Scarabaeidae larvae are neglected greenhouse gas sources in soils

A precise knowledge of the sink and source distributions of greenhouse gases (GHG) in regional and global carbon and nitrogen budgets, and of the processes governing them, is a necessary prerequisite for the development and assessment of climate change adaptation and mitigation strategies1-3. Certain soil-inhabiting Arthropoda groups are known producers of GHG, namely methane (CH4), but apart from termites, their emissions have never been studied in the field and quantified at different scales4,5. Here we report the first field GHG emission data of soil-dwelling Scarabaeidae larvae, focusing on pest insects in a temperate climate region (Melolontha melolontha and M. hippocastani). Variations in larval biomass explained variations in larval field CO2 and CH4 emissions well at the individual and site level. This correlation disappeared after transferring larvae from the field to a laboratory setting. We show that GHG emissions of soil-inhabiting Scarabaeidae larvae are comparable to those from termites, thus questioning the neglect of Scarabaeidae larvae in GHG flux research, and we demonstrate the importance of field-based emission estimates for soil biota.


Introduction
A major challenge in climate change research is to fully understand and accurately quantify interannual to decadal variabilities in atmospheric GHG concentrations driven by natural processes 1,6 .
Soils can act as both, important sinks and sources of atmospheric carbon dioxide (CO2), CH4, and nitrous oxide (N2O), but are also components in the global carbon and nitrogen budget with large uncertainty estimates [1][2][3] . In recent years, it has been proposed to reduce these uncertainties by shifting from an implicit to an explicit representation of soil biota in ecosystem and, ultimately, Earth system models 7,8,9 . However, a major constraint for developing such new biogeochemical models is the lack of field data, especially for soil biota groups other than microorganisms 8,9 . Soil biota can substantially influence the spatial and temporal variability of GHG sinks and sources in the field, but the majority of our current knowledge comes from laboratory experiments, is often controversial and has been limited to only a few regions and species. As a result, the magnitude of the effect of soil biota on the GHG sink and source capacity of soils remains poorly quantified. For example, termites are the only soilinhabiting Arthropoda group, which has been explicitly considered as significant GHG source on a global and regional level thus far 4,10 . They are assumed to contribute about 1 -3 % to the global annual CH4 budget, but the large variation in emission estimates in the literature (0.9 to 150 Tg CH4 y −1 ) underlines the uncertainty in the available data sets 2,4,11 .
We conducted the first GHG field measurement study on soil-inhabiting larvae of the Scarabaeidae family. Throughout the vegetation period of 2017, larvae of Melolontha melolontha (Common cockchafer) and M. hippocastani (Forest cockchafer) were excavated at randomly chosen plots (0.25 m² each, excavation depth ~50 cm) at six sites (Site 1 -Site 6) in west-central and southern Germany covering different larval developmental stages and activity levels, and vegetation types.
Greenhouse gas emissions of individual larvae were measured directly in the field together with larval biomass and various environmental variables. One site was selected for a comparison of field-and laboratory-measured GHG emissions. For this, larvae were randomly separated into three groups. For one group, larval GHG emissions were measured directly in the field. The other two groups were transferred to the laboratory, where the larvae were kept individually in soil-filled containers with ample food supplies for 7 and 18 days, respectively, before GHG emission measurements were carried out (Methods). Larval biomass was the main variable used to describe inter-and intra-site GHG emission variability, and to subsequently model total larval plot-level CO2 and CH4 emissions using linear regression analysis on the pooled field dataset. Upscaled plot-level emission and biomass estimates represented the cumulated individual measurement values per plot standardized to 1 m². To derive an annual European CO2 and CH4 emission estimate, the range of plot-level emissions was multiplied by the available literature value on area colonized by Melolontha spp. in Europe 12 .
Individual larval N2O emission were not upscaled, due to their infrequent occurrence (Methods).

