An Exploration on Greenhouse Gas and Ammonia Production by Insect Species Suitable for Animal or Human Consumption

Background Greenhouse gas (GHG) production, as a cause of climate change, is considered as one of the biggest problems society is currently facing. The livestock sector is one of the large contributors of anthropogenic GHG emissions. Also, large amounts of ammonia (NH3), leading to soil nitrification and acidification, are produced by livestock. Therefore other sources of animal protein, like edible insects, are currently being considered. Methodology/Principal Findings An experiment was conducted to quantify production of carbon dioxide (CO2) and average daily gain (ADG) as a measure of feed conversion efficiency, and to quantify the production of the greenhouse gases methane (CH4) and nitrous oxide (N2O) as well as NH3 by five insect species of which the first three are considered edible: Tenebrio molitor, Acheta domesticus, Locusta migratoria, Pachnoda marginata, and Blaptica dubia. Large differences were found among the species regarding their production of CO2 and GHGs. The insects in this study had a higher relative growth rate and emitted comparable or lower amounts of GHG than described in literature for pigs and much lower amounts of GHG than cattle. The same was true for CO2 production per kg of metabolic weight and per kg of mass gain. Furthermore, also the production of NH3 by insects was lower than for conventional livestock. Conclusions/Significance This study therefore indicates that insects could serve as a more environmentally friendly alternative for the production of animal protein with respect to GHG and NH3 emissions. The results of this study can be used as basic information to compare the production of insects with conventional livestock by means of a life cycle analysis.


Introduction
Production of greenhouse gasses (GHG) is considered as an important cause of climate change [1,2]. The most important GHGs are carbon dioxide (CO 2 ), methane (CH 4 ) and nitrous oxide (N 2 O). Since the end of the 18 th century the atmospheric carbon-dioxide concentration has increased by 30% and CH 4 concentrations by 50% [3]. CH 4 and N 2 O have considerably greater global warming potentials (GWPs) than CO 2 . By assigning CO 2 a value of 1 GWP, the warming potentials of these other gases can be expressed on a CO 2 -equivalent basis: CH 4 has a GWP of 25, and N 2 O has a GWP of 298 [1]. The relative contribution of CO 2 equivalents (CO 2 eq.) of the livestock sector is large, amounting up to 18% of total anthropogenic GHG emissions [2]. Based on a Life Cycle Analysis (LCA) that takes the entire production process of animal products into account, the global contribution to GHG emissions by the animal sector are: 9% for CO 2 (fertilizer production for feed crops, on-farm energy expenditures, feed transport, animal product processing, animal transport, and land use changes), 35-40% for CH 4 (enteric fermentation in ruminants and from farm animal manure) and 65% for N 2 O (farm manure and urine) [2]. Direct CO 2 production through respiration is not relevant when determining the impact of GHGs as respiration by livestock is not considered a net source of CO 2 [2]. The respired carbon, which comes from the feed, was first taken up from CO 2 in the air and stored in an organic compound during the production of the feed. However, the ratio between body growth realised and CO 2 production is an indicator of feed conversion efficiency and thereby a relevant indicator for the environmental impact [4].
Livestock is also associated with environmental pollution due to ammonia (NH 3 ) emissions from manure and urine, leading to nitrification and acidification of soil [5]. Although not considered a GHG, NH 3 can indirectly contribute to N 2 O emission [2], as conversion takes place by specialized soil bacteria [6]. Livestock is estimated to be responsible for 64% of all anthropogenic NH 3 emissions [2]. The main source of gaseous NH 3 is bacterial fermentation of uric acid in poultry manure [7,8] and bacterial fermentation of urea in mammals [9]. Besides these environmental problems the livestock sector faces challenges regarding resistance to antibiotics, zoonosis and animal welfare [10].
All these problems together illustrate the need to find alternatives for conventional sources of animal protein. Minilivestock, for instance edible insects, have been suggested as an alternative source of animal protein [11]. Production of animal protein in the form of edible insects supposedly has a lower environmental impact than conventional livestock [12,13,14]. When evaluating the total environmental impact of animal protein production, a LCA, in which all production factors are taken into account, is needed. Differences in environmental impact in a LCA can be explained mainly by three factors: enteric CH 4 emissions, feed conversion efficiencies and reproduction rates [4].
Before performing a LCA, it is necessary to know the GHG production by edible insects. This information is lacking in literature. Therefore, in this study we experimentally quantified the direct production of the GHGs CH 4 and N 2 O for five insect species. CO 2 production and average daily gain (ADG) were quantified to provide an estimation of feed conversion efficiency. Additionally, NH 3 emissions were quantified. The results of this study represent a quantification of the insect physiological contribution to GHG production by insects and can in turn be used to create a LCA for insect-derived products.

