Radionuclides in fruit systems: Model prediction-experimental data intercomparison study

Science of the Total Environment 366 (2006) 514 – 524 www.elsevier.com/locate/scitotenv

Radionuclides in fruit systems: Model prediction-experimental data intercomparison study Z. Ould-Dada a,⁎, F. Carini b , K. Eged c , Z. Kis c , I. Linkov d , N.G. Mitchell e , C. Mourlon f , B. Robles g , L. Sweeck h , A. Venter i a

Food Standards Agency, Radiological Protection and Research Management Division, Aviation House, 125 Kingsway, Room 715B, London WC2B 6NH, United Kingdom b Università Cattolica del Sacro Cuore, Faculty of Agricultural Sciences, Institute of Agricultural and Environmental Chemistry, Via Emilia Parmense, 84, I-29100 Piacenza, Italy c Department of Radiochemistry, University of Veszprém, P.O. Box 158 H-8201, H-8200 Veszprém, Hungary d ICF Consulting, Inc., 33 Hayden Ave, Lexington, MA 02421, USA e Mouchel Consulting Ltd., West Hall, Parvis Road, West Byfleet, Surrey, KT14 6EZ, United Kingdom f Institute for Radiological Protection and Nuclear Safety (IRSN)/Environment and Emergency Operations Division (DEI), Laboratory of Environmental Modelling (LME), CE/Cadarache, 13 108 St Paul-lez-Durance Cedex, France g CIEMAT, Dept. de Impacto Ambiental (DIAE), Edif. 3A, Avenida Complutense 22, E-28040 Madrid, Spain h SCK•CEN, Boeretang 200, 2400 Mol, Belgium i Enviros Consulting Ltd, Telegraphic House, Waterfront Quay, Salford Quays, Greater Manchester, M50 3XW, United Kingdom Received 29 November 2004; received in revised form 7 October 2005; accepted 10 October 2005 Available online 18 January 2006

Abstract This paper presents results from an international exercise undertaken to test model predictions against an independent data set for the transfer of radioactivity to fruit. Six models with various structures and complexity participated in this exercise. Predictions from these models were compared against independent experimental measurements on the transfer of 134Cs and 85Sr via leaf-to-fruit and soil-to-fruit in strawberry plants after an acute release. Foliar contamination was carried out through wet deposition on the plant at two different growing stages, anthesis and ripening, while soil contamination was effected at anthesis only. In the case of foliar contamination, predicted values are within the same order of magnitude as the measured values for both radionuclides, while in the case of soil contamination models tend to under-predict by up to three orders of magnitude for 134 Cs, while differences for 85Sr are lower. Performance of models against experimental data is discussed together with the lessons learned from this exercise. © 2005 Elsevier B.V. All rights reserved. Keywords: Radionuclides; Model-testing; Biomass; Fruit;

134

Cs;

85

Sr

⁎ Corresponding author. Current address: Department for Environment, Food and Rural Affairs (Defra), Europe Environment Division, Room 5/H15 Ashdown House, 123 Victoria Street, London SW1E 6DE, United Kingdom. E-mail address: [email protected] (Z. Ould-Dada). 0048-9697/$ - see front matter © 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.scitotenv.2005.10.008

1. Introduction One of the objectives of the programme of the Fruits Working Group (preface of Carini et al., this issue), was to undertake testing and validation of existing or new models against independent data sets. The objectives of

Z. Ould-Dada et al. / Science of the Total Environment 366 (2006) 514–524

the exercise were to intercompare the predictions generated by the models and to compare predictions with the observed data. This paper presents the results of the model-testing exercise that was undertaken by the participants of the Group. The exercise involved comparison of outputs from six different models with a data set for which modellers had no prior information. Various experimental data sets were offered by participants of the Group, but it was difficult to find a complete series of data describing the fluxes of radionuclides, from deposition to distribution with time within the compartments of fruit ecosystems, with supporting yield data. After discussion of possible scenarios, a data set on the transfer of 134Cs and 85Sr via leaf-to-fruit and soil-to-fruit in strawberry plants after an acute release was chosen and finalised for the validation exercise. This was the first model validation exercise for fruit crops. It offered the opportunity to test whether models concerned with the assessment of the transfer of radionuclides to fruits adequately describe the system modelled. 2. Description of the test scenarios The model scenarios are based on experimental work carried out at Università Cattolica del Sacro Cuore of Piacenza (Italy) to investigate the short-term transfer of 134 Cs and 85Sr via leaf-to-fruit and soil-to-fruit in strawberry plants after an acute release. Strawberry plants were grown in pots filled with peat substrate and placed beneath a ventilated tunnel in a field representative of horticultural growing conditions in Italy. Three groups of plants were contaminated by application of 134Cs and 85 Sr in the form of chlorides (134CsCl and 85SrCl2) in aqueous solution, either to the above-ground part of the plant (foliar contamination) or to the soil surface (soil

