Global Analysis of the Riverine Transport of 90Sr and 137Cs

Dec 30, 2003 - ... load: https://cdn.mathjax.org/mathjax/contrib/a11y/accessibility-menu.js ..... It may be possible to account for inhomogeneous dist...
0 downloads 0 Views 213KB Size
Environ. Sci. Technol. 2004, 38, 850-857

Global Analysis of the Riverine Transport of 90Sr and 137Cs J I M T . S M I T H , * ,† S I M O N M . W R I G H T , ‡ MATTHEW A. CROSS,§ LUIGI MONTE,| A N A T O L Y V . K U D E L S K Y , ⊥ R I T V A S A X EÄ N , ∇ SERGEI M. VAKULOVSKY,# AND DAVID N. TIMMS§ Centre for Ecology and Hydrology, Winfrith Technology Centre, Dorchester, Dorset, DT2 8ZD, U.K., Centre for Ecology and Hydrology, Merlewood Research Station, Grange-over-Sands, Cumbria, LA11 6JU, U.K., School of Earth, Environmental and Physical Sciences, University of Portsmouth, Burnaby Building, Burnaby Road, Portsmouth, PO1 3QL, U.K., ENEA, CP2400, 0100 Rome, Italy, Institute of Geological Sciences of the Academy of Sciences of Belarus, Zhodinskaya Strasse 7, Minsk 220141, Belarus, STUK-Radiation and Nuclear Safety Authority, P.O. Box 14, FIN-00881, Helsinki, Finland, and Institute of Experimental Meteorology, SPA “TYPHOON”, 82 Lenin Avenue, Obninsk, Kaluga Region, 249020, Russia

Atmospheric nuclear weapons explosions and large-scale nuclear accidents may contaminate large areas of land with the long-lived radionuclides 137Cs and 90Sr. The mobility and bioavailability of these radionuclides in the environment is dependent primarily on soil characteristics and changes significantly over time after fallout (1-4). Radioisotope concentrations in different rivers and at different times after fallout vary over 2-3 orders of magnitude. Many previous studies have concentrated on the interactions of radiocesium and radiostrontium with various environmental components, but there are currently no operative models for their transport over large spatial areas. We collected timeseries measurements of 90Sr and 137Cs in 25 major European and Asian rivers and (using digital data sets with global coverage) determined characteristics of each of the rivers’ catchments. This work has established, for the first time, a quantitative link between riverine transport of these radioisotopes and catchment and soil characteristics at a global scale. A generalized predictive model accounting for time changes in river concentrations and variation in catchment characteristics is developed. This can be used to predict the long-term riverine transport of these radiologically important radionuclides following any large-scale nuclear incident in North America, Europe, or (European and Asian) Russia.

Introduction The long-lived radionuclides 90Sr and 137Cs (physical halflives of 28.6 and 30.2 years, respectively) were deposited * Corresponding author phone: 44 1305 213607; fax: 44 1305 213600; e-mail: [email protected]. † Centre for Ecology and Hydrology, Dorset. ‡ Centre for Ecology and Hydrology, Merlewood. § University of Portsmouth. | ENEA. ⊥ Institute of Geological Sciences of the Academy of Sciences of Belarus. ∇ STUK-Radiation and Nuclear Safety Authority. # Institute of Experimental Meteorology. 850

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 3, 2004

globally following atmospheric nuclear weapons testing (NWT), primarily during the period 1954-63. The 1986 Chernobyl accident deposited 137Cs over large areas of Europe, the fallout of 90Sr being mainly confined to the immediate vicinity of the power plant. Long-range postdepositional transport of these radionuclides occurs primarily via the hydrological system, and it is known that their environmental mobility is dependent on catchment characteristics. For radiocesium, highly organic peat bog soils are known to be associated with high activity concentrations in runoff water (4, 5). For radiostrontium, the influence of soil type on runoff concentrations is not well understood, though it is believed that sorption of radiostrontium to soils is determined by the cation exchange capacity of the soil and the concentration of competing Ca and Mg ions (2). Models have previously been formulated (6-9) which characterize the changes in transport of 90Sr and 137Cs as a function of time after fallout. In addition, over a limited spatial area (Cumbria, U.K.) Smith et al. (10) showed that runoff of 137 Cs from catchments was correlated with land cover class, in particular, the catchment’s degree of coverage with organic, boggy soils. For 90Sr, to our knowledge, rates of runoff have not thus far been quantitatively related to catchment characteristics, although it is known that runoff rates vary by more than an order of magnitude between different catchments and are believed to be higher in areas with peaty soils (4). The development of predictive models for pollutant transport which have wide spatial application is severely limited by the difficulty of obtaining input data for large spatial areas. In recent years, however, topographic, hydrological, soils, and land use data sets with global coverage have become available to the scientific community. In the present study we use these data to develop a global model for 137Cs and 90Sr transport in rivers which, for the first time, accounts for environmental influences on their runoff rates.