Animals and housing
Five insect species were studied: fifth larval stage mealworms Tenebrio molitor L. (Coleoptera: Tenebrionidae), fifth and sixth nymphal stage house crickets Acheta domesticus (L.) (Orthoptera: Gryllidae), third and fourth stage nymphs of migratory locusts Locusta migratoria (L.) (Orthoptera: Acrididae), third larval stage sun beetles Pachnoda marginata Drury (Coleoptera; Scarabaeidae) and a mix of all stages of the Argentinean cockroach Blaptica dubia (Serville) (Dictyoptera: Blaberidae). Currently, T. molitor, A. domesticus and L. migratoria are considered edible, while P. marginata and B. dubia are not. The latter two species were included since they are a potential source of animal protein, for instance by means of protein extraction. These two species can be bred in large numbers with little time investment and are able to utilise a wide range of substrates as feed [15,16].
Per species three to six repetitions were conducted each for a period of three days. Animals were housed per species in two cages or containers per respiration chamber. These containers were placed in one of two, identical, open circuit climate respiration chambers measuring 80*50*45 cm, with a total volume of 265 L [17]. Within these climate respiration chambers, T. molitor and P. marginata were housed in two stacked plastic containers (50*30*8.7 cm). The three other species were housed in metal wire cages (45*37.5*41 cm; mesh width 1 mm) with a glass cover plate. To increase surface area for A. domesticus and B. dubia, hollow plastic tubes (20 cm long and 3 cm in diameter), were stacked to a height of 30 cm in the wired cages, while for L. migratoria, two Vshaped-folded metal screens (70*15 cm) were entered per cage. Humidity, temperature, and day length were based on rearing conditions used by commercial insect rearing companies ( Table 1). All animal masses reported are averages of fresh mass per cage. The starting and final animal mass per cage are provided in Table 1.

Diet
Food was provided for each species at the beginning of each repetition, except when mentioned otherwise.
Tenebrio molitor larvae were reared in 300 g mixed grain substrate (wheat, wheat bran, oats, soy, rye and corn, supplemented with beer yeast) with on top pieces of carrot (615*2 cm) weighing a total average of 637 g per repetition.
Acheta domesticus was provided with chicken mash (501 g) with carrot pieces (784 g) on top for each repetition.
Locusta migratoria was provided with wheat bran (70 g; Arie Blok Animal Nutrition, Woerden, The Netherlands) in a metal bowl at the beginning of each repetition. Fresh Perennial ryegrass (Lolium perenne) was provided daily (463 g in three days). The grass was grown by Unifarm, Wageningen University and Research centre, Wageningen, The Netherlands.
P. marginata larvae were kept in a peat moss substrate (2.0 kg per respiration chamber) in which chicken mash (285 g) was mixed at the beginning of each three-day repetition. Pieces of carrot (615*2 cm) with an average total mass of 161 g per repetition were put on top of the substrate.
B. dubia was provided with a chicken mash diet (199 g) and carrots (559 g), fresh carrot being added during the repetitions.
Peat moss, chicken mash, and carrots, offered to A. domesticus, P. marginata and B. dubia were provided by Kreca V.O.F, Ermelo, The Netherlands. The carrots and mixed grains substrate offered to T. molitor were provided by Insectra, Deurne, The Netherlands.