515

contamination). A synopsis of the scenario is presented below. A full description of the experimental design is given in IAEA (2003). Details and results of foliar contamination are described in Carini et al. (2003). 3. Foliar contamination scenarios Foliar contamination was carried out on two groups of plants (9 replicates for each group) at two phenological stages, anthesis and ripening. The two treatments were named: “first foliar contamination” and “second foliar contamination”. Plants were flowering (anthesis) at the first contamination, had well-developed leaves and a few small immature green fruits. Only 134 Cs was included in this scenario. Plants had aged by a further 26 days by the time of the second foliar treatment, the crop was ripening, although bearing green and red fruits, and very few flowers remained. Contamination was effected with both 134Cs and 85Sr. Radionuclides were applied to plants as small droplets using an aspirated spray to simulate wet deposition, while the soil surface of each pot was protected. The radioactivity sprayed over plants at each phenological stage, expressed as kBq m− 2, is reported in Table 1. In order to determine the activity intercepted by plant components, the above-ground parts of four plants were harvested as soon as dry after spraying, and separated into leaves and fruits. The activity intercepted by leaves and fruits, expressed as percentage of that applied, is reported in Table 1. 4. Soil contamination scenario Soil contamination was carried out only at the anthesis stage, on 12 replicates. The soil of each pot was moistened over the entire surface with 150 ml of an

Table 1 Radioactivity sprayed over plants (kBq m− 2 or kBq plant− 1) and intercepted (% of the sprayed) in the case of foliar contamination, or deposited onto the soil (kBq m− 2 or kBq plant− 1) in the case of soil contamination Contaminated compartments

Above-ground plant part

Above-ground plant part

Soil

Treatment code

First foliar

Second foliar

Soil

Phenological stage at time of contamination Date of contamination Radionuclide Sprayed activity Intercepted activity (% of the sprayed)

Anthesis

Ripening

Anthesis

Deposited activity

22 April 1998 Cs 805.1 36.7 ± 0.9 0.23 ± 0.07 36.9 ± 1.0 – –

134

kBq m− 2 Leaves Fruits Whole above-ground part kBq m− 2 kBq plant− 1

18 May 1998 Cs 890.8 29.2 ± 1.7 1.16 ± 0.22 30.3 ± 1.8 – –

134

85

Sr 776.6 30.3 ± 1.5 1.20 ± 0.24 31.5 ± 1.6 – –

27 April 1998 85 Cs Sr – – – –

134

765.5 147.5

1698.2 327.2

516

Z. Ould-Dada et al. / Science of the Total Environment 366 (2006) 514–524

aqueous solution containing 134CsCl and 85SrCl2. After treatment, the soil surface was covered with a layer of expanded clay to separate the leaves from the soil and prevent their direct contamination. The deposited activity, expressed as kBq m− 2 is reported in Table 1. 5. Harvests Fruits were picked when ripe, weighed as collected and frozen. They were grouped into two harvests: 20 May 1998 and 2 June 1998. Each sample was defrosted and homogenized before being analysed by direct gamma spectrometry. The harvest of each plant was analysed separately as a single replicate. The dry matter content was obtained drying fruits in a fan oven at 60 °C until constant weight, after gamma measurement. The whole plant was harvested at the end of the fruit season (from 1 July to 14 July). After separation of fruits, the above-ground part of each plant was divided into leaves and crowns, weighed as collected and dried at 60 °C until constant weight. Each dried sample was minced and homogenized before being analysed by direct gamma spectrometry. Yields were expressed in g wet weight/plant for fruits and g dry weight/plant for leaves. The arithmetic average yields and standard errors of 9 replicates for foliar contamination and 12 replicates for soil contamination were calculated using a plant density of 5.19 plants m− 2. Yield values of the two fruit harvests and of leaves are reported in Table 2. The concentration of 134Cs and 85Sr was measured by direct gamma spectrometry of the samples. A HpGe detector was used, with an efficiency of 38% and a FWHM resolution of 1.76 keV at 1.33 MeV of the 60 Co. The gamma spectrum analysis was performed using the programme GENIE 2000 (Canberra Nuclear). Each sample was analysed for a time sufficient to collect at least 4000 net counts in the peaks of interest to reach a maximum error of 3–4% at 95% confidence. Different counting geometries were employed, depending on the size of the analytical sample. All results were