Methods Fallout of 137Cs and 90Sr to each of the study catchments (Figure 1) resulted from atmospheric nuclear weapons testing (NWT) and the Chernobyl accident. Time-series measurements taken at Tromsø, Norway, of deposited 137Cs were used to determine the input of NWT fallout over the period 1954-85. The deposition at Tromsø was scaled to estimate deposition on each of the 25 catchments using the CLIMATE annual precipitation data set (11). It was assumed that 137Cs deposition (1955-85) is directly proportional to precipitation and is 3.69 kBq m-2 ( 26% per 1000 mm precipitation (12). The ratio of 137Cs to 90Sr deposited from NWT sources has been calculated to be 1.6 (13), allowing the annual deposition of 90Sr from NWT sources to be estimated. Uncertainties in estimates of radioactive fallout are discussed in detail in ref 12. Comparison of predicted NWT fallout with measurements made using individual soil cores (12) showed that 95% of predictions were within a factor of 2 of measured values. It is, however, expected that uncertainty in predictions at the catchment scale (as made here) will be significantly lower than these estimates since variability at a catchment scale is significantly lower than that of individual soil cores (14). Deposition of 137Cs emitted from Chernobyl was obtained from direct measurements of deposition (15) or from the digital version of the Atlas of Caesium Deposition on Europe after the Chernobyl Accident (16). A deposition map was not available for Chernobyl 90Sr, and there was not a constant fallout ratio between Chernobyl-derived 137Cs and 90Sr. The 10.1021/es0300463 CCC: $27.50

 2004 American Chemical Society Published on Web 12/30/2003

FIGURE 1. Location of the 25 European and Asian river catchments. The spatial analysis of catchment characteristics was performed over the area contributing to flow at the radionuclide sampling point. Chernobyl fallout of 90Sr across Finland was deduced by interpreting data from measurement stations taken following the accident (9). However, no such information was available for the other catchments, so analyses for 90Sr were mainly limited to the NWT fallout. In addition to inputs from NWT, 90Sr in the Ob river in Siberia was also influenced by releases from Soviet nuclear facilities. The Ob River will therefore be treated as a special case in this study. Despite the presence of the Krasnoyarsk Mining, Chemical and Industrial Complex releasing radionuclides to the Yenisei River, 90Sr in the lower reaches of the river was due primarily to NWT fallout (from data in refs 17 and 18), so the Yenisei was not excluded. Activity Concentrations of 137Cs and 90Sr in Rivers. Activity concentrations of 137Cs and 90Sr in river waters were assembled from the literature (Table 1). Measurements were totals of dissolved and particulate phases. Typically 95% of 90Sr is in the dissolved phase (19), and the majority of 137Cs is expected to be found in the dissolved phase in most of these rivers (8, 19). Example data sets of 137Cs and 90Sr in river water are given in the Supporting Information, illustrating the significant variation in concentrations during the year. In this study we focus on mean annual concentrations: standard errors illustrate potential variability in mean annual values (see Supporting Information). These may overestimate errors in the mean value in this case since a proportion of within-year variation may be due to real seasonal changes as well as measurement and sampling uncertainty. Activity concentrations in river waters were normalized by dividing the activity concentration in water (Bq m-3) by the total integrated fallout from atmospheric nuclear weapons testing (over the period 1954-85) or by the fallout from Chernobyl. It was assumed that fallout from NWT was negligible after 1985. Catchment Data. The 25 river catchments studied range in size from 3 360 to 2 950 760 km2 and cover a significant proportion of the land mass of the Northern Hemisphere (Figure 1). Small-scale digital data sets with global coverage were used to characterize the catchments. The data sets

provide consistent coverage of the area over which the catchments are located; they are in the public domain and provided free of charge via the Internet (Table 2). The characteristics of each of the river catchments were assessed by programming the Geographical Information System (GIS) to generate statistical summaries (minimum, maximum, and mean values and standard deviations) or distributions of the catchment characteristics shown in Table 2. These were derived for the catchment area above the point of radionuclide measurement in the river (see Figure 1). The Compound Topographic Index (CTI), commonly known as the wetness index, is a function of the upstream area contributing to water flow (FA, km2) and the slope of the landscape. On the basis of the work of Moore et al. (32), the CTI is defined as

CTI ) ln

(

FA tan(slope)

)

(1)

Higher estimated CTI values indicate a greater degree of soil wetness. In flat areas, i.e., areas of zero slope, a CTI value is obtained by substituting a slope of 0.001. Land cover within each of the catchments was identified using the Global Ecosystems land cover classification (33,34) of the Global Land Cover Classification (GLCC) database. This identifies 94 different land cover types which were further grouped to map the major land uses within the river catchments: forest, inland water, agricultural use, and barren areas. The organic carbon content is commonly used to characterize the amount of organic matter in soils, assuming that approximately 58% of organic matter is composed of organic carbon (35). Soils of greater than 30% organic matter are classified as ‘organic’ for the purposes of this study. The soil and land cover characteristics for each catchment were calculated on an area basis. For example, the “% of soils classed as organic” category used in the correlation analysis (Table 3) represents the percentage of the total surface area of each catchment which is covered by soils classed as organic. VOL. 38, NO. 3, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

