- Email: [email protected]
Derivation of 137Cs deposition density from measurements of 137Cs inventories in undisturbed soils
Journal of Environmental Radioactivity 62 (2002) 295–303 www.elsevier.com/locate/jenvrad
Derivation of 137Cs deposition density from measurements of 137Cs inventories in undisturbed soils P.D. Hien ∗, H.T. Hiep, N.H. Quang, N.Q. Huy, N.T. Binh, P.S. Hai, N.Q. Long, V.T. Bac Vietnam Atomic Energy Agency, 59 Ly Thuong Kiet, Hanoi, Vietnam Received 5 October 2001; received in revised form 18 December 2001; accepted 20 December 2001
Abstract The 137Cs inventories in undisturbed soils were measured for 292 locations across the territory of Vietnam. The logarithmic inventory values were regressed against characteristics of sampling sites, such as geographical coordinates, annual rainfall and physico-chemical parameters of soil. The regression model containing latitude and annual rainfall as determinants could explain 76% of the variations in logarithmic inventory values across the territory. The model part was interpreted as the logarithmic 137Cs deposition density. At the 95% confidence level, 137Cs deposition density could be predicted by the model within ± 7% relative uncertainty. The latitude mean 137Cs deposition density increases northward from 237 Bq m–2 to 1097 Bq m–2, while the corresponding values derived from the UNSCEAR (1969) global pattern are 300 Bq m–2 and 600 Bq m–2. High 137Cs inputs were found in high-rainfall areas in northern and central parts of the territory. 2002 Published by Elsevier Science Ltd. Keywords: 137Cs nuclear tests fallout; Gamma spectrometry; Multiple regression relationship; Latitude; Annual rainfall; 137Cs redistribution
∗
Corresponding author. Fax:+84-4-8266133. E-mail address: [email protected] (P.D. Hien).
0265-931X/02/$ - see front matter 2002 Published by Elsevier Science Ltd. PII: S 0 2 6 5 - 9 3 1 X ( 0 2 ) 0 0 0 1 2 - 7
296
P.D. Hien et al. / J. Environ. Radioactivity 62 (2002) 295–303
1. Introduction Fallout 137Cs from past atmospheric nuclear tests has been utilized as a marker for the study of soil erosion and sediment deposition processes (Ritchie & McHenry, 1975; Loughran et al., 1987; Walling & Quine, 1991). So far, most studies have been carried out in the mid-latitude regions (see bibliography of publications in Ritchie & Ritchie (1989)), where maxima of fallout deposition were observed in both hemispheres (UNSCEAR, 1969). In lower latitude environments, few studies were conducted (Forsyth, 1994; Garcia-Oliva et al., 1995; Owens et al., 1996; Hai et al., 1999). Among the reasons may be the low 137Cs inputs, especially in areas near the equator, where the global fallout deposition exhibits a minimum (UNSCEAR, 1969). A basic indication needed in designing erosion and sedimentation studies is the 137 Cs deposition density, i.e., the decay-corrected cumulative activity of deposited 137 Cs per unit ground surface area (Bq m–2). As 137Cs fallout deposition was not as widely documented in the 1950s and 1960s as 90Sr, the 137Cs deposition density can be estimated from 90Sr data (UNSCEAR, 1969) using a 137Cs/90Sr fission yields ratio of 1.6 (UNSCEAR, 1993). The global deposition patterns of these long-lived debris exhibit maxima in temperate regions and minima in the equatorial and polar regions reflecting the preferential exchange of air between the stratosphere and troposphere in the mid-latitudes of the hemisphere and the large-scale air circulation patterns in the troposphere (UNSCEAR, 2000). According to the global fallout deposition pattern (UNSCEAR, 1969), the present level of 137Cs deposition density in Vietnam — a 320,000 km2 territory extending from 9°N to 23°N along the West coast of the Pacific (Fig. 1) — is expected to increase northward from about 300 Bq m–2 to 600 Bq m–2. The contribution from the Chernobyl accident was insignificant based on the results of fallout radioactivity measurements carried out during 1986–1990 (Hien et al., 1994). A mean 137Cs deposition density of (330 ± 67) Bq m–2 was found for some drainage basins in the Central Plateau of South Vietnam (~11.5°N, 107.5°E) by Hai et al., (1999), thus likely confirming the above lower limit (300 Bq m–2), which is about 7–15 times as low as in North America and Europe (e.g., Ritchie & McHenry, 1975; Walling & Quine, 1991). Meanwhile, the results of radioactivity surveys of surface soils carried out in North Vietnam (Vietnam Atomic Energy Commission, 1998) showed a number of areas having inventories much higher than the above upper limit (600 Bq m–2). In view of the above situation, the 137Cs deposition density distribution over the territory should be established. For this purpose, soil samples were taken from 292 undisturbed sites and the 137Cs inventories (in Bq m–2) at these sites were measured using low-background gamma spectrometers. This approach is similar to that used for establishing the 137Cs reference (or baseline) input in soil erosion studies. The multiple regression of measured inventory values upon characteristics of sampling locations such as latitude, annual rainfall and physico-chemical parameters of soil enabled the construction of models describing the distributions of 137Cs deposition density across the territory.
P.D. Hien et al. / J. Environ. Radioactivity 62 (2002) 295–303
Fig. 1.
