Spatiotemporal Analysis of Environmental Radiation in Korea

Transcription

1 WM 0 Conference, February 25 - March, 200, Tucson, AZ Spatiotemporal Analysis of Environmental Radiation in Korea J.Y. Kim, B.C. Lee FNC Technology Co., Ltd. Main Bldg. 56, Seoul National University Research Park, San4-2, Bongcheon-Dong, Gwanak-gu, Seoul, 5- Korea H.K. Shin Korea Institute of Nuclear Safety, P.O. Box 4, Yuseong-gu, Daejeon, Korea ABSTRACT Geostatistical visualization of environmental radiation is a very powerful approach to explore and understand spatiotemporal variabilities of environmental radiation data. Spatial patterns of environmental radiation can be described quantitatively in terms of variogram and kriging, which are based on the idea that statistical variation of data are functions of distance. INTRODUCTION An Integrated Environmental Radiation monitoring Network (IERNet, has been maintained to monitor the environmental radiation level and radioactivity concentration in South Korea since 6 [,2]. The roles of IERNet are to monitor environmental radiation nationwide during normal times and to serve as an early warning system for possible radiation risk to the public in case of nuclear terrorism or severe accidents at nuclear power plants (NPPs). Environmental radiation monitoring facilities consist of Central Radiation Monitoring Station (CRMS), 2 Regional Radiation Monitoring Stations (RRMSs), and 26 Unmanned Monitoring Posts (UMPs) as shown in Fig.. Geostatistics is a body of methods useful for understanding and modeling the spatial variability inherent in a process of interest. Although it has its origins in mining, geostatistics is now a basic part of many scientific disciplines including soil science, hydrology, agriculture, and environmental engineering. Central to the theory of geostatistics is the idea that measurements taken at locations close together tend to be more alike than values observed at locations farther apart. Geostatistics provides methods for quantifying this spatial correlation and for incorporating it in statistical estimation and inference. A fundamental concept of geostatistics is the use of quantitative measures of spatial correlation, commonly expressed by variogram. The interpolation technique, known as kriging, is a method for making the best linear unbiased estimation (BLUE) of regionalized variables at unsampled locations using the structural properties of the variogram and the initial set of measured data. Here, the emphasis is set on local accuracy, i.e., closeness of estimate to the actual, but unknown value, without any regard for the global statistical properties of the estimates.

2 WM 0 Conference, February 25 - March, 200, Tucson, AZ Chuncheon Seoul Gangneung Longitude Gunsan Yeonggwang Suwon Cheongju Daejeon Gwangju Andong Daegu Busan Uljin Wolsung Kori Jeju Latitude Fig.. Environmental radiation monitoring locations in South Korea (circles are RRMSs, crosses are UMPs, and squares are NPPs). METHODS A data set which seems random to the eye may have a property of strong spatial correlation. Geostatistical approach utilizes this fact that variations of data are not always random, but have some spatial structures. It is based on the regionalized variable (RV) theory which takes into account the random and structured characteristics of spatially distributed variables. Consider a field of A, for which a set of n values are measured [Z(x i ), i=,,n], in which each x i identifies a coordinate position in the space. Each Z(x k ) can be considered a particular realization of a certain random variable for a particular fixed point, x k. The regionalized variable Z(x i ), for all x i inside A, can be considered a realization of the set of random variables [Z(x i ), for all x i inside A]. This set of random variables is called a random function. Application of RV theory assumes that the covariance between any two locations in A depends only on the distance and direction of separation between the two locations and not on their geographic location (second-order stationarity). The assumption of second-order stationarity implies that the mean and variance are functions of separation distance, h, only. However, most of the physical data do not satisfy this condition, because the variance increases with the size of the domain. Therefore another less stringent hypothesis, known as the intrinsic hypothesis, is used. This intrinsic hypothesis does not require the variance of Z to be finite, but the variance of the first-order increment, [Z(x+-Z(x)], to be finite and second-order stationary; [ Z( x x) ] m( E + = (Eq. ) Var [ Z( x x) ] = 2γ ( + (Eq. 2) where m ( is the mean and γ ( is the variance of the increment known as a semivariogram. The application of variogram modeling allows a quantification of the spatial variability. The experimental semivariance γ as a measure for the spatial correlation structure at a given distance h is: N ( 2 γ e( = [ Z( xi + xi )] (Eq. 3) 2N( i=