851

TABLE 1. Measurements of 90Sr and

137Cs

in European Riversa

a Ticks indicate which fallout event and radionuclide the data covers. As some data were reported as monthly or quarterly samples and others as annual averages, all measurements were averaged to produce mean annual radionuclide concentrations. Nuclear weapons test (NWT) 90Sr and Chernobyl 137Cs (grey shaded) data were used for model development; the other data were retained for model testing. NR - not reported.

TABLE 2. Information Concerning the Digital Data Sets Used to Derive Summaries of the Characteristics of the River Catchments physical characteristic

spatial resolution

source of data set

physical size mean slope land cover classification

HYDRO1K dataset (27) HYDRO1K dataset (27) Global Land Cover Classification database (28)

1 × 1 km 1 × 1 km 1 × 1 km

annual precipitation compound topographic index soil texture

CLIMATE database 2.1 (11) HYDRO1K dataset (27) FAO Digital Soil Map of the World version 3.5 (28)

0.5° × 0.5° 1 × 1 km 1 × 1 km

soil carbon content

World-wide organic soil carbon and nitrogen data-set (29) World Minerals Geoscience Database Project (30) WISE-ISRIC soil data set (31) WISE-ISRIC soil data set (31)

1 × 1 km

geology exchangeable Ca content exchangeable K content

Modeling the Temporal Behavior of 90Sr and 137Cs. Runoff of deposited activity from contaminated catchments can be quantified by the use of an exponential “transfer function” model (6-9)

WC(t) ) θ



t

-∞

D(τ)(A1e-(λ+k1)(t-τ) + A2e-(λ+k2)(t-τ) + A3e-(λ+k3)(t-τ)) dτ (2)

where WC(t) is the concentration of the radionuclide in the runoff water (Bq m-3) at time t and D(t) (Bq m-2) is the (timedependent) radionuclide deposition. A1, A2, and A3 are dimensionless coefficients representing, respectively, the fast flush of activity declining during the first few weeks after fallout (at rate k1), the slow decline (at rate k2) as a result of 852

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 3, 2004

1 × 1 km 0.5° × 0.5° 0.5° × 0.5°

units/information km2 integer degrees (0-90°) each cell categorized as forest, inland water, agricultural use, or barren areas mm day-1 dimensionless textural classes reflect the relative proportions of clay, silt and sand in the soil: (1) coarse sands, (2) sandy loams, and (3) fine clays. percentage carbon in surface soils geological data, detailing the age ranges and predominant rock types cmol kg-1 cmol kg-1

soil fixation processes (3), and the very long term (possibly also slowly declining with rate k3) runoff fraction (36). The values of A1, A2, and A3 are estimated such that A1 + A2 + A3 ) 1. The scaling factor, θ (m-1), is used to correlate radionuclide runoff with catchment characteristics. Inspection of eq 2 shows that θ is equal to the normalized radionuclide concentration at time zero for a “spike” type input such as that from Chernobyl (since A1 + A2 + A3) 1, θ ) Wc(0)/D(0) in this case). λ is the decay constant of the radionuclide. The form of eq 2 is illustrated in Figure 2 for the NWT and Chernobyl fallout events. It was observed that the change in radionuclide runoff with time was not significantly correlated with catchment characteristics, so general values of the change of radionuclide concentrations over time were determined and applied to

TABLE 3. Correlation Coefficients between Level of Radionuclide Runoff, θ, and Various Characteristics of the Study Catchmentsa NWT Sr-90 n ) 18 parameter

range in parameter

km2

catchment surface area, mean exch. calcium in soils, cmol kg-1 mean exch. potassium in soils, cmol kg-1 average slope of catchment, deg % of soils with carbon content > 20% mean % carbon content of soils mean annual precipitation, mm % of forested areas in catchment % of “inland water” in catchment % of agricultural lands in catchment % barren areas in catchment % of soils classed as organic % of soils classed as sandy % of soils classed as clay Compound Topographic Index % of catchment with CTI > 13 % of catchment with CTI > 10 % of catchment with CTI < 5 % of catchment with CTI < 3

3360-2.56 × 0.80-45.5 0.025-1.17 0.26-9.65 0-93.6 0.22-23.3 418-1730 19.3-90.0 0.38-22.9 0.22-63.3 0.10-58.2 0-52.4 0-90.9 0-89.4 3.63-7.60 0.89-2.64 2.86-10.1 19.1-78.8 0.39-66.6

