Reduced soil fauna decomposition in a high background radiation area

Decomposition of litter and organic matter is a very important soil ecosystem function where soil fauna play an important role. Knowledge of the responses in decomposition and soil fauna to different stressors is therefore crucial. However, the extent to which radioactivity may affect soil fauna is not so well known. There are some results showing effects on soil fauna at uranium mines and near Chernobyl from relatively high levels of anthropogenic radionuclides. We hypothesize that naturally occurring radionuclides affect soil fauna and thus litter decomposition, which will covary with radionuclide levels when accounting for important soil parameters. We have therefore used standardised litterbags with two different mesh sizes filled with birch leaves (Betula pubescens) to assess litter decomposition in an area with enhanced levels of naturally occurring radionuclides in the thorium (232Th) and uranium (238U) decay chains while controlling for variation in important soil parameters like pH, organic matter content, moisture and large grain size. We show that decomposition rate is higher in litterbags with large mesh size compared to litterbags with a fine mesh size that excludes soil fauna. We also find that litter dried at room temperature is decomposed at a faster rate than litter dried in oven (60⁰C). This was surprising given the associated denaturation of proteins and anticipated increased nutritional level but may be explained by the increased stiffness of oven-dried litter. This result is important since different studies often use either oven-dried or room temperature-dried litter. Taking the above into account, we explore statistical models to show large and expected effects of soil parameters but also significant effects on litter decomposition of the naturally occurring radionuclide levels. We use the ERICA tool to estimate total dose rate per coarse litterbag for four different model organisms, and in subsequent different statistical models we identify that the model including the dose rates of a small tube-shape is the best statistical model. In another statistical model including soil parameters and radionuclide distributions, 226Ra (or uranium precursory radionuclides) explain variation in litter decomposition while 228Ra (and precursors) do not. This may hint to chemical toxicity effects of uranium. However, when combining this model with the best model, the resulting simplified model is equal to the tube-shape dose-rate model. There is thus a need for more research on how naturally occurring radionuclides affect soil fauna, but the study at hand show the importance of an ecosystem approach and the ecosystem parameter soil decomposition.

Yes -all data are fully available without restriction

112
To perform a field experiment within Mining hill, we prepared 100 litterbags. Half of these were 113 made from two layers fine-meshed (0.1 mm) and the other half from coarse-meshed (2 mm) nylon 114 Litterbag studies have shown that different quality types of litter with different N:C ratios are 119 decomposed differently [28]. We used standardised amounts (d.w) of newly naturally shed leaves 120 from birch (Betula pubescens), which is present at each of the experimental localities. This birch 121 species has C:N ratios ranging from 20 to 33 across Europe [46], which is the C:N ratio range that is 122 most affected by soil fauna decomposition effects [23]. Freshly abscised leaves were sampled in an 123 area with normal background radioactivity during litter-fall on September 28 from a regularly 124 maintained lawn. Low levels of radionuclides associated with the sampled leaves were verified for 125 two subsamples by gamma spectrometry (HPGe). 126 Other studies of litter decomposition have used litter dried at either at room temperature or by 127 higher temperatures in oven. We hypothesize that heat treatment will involve different nutrient 128 availability, which will increase decomposition rate. To assess this, we dried half of the leaves on 129 newspapers at room temperature for five days, while the other half was dried in aluminium trays 130 within paper bags in an oven at 60 °C for 48 hours. This is a commonly applied temperature which 131 involves maximum water removal but also denaturation of one third of leaf proteins [47]. We wanted 132 to assess whether oven-dried litter is decomposed / lost from litterbags faster than litter dried at 133 room temperature. The two forms of dried litter were therefore applied in the same amount in a 134 balanced experimental design. Afterwards, the fine mesh-sized litterbags were filled with 7.99-8.03 135 grams of dried litter (mean: 8.00, SD: 0.01), and the coarse-sized litterbags were filled with from 7.77 136 to 8.00 grams of dried litter (mean:7.86, SD:0.05, Fig. 2). The small but significant (t=22, p<<0.01) 137 difference in litter fill was due to loss of small fragments through the meshes during filling. 138