decay-corrected to the same reference time, arbitrarily chosen as the average date of fruit harvest. 6. Endpoint calculations Modellers were requested to estimate 134Cs and 85Sr concentrations in fruit, expressed as Bq g− 1 wet weight, at the two times of harvest: 20 May and 2 June 1998. Radionuclide concentrations in leaves, expressed as Bq g− 1 dry weight, were to be given at 1 July 1998 for the first foliar and the soil contamination scenarios and at 14 July 1998 for the second foliar contamination scenario. Generally speaking, the radionuclide concentration in edible products or in other plant components is expressed on a dry weight basis, to allow comparisons between values derived from different experimental or climatic conditions and/or from products with different water content. At the inception of the activities of the Fruits Working Group, the participants agreed to express the radionuclide concentration in fruit, as well as yield values, on a wet weight basis, given that fruit consumption is a fresh product, and radionuclide concentration in leaves on a dry weight basis. 7. Participating models Six models with various structures and complexity participated in this exercise: SPADE (United Kingdom), FRUTI-CROM (Spain), FRUITPATH (USA), RUVFRU (Hungary), DOSDIM (Belgium) and ASTRAL (France). A compendium of model characteristics and of the representation of major processes is given in Linkov et al. (this issue, Tables 2 and 3). All these models were designed for atmospheric deposition, but very few would be capable of simulating a terrestrial source such as a nuclear waste repository. Five of them were developed by Government agencies for regulatory assessment of radionuclide concentration in fruits and radiation dose resulting from their consumption. Different approaches have been used to simulate the transfer processes (e.g. deposition, translocation, root uptake) of

Table 2 Yield factors (kg wet weight or kg dry weight m− 2) and corresponding dry matter (%) Treatment code Endpoint

First foliar Date of harvest Unit

Yield

Second foliar Dry matter (%) Yield

Fruit, first harvest 20 May 1998 kg ww m− 2 1.169 ± 0.077 6.2 Fruit, second 2 June 1998 kg ww m− 2 1.621 ± 0.137 7.2 harvest Leaf 1–14 July 1998 kg dw m− 2 0.214 ± 0.013 36.2

1.109 ± 0.055 1.595 ± 0.154

Soil Dry matter (%) Yield 6.4 8.1

0.221 ± 0.025 36.9

1.088 ± 0.109 0.888 ± 0.155

Dry matter (%) 6.7 8.0

0.195 ± 0.020 39.5

Z. Ould-Dada et al. / Science of the Total Environment 366 (2006) 514–524

radionuclides to fruit (see discussion below). Most of them are supported by very limited experimental data derived from the literature or from observations and reflecting the specific model structure. Soil Plant Animal Dynamic Evaluation (SPADE) (Thorne and Coughtrey, 1983) models radionuclide uptake by three types of fruit crops: herbaceous, shrubs and trees. The quantity of radionuclides reaching the above-ground compartments of the plant from atmospheric sources is determined according to the interception fraction, which takes account of changes in plant biomass with season. Transfers from soil to plant via root uptake are assumed to vary with soil layer depth, as a function of the root distribution throughout the soil profile. Models are implemented in SPADE for 20 elements, whose original default parameters (Coughtrey and Thorne, 1983) were based largely on data for cereals, but were modified in the case of tree and shrub fruits. Two experimental programmes undertaken in connection with the development of the SPADE fruit models for herbaceous and shrubby fruit crops (Kirton et al., 1987; Donelly and Carini, 1998) provided information for model validation. Full description of SPADE is given in Mitchell (2001). FRUTI-CROM is a fruit-specific model that was developed from an existing vegetable sub-model CROM. Parameters as the dry deposition velocity (vg), the washout rate (ω), the fraction of deposited activity intercepted by edible portion of plant (α), the standardised surface density for the effective root zone in soil (P), the rate constants for reduction of concentration due to processes other than radioactive decay (λw and λs) are derived from Till and Meyer (1983). The concentration factor (Fv) for uptake from soil by edible part and soil adhesion is derived from ECOSYS (Müller and Pröhl, 1993). FRUITPATH (Linkov and Burmistrov, 2003) focuses on a generic ecosystem application. It is a wholly probabilistic model that incorporates uncertain model parameters as probability distributions and predicts distribution for the output radionuclide concentrations in fruit compartments. For generic model application, uncertain model parameters are estimated from literature that includes different fruit and soil types. For site-specific applications, the available literature data are limited to the ecosystem similar to the site under consideration; site-specific parameters are thus estimated. Further model calibration, based on site-specific measurements, can be accomplished by using Bayesian updating procedures. Most of the parameter values used by RUVFRU originate from IAEA (1994, 1995, 1996, 1999) and

517

Hungarian publications presenting results of post Chernobyl measurements carried out in Europe (Hungarian Atomic Energy Office, 1996). Several values derive from generic models as FARMLAND (Mayall, 1995 cited by Mitchell, 2001) and SPADE. Whatever the origin, a seasonal change based on temperature is taken into consideration for the majority of rate constants. DOSDIM is an example of a non-fruit specific model used to calculate the transfer of radionuclides to fruits. For plant-specific parameters, those for leafy vegetables were used to estimate interception by the strawberry plant and those for root vegetables to calculate the translocation rate. The parameter for root uptake was derived from the Transfer Factor values given in the Fruit Review (Carini, 2001). ASTRAL has no fruit-specific sub-model, but there is a sub-model that is used for fruit vegetables: it is assumed that fruit are produced throughout the year (market garden scenario). The model and parameters for the fruit vegetable class have been chosen, as this class covers a wide variety of plants, from vegetables such as tomatoes and beans to strawberries. A detailed discussion of the data that support these models is given in the Fruits Working Group final report (IAEA, 2003). Performance of some of the above models against the present scenario has been published and discussed in detail by Oncsik et al. (2002) for RUVFRU and Ould-Dada et al. (2003) for SPADE. SPADE, FRUTI-CROM, RUVFRU and ASTRAL were designed to provide point estimates for activity concentrations and dose. DOSDIM is capable of incorporating stochastic calculations, while FRUITPATH is the only model that incorporates probabilistic MonteCarlo simulation and predicts probability distribution for radionuclide concentrations at different time scales. 8. Results and discussion Results of the model predictions and comparison of predictions with the test data are presented graphically and described in the following sections. All the modellers reported results as best estimates, except for the results from FRUITPATH model, which were also reported as 95% confidence interval. In this paper, however only the median value for FRUITPATH is reported. The error bars reported for the experimental values indicate the standard errors. 9. First foliar contamination This first scenario considered foliar contamination with 134Cs at anthesis. Contamination with 85Sr was