106

Chernobyl Cs-137 n ) 12

corr. coeff., r 0.18 -0.54 0.14 -0.64 0.63 0.57 -0.67 0.57 0.84 -0.61 -0.22 0.31 0.61 -0.58 0.61 0.65 0.65 -0.56 -0.71

range in parameter 4018-1.8 × 4.16-28.4 0.20-1.08 0.021-12.4 0-93.6 1.35-23.3 467-987 9.32-90.0 0-22.9 0.22-89.1 0.036-30.3 0-52.4 5.03-90.9 0-14.0 3.17-8.17 0.97-2.67 2.41-11.0 2.00-84.6 0-73.4

105

corr. coeff., r -0.22 -0.34 0.58 -0.38 0.60 0.53 -0.67 0.46 0.75 -0.41 -0.19 0.15 0.78 0.08 0.30 0.58 0.36 -0.16 -0.43

a Bold font indicates correlations which are significant at the 5% level; bold and underlined font indicates correlations which are significant at the 1% level. By chance, with 19 variables, we would expect to see one 5% significant correlation and no 1% significant correlations.

all the catchments. The value of the scaling factor, θ, was estimated for each of the catchments by varying its value to fit eq 2 to the radionuclide activity concentration measurements for each of the catchments individually. This factor was then correlated with the mean values of the catchment characteristics to quantify the environmental influences on radionuclide runoff.

Results and Discussion Variation in Radionuclide Activity Concentrations in Large Rivers. The activity concentrations of 90Sr and 137Cs in large rivers (normalized to concentration per unit of surface deposition) are shown in Figure 2. The graphs illustrate the large variation in activity concentrations: the ratio between the highest and lowest normalized 90Sr activity concentrations is 55, and the ratio between highest and lowest normalized 137Cs concentrations is 1440. The smaller range in 90Sr concentrations is partly because most of the 90Sr measurements are from the NWT period where the longer time scale of releases has a dampening effect on short-term variation in activity concentrations. Time Dependence of Radionuclide Transport. As illustrated by the data in Figure 2, the change in activity concentrations over time is similar for the different rivers studied. We could detect no significant correlation between time-dependent parameters (k values in eq 2) and catchment characteristics. We therefore assume that the parameters describing the time dependence of activity concentrations are invariant across rivers. The following parameter values to describe this time dependence were determined from the measurements of 90Sr from NWT fallout (Figure 2a)

WC(t) ) θ



t

-∞

D(τ)(0.984e-(λ+16)(t-τ) + 0.0123e-(λ+0.09)(t-τ) + 0.0037e-(λ)(t-τ)) dτ (3)

and for Chernobyl-derived

WC(t) ) θ



t

-∞

137Cs

in rivers (Figure 2b)

D(τ)(0.905e-(λ+13)(t-τ) + 0.090e-(λ+0.41)(t-τ) + 0.005e-(λ+0.02)(t-τ)) dτ (4)

Notice that the first exponential term has a characteristic time scale (∼1/k1) of a few weeks: changes on this time scale

FIGURE 2. Data sets assembled for model development. The solid lines illustrate the temporal trend in concentrations (eqs 3 and 4) for different values of the scaling factor, θ. The temporal trends for each radionuclide are assumed to be the same for all catchments, but the scaling factor varies according to catchment characteristics. Values of θ are around an order of magnitude greater for 90Sr than for 137Cs, reflecting the relatively much greater mobility of 90Sr. VOL. 38, NO. 3, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

853

FIGURE 3. Correlations between radionuclide runoff in different rivers (quantified by the scaling factor, θ) and the Inland Water land cover class. Values of θ are around an order of magnitude greater for 90Sr than for 137Cs, reflecting the relatively much greater mobility of 90Sr. cannot be detected in the (annual average) measurements assembled for this study, though there is a detectable increase in the annual average for the first year due to short-term runoff. Therefore, estimates of k1 from studies of short-term changes in 90Sr and 137Cs (6, 37) were used. Although the time-dependent behavior in different rivers was similar, the levels of activity concentrations varied significantly (Figure 2). Thus, the scaling factor, θ, was different for each river as illustrated in Figure 2. The value of θ was determined by fitting eqs 3 and 4 to the measurements of 90Sr (NWT) and 137Cs (Chernobyl) using a leastsquares fitting procedure. The values of θ determined for each river were then correlated with the various environmental characteristics of the catchments to quantify their influence on 137Cs and 90Sr runoff. The more limited measurements of 137Cs from NWT and 90Sr from Chernobyl were retained for testing the models. Correlation of Radionuclide Runoff with Catchment Characteristics. The results of the correlations between the catchment characteristics and the river water scaling factors, θ, are summarized in Table 3. In the presentation of results, we term correlations as “significant” if they are significant at the 5% level and “highly significant” if they are significant at the 1% level. It can be seen from Table 3 that there were a number of significant and highly significant correlations between catchment characteristics and radionuclide activity concentrations in rivers. There were (and were expected to be) many co-correlations between the variables in the data set (see Supporting Information). For example, mean catchment slope and precipitation were highly significantly cocorrelated. For both radionuclides highly significant positive correlations in normalized activity concentrations were observed with coverage by the land cover category “Inland Water”. Figure 3 shows graphs of the scaling factor, θ, vs inland water coverage for both 90Sr and 137Cs. The Eurasia land cover characteristics database was compiled from Advanced Very High Resolution Radiometer (AVHRR) data. Reflectance levels from water and very wet soils (i.e., boggy soils) are relatively similar. Hence, the categorization of data from the AVHRR could class areas of very wet soils as inland water. The Inland Water category therefore indicates all forms of surface standing or running water in the catchments. Strontium-90 NWT Data. A number of significant and highly significant correlations were observed between catchment characteristics indicating the wetness of the catchments’ soils and the runoff rate of 90Sr. Highly significant 854

