equations are illustrated in Fig. 5. The equations for the runoff are as follows: (S i, t − h i, j ) r i, j = q i, j (1) Σ mesh Σ j q i, j = Q (2) where S is the storage amount in the tank, h is the hole height, r is the coefficient of discharge, q is the 22 22 223 3223 2231 322311 3222311 32233314 2222311331 11223222333111 1222335222311 3323333323311 3333333323441 15315332231111 1111222221111 1113332323311 331142322311 111122522331 115223322441 111333323141 113311123411 111311133411 15511111351 1111111111 111111 03 24 535 75430 861230 11 9 7 12 30 31 11 9 8 11 30 31 32 13 12 10 12 13 30 31 32 13 12 11 12 14 30 31 32 33 34 35 34 16 14 14 13 13 14 30 31 32 33 34 35 35 34 16 15 16 16 15 15 16 31 32 33 34 35 34 17 18 17 17 17 17 17 31 32 33 34 35 34 19 20 31 31 30 26 18 31 32 33 34 35 34 31 32 32 31 30 26 19 31 32 33 34 35 34 33 32 30 26 25 23 20 21 32 33 34 35 34 33 31 29 27 27 24 25 22 32 33 34 35 33 32 30 28 29 26 26 23 24 33 34 35 34 34 33 31 29 28 28 27 25 33 34 35 35 33 32 31 30 32 32 27 26 33 34 35 36 34 34 34 34 31 30 28 32 33 34 35 37 35 35 35 35 35 30 29 30 31 34 35 38 36 36 36 36 36 31 32 32 32 34 35 37 37 37 37 37 33 33 33 33 34 35 38 38 38 38 36 36 36 36 36 36 37 37 37 37 37 37 Elevation data (m) Land use data Drainage data Fig. 3 Mesh data 1: upland field 2: paddy field 3: forest 4: residence 5: livestock farm 0 10 20 30 40 50 60 8/10 9/10 10/10 11/10 12/10 1/10 2/10 3/10 4/10 5/10 6/10 7/10 濃度 (mg/L) 0 50 100 150 200 250 降水量 (mm) 降水量 T-N NO3-N NH4-N Concentration (mg · L -1 ) Rainfall (mm) NO 3 -N NH 4 -N Rainfall TN Fig. 2 Results at sampling point A Aug Oct Dec Feb Apr Jun 50 0 100 150 200 250 0 40 60 20 ABCD TN 30.45 5.82 10.70 46.73 NO 3 -N 25.85 4.60 9.38 43.24 NH 4 -N 1.01 0.16 0.11 0.49 NO 2 -N 0.43 0.01 0.02 0.02 Table 1. Average measured data (mg·L -1 ) Journal of Water and Environment Technology, Vol.3, No.2, 2005 - 238 - outflow discharge from each hole, Q is the total outflow discharge to the adjacent tank, i is the tank number (1: upper tank, 2: lower tank), j is the hole number (0: bottom), t is the time step and mesh is the item (0–3: the number of cells from which water and nitrogen load flow). The equations for the storage amount are as follows: S 1, t = S 1, t-1 + R + I − E + Q 1 /A − Σ j q 1, j (3) S 2, t = S 2, t-1 + q 1,0 + Q 2 /A − Σ j q 2, j (4) where R is rainfall, I is the quantity of irrigation water, E is the amount of evapotranspiration and A is the total number of cells through which water flows. The equations for the nitrogen load are as follows: (((1 − β) (R N + I N + αPL + D 1 ) + S N1 )/S 1, t + c 1, j B 1, j ) q 1, i = d 1, i (5) (((1 − β) (αd 1, 0 + D 2 ) + S N2 )/S 2, t + c 2, j B 1, j ) q 2, i = d 2, i (6) where R N is the nitrogen load in rainfall, I N is the nitrogen load in irrigation water, PL is the pollutant load for each type of land use, D is the total amount of load flowing to the adjacent tank, S N is the load in the tank, c is the coefficient of cumulative load, B is the amount of cumulative load, d is the outflow load from each hole, α is the dissolution rate and β is the purification rate (Kato et al., 2003). The equations for the cumulative load are as follows: B 1, j, t+1 = B 1, j, t + (1 − α) PL/3 (7) B 2, j, t+1 = B 2, j, t + (1 − α) d 1, 0 /2 (8) Input data The input data are rainfall, TN concentration in rainfall (1.1 mg·L –1 ), quantity of irrigation water, land-use area, human population, livestock population, fertilizer (amount, absorption rate for each crop and schedule) and unit effluent load. Hourly rainfall data were drawn from the data collected at the Hokota observatory. The evapotranspiration, measured at Ibaraki University, was found to be 3 mm·d –1 . The quantity of irrigation water was calculated based on the pump operation records; the average TN concentration of irrigation water was 9.5 mg·L –1 for the period from 20 May to 24 July 2002. Data for fertilizers were obtained from literature (CAES, 1976; Nishio, 2001). The TN load per