139
Within Mining hill, five localities were established through gamma spectrometry to represent a 140 gradient of soil radionuclides in the 232 Th decay chain. Each locality consisted of a 5x5 meter square 141 within a pine stand of approximately the same height and density. Twenty large litterbags were 142 placed at each locality, two by two into quartets of four litterbags, with five quartets per locality at 143 the centre and corners of the locality square, except at one location (Loc26). At Loc26, one quartet 144 was placed at another location within the square and another quartet was placed right outside the 145 square, and four litterbags in one of the corners were placed four by one (next to each other) due to 146 rocks and very uneven substrate. The pattern of litterbags across localities were situated according 147 to litterbag treatment to achieve a balanced random experimental design. Each quartet consisted of 148 two fine-meshed and two coarse-meshed, each of which contained one litterbag with air-dried and 149 one with oven-dried litter. Fine and coarse litterbags always lay side by side, but elsewise placements 150 within each quartet was random. Within quartets, distances were the same, with diagonal litterbags 151 (centres) separated by 0.4 meters and adjacent ones (centres) separated by 0.2 meters. Within 152 localities, the quartet centres were separated by 1.4 to 5.7 meters (mean: 3.6, SD:1.1). Within the 153 whole study area, the centres of localities were separated by 35 to 340 meters (mean: 180, SD:115). 154 Litterbags positions therefore represents three different spatial scale levels. 155 All litterbags were deployed exactly a year from October 17, which correspond with the period when 156 litterfall ends at this altitude and latitude. After deployment the litter samples were dried in oven at 157 105 ⁰ C for 24 h and the difference in dry litter weight before and after deployment was used to 158 calculate the amount (grams) of litter loss per litter bag. Since variation among litterbags in initial 159 litter mass was very small and deployment period was equal, the mass loss was used directly in 160 analyses rather than lost fraction of litter per litterbag. 161

162
To be able to account for spatial heterogeneity, soil core (Ø=5.4 cm with an entrance of 4.8 cm) 163 samples were taken beneath each of the coarse litterbags (0-6 cm) per litterbag quartet. Soil samples 164 varied in wet weight from 36.4 to 138.5 g (mean:86.2, SD:26.9). Each soil sample was dried at 105 ⁰C 165 for 24 hours, and the mass difference between wet and dry weight of each sample was used to 166 assess its fraction of soil moisture. After drying, samples were sieved (2 mm) and soil particles larger 167 than 2 mm were retained to calculate their fraction of the samples total mass. From the sieved soil 168 (<2 mm), subsamples were taken for assessment of pH, organic matter content and activity 169 concentrations of radionuclides. Soil pH was measured using an inoLab 7110 pH meter XXX () localities. In addition, soil samples were analysed for 137 Cs (at energy 661.65 KeV). These HPGe 181 detectors have relative efficiencies of 23% to 50% and cover energies from 20 keV to 3000 keV. They 182 are placed in a low background laboratory and are regularly controlled against a traceable source. 183 Each analysed sample was placed within a circular plastic measurement geometry (36.4 mL), and 184 results corrected for decay since sampling. Within each geometry, the density of each sample was 185 calculated as mass (kg d.w. soil) per volume (36 mL). Finally, organic matter (OM) content was 186 measured by loss on ignition in an oven, using 5 hours to reach and 12 hours at 550 ⁰C, as the 187 difference in mass before and after ignition compared to the dry weight before ignition. 188

189
The ERICA tool [48] was used for dosimetry, modelling organisms in a terrestrial ecosystem with 190 100% occupancy below ground with the measured soil samples of 226 Ra, 228 Ra and Cs137 as input. 191 The measurements of these radionuclides were used as proxy for all the modelled radionuclides,

268
The soil samples from Mining hill contained fragments larger than 2 mm (sieved off) in a fraction that 269 ranged from less than 0.01 to 0.44 (mean: 0.11, median: 0.08, SD: 0.10), but with some variation 270 among the localities (Fig. 3). Soil pH (measured in buffer) ranged from 3.3 to 5.5 (mean: 4.8, median: 271 4.9, SD: 0.5), with especially one locality standing out (Fig. 4). The fraction of organic matter (OM) 272 content (after drying and sieving) ranged from 0.10 to 0.69 (mean: 0.30, median: 0.32, SD: 0.11), also 273 with substantial variation among localities (Fig. 5). The soil moisture fraction ranged from 0.29 to 274 0.55 (mean: 0.38, SD: 0.07), involving similar variation among localities (Fig. 6) (Fig. 7). Levels of 226 Ra ranged from 24 to 120 Bq kg -1 soil dry weight (mean: 45, median: 32, 281 SD: 24), with some variation among localities and especially much within locations 26, 30 and 32 (Fig.  282   8). Levels of 210 Pb (n=20) were for all except one locality higher than 226 Ra levels (Fig. 9), and for 283 samples with no analysed 210 Pb, the location-wise ratio of 210 Pb to 226 Ra was used to estimate 210 Pb 284 from the analysed 226 Ra value. 285 The log10-transformed 228 Ra levels were weakly correlated to log10-transformed levels of 226 Ra 286 (r=0.41, p<0.01), negatively weakly correlated to log10-transformed OM (r=-0.36, p<0.02) but not to 287 log10-transformed soil pH (r=-0.13, p>0.4). The log10-transformed 226 Ra was also weakly negatively 288 correlated to log10-transformed OM (r=-0.29, p<0.05) but neither to log10-transformed soil pH 289 (r=0.09, p>0.5). Log10 transformed soil particle fraction >2mm was negatively weakly correlated to 290 log10 transformed 226 Ra (r=-0.31, p<0.05) and almost to 228 Ra (r=-0.26, p=0.07). Soil levels of 137 Cs 291 ranged from 15 to 120 Bq kg -1 soil dry weight (mean: 35, SD: 19), also with considerable variation 292 among localities but at very low levels (Fig. 10). 293 Regarding spatial autocorrelation, all soil parameters had significant but at varying strengths 294 correlations to the physical distances between pairs of soil samples, while litter loss itself did not 295 (Table 1). This included a weak negative spatial autocorrelation for log10 of soil pH and a weak 296 positive spatial autocorrelation for log10 of both soil 228 Ra and soil 226 Ra, while no spatial 297 autocorrelation at all for litter loss (only coarse litterbags). The pattern is also clear among the three 298 levels of spatial scales, as the means and standard deviations of soil 228 Ra pairwise absolute 299 differences increase by doubling and tripling, respectively, with spatial scale (from within quartets, to 300 within locations and within study area, Table 2). By comparison, the mean and standard deviation of 301 pairwise absolute difference of soil 226 Ra also increased from within quartets to within localities but 302 not from within localities to within the study area, expressing spatial autocorrelation but a less clear 303 gradient among localities. The 137 Cs absolute differences between soil samples have similar means 304 and standard deviations between the three scale levels ( Table 2). 305