Results
Gaseous carbon emissions of individual Melolontha spp. larvae showed a large inter-and intra-site variability which could not be explained by differences in soil temperature (range: 11.4 -29.3 °C) and soil moisture (range: 3.2 -32.7 vol%). There was a clear tendency for emissions to increase with larval biomass at the site level, especially for CO2 (Fig. 1). Average larval biomass ranged between 0.5 and 2.2 g larva -1 . When pooling all data regardless of site and species, there was a strong positive correlation between CO2 and CH4 emissions (rs = 0.76, p<0.001), and between larval biomass and CO2 emissions (rs = 0.84, p<0.001). The correlation between larval biomass and CH4 emissions was less pronounced, but still significant (rs = 0.68, p<0.001). The excavation depth of the larvae correlated negatively with larval biomass (rs = -0.51, p<0.001), CO2 emissions (rs = -0.58, p<0.001), and CH4 emissions (rs = -0.48, p<0.001), respectively (Supplementary information S1). It could be seen as an indicator for larval access to fresh plant root material, which was higher the closer the larvae were to the soil surface, i.e. the lower the excavation depth was. Two-thirds of the larvae were found at 0 -15 cm soil depth. Nitrous oxide emissions were only occasionally observed. Out of the 64 field larvae tested for N2O (sites 2 -6), 13 individuals emitted significant amounts, ranging between 1.3 and 90.4 ng N2O h -1 larva -1 .
Since Scarabaeidae larvae need to reach a certain biomass to be able to pupate and since their food intake increases with size, larval biomass was a good proxy for larval age, sampling time, and food availability. Larval biomass could also be used to differentiate between species and to encode larval abundances at the plot scale. Larval abundances ranged between 4 and 68 larvae m -2 (Supplementary information S2). The correlation between gaseous carbon emissions and larval biomass persisted when upscaling emissions to the plot scale, and a large proportion of the inter-and intra-site emission variability could be explained by variations in total larval biomass (Fig. 2). Across all sites, CO2 emissions increased on average by 0.51±0.03 mg CO2 h -1 m -2 with each g total larval biomass increase (p<0.001). The relationship between CH4 emissions and total larval biomass was best fitted with a A comparison of field and laboratory measurements from larvae excavated at Site 3 revealed a strong impact of laboratory conditions on CO2 and CH4 emissions. Despite no significant differences in larval biomass between the three measured groups (p=0.12) and ample food supply, overall emission strength and variability decreased rapidly with prolonged time at the laboratory (Fig. 3). In contrast to the field observations, no significant correlation between larval biomass and CO2 and CH4 emissions was found, respectively, after two and a half weeks in the laboratory. N2O emissions tended to be lower under laboratory conditions in comparison to field measurements as well; however, it was not possible to discern a statistically significant effect of the laboratory conditions on larval N2O emissions. Of the 65 larvae incubated in the laboratory, only 13 emitted N2O, with emissions ranging between 1.81 and 43.70 ng N2O h -1 larva -1 (Supplementary information S2).

Discussion
There are no field studies on direct CO2 and CH4 emissions from soil-inhabiting Scarabaeidae larvae yet to which we can compare our data 4 , but both the field-and laboratory-measured emission rates fall within the range of emission rates known from the few available laboratory studies on Scarabaeidae larvae and other soil Arthropoda groups in temperate regions 5,13,14 , or field and laboratory studies on termites in temperate, subtropical and tropical regions 4,15,16 . However, our study demonstrates how careful we have to be in interpreting GHG emission rates derived from laboratory studies. It has been known for termites that emission rates can decline over the course of a laboratory experiment 17 . In addition, we show that such a trend can also coincide with a considerable reduction of the emission rate variability between individual larvae and a disappearance of correlations between emission rates and environmental variables in comparison to field measurements.
Large variations in emission rates and the use of larval biomass for emission rate upscaling are features well known from termite studies 10,16,18,19 . Our biomass-based European CO2 and CH4 emission estimates for Melolontha spp. larvae are two orders of magnitude lower than the corresponding emission estimates for termites in temperate regions 18 . Termites are considered as a significant, but quite small global GHG source with the majority of these emissions stemming from subtropical and tropical regions 4,10,20 . Greenhouse gas emissions of soil Arthropoda groups other than termites are seen as too low to significantly affect regional budgets 4,5 , which our study seems to confirm at the first glance. However, in contrast to many termite studies, we did not attempt to use the emission rates of a few species to infer the emissions for this entire Arthropoda group. Our estimates are strictly for the genus Melolontha only, and thus, show only a fraction of the potential GHG emissions of European Scarabaeidae larvae. Worldwide, larvae of several Scarabaeidae species are regarded as economically important pest insects 21,22 . Regionally, these pest insects can reach biomass levels comparable to or considerably higher than those used for upscaling termite GHG emissions 23 , but we have no estimates of the total biomass of soil-dwelling Scarabaeidae larvae in Europe. Furthermore, for CH4 it is important to differentiate between gross and net soil fluxes. Purely biomass-based CH4 emission rates like ours represent gross soil CH4 fluxes, as they do not account for simultaneously occurring gross CH4 consumption in soils 11,18,20 . Recent studies suggests that CH4 emissions of soil-inhabiting Scarabaeidae larvae can stimulate soil CH4 consumption, and thus, potentially increase the overall net CH4 sink capacity of soils 24,25 . For regional and global CH4 budgets, it might therefore be more important to quantify the effect of soil faunal CH4 emissions on the net soil CH4 flux, instead of just quantifying total soil faunal CH4 emissions.
Regarding N2O, the emission rates measured in this study were of the same magnitude as those observed from earthworms 26 . Earthworms are the only faunal group for which a considerable amount of literature on soil N2O emissions is available 27 . In their presence, emissions can increase by more than 40 % due to the activation of nitrate-and nitrite-reducing bacteria during earthworm gut passage 28 . It is unclear if Scarabaeidae larvae are capable to affect soil N2O emissions in a similar manner as our data base it too inconsistent and no other data are available yet.
Overall, our data show that Scarabaeidae larvae should not be neglected as sources of CO2, CH4, and N2O in soil GHG flux research. However, to assess the impact of Scarabeidae larvae on regional and global GHG budgets and to better understand seasonal and interannual variations in GHG emissions, including the possibility of increased CH4 consumption in soils, it is mandatory to gather more field data on emission rates and species-dependent spatial larval biomass distributions and activities. These are exactly the same challenges, which are known from termite GHG emission research 18 , but we have to address these challenges if we want to explicitly include these soil faunal groups in future ecosystem and Earth system models.