Gas measurements
During the experiment concentrations of CO 2 and CH 4 were measured every 9 min in the ingoing and outgoing air stream of the respiration chambers. The difference in CO 2 and CH 4 concentrations between ingoing and outgoing air thus represents the total production of CO 2 and CH 4 of insects, feed, and substrate. The exact air volumes were measured with a calibrated Schlumberger G1.6 dry gas meter and corrected for measured air temperature and pressure. CO 2 and CH 4 concentrations were measured in dried gas. Gas was dried in a +2uC dew-point cooler.
Nondispersive infrared analyzers were used to measure CO 2 (type Uras 3G, Hartmann and Braun, Frankfurt, Germany) and CH 4 (type Uras 10E, Hartmann and Braun, Frankfurt, Germany). The refreshed air volume was set so that CO 2 levels did not exceed 1%. From each climate respiration chamber, as well as from the incoming air, an air sample was taken for N 2 O analysis after 24, 48, and 72 h with a 60 ml syringe. The syringes were sealed by a shutoff valve and stored at 20uC until analysis (within 48 h). The N 2 O concentration was analysed by a gas chromatograph (CE instruments GC8000 Top, Interscience, Breda, The Netherlands) using a Haysep Q 80-100 mesh 2 m61/80 SS column, at a constant temperature of 60uC. N 2 O was detected with an electron capture detector (ECD). Injection volume was 5.0 ml in a fixed loop. NH 3 concentrations in the climate respiration chambers were determined twice daily (at 12.00 and 24.00 h) by means of a gas detection tube system (Kitagawa, type AP-20; Komyo rikagaku kogyo, Tokyo, Japan; type 105 NH 3 gas detector tubes with a range of 1-20 ppm).

Calculations
Production of N 2 O was calculated by subtracting the N 2 O concentration from the incoming air from that in the outgoing air. These differences were then used in a formula adapted from Wheeler et al (2003) [18]: where VV = ventilation volume of air in a specified time period. The average concentration difference of the three samples taken during the three-day period was used to determine the average N 2 O production in a repetition.
The formula used by Wheeler (2003) was also used for the calculation of NH 3 production. A molecular mass of 17 was used and instead of a difference in concentration, the measured concentration was used, leading to a slight overestimation of the actual NH 3 production (between 0 and 0.1 mg/kg BM/day). CO 2 equivalents were calculated by adding the multiplications of the produced amounts of CH 4 and N 2 O with their global warming potential; 25 for CH 4 , and 298 for N 2 O [1].
Mean body mass was calculated by averaging the body mass at the start of the experiment and the body mass at the end of the experiment. Average daily gain (ADG) was calculated as follows: (((End mass -Start mass)/Start mass)/3)*100%, in which 3 is the number of days the experiment was running.
The ratio between CO 2 production per unit biomass per day and ADG gives an indication of the feed conversion efficiency, in which higher values indicate lower efficiencies.
To determine CO 2 production from feed and substrate, all feeds were independently tested in the same respiration chambers, without the animals. A linear time course of consumption was assumed and CO 2 production was recalculated to kg of live insect.

Statistics
The N 2 O and NH 3 assay data were subjected to a two-way analysis of variance (ANOVA) with species and time of sampling (24, 48, or 72 h) as fixed factors to determine whether the time of sampling had an effect. No significant effect of the time of sampling was found for N 2 O (Pillai's trace: F = 1.467, P = 0.199). Therefore, the average of the three samples taken during the 3-day trial period was used to determine the change per repetition and to calculate total production. However, NH 3 production was significantly affected by the time of sampling (day or night; Pillai's trace: F = 4.065, P = 0.019) and the day of the repetition (first, second or third; Pillai's trace: F = 17.170, P,0.001). CO 2 and CH 4 production for all five species were analyzed by means of a one way analysis of variance (ANOVA) followed by a Tukey post hoc test. Statistical analysis of all data was done by means of SPSS 15.0.

Results
Production of CO 2 is expressed per kilogram of mean live body mass (BM) per day (24 hours) and per kilogram of mass gain (Table 2) and the average daily gain (ADG) is reported ( Table 2). Production of CH 4 , N 2 O, CO 2 equivalents, and NH 3 , are expressed per kilogram of mean live body mass (BM) per day (Table 3) and per kilogram of mass gain (Table 4).