Sampling map and spatial distribution of
137
297
Cs deposition density
2. Soil sampling and analysis Soil samples were taken at flat, spacious and grass-covered terrains, which were believed to be mostly undisturbed for the last five decades. Virgin plots in preserved forests, meteorological gardens, historical places and air-landing fields are among most suitable sampling sites. A total of 292 sites were chosen throughout ten geographical–climatic regions of Vietnam. Due to logistic constraints, some mountainous
298
P.D. Hien et al. / J. Environ. Radioactivity 62 (2002) 295–303
and remote areas were not sampled. On a (0.5° x 0.5°) grid, only 100 out of 140 grid cells were sampled (Fig. 1), and the average number of samples per grid cell was 2.9 ± 1.6. About 80% of soil samples collected belong to acrisols, fluvisols and ferralsols; the three major soil groups accounting for 85% of Vietnam’s territory (Vietnam Soil Science Association, 1996). Geographical coordinates of sampling locations were determined using a Magellan GPS tracker. Data on annual rainfall were taken from the long-term records at nearby meteorological stations. If this was not the case, the annual rainfall value was derived from the map constructed by the Central Meteorological Service using long-term records available since the 1950s for 167 meteorological stations from all over the country. Soil samples were collected to a depth of 30 cm using corers of 4 cm inner diameter. Three cores were taken from each sampling terrain, the materials were stored in plastic bags for further sample preparation and analysis. In the laboratory, the airdried soil was gently ground to pass through a 2-mm sieve for the removal of stones and roots, and carefully mixed. For gamma spectrometry analysis, a portion of 500– 600 g was taken, oven-dried at 105 °C for 24 h and the bulk density of dry soil was determined by weighing. Activity concentrations of radionuclides (in Bq kg–1) were measured using lowbackground gamma spectrometers with active volumes of the HP-Ge detectors from 90 to 140 cm3 and peak resolutions (FWHM) of around 1.5 keV at 662 keV. 137Cs inventory was calculated by multiplying activity concentration by soil density (kg m–2). Besides 137Cs, activity concentrations of naturally occurring isotopes (40K, U and Th series) also were determined; the results will be reported elsewhere. By using Marinelli beakers containing about 400 g of dry soil, the minimum detectable activity of 137Cs, defined as three background areas at the 662 keV photopeak, was around 0.3 Bq kg–1 for a 24-hr counting time. The precision of gamma spectrometry measurements, calculated as one standard error of the net area of the 137Cs 662 keV photopeak, was from 7% to 15% . Physico-chemical properties of soil including pH, organic matter, cation-exchange capacity and granulometric compositions were analysed according to the FAO guidelines (FAO, 1995).
3. Results and discussions 3.1. Regression analysis The summary statistics of the measured data and characteristics of sampling locations are given in Table 1. The measured 137Cs inventories increase with latitude resulting in a mean value of about two times higher in the northern (⬎ 16°N) than in the southern (⬍ 16°N) part of the territory (Table 1). Within any latitude band of several degrees, 137Cs inventory tends to increase with annual rainfall. The multiple regression of measured inventories upon characteristics of sampling locations was applied to establish their relationships. In regression analysis, independent variables were latitude, longitude, annual rainfall and four soil parameters; namely pH, clay
P.D. Hien et al. / J. Environ. Radioactivity 62 (2002) 295–303
299
Table 1 Summary statistics of measured data (total number of samples: 292)
Latitude, °N Longitude, °E 137 Cs activity concentration, Bq kg–1 137 Cs inventory, Bq m–2 South Vietnam, ⬍16 °N North Vietnam, ⬎16 °N Bulk density, kg m–3 Mean annual rainfall, m Content of soil organic matter, % Clay, % Total cation exchange capacity,cmol kg–1 pH
Mean
Stand. dev.
Median
Range
16.8 106.5 2.03 643 441 808 1.22 1.97 1.97 19.1 7.5 4.90
4.34 1.5 2.20 384 315 359 0.30 0.56 1.11 12.6 8.2 1.10
17.1 106.3 1.58 590 361 750 1.23 1.88 1.81 15.9 4.5 4.39
9.2–23 103–110 0.5–18.0 129–3294 129–1603 314–3294 0.43–2.1 0.79–3.8 0.30–8.5 0.40–60.7 0.70–54.8 3.3–7.7
component, content of organic matter and total cation exchange capacity. The dependent variable was inventory values, which were logarithmically transformed to enable the construction of regression models with higher goodness-of-fits (R2) and constant error variances (homoscedasticity). The physical interpretation of the logarithm transformation of the dependent variable will be discussed later. The stepwise multiple regression method was applied to select among eight characteristics of sampling locations those which are most significant in explaining the variations of 137Cs inventories across the territory. The stepwise regression analysis was performed using SPSS version 7.5. If the significance level of the regression coefficients was set at 0.01, only latitude L and annual rainfall AR were selected, i.e., they appeared as predictors of the 137Cs inventories: Ln(I) ⫺ e ⫽ (3.53 ± 0.09) ⫹ (0.092 ± 0.004)L
(1)
⫹ (0.62 ± 0.03)AR The intercept and regression coefficients are given in Equation 1 with their standard errors. The regression model is free from multi-colinearity because AR and L do not correlate with each other (r = 0.001). The two predictors L and AR could explain 76% (R2 = 0.76) of the total variance of Ln(I), leaving the 24% remaining variance to the residual e. The latter follows a normal distribution with zero mean (e¯ = 0) and a standard deviation of 0.30, which is apparently greater than that associated with measurement errors. The contribution from measurement errors to e can be estimated as Ln(1 + α) ~ 0.1, where α is the typical relative standard error of the inventory measurements (α ~ 10%). Physico-chemical parameters of soil did not appear in the regression model (Equation 1). Soil parameters, however, would become predictors in the regression models for small regions over which the variations in the latitude and annual rainfall are insignificant.