Methods
Sampling sites and species. This study was conducted in central and southern Germany -a temperate climate region with average annual air temperatures between 8 and 12 °C and average annual precipitation between 600 and 1000 mm (reference period 1961 -1990) 29  Gas emission measurements. Immediately following soil excavation, the larvae were individually placed in 110 ml glass tubes, which were sealed air-tight with butyl rubber stoppers. The larvae were incubated in the field for about an hour and a blank was included at each plot. No soil particles adhered to the larvae's skin, but larvae could defecate during incubation. At the end of the incubation period, 25 ml of air were extracted from each glass tube with a plastic syringe and transferred to evacuated 12 ml glass vials sealed with grey chlorobutyl rubber septa (Labco Exetainers 839W, Labco Limited, Lampeter, United Kingdom), while control vacutainers were filled with ambient air during field incubations. At site 3, in addition to 18 larvae incubated directly in the field, 65 larvae were excavated and transferred to the laboratory, instead of being field-incubated. These larvae were kept individually in small plastic containers filled with ~250 ml soil from the excavation site, and were supplied with ample amounts of fresh grass roots and carrot slices as food sources. Storage temperature was 18 °C and soils were sprayed with tab water once per week to keep them moist. After 7 days, 39 of these larvae were incubated in the laboratory (incubation temperature 24 °C) following the same protocol as in the field. After another 11 days, the remaining 26 larvae were incubated for gas sample collection. Gas samples were analysed with a SRI 8610C gas chromatograph with autosampler (SRI Instruments Europe GmbH, Bad Honnef, Germany) equipped with a flame ionisation detector (FID) coupled to a methanizer for CO2 and CH4 measurements, and an electron capture detector (ECD) for N2O measurements. Each detector was equipped with a Porapak Q pre- A test for correlations between paired samples was performed on the entire pooled larval field dataset using Spearman's rho (rs) statistic. Input variables were instar, larval excavation depth, larval weight (= biomass), larval abundance at the respective plot, individual CO2, CH4, and N2O emissions, air temperature, soil surface temperature and moisture, as well as soil temperature and moisture at the bottom of the respective excavated plot. For the comparison of larval field-and laboratory-measured emissions from Site 3, the same test statistic was also applied to the three flux data subsets (0, 7 and 18 days after excavation). The test was performed separately on each data subset with larval weight and larval CO2, CH4 and N2O emissions as input variables. Across-group comparisons on the data subsets were only carried out on larval weight to check for significant differences in mean biomass (Kruskal-Wallis rank sum test).
Larval CO2 and CH4 emissions were scaled up in two steps: from individual larvae to the plot level and from plot level to European level. Total larval emissions and larval biomass per plot were calculated by summing up the individual emissions or biomass and multiplying by four to scale to 1 m². Larval biomass was subsequently used as independent variable in linear regression analysis for modelling m²-level larval CO2 and CH4 emissions. The obtained plot level larval CO2 and CH4 emissions were upscaled for 6 months per year (excluding larval winter rest) with the available literature data on European land area colonised by Melolontha spp. to derive a first rough annual CO2 and CH4 emission range estimate for Europe. A more precise upscaling to the European level is not yet possible, due to lack of field data. It needs to be considered that the available literature values on colonised land area are likely very conservative 34 . Due to their scattered occurrence, N2O emissions were not upscaled.
All test statistics and regression analysis were performed with the software R (version 3.4.3) 35 .
In addition to the software's standard library, the function 'chart.Correlation' (package: PerformanceAnalytics) 36 was used. and "Lab2" were kept 7 and 18 days in the laboratory, respectively, before emissions were measured. The boxplots for "Field" are identical to the boxplots for site 3 in Fig. 1. Numbers above the CO2 and CH4 emissions box plots are Spearman correlation coefficients between the respective gas emissions and the batch's larval weight.