ADG and CO 2 production
ADG varied between 4.0% (P. marginata) and 19.6% (L. migratoria) with the three other species having an ADG of 6-7%. CO 2 production among the five insect species differed significantly and ranged from 19 (B. dubia) to 110 (L. migratoria) g per kg BM/ day. Also, the CO 2 production per kg of metabolic weight (i.e. the weight of metabolically active body tissue) differed greatly between species (Table 5). CO 2 production expressed per kg of mass gain was intermediary for L. migratoria due to the high ADG. Still, the CO 2 production of L. migratoria per kg of mass gain was more than double the production of CO 2 by B. dubia. Pachnoda marginata had the highest production of CO 2 per kg of mass gain (1,539 g/kg), which was more than double the amount of L. migratoria.

CH 4
Production of methane was detected for P. marginata and B. dubia, but not for the three other species. Pachnoda marginata produced more than three times as much CH 4 per kg of mass gain than B. dubia (4.9 vs 1.4 g). This difference was caused by a higher production of CH 4 per kg BM (0.16 g vs 0.08 g) and a lower ADG (4.0% vs 6.1%).

N 2 O
N 2 O was produced only in significant amounts by T. molitor and L. migratoria (1.5 and 8.0 mg/kg BM/day, respectively). Production of N 2 O by L. migratoria per kg BM was more than 5-fold the production by T. molitor, this difference decreased to almost 2.5fold when expressed per kg of mass gain, due to a much higher ADG of L. migratoria.