Z. Ould-Dada et al. / Science of the Total Environment 366 (2006) 514–524

only effected in the second foliar scenario. Fruits were then harvested at 28 and 41 days after deposition. Measured values show that the total activity was 40 times higher than the activity directly intercepted by fruitlets, supporting the hypothesis that fruit activity for 134Cs after deposition at anthesis is mainly ascribable to the process of leaf-to-fruit translocation. Predicted values for fruit activity, both for the first and the second harvest, covered two orders of magnitude: from 1.1 to 1.6 · 102 Bq g− 1 ww (Fig. 1). Comparison of predicted with observed values, (2.5–2.7 ± 0.2) · 101 Bq g− 1 ww, revealed, however, differences of only one order of magnitude. In particular all models, except FRUITPATH, under-predicted 134 Cs concentration in fruit by a factor of 0.6–0.2 (SPADE), 0.4–0.6 (RUVFRU), 0.8 (DOSDIM) and 0.3–0.2 (ASTRAL). FRUITPATH predictions were higher than the observed value by a factor of 3.1–5.6. All models, except RUVFRU, predicted a decrease of 134Cs concentration in fruit from the first to the second harvest, by a factor of 2.7 for SPADE to 1.2 for DOSDIM. A possible reason is that RUVFRU can consider a part of the second harvest not yet ripened but already present at the first harvest. This part is not removed at the first harvest. Experimental values show, however, a non significant decrease (factor of 1.1) from first (27.0 ± 2.2 Bq g− 1 wet weight) to second (25.0 ± 1.6 Bq g− 1 wet weight) harvest. Leaf activity for 134Cs at harvest results from the contribution of the various processes of interception, loss, absorption, resuspension and translocation to other plant components, which are considered very differently in the participating models. Fig. 2 shows predicted 134Cs in leaf from all models (from 5.0 · 101 to 6.7 · 102 Bq g− 1 dw) being in the same order of

Bq g-1 dry weight

518

700

Measured

600

DOSDIM SPADE

500 400

FRUITPATH FRUTI-CROM

300

RUVFRU

200 100 0 01/07/1998 Date of harvest

Fig. 2. Measured and predicted after first foliar contamination.

134

Cs activity concentration in leaf

magnitude as the measured values (5.1 ± 0.3) · 102 Bq g− 1 dw. FRUTI-CROM prediction is in a very good agreement with the measured value and this may be explained by the transfer processes considered in the model. This model takes into account interception fraction by leaf component, loss due to growth and pruning, and translocation from external to internal plant. RUVFRU also performed well although it over-predicted the measured value by a factor of 1.3 (Fig. 2). It takes into account interception by leaves and loss as weathering from leaves and transport towards the internal part of the leaves. The inner part of leaves is connected to the surface of leaves and inner part of plant during the whole vegetation period. The interception factor depends basically on the plant growing stage. 134 Cs concentration in leaf was under-predicted by both SPADE (factor of 0.2) and FRUITPATH (factor of 0.4). FRUITPATH considers the external loss of deposit from leaves, whereas SPADE considers both external and internal (i.e. translocation) losses. The lowest value

160

Bq g-1 wet weight

140 120 100 80 60

Measured FRUTI-CROM ASTRAL RUVFRU SPADE DOSDIM FRUITPATH

40 20 0 20/05/98

Date of harvest

2/06/98

Fig. 1. Measured and predicted 134Cs activity concentration in fruit at two harvests after first foliar contamination.