306
External dose rates estimated for the default annelid with ERICA ranged from 0.7 to 6.8 μGy h -1 307 (mean: 3.1, SD: 1.8) with clear differences between the localities (Fig. 11), and were very similar in 308 the other three modelled organisms, differing maximally by 0.016 μGy h -1 for any radionuclide or on 309 average only 0.001 μGy h -1 across radionuclides. Including internal dose rates, the total dose rates 310 varied between the four modelled organisms and ranged from 1.8 to 17 μGy h -1 (mean: 8.2, SD: 4.6) 311 for the annelid, from 1.4 to 13 μGy h -1 (mean: 6.1, SD: 3.4) for the detrivore arthropod, from 5.4 to 32 312 μGy h -1 (mean: 16, SD: 8.0) for the small tube shape and from 3.8 to 20 μGy h -1 (mean: 10, SD: 4.9) for 313 the box shape. The pattern across localities was relatively similar for the four organisms ( Fig. 12) but 314 with higher levels and more variation in the two additional shapes. Among radionuclides, the main 315 contributions to dose rate came from the radionuclides in the 232 Th chain and from 226 Ra (Table 3). 316 Differences between the four organisms in the radium isotopes are due to the different applied CR's. 317

318
Within Mining hill, large litter bags had lost from 2.5 to 6.1 grams (d.w.) litter (mean: 3.6, SD: 0.7) 319 after one year of deployment during the field experiment. There were clear differences in litter loss 320 between coarse and fine meshed litterbags (t=8, p<<0.01), as well as between litterbags with litter 321 dried in oven and room temperature (t=3, p<0.01, Fig. 13 with air-dried litter with either coarse or fine mesh differed with from 0.1 to 2.6 grams of litter loss 330 (d.w., loss mean: 1.1, SD: 0.7). Differences between litterbag litter loss was thus significant within 331 quartets (relating to mesh size and drying regime) but larger among quartets and localities. 332 For the linear statistical model of coarse meshed litterbag loss including radionuclide distributions, 333 the radium isotopes were chosen since these contribute most to dose rates and covary with those 334 other radionuclides that also contribute significantly (  (Table 3), 338 an alternative model including only 228 Ra was therefore also subsequently explored through AIC. For 339 the four linear statistical models of litter loss including the estimated dose rates of organisms types 340 (adj R 2 =0.33-0.40, F(36, 7)=4.5-6.0, p<0.002), non-significant terms included moist (t=1.2-1.3, 341 p>0.21) for all models while size particle fraction > 2mm (log10 transformed) was non-significant (t=-342 1.6, p>0.11) for the two default organism models, nearly significant for the small tube-shape (t=-2.0, 343 p<0.052) and significant for the box shape (t=-2.2, p<0.04). After model simplification, comparison of 344 the six explored models show that their ability to explain variation in loss from litterbags (log10 345 transformed) is relatively similar but that the two best models are the one including the dose rate of 346 the tubes-shaped organism and the model including the 226 Ra distribution (Table 4). For all six 347 models, the effect of all significant parameters was negative and relatively similar. Among parameter 348 estimates the effect is clearly larger for the soil parameters organic matter fraction (OM) and pH 349 (both log10 transformed) but somewhat less for log10 226 Ra and especially the dose rate (Table 5). By 350 comparison, the parameter estimate of log10 228 Ra was even lower (β=-0.07 SE=0.04). Combining the 351 two best models into one model yielded the non-significant terms log10 226 Ra (t=-1.6.0, p<0.128), 352 log10pH (t=-1.6.0, p<0.122), log10tubeshapeDR (t=-1.7.0, p<0.103) and log10>2mm (t=-2.0, p<0.056), 353 and ended after model simplification again with the best dose rate model (tubeshape DR). 354

