Geophysical Research Letters

Inter- and intra-continental transport of radioactive cesium released by boreal forest fires

Authors


Abstract

[1] A high-precision radionuclide monitoring site was established in Yellowknife/Canada in 2003. Far away from nuclear activities, regular signals of 137Cs were found there during the summers of 2003 and 2004. We show that these signals can be explained by transport from fires burning in the boreal forests of North America and Asia. This finding has important implications. It demonstrates that 137Cs deposited world-wide from past nuclear testing is re-injected into the atmosphere by combustion to a significant extent and on a large scale, and is subsequently transported across great distances. Besides this, the analysis shows how efficiently a new receptor-oriented atmospheric transport modeling technique can be used to check whether a 3D emission inventory is consistent with discrete point measurements.

1. Introduction

[2] During the period of atmospheric nuclear testing between 1945 and 1980 nearly 1 EBq (1018 Bq) of cesium (137Cs) was deposited globally, with a maximum in the latitude band between 40° and 50° N [United Nations Scientific Committee on the Effects of Atomic Radiation (UNSCEAR), 2000a]. Due to the long half-life of 137Cs (30 years), there is still on the order of 2 kBq m−2 in the Northern Hemisphere mid-latitudes. Another 0.1 EBq of 137Cs was released during the Chernobyl accident in 1986 [United Nations Scientific Committee on the Effects of Atomic Radiation (UNSCEAR), 2000b; OECD Nuclear Energy Agency, 2002]. The strong and regular emissions during the atmospheric testing period combined with the upholding effect of the stratosphere caused a hemispheric 137Cs airborne background that reached its maximum in the mid-1960s and returned to zero by the mid-1980s, reflecting that the 137Cs has by then been removed and stored in reservoirs of the biosphere and lithosphere. In areas of the northern hemisphere not particularly influenced by Chernobyl, detection of airborne 137Cs has occurred only infrequently during the last decade.

[3] In 2003, a radionuclide (RN) monitoring station was established in Yellowknife/Canada (114.48° W, 62.45° N) as part of the International Monitoring System (IMS). The RN network is built by the Provisional Technical Secretariat (PTS) of the Comprehensive Nuclear-Test-Ban Treaty Organization (CTBTO) [see Hoffmann et al., 2000] to detect nuclides from nuclear explosions. All PTS RN sites are equipped with high-volume aerosol samplers and high-resolution germanium gamma spectrometers [Schulze et al., 2000]. They yield a minimum detectable concentration (MDC) for 137Cs on the order of 1 μBqm−3. This is about one order of magnitude more sensitive than typical national measurement networks [e.g., Beak Consultants Limited, 1995] and allows sampling durations of as short as 24 hours. The Yellowknife station commenced sending data to the PTS in January 2003. Starting in May 2003, 137Cs activity was regularly found in samples. This persisted until mid-September. In 2004, a detection episode started end of June and again stretched until mid-September. Between these two periods, only one detection occurred. The observed concentrations varied between 1 and about 15 μBqm−3, with the highest values occurring in 2004 (Figure 1). The PTS operates two other remote sites at high northern latitudes, in Iceland and in Spitsbergen. Both stations, in contrast to Yellowknife, are rarely affected by 137Cs. Therefore, the summertime 137Cs signals detected in Yellowknife were subject to further investigation.

Figure 1.

137Cs activity concentrations detected at the radionuclide monitoring station Yellowknife (Canada) between April 2003 and September 2004.

2. Forest Fire Situation and Cesium Resuspension Emission Estimate

[4] The year 2003 was among the worst with respect to forest fires in southern and eastern Russia. An estimated area of 20 million ha (2 · 1011 m2) of forest land burnt starting in May and continuing through the summer. Country reports for Russia are available on the web page of the Global Fire Monitoring Centre (http://www.fire.uni-freiburg.de). Large fires consuming about 9 million ha were reported from regions around Lake Baikal [Goldammer et al., 2004]. On the other hand, the forest fire season 2003 in Alaska and Canada was below average, with most of the burning activity in regions downwind of Yellowknife (Manitoba, Ontario). In 2004, the situation reversed. In Alaska, 2.6 million ha forest burned, making this season the worst on record. Large fire activity started by mid-June and stopped in September. An additional 1.8 million ha burned in the neighboring Yukon Territory. Burning reports for Alaska are available at the web page of the National Interagency Fire Centre (http://www.nifc.gov/index.html), for Canada at the page of the Canadian Forest Service (http://www.nrcan-rncan.gc.ca/cfs-scf). The total burning in Siberia, on the other hand, was on the order of 5 million ha and thus only 25% of the 2003 value. The regional differences in burning activity are well reflected in the monthly fire pixel count numbers for July 2003 and July 2004 (see Figure 2) as obtained with the MODIS instrument onboard the TERRA satellite [Giglio, 2005].

