JpGU-AGU Joint Meeting 2017

講演情報

[EJ] ポスター発表

セッション記号 H (地球人間圏科学) » H-CG 地球人間圏科学複合領域・一般

[H-CG31] [EJ] 福島第一原子力発電事故からの地域復興に貢献できること

2017年5月23日(火) 13:45 〜 15:15 ポスター会場 (国際展示場 7ホール)

コンビーナ:西村 拓(東京大学大学院農学生命科学研究科生物・環境工学専攻)、溝口 勝(東京大学大学院農学生命科学研究科)、登尾 浩助(明治大学)

[HCG31-P02] Quantitative estimate of 137Cs load characteristics in Kuchibuto river watershed

*内藤 舜斗1大西 健夫1保高 徹生2中村 公人3宮津 進4 (1.岐阜大学、2.産業技術総合研究所、3.京都大学、4.農業・食品産業技術総合研究機構)

キーワード:福島第一原子力発電所事故、セシウム137負荷量、SWAT

137Cs has major impact on the environment due to its long half-life (30.1 years). To understand 137Cs load characteristics in the watershed, we estimated the effective 137Cs half-life in the Kuchibuto river watershed by calculating the 137Cs load in the river. Watershed area is 22km2, elevation is from 329m to 1,050m. Annual precipitation and average temperature is 1,158mm and 14°C respectively. The research watershed was covered with forest (74%), agricultural land (17.4%) and paddy field (7.6%). And, soil types are brown forest soil (51%) and Andosol (49%).
During the period from 6 July 2014 to 24 August 2015, we measured river discharge, SS, and 137Cs concentration. Utilizing measured data, we attempted to estimate the 137Cs load during non-observed period. First, Soil and Water Assessment Tool (SWAT) was utilized for river discharge estimate. Model warmup period was from 2008 to 2010. For the calibration, 2000 times simulation was conducted with Latin Hyper-cubic method from 1 October 2014 to 31 May 2015. Validation was also conducted from 6 July 2014 to 30 September and from 1 June 2015 to 24 August 2015. The model performance was assessed by Nash-Sutcliff efficiency (NSE) and regression coefficient (R2). Particulate 137Cs concentration was calculated by regression curves between discharge and suspended solid (SS), and SS and 137Cs concentration. Regression curves were constructed from observed discharge[m3 s-1], SS[g L-1] and 137Cs concentration[Bq L-1], and bias was compensated. Dissolved 137Cs was also calculated by using the partition coefficient which was ratio of particulate and dissolved 137Cs in the river. For uncertainty analysis, 95% confidential interval of 137Cs load was estimated by using the composition of Gaussian distribution of each regression curves from1 October 2014 to 31 May 2015. Discharge uncertainty was estimated by the sequential uncertainty fitting (SUFI).
During the observed period (6 July 2014 - 24 August 2015), particulate and dissolved 137Cs load was calculated at 6.1×108 and 1.5×106 Bq km-2 and these values were equal to 0.26% and 0.00065% of total 137Cs deposition on the watershed (5.13TBq, 28 December 2012 at present). Through the hydrological simulation by SWAT, total load of 137Cs from 2013 to 2015 were estimated. For about the model performance, NSE and R2 for calibration and validation were 0.75, 0.76 and 0.50, 0.54 respectively. Especially, in September 2015, large scale rainfall event(165mm day-1) was occurred and this event contributed to huge amount of 137Cs discharge in 2015. Annual total 137Cs load excluding this rainfall event in September 2015 was 2.41×108 - 2.86×108Bq yr-1 km-2 which was equal to about 0.1% of total 137Cs deposition. Otherwise, annual particulate and dissolved load including the large scale rainfall event in September 2015 were estimated at 2.41×108 - 6.8×1010Bq yr-1 km-2 and 8.7×105 - 5.76×107Bq yr-1 km-2 respectively and these values were equal to 0.1 - 29% of total 137Cs deposition in this watershed. However, it needs to be paid attention that the estimation of this huge scale rainfall might have large uncertainty because our observed period did not cover such large event. Lastly, effective 137Cs half-life with consideration of 137Cs load was calculated at 4.33 years according to our point estimation, and it appears that total amount of 137Cs in the watershed is decreasing to 0.82% of initial 137Cs amount within next 30 years. However, according to our uncertainty analysis, uncertainty range of 137Cs load was crossing over 2 - 3 orders, thus, effective 137Cs half-life is also probably highly uncertain. Thus, to obtain more accurate estimate, we need to improve the model performance during the extremely high flow events.