355
As expected, there was a much larger loss of litter from litterbags with a coarse mesh size (2 mm) 356 than from litterbags with a fine mesh size (0.1 mm, Figure 2) The most important soil parameters were assessed to avoid confounding factors and showed like 375 radionuclide distributions different variation and gradients across locations and litterbags (Figures 3-376 10), involving an ideal opportunity to separate any effects in a statistical analysis. Within the area 377 bedrock is redrock, which is a carbonatite, probably involving less variation in soil parameters than if 378 more bedrock were present. This field experiment can thus with its limited spatial extent be viewed 379 as a common garden experiment, where the design with systematised different distances between 380 litterbags allowed assessment and verification of spatial autocorrelation in the most important 381 parameters. The carbonatite bedrock also explains the relatively high soil pH. Also, the negative 382 effect of soil pH on soil fauna is well known [64]. The fraction of organic matter (OM) observed in the 383 study area is relatively high and probably does not per se involve a negative effect on soil fauna. The 384 negative relationship between litter loss and OM in statistical models demonstrate a higher level of 385 OM beneath litterbags with the lowest degrees of decomposition. This is probably due to a lower 386 abundance or activity of soil fauna, as organic matter will build up when decomposition is low. 387 Finally, and most tantalising, when taking into the account the covariation of important soil 388 parameters, there was a clear statistical effect of the radionuclide levels in this area. 389 However, the results of the statistical model for radionuclide distribution were surprising given their 390 activity concentrations in the study area and the associated dose rates (Table 3). It was surprising 391 that 226 Ra was a highly significant term while 228 Ra was not. This strongly hints to other effects on 392 litter decomposition than just dose rates. The sizes of the effect of terms on litter loss also support 393 this notion, with the 226 Ra term having an effect around five times larger than both the dose rate 394 terms and the 228 Ra term in the statistical models. However, when the two best models were 395 combined and model simplification performed, the subsequent model was the same as the best dose 396 rate model, highlighting an effect of radionuclide dose rate on soil fauna in the area. Interestingly, 397 removal of the 226 Ra term resulted in the pH term becoming significant, suggesting these terms 398 explaining similar parts of variation in litter loss. 399 A potential explanation for the significance of the term on activity concentration of 226 Ra but not 400 228 Ra, may be that radionuclide toxicity also is important to soil fauna. Within the study area the 401 activity concentrations of 228 Ra are from 20 to 30 times greater than for 226 Ra, while the half-life of 402 226 Ra (1600 y) is 280 times longer than for 228 Ra (5.7). By comparison, progenitor radionuclides of 403 226 Ra, uranium isotopes 234 U and 238 U, have half-lives of 250 ky and 4.5e9. The number of radionuclide 404 atoms present and thus toxicity of these radionuclides with very long half-lives may therefore be 405 important. A recent epidemiological study suggest a higher chemical than radiological risk from 226 Ra 406 [65], but it is also well known that uranium has a pronounced chemical toxicity that is more acute 407 than its radiotoxicity occurring radionuclides on variation in litter decomposition but also hints at effects of radionuclide 442 toxicity. In the study at Chernobyl that failed to identify any effects of anthropogenic radionuclides, 443 137 Cs and 90 Sr were the dominant ones and assessed dose rates ranged from less than 0.5 to 75 μGy 444 h -1 [34], which is well below up to four times as high dose rates as the range in the study at hand. 445 However, stable strontium is not considered toxic [72], and stable caesium at environmental levels is 446 neither toxic to animals, even though both stable and radiocaesium may be somewhat chemically 447 toxic at higher concentrations due to its similarity to kalium [73,74]. Radioceasium is however 448 probably much less chemically toxic than uranium and radium, and in Chernobyl it is probably mainly 449 dose rates that are affecting soil fauna. 450 Regardless, the study at hand has identified effects on decomposition rate shown by how common 451 soil parameters but also the occurrence of naturally occurring radionuclides explain variation in litter 452      Table 5 Parameter estimates (β±SE) with the two best simplified linear models of log10 litter 509 loss, with log10 transformation of fraction of organic matter (OM), pH, fraction of soil 510 particle size>2mm (>2mm), 226 Ra and dose rate of organism (tube-shape DR), but not 511 for litter drying regime (dried_oven).