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 3, 2004

positive correlations were observed with the Inland Water land cover class and high Compound Topographic Index (CTI) values. Negative correlations were observed for low CTI values. Somewhat unexpectedly, a highly significant negative correlation was observed with mean annual precipitation, though soil wetness (CTI) was significantly negatively co-correlated with precipitation. The catchment slope showed a negative correlation with 90Sr runoff. This may be due to the fact that high slope leads to low soil wetness. Soils of high carbon content showed a highly significant positive correlation with 90Sr runoff, though this was not observed in the correlation with the organic soil class from the soil texture data. High organic matter content would not necessarily be expected to correlate with high 90Sr runoff since sorption of 90Sr to soils is proportional to soil cation exchange capacity (2), generally higher in organic soils than in mineral. A significant negative correlation was observed between mean exchangeable calcium in the soil and 90Sr runoff. Previous work (38) reported that 90Sr was less mobile in soils of high exchangeable calcium. Strontium can substitute with calcium at the structural sites of calcite (39), thus reducing its mobility (40). In a study such as this, it is inevitable that parameters will be co-correlated, making individual processes difficult to distinguish. It appears, however, that in general high 90Sr runoff is positively associated with wet areas in a catchment and negatively associated with mean exchangeable calcium in the soils. Cesium-137 from Chernobyl. Results presented in Table 3 show that 137Cs runoff is highly significantly correlated with the Inland Water land cover class and the sandy soils texture class and significantly correlated with soil carbon content. Previous studies have shown that high 137Cs mobility is associated with organic, boggy soils. In soils high in clay minerals, 137Cs is sorbed to highly selective “Frayed Edge Sites” (FES) on the illitic clay fraction (41). Work has shown (42), however, that in highly organic soils FES concentrations are low, leading to a reduced binding of 137Cs to the solid phase. In field studies on small catchments (5) it was found that highly organic soils (particularly saturated peats) continue to release up to an order of magnitude more radiocesium to surface waters than some mineral soils. Runoff of 137Cs is significantly positively correlated with exchangeable potassium, which may reflect competition for sorption sites by potassium (41). As with 90Sr, there is a significant negative correlation of 137Cs runoff with precipitation, though, as discussed above, precipitation is highly significantly negatively co-correlated with soil wetness (CTI), see Supporting Information. Limitations of Large-Scale Data Analysis. Our use of large-coverage data sets to study the environmental mobility of 90Sr and 137Cs has allowed the evaluation of their behavior over large areas but at low resolutions. The coarse resolution of the data sets can result in the loss of fundamental information or the misrepresentation of data. Information is lost due to the relatively coarse resolution of 1 km2 grids, the averaging of data required to obtain a mean soil texture for the river catchment, and classifying a variable within a category. Despite the acknowledged limitations of the approach, it should be noted that no “process”-based model of radionuclide transport is able to predict the influence of environmental parameters at the spatial scale required for assessing the impact of large-scale nuclear incidents. The data are simply not available at the spatial scale required for process level models. We believe that the general characterization of catchments using data sets with global coverage is the only way to carry out such an assessment. Although data sets with global coverage do not allow a detailed investigation of the behavior of 90Sr and 137Cs, this analysis

TABLE 4. Driving Variables for the Generalized Predictive Model and Description of How They Are Obtained description of how parameter value is obtained parameter m-2)

deposition, D (Bq A1, A2, A3 ([dimensionless]) k1, k2, k3 (yr-1) scaling factor, θ (m-1)