Figure 2.

Integrated (May to September) fire pixel counts based on data from the MODIS instrument onboard the TERRA satellite for the boreal region in (a and b) Asia and (c and d) North America in 2003 (Figures 2a and 2c) and 2004 (Figures 2b and 2d). The images show the intense burning in Asia in 2003 as well as the exceptional fire activity in Alaska and Yukon Territory in 2004.

[5] For modeling purposes, a gridded, hemispheric-scale, emission inventory of 137Cs due to fire re-suspension is needed. Such an inventory, however, is not available. Hence, to estimate the spatial and temporal distribution of forest fires, monthly 1° × 1° MODIS/TERRA fire pixel counts [see Giglio, 2005] were employed. When these counts were integrated over the areas of North America (170°W ≤ λ ≤ 60°W; 50°N ≤ ϕ ≤ 90°N) and Russia (50°E ≤ λ ≤ 180°E; 50°N ≤ ϕ ≤ 90°N) for the whole season, it turned out that they correlate extremely well with the independent burning reports as quoted above (r2 = 0.995; see Table 1). Hence, we assume a linear relationship between fire pixel count and area burned at each grid cell within the two regions. We furthermore assume a linear relationship between area burned and the re-suspension emission term of 137Cs. Doing so, the fire pixel counts can be directly translated into a gridded emission inventory. As a first-guess estimate, we assume that 5 · 106 Bq of 137Cs was released per ha area burned in both regions, which would translate into 25% of all 137Cs available [UNSCEAR, 2000a]. This first guess can later be refined based on the observations.

Table 1. Modis/Terra Total Fire Pixel Counts (May–September) for the Years 2003 and 2004 for the Boreal Forest Regions of Asia and North Americaa
 Fire Pixel CountLinear Est. of Area Burned, 106 haFirst-Guess 137Cs Em. Est., 1012 BqBest-Guess 137Cs Em. Est., 1012 Bq
  • a

    Related to that, the linear-regression based estimated area burning and the estimated 137Cs emissions (first-guess estimate, best-guess estimate) are shown.

Boreal Asia 200310303719.999.64.0
Boreal North America 2003106291.78.30.3
Boreal Asia 2004314665.828.91.2
Boreal North America 2004268634.924.31.0

3. Model Simulation of Cesium Transport

[6] To test our 137Cs emission hypothesis, an atmospheric transport modeling study was conducted. This study was based on the receptor-oriented Source Receptor Sensitivity (SRS) concept as introduced by Wotawa et al. [2003], Stohl et al. [2003] and Seibert and Frank [2004]. To define it, let us consider the activity concentration c of one specific radionuclide measured within one sample taken during 24 hours at a radionuclide station. Then, c [Bqm−3] can be expressed as sum of the products of a spatio-temporal source field S [Bq] and a corresponding SRS field M [m−3] at discrete locations (i, j) and time intervals n as follows:

equation image

The PTS routinely computes the SRS fields pertaining to all RN samples by applying the transport model FLEXPART [Stohl et al., 1998, 2005] in backward (adjoint) mode [see Wotawa et al., 2003] using analyzed global meteorological data provided by the European Centre for Medium-Range Weather Forecasts (ECMWF) with a resolution of 1° × 1° and three hours. To account for elevated release heights of the fires [FIRESCAN Science Team, 1996; Wotawa and Trainer, 2000; Fromm and Servranckx, 2003], all SRS fields for Yellowknife in 2003 and 2004 were recalculated assuming emission heights equally distributed between surface and 5000 m. Based on these fields, equation (1) can be applied on each daily measurement, together with the 137Cs emission inventory, to compute the 137Cs concentrations that would result from the underlying emission assumption. This is separately done for both emission regions (North America, Siberia), yielding two concentration predictions. To account for the fact that the satellite-based fire hot spots counts are only updated every month, we create monthly averaged 137Cs concentration sets for the subsequent analysis.