3 WM 0 Conference, February 25 - March, 200, Tucson, AZ where h (called lag) is the distance between two sampling positions, Z ( x i ) is the measured value at the position x i and N ( is the number of pairs with the distance h. The spatial distributions of measured values are visualized in contour maps using ordinary kriging. Kriging is a method for making optimal, unbiased estimates of RVs at unsampled locations using the structural properties of the semivariogram and the initial set of measured data. A useful advantage of kriging compared with the other traditional linear interpolators is that an error term, expressing the estimation variance or uncertainty in estimation, is calculated for each interpolated value. Moreover, kriging has the property of exactitude, i.e., it returns the datum value for the estimate if the location to be estimated coincides with a sampled location. Kriging is also known as the best linear unbiased estimator. It is linear because the kriging estimate at an arbitrary point X based on N neighboring sampled values is given as: Z * ( x0) λ iz( x i ) (Eq. 4) = N i= where Z are measured value at the observed location x i and Z * are estimated value at the interpolated location x 0. The theory of geostatistics has been described and well documented in a wide number of texts [3,4,5,6,]. Geostatistical routines from the software package GSLIB [3] are used to analyze spatial variability of external gamma-dose rate and annual collective dose rate, which are the important environmental radiation metrics. GSLIB routines are also used to analyze spatial variability of environmental radioactivity of Cs-3, which is an artificial gamma radioisotope. RESULTS The plots in Fig. 2 and Fig. 3 display contour maps for the spatial variabilities of external gamma-dose rates and annual collective dose rates, respectively. Regional distribution of external gamma-dose rate shows that higher values are located in the midsection of the country than in the southern part because of different component parts of lithosphere; Granite and gneiss are the dominant rocks in the midsection, while sedimentary rock is main component in the south. No peculiar increase of radiation level around NPPs is observed and all the monitored dose rates are below the limit of annual natural radiation exposure rate of 2.4 ( 3 ). Fig. 4 displays contour maps for the spatial variability of concentration of Cs-3 around Kori units. Regional distribution of Cs-3 shows that any peculiar increase of artificial radioactivity level around nuclear site is not observed during 2003~2004. Cs-3 generally shows a nationwide distribution in Korea because it is known to be a leftover due to fallout of an atmospheric nuclear test of the powers in past 50~60 s. CONCLUSION A visualization tool for the spatiotemporal analysis and prediction of environmental radiation is developed using geostatistics and Matlab. It can be used as the web-based real-time information system along with IERNet of Korea Institute of Nuclear Safety (KINS). The developed technique can be also applied to the radiation environmental impact assessment and the formulation of risk map of the public.

4 WM 0 Conference, February 25 - March, 200, Tucson, AZ (a) 5 (b) (c) (d) Fig. 2. Kriging maps for annual means of external gamma-dose rates in (a), (b) 2000, (c) 2002, and (d) 2003; crosses are monitoring stations and squares are NPP sites. (a).5. (b) (c) (d) Fig. 3. Kriging maps for annual collective dose rates in (a), (b) 2000, (c) 2002, and (d) 2003; crosses are monitoring stations and squares are NPP sites.

5 WM 0 Conference, February 25 - March, 200, Tucson, AZ Fig. 4. Geostatistical prediction maps of Cs-3 in soil around Kori nuclear power plant; a, b, c, and d above correspond to the date of spring and fall of 2003, spring and fall of 2004, respectively. REFERENCES. Korea Institute of Nuclear Safety, Environmental Radioactivity Survey Data, KINS/ER-2, Vol. 35, Korea (2003). 2. Korea Institute of Nuclear Safety, The Annual Report on the Environmental Radiological Surveillance and Assessment around the Nuclear Facilities, KINS/AR-40, Vol.4, Korea (2003). 3. Deutsch, C.V. and A.G. Journel, GSLIB: Geostatistical Software Library and User s Guide, 2 nd ed., Oxford University Press, New York (). 4. Journel, A.G. and Ch.J. Huijbregts, Mining Geostatistics, Academic Press, New York (). 5. Isaaks, E.H. and R.M. Srivastava, An Introduction to Applied Geostatistics, Oxford University Press, New York (). 6. Goovaerts, P, Geostatistics for Natural Resources Evaluation, New York: Oxford University Press ().. Webster, R., and M. Oliver, Geostatistics for Environmental Scientists, Chichester, England: John Wiley & Sons (200).