Cs-137 model

Sr-90 model

estimated from fallout data fixed values, as given in eq 4 fixed values, as given in eq 4 estimated from Global Land Cover Classification database using θ ) 0.0108 × (% inland water) + 0.063 (Figure 3)

estimated from fallout data fixed values, as given in eq 3 fixed values, as given in eq 3 estimated from Global Land Cover Classification database using θ ) 0.141 × (% inland water) + 0.55 (Figure 3)

has identified a number of significant and highly significant correlations which are consistent with process-level knowledge of their behavior and that enable predictions of their environmental transport to be made. Development of a Predictive Model. The relationships between 90Sr and 137Cs runoff (i.e., θ - values) and the percentage of Inland Water land cover class (Figure 3) can be used to predict θ for new rivers. Inputting the estimated value of θ into the time-series relationships (eqs 3 and 4) allows predictions of 90Sr and 137Cs in the river at any time after a fallout event. The required model-driving variables are presented in Table 4. Note that the model assumes either an approximately homogeneous distribution of fallout in the catchment or (if fallout is inhomogeneous) that the distribution of key catchment characteristics in the contaminated area is similar to the average for the whole catchment. For example, we excluded data from the Pripyat-Dnieper River system below the Chernobyl 30 km zone from the study since radionuclide concentrations are dominated by the relatively small “hot spot” around the plant. It may be possible to account for inhomogeneous distributions of fallout by breaking down the catchment characteristics distribution to smaller units; however, this is beyond the scope of our present study. Model Testing. The comparison of predictions of the generalized models (Table 4) against data used to develop the model is shown in Figure 4. Model errors were determined by an analysis of residuals against the model development data set: Figure 4 shows error ranges within which 90% of measured data fall. For the 90Sr model, 90% of measurements were within a factor 2.5 of model predictions; for 137Cs, 90% of measurements were within a factor 4.0 of predictions. The models were then tested (Figure 5a) against data from the Chernobyl accident for 90Sr and against data from NWT for 137Cs (i.e., against data not used in model development, see Table 1). The test in Figure 5a shows good agreement between model predictions and measured values, though the test is somewhat limited since the catchments are a subset of those which were used to develop the models. However, the tests do show that models developed from one fallout event (NWT or Chernobyl) can be applied to make predictions for a different incident. A stronger “blind” test of the radiocesium model was carried out against measurements of NWT 137Cs in the Thames (Figure 5b), again showing good agreement. Data from the Thames catchment was not used to develop the radiocesium model. The model parameters (Table 4) could, of course, be “tuned” to better fit the data, but our primary purpose here is to develop a general predictive model rather than to fit the model to a particular river. Radiostrontium in the Ob River. The transport of radionuclides from Siberian nuclear facilities to the Arctic seas is an issue of international concern. As well as having inputs from global nuclear weapons fallout, the Ob River has been contaminated by activities at the Mayak Production Association and Tomsk-7 plutonium production facilities situated on the Techa and Tom Rivers, respectively (both are tributaries of the Ob). The main releases of radiostrontium

FIGURE 4. Validation of the generalized model (Table 4) against all data used for model development: (a) NWT 90Sr data; (b) Chernobyl 137Cs data. The dotted lines illustrate the model uncertainty estimate: 90% of points lie within the ranges shown on each graph. to the Ob and its catchment were (1) direct discharges from Mayak to the river (12 PBq, 1949-56); (2) an explosion of a high-level waste tank in 1957 (the “Kyshtym” accident, 2 PBq released to Techa/Ob catchment); (3) wind dispersion of sediments from Lake Karachay in 1967 (0.0037 PBq) (43). From this study we estimate total 90Sr deposition to the catchment of 5.4 PBq from NWT fallout. There is (to our knowledge) no data on the direct discharges to the river from the Tomsk-7 facility or of releases from the Mayak site following the construction of a holding reservoir system in 1957. The Semipalatinsk nuclear test site is also in the Ob VOL. 38, NO. 3, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

855

event (NWT or Chernobyl) and of the characteristics of particular river catchments. Thus, time changes in the concentrations of these radionuclides in rivers can be predicted using eqs 3 and 4. It is possible that these timedependent relationships will also apply to many terrestrial systems (see refs 3, 36). The considerable variation in radionuclide concentrations in different rivers (illustrated in Figure 2) is not due to differences in time-dependent behavior between rivers but is shown to correlate with a number of characteristics of the catchment. We have shown that the Inland Water land cover class can be used to predict this variation (Figure 3). In the event of a large-scale nuclear incident, the general model presented in Table 4 could be used to predict 90Sr and 137Cs in any river in North America, Europe, or (European and Asian) Russia.

Acknowledgments We acknowledge financial support from the European Commission Inco-Copernicus (STREAM IC15-CT98-0219; AQUASCOPE IC15-CT98-0205) and INTAS (RESPOND, 010556) programs. We would like to thank three anonymous referees for their constructive criticisms.

Supporting Information Available Tables containing example data sets from three catchments (Table S1), giving co-correlations between the catchment characteristics (Table S2). This material is available free of charge via the Internet at http://pubs.acs.org.