[7] We now use the model predictions together with the measurements to check the validity of the satellite-based emission estimate, and to refine the first-guess emission factor for linear errors. Therefore, the following multiple linear regression model is formulated:

equation image

with C being the monthly averaged observed 137Cs concentrations, Cmod,NAM the (averaged) model results pertaining to North America, Cmod,ASIA the model results for Siberia, and C0, a and b the parameters to be specified by the regression analysis. In case of a tight and significant relationship, these parameters can finally be used to correct the first-guess emission estimate.

4. Results

[8] In total, the two model predictions driven by our 137Cs emission hypothesis explain remarkably well the monthly-averaged observed 137Cs concentrations in Yellowknife from May to September 2003 and 2004 (see Figure 3, top). The multiple linear regression analysis yields a correlation of r2 = 0.89, and the model results are a significant predictor on the 99.9% confidence level. Our hypothesis thus explains about 90% of the variance of the observed 137Cs signals. 84% of the total variability in northern Canada is explained by re-suspension emissions in North America, and about 5% by emissions in Russia. Siberian burning, however, was according to the model the dominating factor for Yellowknife in early summer 2003. As far as the parameters specified by the analysis are concerned, the parameter C0 is about 0.4 μBqm−3 and corresponds well with the measurement system MDC. The multiplication factors a and b for the model predictions related to emissions in North America and Siberia are 0.035 ± 0.005 and 0.044 ± 0.023, respectively. The resulting best-fit linear model well explains the measurements at the station (Figure 3, bottom). Taking into account the statistical significance of the parameters at the 95% confidence level, the best-guess emission term for 137Cs re-suspension lies between 1.2 and 2.3 ·105 Bq per ha forest burned in North America (p-value 0.0001, highly significant), and between 0 and 5 · 105 Bq per ha forest burned in Siberia (p-value = 0.1, not significant). This indicates that re-suspension emission factors in both regions are about the same. Taking only into account the significant relationship for North America, our best estimate would be about 2 · 105 Bq/ha.

Figure 3.

(top) Comparison between monthly-averaged measured 137Cs activity concentrations (μBqm−3) and model simulations fed with an 137Cs emission inventory based on monthly fire pixel counts for the boreal regions of North America and Asia and a first-guess emission estimate (5 106 Bq/ha). (bottom) Comparison between the monthly-averaged measured 137Cs concentrations and the best-fit multiple linear regression model fed with the model simulations for North America and Asia. For the parameters of the fit refer to the text.

5. Conclusions

[9] It was clear from previous studies that 137Cs with a boiling point of about 670°C [Amiro et al., 1996] is re-suspended by forest fires. These studies, however, have been localized to the general vicinity of specific forest fire episodes [e.g., Melin and Wallberg, 1991; Paliouris et al., 1995; Kashparov et al., 2000]. Our analysis demonstrates, for the first time, that the whole boreal burning activity re-suspends and redistributes about 1% of all 137Cs totally available within the burnt region, or about 10% of the 137Cs fraction presumably available in the biomass (assuming that the biomass stores 10% of the total 137Cs, see Melin and Wallberg [1991]). This underpins the important role of the boreal forest as a reservoir of this nuclide and as quasi-continuous source, and extends our knowledge of the environmental circulation and transport of radio-cesium on a hemispheric-scale and in a longer-term context. For the first time, we are now able to estimate a large-scale average 137Cs re-suspension emission factor from burning, which is about 2 · 105 Bq/ha. During the two years under investigation, the typical 137Cs release from boreal burning was on the order of TBq per season (Table 1). This means that low-level 137Cs signals may be found at a number of monitoring sites to be established under the CTBT. Last but not least, the analysis demonstrates that the receptor-oriented modeling concept as developed by the PTS can be used for a variety of applications related to data from measurement networks, including the validation and improvement of emission inventories.

Acknowledgments

[10] As input for the atmospheric transport modelling, data from the European Centre for Medium-Range Weather Forecasts (ECMWF) were utilized. We furthermore wish to acknowledge provision of data from U.S. NCEP (meteorological analyses) and from NASA (MODIS TERRA fire pixel data) through their public web services. The views expressed in this publication are those of the authors and do not neccesarily reflect the views of the CTBTO Preparatory Commission.

Ancillary