Discussion
Insects, being poikilotherms, do not use their metabolism to maintain a body temperature within narrow ranges, contrary to homeothermic animals. This is expected to result in higher feed conversion efficiencies. CO 2 production related to growth, has an inverse relationship with feed conversion efficiency in a given situation. CO 2 production by insects depends on the species, stage of development [19,20], temperature [21], feeding status [22], and on activity level [23,24]. A production of 37 g CO 2 /kg BM/day was reported for Anabrus simplex (Orthoptera, Tettigoniidae), 40 g CO 2 /kg BM/day for the locust Schistocerca americana (Orthoptera; Acrididae) [25] and 94 g/kg BM/day for adult Tribolium castaneum (Coleoptera; Tenebrionidae) [26]. All five species in the current Table 3. CH 4 , N 2 O, CO 2 eq. and NH 3 production (average 6 standard deviation) per kilogram of bodymass per day for five insect species, pigs and beef cattle.  study had a fairly high production of CO 2 . This might to a large extent be explained by ad libitum feeding during the experiment that has been reported to increase oxygen consumption fivefold [22]. Reported CO 2 production for inactive, unfed, Tenebrionid adults ranged between 5.4-13.3 g/kg BM/day [27], which is 5-10 times lower than observed for T. molitor in this experiment. This can partially be explained by the locomotory activities of T. molitor larvae in this experiment [37]. Furthermore, growing larvae are expected to have a higher CO 2 production than adults. The range of CO 2 production for T. molitor is comparable to the factorial metabolic scope reported for tiger beetles (Cicindela spp: Coleoptera; Cicindelidae) of 6.1-16.5 [28]. Size differences in animals account for a difference in metabolic rate, and thereby CO 2 production. The relation between metabolic rate (B) and body mass (M) was described by Kleiber [29] as B = aM b , in which a is a constant and b = 0.75. The value of b has been much debated since [30,31,32]. For poikilotherms values between 0.67 and 1.0 have been reported and a comparison of several arthropod species suggested b approximates 0.82 [33,34]. The value chosen for b has a large impact on the metabolic weight and thereby the calculated CO 2 production (Table 5). Applying b = 0.75 for pigs and beef cattle and b = 0.82 for insects, resulted in a lower CO 2 production based on metabolic weight for the studied insect species (Table 5). For L. migratoria CO 2 production was only slightly lower than for beef cattle, however, for the other four species production was between 18% and 54% of that for beef cattle and between 11% and 34% of the CO 2 production of pigs.
The CO 2 production per kg BM of insect species investigated in this study was higher than for pigs or cattle (Table 3). This concurs with Prothero et al. (1979) [35], who reported a higher oxygen consumption per kg of BM for insects than for mammals, assuming the respiratory quotient (CO 2 production/O 2 consumption) has similar values (0.7-1.0) for both animal groups. However, the CO 2 production per kg of mass gain for the five insect species in the current study (337-1,539 g/kg) was either 39% (minimum values) or 129% (maximum values) when compared with pigs (865-1,194 g/kg) and much lower (12%-54% respectively) than cattle (2,835 g/kg). Therefore, CO 2 production per kg of mass gain suggests higher feed conversion efficiencies for insects than for mammalian livestock. These results concur with those of other authors [13,14,36,37].
A similar trend was visible for ADG; the ADG for the five insect species studied was 4.0-19.6%, the minimum value of this range being close to the 3.2% reported for pigs, whereas the maximum value was 6 times higher. Compared to cattle (0.3%), insect ADG values were much higher. In general, the rate of ADG depends, amongst others, on life phase. Therefore, where available, literature data on growing animals were used. The fundamental biological differences in growth and development processes between pigs and cattle and the studied insects impeded further synchronization. CH 4 production for the species studied was in agreement with Hackstein and Stumm (1994) [38]; for insects, only representatives of cockroaches, termites, and scarab beetles produce CH 4 . This originates from bacterial fermentation by methanobacteriaceae in the hindgut [39].
We found large variability for the N 2 O emission rates. Earlier studies in laying hens using a similar method for determining N 2 O production, concluded that production was either negligible or undetectable [7,40]. However, other authors [41,42] determined a production of 28 mg N 2 O/kg BM/day and 52 mg N 2 O/kg BM/ day, respectively, indicating the difficulty of accurately determining N 2 O production [43].
In earlier studies respiration of feed was considered to have a negligible effect on utilisation of dry mass as determined gravimetrically [44] and therefore on CO 2 production. Later studies suggested that respiration by plant leaves can be an important source of error in the calculation of insect feed intake using gravimetric methods [45] and can cause major errors in energy budget studies of plant-feeding insects [46]. Our reported CO 2 production includes the respiration of the feed ( Table 6). The extremely high contribution to total CO 2 production by the substrate of P. marginata (92.5%) was most likely due to large amounts of fungal biomass observed in the mixed feed and substrate when insects were absent in the experiments aimed to obtain correction values for CO 2 -production by the substrate. No fungal growth was apparent during the experiments on feeding P. marginata larvae, suggesting that the contribution of the substrate to total respiration during the experiment was much lower. We conclude that the interaction between actively feeding P. marginata larvae and the substrate suppressed fungal growth through either consumption by the beetle larvae [47] of fungal biomass or through unknown chemical or combined chemical/mechanical mechanisms. Such interactions hinder the application of realistic corrections for the contribution of feed and substrate to the total CO 2 production and thus to quantify the CO 2 production arising from insect metabolism separately.