Z. Ould-Dada et al. / Science of the Total Environment 366 (2006) 514–524

was predicted by DOSDIM which is a factor of 0.1 lower than the measured value. DOSDIM does not consider interception fraction by leaves and only considers translocation to fruit. No results are given for ASTRAL for the leaf compartment as the model considers only translocation to the edible organ, as an aggregated transfer factor. Figs. 1 and 2 show that models performed better in predicting 134 Cs in leaf than in fruit, with lower variation in predicted values. The pathway leaf-to-fruit has been recognized to have one of the strongest interactions on the system as a whole, but the knowledge of this pathway is still poor. The translocation of radionuclides within crop plants is the main process that is not yet adequately modelled. 10. Second foliar contamination In this scenario, deposition of 134Cs and 85Sr was applied to strawberry leaves at the stage of fruit ripening. Fruits were then harvested at 2 and 15 days after deposition. There is evidence that fruit contamination is affected to a very variable extent by the activity directly deposited onto its surface and by that translocated to it after absorption into the leaf. Many variables contribute to these processes, but information to evaluate their importance is scarce. Measured values for 134Cs show that the process of loss affects fruit activity (that is reduced to 27% of the intercepted one) in the first 2 days after deposition, but is then overwhelmed by that of leaf-to-fruit translocation, so that fruit activity increases significantly (by a factor 7) in the second harvest. Fig. 3 shows that all models, except SPADE, over-predicted 134Cs concen-

tration in fruit at the first harvest by a factor of 3.6 (FRUTI-CROM), 4.4 (DOSDIM), 9.2 (ASTRAL), 16.8 (RUVFRU) and 38 (FRUITPATH), probably underestimating the process of loss (i.e. the transfer of radioactivity to surrounding environmental compartments or to other parts of the plant). Most models predicted a decrease of 134Cs activity in fruit from the first to the second harvest, with the exception of DOSDIM and SPADE which predicted an increase by a factor of 1.2 and one order of magnitude, respectively. However, for both scenarios, the smaller range of predicted values shows a better agreement between predicted and measured 134Cs concentrations in fruit in the second than in the first harvest (Figs. 1 and 3). Measured values show a loss of 85Sr from fruit following deposition, whose activity is reduced to 20% of the intercepted one in the first 2 days. The process of leaf-to-fruit translocation is lower for 85Sr than for 134Cs, producing a nonsignificant increase (a factor of about 2) in the activity of the second harvest as compared with the first. All models, except SPADE in the first harvest, over-predicted 85Sr concentration in fruit by a factor of 1.8–3.8 (FRUTI-CROM), 2.1–5.6 (DOSDIM), 2.5–5.5 (ASTRAL) and 10.9–21.8 (RUVFRU) (Fig. 4). All models predicted a decrease of 85Sr concentration in fruit from first to second harvest, except for SPADE that, similarly to 134Cs predictions, predicted an increase of one order of magnitude. There is a better agreement between predicted and measured values for the second harvest than for the first harvest. For all models, the range of predicted 85Sr in fruit is smaller than that of predicted 134Cs values, although the trend for 85Sr is very similar to that of 134Cs (Figs. 3 and 4). All models predicted lower concentrations in fruit

100 Measured 80

SPADE

Bq g -1 wet weight

FRUTI-CROM DOSDIM 60

ASTRAL RUVFRU

40

519

FRUITPATH

20

0 20/05/98

2/06/98 Date of harvest

Fig. 3. Measured and predicted 134Cs activity concentration in fruit at two harvests after second foliar contamination.

520

Z. Ould-Dada et al. / Science of the Total Environment 366 (2006) 514–524

Bq g-1 wet weight

40 Measured SPADE FRUTI-CROM ASTRAL DOSDIM RUVFRU

30

20

10

0 20/05/98

Date of harvest

2/06/98

Fig. 4. Measured and predicted 85Sr activity concentration in fruit at two harvests after second foliar contamination.

for 85Sr than for 134Cs, as well as lower concentrations in the second than in the first harvest, except for SPADE, reflecting the very low ability of 85Sr to translocate from leaf-to-fruit. Measured values show a decrease of leaf activity during 57 days of a factor of 1.7 for 134 Cs and 2.4 for 85 Sr. Fig. 5 shows that, as in the first scenario (Fig. 2), all models, except RUVFRU for 134 Cs, underpredicted 134Cs and 85Sr concentrations in leaf. This suggests an overestimation of the loss of deposit from leaves by most modellers. The scenario simulates a growing system where strawberry plants are kept under open tunnels, in order to protect fruit production against sharp temperature ranges during Spring nights. There is evidence that, after cloud deposition, loss from leaves is lower in plants growing under a tunnel than under field conditions (Ertel et al., 1989; 800

Krieger and Burmann, 1969; cited by Pröhl and Hoffman, 1996). The loss of deposit from leaves (i.e. due to weathering) was considered differently by modellers. Some modellers considered this to be less important to reflect the fact that plants were sheltered under a tunnel during the experimental work. Furthermore, in some models the same parameter value was used for 134 Cs and 85Sr transfer from external to internal leaf although different values should have been used. It is known in fact that the rate of penetration through leaf cuticle is higher for Cs+ than for Sr2+ (Swietlik and Faust, 1984). This shows how the interpretation of the test scenario and assumptions used by the modeller can affect model predictions and contribute to the difference between predicted and measured values. Other processes in fruit systems are considered very differently in the

Measured

DOSDIM

SPADE

FRUITPATH

FRUTI-CROM

RUVFRU

700

Bq g-1 dry weight

600 500 400 300 200 100 0 Cs-134

Sr-85

Date of harvest 14/07/1998 Fig. 5. Measured and predicted 134Cs and 85Sr activity concentration in leaf after second foliar contamination.