300
P.D. Hien et al. / J. Environ. Radioactivity 62 (2002) 295–303
3.2. Interpretation of regression model (1) If 137Cs has remained and been attached to soil particles at the point where it was deposited, the inventory measurement would yield the deposition density D. Within the territory of Vietnam, the nuclear tests fallout deposition density varies insignificantly with longitude (UNSCEAR, 1969), so that only two factors control the spatial variability of fallout deposition rate, i.e., latitude and annual rainfall. The product relationship of deposition rate with these two factors results in a linear relationship of logarithmic deposition density with latitude and annual rainfall. We would have in this case a relationship similar to Equation 1, LnI⫺eexp ⫽ LnD
(2)
where D is a function of L and AR, and eexp ~ Ln(1 + α) ~ 0.1. In fact, |e| in Equation 1 is apparently greater than eexp. Therefore, besides measurement errors, there are other sources contributing to the residual term in Equation 1. One potential source is associated with redistribution processes taking place at the sampling area with dissolved 137Cs just after its deposition by rain via overland flow (VandenBygaart et al., 1999), as well as with 137Cs adsorbed in soil particles via erosion transport (see Pennock, 1998 and references therein). As a result of redistribution processes the inventory may be either smaller or greater than the deposition density at the sampling site. It is natural to assume that either case were equally likely to occur in our sampling campaign so that the mean logarithmic inventory for sampling locations having similar L and AR tends to converge toward the logarithm of deposition density. A second source giving rise to the residual in Equation 1 is associated with errors of the AR values. In soil erosion studies, variability of rainfall on the study area is thought to be among potential sources of spatial variability of 137Cs reference input (VandenBygaart et al., 1999). In our case, an error in AR will lead to a deviation of the predicted deposition density from the truly expected value. An error of 0.16 m in AR, for example, would lead to a 10% relative deviation of the deposition density predicted by Equation 1. Thus the regression model can be rewritten as follows: Ln(I) ⫺ e ⫽ Ln(D)
(3)
where D is the deposition density and the residual ε has three components associated with redistribution processes, inventory measurement uncertainty and errors in AR values. It is worth noting that the three error components are independent and random with zero means. In spite of a rather large dispersion of experimental values Ln(I) around Ln(D), i.e.. ε ~ ± 0.30, the logarithmic deposition density could be predicted with a much better certainty due to a large number of samples involved in regression analysis. At the 95% confidence level, the prediction interval for Ln(D) varies from ± 0.04 (for areas having mean latitude and annual rainfall) to ± 0.14 (for areas having very high annual rainfall in the northern part of the territory), with the mean being ± 0.067, which corresponds to a 7% relative uncertainty of the 137Cs deposition density.
P.D. Hien et al. / J. Environ. Radioactivity 62 (2002) 295–303
4. Distribution of
301
137
Cs deposition density
The deviations of the measured inventories from the deposition density due to random errors mentioned above would be reduced if experimental data were averaged over an area containing a sufficient number of sampling points. The regression of the mean of Ln(I) upon mean characteristics of the sampling locations would then yield the same relationship as Equation 1 with smaller residuals. If, for example, experimental data were averaged over (0.5° x 0.5°) grid cell (Fig. 1), the regression model for the grid cell mean inventories would be: [Ln(I)]⫺e’ ⫽ (3.51 ± 0.11) ⫹ (0.093 ± 0.005)[L] ⫹ (0.61 ± 0.04)[AR]
(4)
where [Ln(I)], [L] and [AR] are the grid cell averages of Ln(I), L and AR, respectively. The intercept and regression coefficients in Equation 3 remain almost similar to Equation 1, but the new residuals e⬘ have a standard deviation of 0.17, much less than that of the original model. The regression model (Equation 4) could explain 88% of the variance in [Ln(I)] across 100 grid cells. In this case the right-hand side of (Equation 4) would represent the average deposition density in each grid cell [Ln(D)]. The spatial distribution of [Ln(D)] is mapped in Fig. 1, where both the northward-increasing trend of the deposition density and the effect of annual rainfall variability are evident.
5. Latitudinal distribution of
137
Cs deposition density
The latitudinal distribution of deposition density could be obtained by regressing the average of Ln(I) against the mean characteristics of sampling locations with the same latitude. If data were averaged over half-degree latitude bands, the regression model would be: [Ln(I)]L⫺e’’ ⫽ (3.64 ± 0.11) ⫹ (0.094 ± 0.004)[L]L
(5)
⫹ (0.54 ± 0.04)[AR]L where the symbol [...]L denotes the averages. Again, the intercept and regression coefficients remain almost similar to Equations 1 and 4, but the new regression model could explain up to 96% of the variance in [Ln(I)]L, leaving only 4% to the residuals e⬘⬘. The right-hand side of Equation 5 is [Ln(D)]L — the mean logarithmic deposition density at latitude L. In ordinary scale, Equation 5 would provide a relationship between geometric means of inventory and deposition density; both are plotted in Fig. 2. The latitude mean 137Cs deposition density increases northward from 237 to 1097 Bq m–2, while the corresponding values expected from the UNSCEAR (1969) global pattern are 300 and 600 Bq m–2. Deposition density less than 300 Bq m–2 was found in southernmost areas with low annual rainfalls (less than 1.2 m), while high-rainfall (over 3 m) areas in northern and central parts of the territory have received 137Cs inputs considerably exceeding 600 Bq m–2.