For all other species the relative contribution of the feed to total CO 2 production was minor, varying between 1.3% and 3.6%. Although feed respiration did have an impact on production of CO 2 , still the production of CO 2 is much higher for L. migratoria than for the other insect species. A likely explanation for this higher production of CO 2 is the 7uC higher temperature L. migratoria was kept at, as a difference of 10uC is expected to double CO 2 production. Furthermore, the comparatively high ADG of L. migratoria is expected to result in higher production of CO 2. In one of the repetitions for A. domesticus, a lower ADG and increased mortality were observed. Excluding this repetition, the emission of CO 2 per kg BM decreased slightly (68 vs 71 g/kg), but the emission of CO 2 per kg mass gain changed considerably (918 vs 1468 g/kg). This difference can for a large part be explained by a decrease in ADG (from 9.0 to 7.2%). Acheta domesticus did not produce CH 4 , but N 2 O production doubled (from 0.1 to 0.2 mg/ kg BM; 1.9 vs 5.3 mg/kg mass gain). The production of CO 2 eq. also increased (0.04 vs 0.05 g CO 2 eq. /kg BM and 0.57 vs 1.57 g/kg mass gain). It is well possible that the higher N 2 O production measured was caused by saprophytic bacteria utilising the dead A. domesticus and producing N 2 O [6]. Although we included this repetition in the results, it is not clear whether this represents the practical situation best. Large differences in NH 3 emission have been reported for conventional livestock. Pigs for example emit 4.8-75 mg/kg BM/ day [48,49,50], poultry 72-436 mg/kg BM/day [41,49,51] and cattle 14-170 mg/kg BM/day [49,52,53]. Several factors influence NH 3 emission, such as temperature, relative humidity, food type, moisture content, pH, wind speed, housing type, and substrate [54,55].
In the current experiment, a clear NH 3 emission pattern was found; higher amounts of NH 3 were emitted during daytime for A. domesticus, L. migratoria and B. dubia, than during nighttime. Daynight rhythms for NH 3 excretion have been documented for pigs [5] and are strongly correlated with activity levels [56]. Quantitatively the differences between day and night emission levels are small; 7-10% with a maximum difference of 25% [5]. In our study this relative difference was approximately 33%. In all cases NH 3 emission levels were higher during the daytime than during the night-time. For L. migratoria this is the active period, for the nocturnal B. dubia and A. domesticus it is not, indicating that a different, unknown variable might influence NH 3 emission patterns in these insects. NH 3 concentrations in the outgoing air, and consequently calculated NH 3 emission, increased from day one to day three in B. dubia (1.57 to 4.29 mg/kg BM/day) and A. domesticus (2.46 to 8.01 mg/kg BM/day). This could indicate that NH 3 emissions might be underestimated due to the relatively short time frame of our experiments. For L. migratoria NH 3 emission did not increase between day 1 and day 3 (5.57 and 5.05 mg/kg BM/day), suggesting that NH 3 production was stable. This might be caused by the faeces of this species that, contrary to those of B. dubia or A. domesticus, dry quickly after defecation.
We conclude that P. marginata and T. molitor probably did not emit NH 3 . Poultry deep litter systems [57] have higher NH 3 emission rates than battery systems [55], which is explained by the presence of substrate.
The presence of substrates for P. marginata and T. molitor in this study corresponded with lower NH 3 emissions. A possible explanation is that gas exchange in the container is inhibited by the substrate and therefore less emission of NH 3 was measured. However, it could also be that these species produce less NH 3 .
All insect species in this study produced much lower amounts of NH 3 (3.0 to 5.4 mg/kg BM/day for A. domesticus, L. migratoria and B. dubia) than conventional livestock (4.8-75 mg/kg BM/day for pigs and 14-170 mg/kg BM/day for cattle). Further research is needed to determine for which insect species and to what extent NH 3 emissions increase further when a longer time frame is used.

Conclusions
To the authors' knowledge, the study presented here is the first to report on both GHG and NH 3 emissions of edible insect species. An evaluation of the GHG emissions of edible insect species is most relevant when based on CO 2 eq. per kg of mass gain. In that way a comparison of the selected species with each other and with conventional livestock is based on a cost-benefit principle, in which the GHG production (environmental cost) is directly linked to food production (benefit). GHG emission of four of the five insect species studied was much lower than documented for pigs when expressed per kg of mass gain and only around 1% of the GHG emission for ruminants.
The measured NH 3 emission levels of all insect species in this experiment were lower than reported NH 3 emission levels for conventional livestock.
The ADG of all insect species in this study was higher than for conventional livestock, while CO 2 production expressed as g/kg mass gain was comparable or lower, which indicates higher feed conversion efficiencies for insects.
This study therefore indicates that insects could serve as a more environmentally friendly alternative for the production of animal protein from the perspective of GHG and NH 3 emissions. A complete lifecycle analysis for species of edible insects is lacking at this point in time [58] and should be the focus point of further studies to allow a conclusive evaluation of the sustainability of insects as a protein-rich food source. The data presented in this study are indispensable for conducting a lifecycle analysis for edible insects.