unit of livestock and septic tank were 28 and 10.35 g·d –1 , respectively (IPLEDKCS, 1999). Model calibration and verification The trends in the measured data for point A for the period from 10 August to 31 December 2001 were simulated with the parameters manually adjusted to reduce the least squares error (Table 2). Considering that nitrogen effluent rates differ depending on the land use, the dissolution rate (α) is set to 0.5 for the upland fields, 0.0 for livestock and 1.0 for others, that is, the nitrogen load from fertilizers is divided into surface runoff load and cumulative load, and the nitrogen load from animal waste constitutes only cumulative load. The model simulates the runoff and the TN concentration with average relative errors of about 10% and 2%, respectively (Fig. 6). Although the time step was 10 min, the daily data for only noon are considered representative. The verification is shown in Table 3. The monthly relative errors for TN concentration at β d 1, 2 R N , I N , D 1 c 1, 2 q 1, 2 r 1, 2 S 1 Fig. 5 Symbols for the upper tank Runoff Load R, I E B 1, 2 S N1 PL α 1−α 1−β h 1, 2 h 1, 2 A Q 1 Tank Land use Elevation Drainage Drainage tank Fig. 4 Schematic diagram of the model Journal of Water and Environment Technology, Vol.3, No.2, 2005 - 239 - point A are over 30% for the period from January to May, 2002 because this area received little rainfall in winter and the calculated runoff values were lower than the measured values. As for nitrogen load, the relative errors become small—the average relative errors are about 10% for point B, 30% for point C and 20% for point D. In addition, Fig. 7 shows the verification of the TN concentration at point A every 10 min when rainfall occurred (15 and 16 July 2002); the relative error is about 15%. Note that the calculated TN values express the difference between the four points according to the measured values and that the model responds to the 10-min time step. 0.00E+00 5.00E-05 1.00E-04 1.50E-04 2.00E-04 8/10 9/10 10/10 11/10 12/10 Calc ula te d Observed 0 10 20 30 40 50 60 8/10 9/10 10/10 11/10 12/10 Calc ula te d Observed Fig. 6 Simulation of the runoff and the TN concentration TN concentration (mg · L -1 ) Runoff (10 -4 m · 10 min -1 ) 0 2.0 1.5 1.0 0.5 Oct Nov Dec Sep Aug 60 40 20 0 Calculated Observed Calculated Observed Oct Nov Dec Sep Aug Land use ih r cαβLand use ih r cαβ Upland 1 0.16 0.001 4 0.5 0.3 Residence 1 0.08 0.001 7 1 0.3 0 0.01 1 0 0.0005 1.8 0 0.03 0.9 0 0.01 0.8 2 0.001 0.0001 2 2 0.002 0.0001 5.5 0 0.0005 0.5 0.001 0.0007 5.5 Paddy 1 0.16 0.0005 3.5 1 0.6 (ir) Stock 1 0.16 0.001 4 0 0.3 0.005 0.008 0.9 0.3 0.009 0.013 1.2 0 0.005 2.8 (non-ir) 0 0.01 1 2 0.001 0.0001 2.8 2 0.001 0.0001 1.9 0 0.0002 0.4 0 0.0005 0.2 Forest 1 0.16 0.012 4.8 1 0.3 River 1 0.5 0.6 0 0 0.005 0.0005 5.6 0 0.1 0 0.01 6.2 2 0.002 0.0001 5.5 0.001 0.0005 5.5 Table 2. Values of the model parameters Residence Jan Feb Mar Apr May Jun Jul ATN Cal mg·L -1 52 50 63 37 20 24 22 Mea mg·L -1 32 32 - - 29 26 27 Error % 63 56 - - 31 8 19 N load Cal kg 227 222 210 222 228 217 221 Mea kg 259 255 - - 228 290 216 Error % 12 13 - - 0 25 2 BTN Cal mg·L -1 - - -7555 Mea mg·L -1 6556 Error % 17 0 0 17 CTN Cal mg·L -1 ---10655 Mea mg·L -1 12 10 9 9 Error % 17 40 44 44 DTN Cal mg·L -1 - - - 43373533 Mea mg·L -1 49 47 45 46 Error % 12 21 22 28 Table 3. Verification Journal of Water and Environment Technology, Vol.3, No.2, 2005 - 240 - SCENARIO ANALYSIS FOR NITROGEN LOAD REDUCED BY ARTIFICIAL WETLANDS Scenario analysis The TN concentration at point A is calculated for artificial wetlands at different locations. The three scenarios are as follows: no artificial wetland, cell located at the head of the valley and cell located on a flow path through an animal waste cell. The common calculation conditions are as follows: ・Prediction period: 10 August 2002 to 9 August 2003 ・Time step: 10 min ・Precipitation: Data for the period from 10 August 2001 to 9 August 2002 ・The parameters for the