302
P.D. Hien et al. / J. Environ. Radioactivity 62 (2002) 295–303
Fig. 2.
Latitudinal distribution of
137
Cs inventory and deposition density
6. Conclusion The multiple regression of logarithmic inventory values against characteristics of sampling sites enabled the establishment of a model describing the relationship of 137 Cs deposition density with latitude and annual rainfall within the territory of Vietnam. This regression model could explain 76% of the variations in logarithmic inventory values at 292 locations across the territory. The remaining variance contained in the regression residual could be attributed to 137Cs redistribution processes taking place at the sampling area and random errors of inventory measurements and annual rainfall values. Due to a large number of samples investigated, the 137Cs deposition density at any location can be predicted from the regression model with high certainty. At the 95% confidence level, 137Cs deposition density could be predicted by the model within ± 7% relative uncertainty. Within the territory of Vietnam, the latitude mean 137Cs deposition density increases northward from 237 to 1097 Bq m–2. This range is larger than that expected from the UNSCEAR (1969) global pattern (from 300 to 600 Bq m–2). The results of this work are expected to contribute to refining the global distribution pattern of nuclear test fallout deposition in tropical areas, especially in Southeast Asia, for which fallout data are scarce.
Acknowledgements This study was sponsored by the Ministry of Planning, and Investment and was performed under supporting administrative arrangements of the Ministry of Science Technology and Environment. A great number of colleagues of Vietnam Atomic Energy Agency participating in sample collection and preparation are gratefully acknowledged.
P.D. Hien et al. / J. Environ. Radioactivity 62 (2002) 295–303
303
References Food and Agriculture Organization of the United Nations (FAO) (1995). Procedures for soil analysis (3rd ed.). Netherlands: Wageningen. Forsyth, T. J. (1994). The use of 137Cs measurements of soil erosion and farmer’s perceptions to indicate land degradation amongst shifting cultivators in northern Thailand. Mountain Research and Development, 14, 229–244. Garcia-Oliva, F., Lugo, R. M., & Maass, J. M. (1995). Long-term net soil erosion as determined by 137Cs redistribution in an undisturbed and perturbed tropical deciduous forest ecosystem. Geoderma, 68, 135–147. Hai, P. S., Son, P. N., Dien, N. N., Tan, V. H., & Hien, P. D. (1999). Assessment of erosion and accretion in catchment areas based on 210Pb and 137Cs contents in soil and sediment. In Isotope production and applications in the 21st century. In Proceedings of the 3rd International Conference on Isotopes. Vancouver: World Scientific (pp. 415–418). Hien, P. S., Binh, P. N., Truong, N. N., Bac, V. T., & Ngo, N. T. (1994). Variations of caesium isotope concentrations in air and fallout at Dalat, South Vietnam, 1986-91. Journal of Environmental Radioactivity, 22, 55–62. Loughran, R. J., Campbell, B. L., & Walling, D. E. (1987). Soil erosion and sedimentation indicated by 137 Cs: Jackmoor Brook catchment, Devon, England. Catena, 14, 201–212. Owens, P. N., Walling, D. E., & He, Q. (1996). The behavior of bomb-derived 137Cs fallout in catchment soils. Journal of Environmental Radioactivity, 32, 169–191. Pennock, D. J. (1998). New perspectives on the soil erosion–soil quality relationship. In Use of 137Cs in the study of soil erosion and sedimentation. IAEA-TECDOC-1028, Vienna. Ritchie, J. C., & McHenry, J. R. (1975). Fallout 137Cs: a tool in conservation research. Journal of Soil and Water Conservation, 30, 283–286. Ritchie, J. C., & Ritchie, C. A. (1989). Bibliography of publications of 137Cs studies related to soil erosion and sediment deposition. IAEA-TECDOC-1028, Vienna: IAEA. United Nation Scientific Committee on the Effects of Atomic Radiation (UNSCEAR). (1969). 24th Session, Suppl. No 13 (A/7613) New York: United Nations. United Nation Scientific Committee on the Effects of Atomic Radiation (UNSCEAR). (1993). 48th Session, Suppl. No 46 (A/48/46) New York: United Nations. United Nation Scientific Committee on the Effects of Atomic Radiation (UNSCEAR). (2000). 55th Session, Suppl. No 46 (A/55/46) New York: United Nations. VandenBygaart, A. J., King, D. J., Groenevelt, P. H., & Protz, R. (1999). Cautionary notes on the assumptions made in erosion studies using fallout 137Cs as a marker. Canadian Journal of Soil Science, 79, 395–397. Vietnam Atomic Energy Commission. (1998). Internal Reports on the implementation of R&D programmes (in Vietnamese). Vietnam Soil Science Association (1996). Soil of Vietnam. Hanoi: Agriculture Publishing House. Walling, D. E., & Quine, T. A. (1991). The use of 137Cs measurements to investigate soil erosion on arable fields in the UK: potential applications and limitations. Journal of Soil Science, 42, 147–165.