Z. Ould-Dada et al. / Science of the Total Environment 366 (2006) 514–524

models. For example, SPADE and RUVFRU consider transfer among seven vegetation compartments, while DOSDIM and ASTRAL consider translocation to fruits only. FRUITPATH does not consider these processes but consider fruit as a part of the plant. Some models consider direct deposition to fruit surfaces whereas others only consider deposition to the plant. Given the complexity of the processes considered and the poor knowledge of their relative significance under different scenarios, the values derived from model calculations, even when in good agreement with experimental values, can be the result of a combination of under- and over-predictions of individual synergic or opponent processes, like leaf-to-fruit translocation, fruit accumulation or loss from fruit. To give some examples, 137Cs concentration in fruit after foliar contamination at anthesis (Fig. 1) can emerge from the underestimation of 137Cs interception by fruit and/or the concurrent underestimation of the process of leafto-fruit translocation. Similarly, 137 Cs concentration in fruit at the first harvest after foliar contamination at ripening (Fig. 3) can result from the combination of the overestimation of fruit interception and underestimation of loss from fruit. 11. Soil contamination Fruits were harvested 28 and 41 days after soil contamination carried out at the anthesis stage. The processes of radionuclide migration in the soil, root uptake and translocation from roots to the aboveground part of the plant are among those responsible for the concentration of 134Cs and 85Sr in fruit and leaf. Measured activities in fruit are one order of magnitude higher for 134Cs, (1.4 ± 0.3) · 101 and (2.0 ± 0.6) · 101 Bq g− 1 ww for the first and the second harvest, respectively (Fig. 6), than for 85Sr, (1.1 ± 0.5) and (2.4 ± 1.7) Bq g− 1 ww (Fig. 7).

Bq g-1 wet weight

100

10

1

Predicted values in fruit for 134Cs are one to three orders of magnitude lower than measured values for all models (Fig. 6). Predictions for 85Sr are in the same order of magnitude as measured values for three models: FRUTI-CROM, DOSDIM and ASTRAL (Fig. 7), but are lower for SPADE and RUVFRU by a factor of approximately 3 and 30, respectively . The range of predicted values for 85Sr in fruit is much smaller than that for 134Cs. All models under-predicted 134Cs concentration in leaf by two orders of magnitude, although differences between predictions only vary by a factor of 3 (Fig. 8). Only three models predicted 85Sr in leaves with results showing the same pattern as for 134Cs: predicted 85Sr values were lower than measured values by a factor of 15 to 30 (Fig. 8). The difference between predicted and measured values both in fruit and in leaf is greater following soil contamination than foliar contamination, especially for 134 Cs. These results confirm that plant uptake of radionuclides from soil is an area of large uncertainty in modelling the transfer of radioactivity to fruits. Sensitivity analysis was not carried out as part of the Fruits Working Group activities. However discussions on models and scenarios pointed out some factors responsible for the inconsistencies between measured and predicted results. A first problem encountered was the interpretation of the scenario. Strawberry plants were placed in pots and irrigated daily, which makes root growth different to field conditions and plant uptake higher than under dry conditions (Prister et al., 1993 in Frissel et al., 2002). This resulted in the underestimation of the soilto-plant transfer of radionuclides, more evident for Cs, because its transfer to plants is enhanced on peat soils as compared to mineral soils. For instance, transfer factors of Cs for cereals on peat soils at pH N4.8 for accidental releases range from 0.2 to 2.0 (Bq/kg dry crop to Bq/kg

Measured RUVFRU ASTRAL DOSDIM SPADE FRUTI-CROM FRUITPATH

0.1

0.01

20/05/98

521

Date of harvest

2/06/98

Fig. 6. Measured and predicted 134Cs activity concentration in fruit at two harvests after soil contamination.

522

Z. Ould-Dada et al. / Science of the Total Environment 366 (2006) 514–524

Measured

RUVFRU

SPADE

ASTRAL

DOSDIM

FRUTI-CROM

Bq g-1 wet weight

3

2

1

0 20/5/1998

02/06/1998 Date of harvest

Fig. 7. Measured and predicted 85Sr activity concentration in fruit at two harvests after soil contamination.

dry soil), while corresponding expected values on nonpeat soils range from 0.02 to 0.5 (Frissel et al., 2002). The approaches taken to modelling crop contamination following soil deposition, the parameters used to represent the processes, and their mathematical descriptions differ from model to model. For instance, FRUTICROM uses a value of Fv (Bq/kg fresh weight) of 0.001 for Cs, derived from ECOSYS (Müller and Pröhl, 1993), DOSDIM a value of TF (Bq/kg fresh crop to Bq/kg dry soil) of 0.005 derived from the review (Carini, 2001), but adjusted to lower values considering the spike contamination. ASTRAL uses a TF on an area basis (m2/kg fresh weight) of 2.0 · 10− 5 for clay soils, from a IPSN Report (Calmon and Mourlon, 2001), because no specific value is available for peat. FRUIT1000