Literature Cited

FIGURE 5. (a) Model testing against data for 90Sr from Chernobyl and 137Cs from NWT (i.e., data not used for model development). Dotted and thin lines show the uncertainty estimate in model predictions for 137Cs and 90Sr, respectively. (b) “Blind” test of the 137Cs model against measurements in the river Thames and predictions for 90Sr in the Ob assuming inputs from NWT only. Dotted and thin lines show model uncertainty estimates for the Thames and Ob predictions, respectively. Uncertainty estimates, derived from Figure 4, give the estimated range within which measured values will fall in 90% of cases. catchment; however, it has been stated that its direct contribution to radioactive contamination of the Ob is small (43). Measurements of 90Sr in the mouth of the Ob River at Salekhard were made from 1961 onward (Figure 5b) (24). We used the model developed above to test whether we can detect significant additional export of 90Sr from the nuclear facilities and the Semipalatinsk test site (via the Ob) to Arctic seas during this period. Figure 5b shows the comparison of the model predictions of 90Sr in the Ob (assuming only inputs from global NWT fallout) with the measured data. Although 90Sr activity concentrations in the Ob are higher than those predicted for NWT fallout alone, they are within the range of error in model predictions. In addition, the activity concentrations closely follow the temporal pattern predicted to result from the NWT input, implying that the NWT signal dominates the measured values. While elevated 90Sr activity concentrations were observed in the upper parts of the Ob catchment during this period (44), we could not detect significantly elevated levels being discharged at the river mouth to the Kara Sea from 1960 to 1985. Further Applications. The analysis shows (within the limitations of the data set) that the change in 90Sr and 137Cs runoff over time (Figure 2) is independent both of the fallout 856

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 3, 2004

(1) Absalom, J. P.; Young, S. D.; Crout, N. M. J. Eur. J. Soil Sci. 1996, 46, 461-469. (2) Yasuda, H.; Uchida, S. Environ. Sci. Technol. 1993, 27, 24622465. (3) Smith, J. T.; Fesenko, S. V.; Howard, B. J.; Horrill, A. D.; Sanzharova, N. I.; Alexakhin, R. M.; Elder, D. G.; Naylor, C. Environ. Sci. Technol. 1999, 33, 49-54. (4) Hansen, H. J. M.; Aarkrog, A. Water Res. 1990, 24, 1137-1141. (5) Hilton, J.; Livens, F. R.; Spezzano, P.; Leonard, D. R. P. Sci. Total Environ. 1993, 129, 253-266. (6) Monte, L. J. Environ. Radioact. 1995, 26, 71-82. (7) Monte, L. Sci. Total Environ. 1997, 201, 17-29. (8) Smith, J. T.; Clarke, R. T.; Saxe´n, R. J. Environ. Radioact. 2000, 49, 65-83. (9) Cross, M. A.; Smith, J. T.; Saxe´n, R.; Timms, D. J. Environ. Radioact. 2002, 60, 149-163. (10) Smith, J. T.; Howard, D. C.; Wright, S. M.; Naylor, C.; Brookes, A. M.; Hilton J.; Howard, B. J. Sci. Total Environ. 1998, 209, 1-15. (11) http://www.pik-potsdam.de/∼cramer/climate.htm. (12) Wright, S. M.; Howard, B. J.; Strand, P.; Nyle´n, T.; Sickel, M. A. K. J. Environ. Pollut. 1999, 104, 131-143. (13) Cambray, R. S.; Playford, K.; Lewis, G. N. J.; Carpenter R. C. Report of the Atomic Energy Authority AERE R 13226, Harwell, 1989. (14) Smith, J. T.; Appleby, P. G.; Hilton, J.; Richardson, N. J. Environ. Radioact. 1997, 37, 127-142. (15) Saxe´n, R. In The transfer of radionuclides through Nordic Ecosystems to Man; Dahlgaard, E. H., Ed.; Elsevier: Amsterdam, 1994; pp 63-78. (16) De Cort, M.; Dubois, G.; Fridman, Sh. D.; Germenchuk, M. G.; Izrael Yu. A.; Janssens, A.; Jones, A. R.; Kelly, G. N.; Kvasnikova E. V.; Matveenko, I. I.; Nazarov, I. M.; Pokumeiko, Yu. M.; Sitak, V. A.; Stukin, E. D.; Tabachny, L. Ya.; Taturov, S. Yu.; Avdyushin, S. I. European Commission report EUR 16737, Luxembourg, 1998. (17) Vakulovsky, S. M. Report of the ISTC Project 1404, SPA TYPHOON, Obninsk, 2001. (18) Vakulovsky, S. M.; Chumichev, V. B. Proceedings of the 5th Annual Conference on Environmental Radioactivity in the Arctic and Antarctic. Norwegian Radiation Protection Authority: Østerås 2002. (19) Vakulovsky, S. M.; Nikitin, A. I.; Chumichev, V. B.; Katrich, I. Y.; Voitsekhovitch, O. A.; Medinets, V. I.; Pisarev, V. V.; Bovkum, L. A.; Khersonsky, E. S. J. Environ. Radioact. 1994, 23, 103-122.