Derivation of 137Cs deposition density from measurements of 137Cs inventories in undisturbed soils P.D. Hien ∗, H.T. Hiep, N.H. Quang, N.Q. Huy, N.T. Binh, P.S. Hai, N.Q. Long, V.T. Bac Vietnam Atomic Energy Agency, 59 Ly Thuong Kiet, Hanoi, Vietnam Received 5 October 2001; received in revised form 18 December 2001; accepted 20 December 2001
Abstract The 137Cs inventories in undisturbed soils were measured for 292 locations across the territory of Vietnam. The logarithmic inventory values were regressed against characteristics of sampling sites, such as geographical coordinates, annual rainfall and physico-chemical parameters of soil. The regression model containing latitude and annual rainfall as determinants could explain 76% of the variations in logarithmic inventory values across the territory. The model part was interpreted as the logarithmic 137Cs deposition density. At the 95% confidence level, 137Cs deposition density could be predicted by the model within ± 7% relative uncertainty. The latitude mean 137Cs deposition density increases northward from 237 Bq m–2 to 1097 Bq m–2, while the corresponding values derived from the UNSCEAR (1969) global pattern are 300 Bq m–2 and 600 Bq m–2. High 137Cs inputs were found in high-rainfall areas in northern and central parts of the territory. 2002 Published by Elsevier Science Ltd. Keywords: 137Cs nuclear tests fallout; Gamma spectrometry; Multiple regression relationship; Latitude; Annual rainfall; 137Cs redistribution
∗
Corresponding author. Fax:+84-4-8266133. E-mail address: [email protected] (P.D. Hien).
0265-931X/02/$ - see front matter 2002 Published by Elsevier Science Ltd. PII: S 0 2 6 5 - 9 3 1 X ( 0 2 ) 0 0 0 1 2 - 7
296
P.D. Hien et al. / J. Environ. Radioactivity 62 (2002) 295–303
1. Introduction Fallout 137Cs from past atmospheric nuclear tests has been utilized as a marker for the study of soil erosion and sediment deposition processes (Ritchie & McHenry, 1975; Loughran et al., 1987; Walling & Quine, 1991). So far, most studies have been carried out in the mid-latitude regions (see bibliography of publications in Ritchie & Ritchie (1989)), where maxima of fallout deposition were observed in both hemispheres (UNSCEAR, 1969). In lower latitude environments, few studies were conducted (Forsyth, 1994; Garcia-Oliva et al., 1995; Owens et al., 1996; Hai et al., 1999). Among the reasons may be the low 137Cs inputs, especially in areas near the equator, where the global fallout deposition exhibits a minimum (UNSCEAR, 1969). A basic indication needed in designing erosion and sedimentation studies is the 137 Cs deposition density, i.e., the decay-corrected cumulative activity of deposited 137 Cs per unit ground surface area (Bq m–2). As 137Cs fallout deposition was not as widely documented in the 1950s and 1960s as 90Sr, the 137Cs deposition density can be estimated from 90Sr data (UNSCEAR, 1969) using a 137Cs/90Sr fission yields ratio of 1.6 (UNSCEAR, 1993). The global deposition patterns of these long-lived debris exhibit maxima in temperate regions and minima in the equatorial and polar regions reflecting the preferential exchange of air between the stratosphere and troposphere in the mid-latitudes of the hemisphere and the large-scale air circulation patterns in the troposphere (UNSCEAR, 2000). According to the global fallout deposition pattern (UNSCEAR, 1969), the present level of 137Cs deposition density in Vietnam — a 320,000 km2 territory extending from 9°N to 23°N along the West coast of the Pacific (Fig. 1) — is expected to increase northward from about 300 Bq m–2 to 600 Bq m–2. The contribution from the Chernobyl accident was insignificant based on the results of fallout radioactivity measurements carried out during 1986–1990 (Hien et al., 1994). A mean 137Cs deposition density of (330 ± 67) Bq m–2 was found for some drainage basins in the Central Plateau of South Vietnam (~11.5°N, 107.5°E) by Hai et al., (1999), thus likely confirming the above lower limit (300 Bq m–2), which is about 7–15 times as low as in North America and Europe (e.g., Ritchie & McHenry, 1975; Walling & Quine, 1991). Meanwhile, the results of radioactivity surveys of surface soils carried out in North Vietnam (Vietnam Atomic Energy Commission, 1998) showed a number of areas having inventories much higher than the above upper limit (600 Bq m–2). In view of the above situation, the 137Cs deposition density distribution over the territory should be established. For this purpose, soil samples were taken from 292 undisturbed sites and the 137Cs inventories (in Bq m–2) at these sites were measured using low-background gamma spectrometers. This approach is similar to that used for establishing the 137Cs reference (or baseline) input in soil erosion studies. The multiple regression of measured inventory values upon characteristics of sampling locations such as latitude, annual rainfall and physico-chemical parameters of soil enabled the construction of models describing the distributions of 137Cs deposition density across the territory.
P.D. Hien et al. / J. Environ. Radioactivity 62 (2002) 295–303
Fig. 1.