PATH considers half-times of 2 years for berries and 5 years for apples (unpublished reference) and, as RUVFRU, uses plant uptake rates from the labile soil compartment. FRUTI-CROM, ASTRAL and DOSDIM use parameters assumed to be calculated at equilibrium and SPADE uses a transfer rate from soil solution. Generally speaking, many of the parameter values used by modellers are derived from equilibrium models in the absence of detailed dynamic data. This increases the uncertainties associated with predictions when calculations, as in this exercise, are made for accidental releases, where the soil to crop transfer process is higher than in equilibrium conditions. To give an example, transfer factor values of Cs for cereals reported in literature range from 0.01 to 0.1 (Bq/kg dry crop)/(Bq/kg

Measured

RUVFRU

FRUITPATH

SPADE

Bq g-1 dry weight

FRUTI-CROM 100

10

1 Cs-134

Sr-85 Date of harvest 1/07/1998

Fig. 8. Measured and predicted 134Cs and 85Sr activity concentration in leaf after soil contamination.

Z. Ould-Dada et al. / Science of the Total Environment 366 (2006) 514–524

soil in the upper 20 cm) in equilibrium conditions, against 0.05 to 0.5 in accidental release conditions (Frissel et al., 2002). 12. Conclusions and recommendations Model predictions of 134 Cs and 85Sr concentration in fruit and leaf of strawberry contaminated via leaves or via soil were tested against experimental results, describing scenarios of foliar contamination at anthesis or at ripening, or soil contamination at anthesis. In general, models performed reasonably well within the constraint of the foliar contamination scenarios. Most models tended to over-predict 134Cs concentration in fruit after foliar contamination at ripening, overestimating direct deposition to fruit or underestimating loss from fruit. Most of the models under-predicted 134Cs concentration in leaf and all models under-predicted that of 85Sr in leaf, probably overestimating the process of weathering, reduced by the fact that plants were sheltered under a tunnel during the experimental work. In the case of soil contamination scenario, the agreement between modelled and measured values was not so good, as models tended to underestimate fruit and leaf activity, especially for 134Cs. This could be due to an erroneous interpretation of the scenario. Strawberry plants were grown in pots which makes root growth and leaching different to field conditions and plant uptake higher than under dry conditions. Furthermore many model parameters, like distribution coefficients and transfer factors, refer to clay or mineral soils, characterised by lower transfer values for caesium than organic soils, like the peat substrate used in the experimental scenario. The tendency for the models to under-predict leaf concentrations for both radionuclides after foliar contamination and fruit concentration for 134 Cs after soil contamination emphasizes the need for a better understanding of the processes that influence radionuclide concentrations in fruit and fruit-bearing plants. Results from the present exercise confirm the need for more data to support models and improve their predictive capability. Model predictions may improve significantly when data become available on various processes (e.g. direct deposition, translocation and root uptake) for a range of radionuclides and fruit types. Experimental studies must be designed focusing on understanding the key processes and producing timedependent data, showing how the distribution varies with time.

523

Sensitivity and uncertainty analyses were not carried out as part of the current FWG work activities, but these would be useful in future work to address the great uncertainty connected with short term processes, such as fruit direct deposition, translocation and phenological stage at time of deposition. More general recommendations made by the FWG participants at the completion of the BIOMASS programme (IAEA, 2003) include model-testing exercises to be carried out using a wide range of contamination scenarios for other radionuclides, other than Cs and Sr, and other fruit types, like orange and olive, that are of nutritional importance and economically significant to Mediterranean countries. It is also a priority to address other climatic regions such as tropical and Asiatic countries. More generally, future activities have to be developed with the aim of producing useful information for the evaluation of dose to humans. Concerning this aspect, a reference crop should be included in the fruit system, both in modelling and, where practicable, in experimental studies, to serve as an analogue where there is a lack of data on fruits, in order to model biosphere processes. Acknowledgements The BIOMASS programme was organised by the International Atomic Energy Agency (IAEA) in Vienna. The authors are very grateful for the support and resources made available by IAEA for meetings and report production. The authors are also grateful for the financial support to the Fruits Working Group by the FSA (formerly Ministry of Agriculture, Fisheries and Food—MAFF) and the Environment Agency for England and Wales. References Calmon P, Mourlon C. Equations et paramètres du logiciel Astral V2. IPSN Report: IPSN/DPRE/SERLAB 01-19 2001 (in French). Carini F. Radionuclide transfer from soil-to-fruit. J Environ Radioact 2001;52:237–79. Carini F, Brambilla M, Ould-Dada Z, Mitchell NG. 134Cs and 85Sr in strawberry plants following wet aerial deposition. J Environ Qual 2003;32:2254–64. Carini F, Green N, Spalla S. Radionuclides in fruit systems: a review of experimental studies. Sci Total Environ (this issue). doi:10.1016/j. scitotenv.2005.05.034. Coughtrey PJ, Thorne MC. Radionuclide distribution and transport in terrestrial and aquatic ecosystems. A Critical Review of Data, vol. 1. Rotterdam: Balkema; 1983. 495 pp. Donelly CE, Carini F. Modelling and environmental study on the transfer of deposited radioactivity to fruit. Mouchel technical note 48100.001-TN1. West Byfleet, UK: Mouchel Consulting Limited; 1998 (April).