artificial wetlands are the same as those of the paddy fields, but the purification rate is 0.6 for the entire year. ・Artificial wetlands: The total number of cells in the paddy field cells is 60, of which three cells, i.e., 5%, are used. The symbols (i, ii, iii, iv and v) indicate the location of the cell in Fig. 8. ・Equations: Eq. (5) and (6) are modified to express the purification of the nitrogen load in a tank. ((1 − β) (R N + I N + αPL + D 1 + S N1 )/S 1, t + c 1, j B 1, j ) q 1, i = d 1, i (9) ((1 − β) (αd 1, 0 + D 2 + S N2 )/S 2, t + c 2, j B 1, j ) q 2, i = d 2, i (10) Results Case 1: No artificial wetland The calculated annual nitrogen load is 9.65 kg and the calculated TN concentration is 31.0 mg·L –1 (Table 4). The values are almost the same as the measured values obtained during the field investigation. Case 2: The cell is located at the head of the valley. The artificial wetlands are located at the head of the valley (i, ii and iii). As the head of the valley is inconvenient to access, paddy fields at this location tend to become fallow. Therefore, this location is a possible site for artificial wetlands. The calculated annual nitrogen load is 9.30 kg; the calculated TN concentration is 29.9 mg·L –1 , which is 4% lower than that of case 1. Case 3: The cell is located on a flow path through an animal waste cell. 0 10 20 30 40 50 21:00 3:00 9:00 15:00 0 2 4 6 8 10 Calculated Observed ． Fig. 7 Verification at point A every 10 min with rainfall Rainfall (mm · 10 min -1 )TN concentration (mg · L -1 ) 24:00 6:00 12:00 18:00 v iii iv ii i Fig. 8 Location of artificial wetlands Drainage cell Livestock farm cell Artificial wetland cell 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 Flow direction − i v Journal of Water and Environment Technology, Vol.3, No.2, 2005 - 241 - The number of cells for which the land use is livestock farms is three for ‘i’ and two each for ‘iv’ and ‘v’. The load from animal wastes is purified by the artificial wetlands. The calculated annual nitrogen load is 8.60 kg; the calculated TN concentration is 27.6 mg·L –1 , which is 11% lower than that of case 1. Table 4. Results of scenario analyses Case 1 Case 2 Case 3 N load 10 3 kg·y -1 9.65 9.30 8.60 TN concentration mg·L -1 31.0 29.9 27.6 Reduction rate % - 4 11 CONCLUSIONS A distributed water quality tank model was developed to simulate nitrogen load reduction by artificial wetlands in the A10 area, located in the Kasumigaura watershed in Japan. The model simulates the measured runoff and TN concentration with average relative errors of about 10% and 2%, respectively. The calculated TN concentration expresses the difference between the four points according to the measured values. Scenario analyses were performed, and the annual TN concentration was found to be lower by 4% and 11% in cases 2 and 3, respectively. The results suggest that, to effectively reduce nitrogen load, the artificial wetlands should be located near flowing water that contains a large nitrogen load. The model is capable of quantitatively evaluating the difference in nitrogen concentration across locations. In Japan, paddy fields, especially those at inconvenient locations, tend to be fallow due to the decrease in rice demand. However, with paddy fields being converted to artificial wetlands, their purification function is maintained. This is one of the methods of conserving the water environment in an agricultural watershed. Simulating the appropriate location of artificial wetlands can provide policymakers with the information necessary to set the target water quality and the size of the artificial wetlands. The following steps need to be taken in the future: - The model must be improved to achieve accurate simulations during no-rain periods. One approach is to consider the ground water level, which might affect the runoff. - Verification of the model response to rates such as the cumulative rate and purification rate is required. The values used in the