Sampling map and spatial distribution of
137
297
Cs deposition density
2. Soil sampling and analysis Soil samples were taken at flat, spacious and grass-covered terrains, which were believed to be mostly undisturbed for the last five decades. Virgin plots in preserved forests, meteorological gardens, historical places and air-landing fields are among most suitable sampling sites. A total of 292 sites were chosen throughout ten geographical–climatic regions of Vietnam. Due to logistic constraints, some mountainous
298
P.D. Hien et al. / J. Environ. Radioactivity 62 (2002) 295–303
and remote areas were not sampled. On a (0.5° x 0.5°) grid, only 100 out of 140 grid cells were sampled (Fig. 1), and the average number of samples per grid cell was 2.9 ± 1.6. About 80% of soil samples collected belong to acrisols, fluvisols and ferralsols; the three major soil groups accounting for 85% of Vietnam’s territory (Vietnam Soil Science Association, 1996). Geographical coordinates of sampling locations were determined using a Magellan GPS tracker. Data on annual rainfall were taken from the long-term records at nearby meteorological stations. If this was not the case, the annual rainfall value was derived from the map constructed by the Central Meteorological Service using long-term records available since the 1950s for 167 meteorological stations from all over the country. Soil samples were collected to a depth of 30 cm using corers of 4 cm inner diameter. Three cores were taken from each sampling terrain, the materials were stored in plastic bags for further sample preparation and analysis. In the laboratory, the airdried soil was gently ground to pass through a 2-mm sieve for the removal of stones and roots, and carefully mixed. For gamma spectrometry analysis, a portion of 500– 600 g was taken, oven-dried at 105 °C for 24 h and the bulk density of dry soil was determined by weighing. Activity concentrations of radionuclides (in Bq kg–1) were measured using lowbackground gamma spectrometers with active volumes of the HP-Ge detectors from 90 to 140 cm3 and peak resolutions (FWHM) of around 1.5 keV at 662 keV. 137Cs inventory was calculated by multiplying activity concentration by soil density (kg m–2). Besides 137Cs, activity concentrations of naturally occurring isotopes (40K, U and Th series) also were determined; the results will be reported elsewhere. By using Marinelli beakers containing about 400 g of dry soil, the minimum detectable activity of 137Cs, defined as three background areas at the 662 keV photopeak, was around 0.3 Bq kg–1 for a 24-hr counting time. The precision of gamma spectrometry measurements, calculated as one standard error of the net area of the 137Cs 662 keV photopeak, was from 7% to 15% . Physico-chemical properties of soil including pH, organic matter, cation-exchange capacity and granulometric compositions were analysed according to the FAO guidelines (FAO, 1995).
3. Results and discussions 3.1. Regression analysis The summary statistics of the measured data and characteristics of sampling locations are given in Table 1. The measured 137Cs inventories increase with latitude resulting in a mean value of about two times higher in the northern (⬎ 16°N) than in the southern (⬍ 16°N) part of the territory (Table 1). Within any latitude band of several degrees, 137Cs inventory tends to increase with annual rainfall. The multiple regression of measured inventories upon characteristics of sampling locations was applied to establish their relationships. In regression analysis, independent variables were latitude, longitude, annual rainfall and four soil parameters; namely pH, clay
P.D. Hien et al. / J. Environ. Radioactivity 62 (2002) 295–303
299
Table 1 Summary statistics of measured data (total number of samples: 292)
Latitude, °N Longitude, °E 137 Cs activity concentration, Bq kg–1 137 Cs inventory, Bq m–2 South Vietnam, ⬍16 °N North Vietnam, ⬎16 °N Bulk density, kg m–3 Mean annual rainfall, m Content of soil organic matter, % Clay, % Total cation exchange capacity,cmol kg–1 pH
Mean
Stand. dev.
Median
Range
16.8 106.5 2.03 643 441 808 1.22 1.97 1.97 19.1 7.5 4.90
4.34 1.5 2.20 384 315 359 0.30 0.56 1.11 12.6 8.2 1.10
17.1 106.3 1.58 590 361 750 1.23 1.88 1.81 15.9 4.5 4.39
9.2–23 103–110 0.5–18.0 129–3294 129–1603 314–3294 0.43–2.1 0.79–3.8 0.30–8.5 0.40–60.7 0.70–54.8 3.3–7.7
component, content of organic matter and total cation exchange capacity. The dependent variable was inventory values, which were logarithmically transformed to enable the construction of regression models with higher goodness-of-fits (R2) and constant error variances (homoscedasticity). The physical interpretation of the logarithm transformation of the dependent variable will be discussed later. The stepwise multiple regression method was applied to select among eight characteristics of sampling locations those which are most significant in explaining the variations of 137Cs inventories across the territory. The stepwise regression analysis was performed using SPSS version 7.5. If the significance level of the regression coefficients was set at 0.01, only latitude L and annual rainfall AR were selected, i.e., they appeared as predictors of the 137Cs inventories: Ln(I) ⫺ e ⫽ (3.53 ± 0.09) ⫹ (0.092 ± 0.004)L
(1)
⫹ (0.62 ± 0.03)AR The intercept and regression coefficients are given in Equation 1 with their standard errors. The regression model is free from multi-colinearity because AR and L do not correlate with each other (r = 0.001). The two predictors L and AR could explain 76% (R2 = 0.76) of the total variance of Ln(I), leaving the 24% remaining variance to the residual e. The latter follows a normal distribution with zero mean (e¯ = 0) and a standard deviation of 0.30, which is apparently greater than that associated with measurement errors. The contribution from measurement errors to e can be estimated as Ln(1 + α) ~ 0.1, where α is the typical relative standard error of the inventory measurements (α ~ 10%). Physico-chemical parameters of soil did not appear in the regression model (Equation 1). Soil parameters, however, would become predictors in the regression models for small regions over which the variations in the latitude and annual rainfall are insignificant.