(20) Salo, A.; Saxe´n, R.; Puhakainen, M. Aqua Fennica 1984, 14, 2131. (21) Saxe´n, R.; Ilus, E. J. Environ. Radioact. 2001, 54, 275-291. (22) Spezzano, P.; Bortoluzzi, S.; Giacomelli, R.; Massironi, L. J. Environ. Radioact. 1994, 22, 77-88. (23) Szerbin, P.; Szabo, G. In Radioactive contamination of rivers and transport through rivers, deltas and estuaries to the sea; NATO, 1998; pp 51-58. (24) Chumichev, V. B. In Environmental radioactivity in the arctic; Strand, P., Cooke, A., Eds.; Norwegian Radiation Protection Authority: Østerås, 1995; pp 79-83. (25) Kudelsky, A. V.; Smith, J. T.; Zhukova, O. M.; Matveyenko, I. I.; Pinchuk, T. M. Proc. Acad. Sci. Belarus 1998, 42, 90-94 (in Russian). (26) Linsley, G. S.; Haywood, S. M.; Dionan, J. 1982 International Atomic Energy Agency report IAEA-SM-257/8, Vienna, 1982; pp 615-633. (27) Verdin, K. L.; Greenlee, S. K. In Proceedings, Third International Conference/Workshop on Integrating GIS and Environmental Modeling, Santa Fe, New Mexico; National Center for Geographic Information and Analysis: Santa Barbara, CA, 1996. Data available from: http:// edcdaac.usgs.gov/gtopo30/hydro/index.html. (28) FAO, Digital soil map of the world and derived soil properties, FAO Land and Water Digital Media Series N°1, Dec 1998, ISBN 92-5-104050-8, 1998. Data set available from http://edcdaac.usgs.gov/glcc/glcc.html. (29) Post, W. M.; Emanuel, W. R.; Zinke, P. J.; Stangenberger, A. G. Nature 1982, 298, 156-159. Data set available on-line [http:// www.daac.ornl.gov] from Oak Ridge National Laboratory Distributed Active Archive Center, Oak Ridge, TN, 1998. (30) Data set available from: http://www.nrcan.gc.ca/gsc/mrd/ wmgdb/index_e.php. (31) Batjes, N. H., Ed.; 1996. Working paper and Preprint 96/05. International Soil Reference and Information Centre (ISRIC), Wageningen, The Netherlands, 1996. Data set available on-line [http://www.daac.ornl.gov] from Oak Ridge National Laboratory Distributed Active Archive Center, Oak Ridge, TN, 2000. (32) Moore, I. D.; Grayson, R. B.; Ladson, A. R. Hydrol. Processes 1991, 5, 3-30.

(33) Olson J. S. Global ecosystem framework-definitions: USGS EROS Data Center Internal Report, Sioux Falls, SD, 1994; 37 p. (34) Olson J. S. Global ecosystem framework-translation strategy: USGS EROS Data Center Internal Report, Sioux Falls, SD, 1994; 39 p. (35) Jarvis, R. A.; Bendelow, V. C.; Bradley, R. I.; Carroll, D. M.; Furness, R. R.; Kilgour, I. N. L.; King, S. J. Soils and their use in Northern England: Bulletin No. 10. Soil Survey of England and Wales, Harpenden, 1984. (36) Smith, J. T.; Comans, R. N. J.; Beresford, N. A.; Wright, S. M.; Howard, B. J.; Camplin, W. C. Nature 2000, 405, p141. (37) Voitsekhovitch, O. V.; Borzilov, V. A.; Konoplev, A. V. In Proceedings of a Seminar on Comparative Assessment of the Environmental Impact of Radionuclides Released During Three Major Nuclear Accidents; Kyshtym, Windscale, Chernobyl; European Commission Radiation Protection Series No. 53; EUR 13574, Luxembourg, 1991; pp 528-548. (38) Squire, H. M. Radiat. Bot. 1966, 6, 49-67. (39) Pingitore, N. E.; Eastman, M. P. Geochim. Cosmochim. Acta 1986, 50, 2195-2203. (40) Bunde, R. L.; Rosentreter, J. J.; Liszewski, M. J.; Hemming, C. H.; Welhan, J. Environ. Geol. 1997, 32, 219-229. (41) Cremers, A.; Elsen, A.; De Preter, P.; Maes A. Nature 1988, 335, 247-249. (42) Valcke, E.; Cremers, A. Sci. Total Environ. 1994, 157, 275-283. (43) NATO Radioactive contamination of rivers and transport through rivers, deltas and estuaries to the sea; NATO report No. 225; NATO: 1998; 136 pp. (44) Chumichev, V. B.; Demyanchenko, Z. A. In Environmental radioactivity in the Arctic and Antarctic; Strand, P., Holm, E., Eds.; Norwegian Radiation Protection Authority: Østerås, 1993; pp 111-114.

Received for review April 3, 2003. Revised manuscript received October 14, 2003. Accepted October 17, 2003. ES0300463

VOL. 38, NO. 3, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

857