developed model are obtained from literature; however, we would like to continuously monitor the study area to obtain values from the measured data. - The number of parameters should be reduced to shorten the calculation time. Acknowledgements We greatly appreciate Dr. Tabuchi Toshio and Dr. Wesley W. Wallender of University of California, Davis for giving us valuable opinions. Thanks go to Mr. Kifune Yasunori who helped develop the model. This research was supported by the Ministry of Education, Culture, Sports, Science and Technology‚’Grants-in-Aid for Scientific Research’ (subject number: A 14206027). References Abbott M. B. and Refsgaard J. C. (1996). Distributed Hydrological Modeling. Kluwer Academic Publishers CAES: Chiba Agricultural Experiment Station (1976). Vegetable Handbook. Chiba, Japan. (in Japanese) Goto A. and Nakamura R. (1993). Long-term Runoff Simulation by a Mesh-type Watershed Model. Bulletin of the College of Agriculture, Utsunomiya University, Vol.15, No.2, 1-10. Ibaraki Prefecture environment white paper H15 (2003). Ibaraki, Japan. (in Japanese) IPLEDKCS: Ibaraki Prefecture life environment department Kasumigaura countermeasure section (1999). The report to the lake water quality preservation plan in Kasumigaura basin (the 3rd term). Japan. (in Japanese). Journal of Water and Environment Technology, Vol.3, No.2, 2005 - 242 - Kato T., Kuroda H. and Nakasone H. (2003). Application of Water Quality Tank Model Classified by Land Use to Nitrogen Load Reduction Plans. Trans. of JSIDRE, No.224, 97-103. (abstract in English) Kuroda H. (1998). A Method of Water Quality Management on Topographical Chain. Journ. of JSIDRE, Vol.66, No.12, 19-23. (in Japanese) Kuroda S. and Tabuchi T. (1996). Change in Nitrate Nitrogen Concentration and Load of the Springwater -Studies on the Characteristics of Nitrate Outflow from the Vegetable Upland Fields (I)-. Trans. of JSIDRE, No.181, 31-38. (abstract in English) Lu M., Koike T. and Hayakawa N. (1989). A rainfall-runoff model using distributed data of radar rain and altitude. Proceedings of JSCE, 411/II-12, 135-140. (in Japanese) Nakasone H., Kuroda H., and Kubota K. (1996). Runoff Characteristics of Water Quality from Small Agricultural Watershed Having Topographical Chain and Irrigation. Journ. of JSWE, Vol.19, No.1, 56-62. (abstract in English) Nakasone H., Kuroda H. and Kato T. (2000). Development of Nitrogen Runoff Model from Watershed. Trans. of JSIDRE, No.205, 95-100. (abstract in English) NIES: National Institute for Environmental Studies, Japan (2002). Investigation Research on Water Environment Restoration System for Making Low Loading-resources Recycling Society. Report of Special Research from the National Institute for Environmental Studies Japan, SR-45-2002. (in Japanese) Nishio M. (2001). Analysis of the Actual State of Nitrogen Application in Arable Farming in Japan. Japan Journ. Soil Sci. Plant Nutr, No.72, 513-528. (in Japanese) Shimura M. and Tabuchi T. (1996). Seasonal Variation of Nitrogen Concentration and Load in Stream Waters in the Area where there are Pig Farms with Unlined Storage Pond -Research on nitrogen outflow from high stocking density area (II) -. Trans. of JSIDRE, No.182, 9-16. (in Japanese). Tabuchi T. (1986). Nitrogen outflow diagram in a small agricultural area. Trans. of JSIDRE, No.124, 53-60. (abstract in English) Tabuchi T. and Kuroda H. (1991). Nitrogen outflow diagram I n a small agricultural area having uplands and lowland. Trans. of JSIDRE, No.154, 65-72. (abstract in English) Vieux B. E. (2001). Distributed Hydrologic Modeling Using GIS. Kluwer Academic Publishers. . 35 37 35 35 35 35 35 30 29 30 31 34 35 38 36 36 36 36 36 31 32 32 32 34 35 37 37 37 37 37 33 33 33 33 34 35 38 38 38 38 36 36 36 36 36 36 37 37 37 37 37 . 2 23 32 23 2 231 32 231 1 32 2 231 1 32 233 314 222 231 133 1 112 232 2 233 3111 122 233 522 231 1 33 233 333 233 11 33 333 333 234 41 1 531 533 2 231 111 1111222221111 11 133 3 232 331 1 33 114 232 231 1