300
P.D. Hien et al. / J. Environ. Radioactivity 62 (2002) 295–303
3.2. Interpretation of regression model (1) If 137Cs has remained and been attached to soil particles at the point where it was deposited, the inventory measurement would yield the deposition density D. Within the territory of Vietnam, the nuclear tests fallout deposition density varies insignificantly with longitude (UNSCEAR, 1969), so that only two factors control the spatial variability of fallout deposition rate, i.e., latitude and annual rainfall. The product relationship of deposition rate with these two factors results in a linear relationship of logarithmic deposition density with latitude and annual rainfall. We would have in this case a relationship similar to Equation 1, LnI⫺eexp ⫽ LnD
(2)
where D is a function of L and AR, and eexp ~ Ln(1 + α) ~ 0.1. In fact, |e| in Equation 1 is apparently greater than eexp. Therefore, besides measurement errors, there are other sources contributing to the residual term in Equation 1. One potential source is associated with redistribution processes taking place at the sampling area with dissolved 137Cs just after its deposition by rain via overland flow (VandenBygaart et al., 1999), as well as with 137Cs adsorbed in soil particles via erosion transport (see Pennock, 1998 and references therein). As a result of redistribution processes the inventory may be either smaller or greater than the deposition density at the sampling site. It is natural to assume that either case were equally likely to occur in our sampling campaign so that the mean logarithmic inventory for sampling locations having similar L and AR tends to converge toward the logarithm of deposition density. A second source giving rise to the residual in Equation 1 is associated with errors of the AR values. In soil erosion studies, variability of rainfall on the study area is thought to be among potential sources of spatial variability of 137Cs reference input (VandenBygaart et al., 1999). In our case, an error in AR will lead to a deviation of the predicted deposition density from the truly expected value. An error of 0.16 m in AR, for example, would lead to a 10% relative deviation of the deposition density predicted by Equation 1. Thus the regression model can be rewritten as follows: Ln(I) ⫺ e ⫽ Ln(D)
(3)
where D is the deposition density and the residual ε has three components associated with redistribution processes, inventory measurement uncertainty and errors in AR values. It is worth noting that the three error components are independent and random with zero means. In spite of a rather large dispersion of experimental values Ln(I) around Ln(D), i.e.. ε ~ ± 0.30, the logarithmic deposition density could be predicted with a much better certainty due to a large number of samples involved in regression analysis. At the 95% confidence level, the prediction interval for Ln(D) varies from ± 0.04 (for areas having mean latitude and annual rainfall) to ± 0.14 (for areas having very high annual rainfall in the northern part of the territory), with the mean being ± 0.067, which corresponds to a 7% relative uncertainty of the 137Cs deposition density.
P.D. Hien et al. / J. Environ. Radioactivity 62 (2002) 295–303
4. Distribution of
301
137
Cs deposition density
The deviations of the measured inventories from the deposition density due to random errors mentioned above would be reduced if experimental data were averaged over an area containing a sufficient number of sampling points. The regression of the mean of Ln(I) upon mean characteristics of the sampling locations would then yield the same relationship as Equation 1 with smaller residuals. If, for example, experimental data were averaged over (0.5° x 0.5°) grid cell (Fig. 1), the regression model for the grid cell mean inventories would be: [Ln(I)]⫺e’ ⫽ (3.51 ± 0.11) ⫹ (0.093 ± 0.005)[L] ⫹ (0.61 ± 0.04)[AR]
(4)
where [Ln(I)], [L] and [AR] are the grid cell averages of Ln(I), L and AR, respectively. The intercept and regression coefficients in Equation 3 remain almost similar to Equation 1, but the new residuals e⬘ have a standard deviation of 0.17, much less than that of the original model. The regression model (Equation 4) could explain 88% of the variance in [Ln(I)] across 100 grid cells. In this case the right-hand side of (Equation 4) would represent the average deposition density in each grid cell [Ln(D)]. The spatial distribution of [Ln(D)] is mapped in Fig. 1, where both the northward-increasing trend of the deposition density and the effect of annual rainfall variability are evident.
5. Latitudinal distribution of
137
Cs deposition density
The latitudinal distribution of deposition density could be obtained by regressing the average of Ln(I) against the mean characteristics of sampling locations with the same latitude. If data were averaged over half-degree latitude bands, the regression model would be: [Ln(I)]L⫺e’’ ⫽ (3.64 ± 0.11) ⫹ (0.094 ± 0.004)[L]L
(5)
⫹ (0.54 ± 0.04)[AR]L where the symbol [...]L denotes the averages. Again, the intercept and regression coefficients remain almost similar to Equations 1 and 4, but the new regression model could explain up to 96% of the variance in [Ln(I)]L, leaving only 4% to the residuals e⬘⬘. The right-hand side of Equation 5 is [Ln(D)]L — the mean logarithmic deposition density at latitude L. In ordinary scale, Equation 5 would provide a relationship between geometric means of inventory and deposition density; both are plotted in Fig. 2. The latitude mean 137Cs deposition density increases northward from 237 to 1097 Bq m–2, while the corresponding values expected from the UNSCEAR (1969) global pattern are 300 and 600 Bq m–2. Deposition density less than 300 Bq m–2 was found in southernmost areas with low annual rainfalls (less than 1.2 m), while high-rainfall (over 3 m) areas in northern and central parts of the territory have received 137Cs inputs considerably exceeding 600 Bq m–2.