524

Z. Ould-Dada et al. / Science of the Total Environment 366 (2006) 514–524

Ertel J, Voigt G, Paretzke HG. Weathering of 134/137Cs following leaf contamination of grass cultures in an outdoor experiment. Radiat Environ Biophys 1989:319–26. Frissel MJ, Deb DL, Fathony M, Lin YM, Mollah AS, Ngo NT, et al. Generic values for soil-to-plant transfer factors of radiocesium. J Environ Radioact 2002;58:113–28. Hungarian Atomic Energy Office. Lessons of Chernobyl accident after 10 years. Budapest; 1996 [in Hungarian]. International Atomic Energy Agency. Handbook of parameter values for the prediction of radionuclide transfer in temperate environments. Technical Report Series, vol. 364. Vienna: IAEA; 1994. 74 pp. International Atomic Energy Agency. Validation of models using fallout data from the Central Bohemia region of the Czech Republic. TecDoc-795. Vienna: IAEA; 1995. International Atomic Energy Agency. Validation of models using fallout data from Southern Finland. TecDoc-904. Vienna: IAEA; 1996. International Atomic Energy Agency. A critical review of experimental, field and modelling information on the transfer of radionuclides to fruit. Working document: BIOMASS/T3FM/WD01. Vienna: IAEA, Division of Radiation and Waste Safety; 1999. International Atomic Energy Agency. Modelling the transfer of radionuclides to fruit. Report of the fruits working group of BIOMASS theme 3. BIOsphere modelling and ASSessment programme. IAEA-BIOMASS-5. Vienna, Austria: IAEA, Division of Radiation and Waste Safety; 2003. Kirton JA, Coughtrey PJ, Mitchell NG. Experimental studies on radionuclide dynamics in soil–plant systems. ANS report no. 5972. Epsom: ANS Consultants Ltd.; 1987. Krieger HL, Burmann FJ. Effective half-times of Sr-85 and Cs-134 for a contaminated pasture. Health Phys 1969;17:811–24. Linkov I, Burmistrov D. Model uncertainty and choices made by modelers: lessons learned from the International Atomic Energy Agency Model Intercomparisons. Risk Anal 2003;23:1335–46. Linkov I, Carini F, Collins C, Eged K, Mitchell NG, Mourlon C. et al. Radionuclides in fruit systems: model–model intercomparison

study. Sci Total Environ, this issue. doi:10.1016/j.scitotenv.2005. 08.002. Mayall A. Farmland: transfer of radionuclides to fruit, National Radiological Protection Board Memorandum NRPB-M545. UK: Chilton; 1995. Mitchell NG. Models for radionuclide transfer to fruits and data requirements. J Environ Radioact 2001;52:291–307. Müller H, Pröhl G. Ecosys-87: a dynamic model for assessing radiological consequences of nuclear accidents. Health Phys 1993;64:232–52. Oncsik BM, Eged K, Kis Z, Kanyár B. Validation study for transport of the 134Cs to strawberry. J Environ Radioact 2002;61(3):319–29. Ould-Dada Z, Carini F, Mitchell NG. Testing of a foodchain model using Chernobyl 137Cs fallout data and considering the effect of countermeasures. Sci Total Environ 2003;301:225–37. Prister BP, Perepelyatnikov GP, Perepelatnikova LV. Countermeasures used in the Ukraine to produce forage and animal food products with radionuclide levels below intervention limits after Chernobyl accident. Sci Total Environ 1993;137:183–98. Pröhl G, Hoffman FO. Radionuclide interception and loss processes in vegetation. In: IAEA, editor. Modelling of radionuclide interception and loss processes in vegetation and of transfer in semi-natural ecosystems. Second report of the VAMP Terrestrial Working Group, IAEA-TECDOC-857. Vienna, Austria: IAEA, Division of Radiation and Waste Safety; 1996 (January). p. 9-47. Swietlik D, Faust M. Foliar nutrition of fruit crops. In: J. Janick Purdue University, editor. Horticultural Reviews, vol. 6. Westport, CT: AVI Publishing Company, Inc.; 1984. p. 299–301. Thorne MC, Coughtrey PJ. Dynamic models for radionuclide transport in soils, plants and domestic animals. In: Coughtrey PJ, Bell JNB, Roberts TM, editors. Ecological aspects of radionuclide release. Oxford: Blackwell Scientific Publications; 1983. p. 127–39. Till JE, Meyer HR. Radiological assessment—a textbook on environmental dose analysis. NUREG/CR-3332, ORNL-5968. Washington: Nuclear Regulatory Commission; 1983.