302
P.D. Hien et al. / J. Environ. Radioactivity 62 (2002) 295–303
Fig. 2.
Latitudinal distribution of
137
Cs inventory and deposition density
6. Conclusion The multiple regression of logarithmic inventory values against characteristics of sampling sites enabled the establishment of a model describing the relationship of 137 Cs deposition density with latitude and annual rainfall within the territory of Vietnam. This regression model could explain 76% of the variations in logarithmic inventory values at 292 locations across the territory. The remaining variance contained in the regression residual could be attributed to 137Cs redistribution processes taking place at the sampling area and random errors of inventory measurements and annual rainfall values. Due to a large number of samples investigated, the 137Cs deposition density at any location can be predicted from the regression model with high certainty. At the 95% confidence level, 137Cs deposition density could be predicted by the model within ± 7% relative uncertainty. Within the territory of Vietnam, the latitude mean 137Cs deposition density increases northward from 237 to 1097 Bq m–2. This range is larger than that expected from the UNSCEAR (1969) global pattern (from 300 to 600 Bq m–2). The results of this work are expected to contribute to refining the global distribution pattern of nuclear test fallout deposition in tropical areas, especially in Southeast Asia, for which fallout data are scarce.
Acknowledgements This study was sponsored by the Ministry of Planning, and Investment and was performed under supporting administrative arrangements of the Ministry of Science Technology and Environment. A great number of colleagues of Vietnam Atomic Energy Agency participating in sample collection and preparation are gratefully acknowledged.
P.D. Hien et al. / J. Environ. Radioactivity 62 (2002) 295–303
303
References Food and Agriculture Organization of the United Nations (FAO) (1995). Procedures for soil analysis (3rd ed.). Netherlands: Wageningen. Forsyth, T. J. (1994). The use of 137Cs measurements of soil erosion and farmer’s perceptions to indicate land degradation amongst shifting cultivators in northern Thailand. Mountain Research and Development, 14, 229–244. Garcia-Oliva, F., Lugo, R. M., & Maass, J. M. (1995). Long-term net soil erosion as determined by 137Cs redistribution in an undisturbed and perturbed tropical deciduous forest ecosystem. Geoderma, 68, 135–147. Hai, P. S., Son, P. N., Dien, N. N., Tan, V. H., & Hien, P. D. (1999). Assessment of erosion and accretion in catchment areas based on 210Pb and 137Cs contents in soil and sediment. In Isotope production and applications in the 21st century. In Proceedings of the 3rd International Conference on Isotopes. Vancouver: World Scientific (pp. 415–418). Hien, P. S., Binh, P. N., Truong, N. N., Bac, V. T., & Ngo, N. T. (1994). Variations of caesium isotope concentrations in air and fallout at Dalat, South Vietnam, 1986-91. Journal of Environmental Radioactivity, 22, 55–62. Loughran, R. J., Campbell, B. L., & Walling, D. E. (1987). Soil erosion and sedimentation indicated by 137 Cs: Jackmoor Brook catchment, Devon, England. Catena, 14, 201–212. Owens, P. N., Walling, D. E., & He, Q. (1996). The behavior of bomb-derived 137Cs fallout in catchment soils. Journal of Environmental Radioactivity, 32, 169–191. Pennock, D. J. (1998). New perspectives on the soil erosion–soil quality relationship. In Use of 137Cs in the study of soil erosion and sedimentation. IAEA-TECDOC-1028, Vienna. Ritchie, J. C., & McHenry, J. R. (1975). Fallout 137Cs: a tool in conservation research. Journal of Soil and Water Conservation, 30, 283–286. Ritchie, J. C., & Ritchie, C. A. (1989). Bibliography of publications of 137Cs studies related to soil erosion and sediment deposition. IAEA-TECDOC-1028, Vienna: IAEA. United Nation Scientific Committee on the Effects of Atomic Radiation (UNSCEAR). (1969). 24th Session, Suppl. No 13 (A/7613) New York: United Nations. United Nation Scientific Committee on the Effects of Atomic Radiation (UNSCEAR). (1993). 48th Session, Suppl. No 46 (A/48/46) New York: United Nations. United Nation Scientific Committee on the Effects of Atomic Radiation (UNSCEAR). (2000). 55th Session, Suppl. No 46 (A/55/46) New York: United Nations. VandenBygaart, A. J., King, D. J., Groenevelt, P. H., & Protz, R. (1999). Cautionary notes on the assumptions made in erosion studies using fallout 137Cs as a marker. Canadian Journal of Soil Science, 79, 395–397. Vietnam Atomic Energy Commission. (1998). Internal Reports on the implementation of R&D programmes (in Vietnamese). Vietnam Soil Science Association (1996). Soil of Vietnam. Hanoi: Agriculture Publishing House. Walling, D. E., & Quine, T. A. (1991). The use of 137Cs measurements to investigate soil erosion on arable fields in the UK: potential applications and limitations. Journal of Soil Science, 42, 147–165.