Our literature search followed the PRISMA-Ecoevo guideline (O'Dea et al. 2021) (Fig. 1). We focused on AM fungal effect on N2O emission from both natural and managed ecosystems, including meadow steppe, paddy field, grassland, forest, farmland and bioretention systems, etc. We ran a topic search in Web of Science Core collection database (WOS, http://www.webofscience.com/) and China National Knowledge Infrastructure (CNKI, http://www.cnki.net/) Professional search database via subscription of the Nanjing Tech University on 4 February 2021. The available timespan of databases of WOS and CNKI was from 1978 to the time of our research. We updated our datasets on 28 June 2021 after we collected newly published articles. Our search terms were: ‘“Arbuscular Mycorrhiz*” AND (ammonia OR NH OR “nitrous oxide” OR N2O OR NO gas OR emission OR loss OR efflux OR flux*) NOT review’.
Our criteria for selecting articles were: original research articles (i.e. not reviews or comments); reporting N2O or NH3 fluxes or denitrification rates which indicated by N2O production or potential; involving an AM fungal treatment and a corresponding control while there were no other differences between these two groups. Finally, we selected 20 publications and 1 unpublished study from our lab, including 16 articles and 5 theses (more details can be found in Note S1).
2.4. Preliminary investigation
We only had 20 publications and 1 unpublished study (this is a relatively small number for a meta-analysis) meeting our criteria so that we tried to cover the broadest possible data to make our analysis robust. We chose four moderators, publication language, gas form, N2O indicator type, and data type (i.e. flux or cumulation) to investigate if and how these factors affect our analysis.
As we were not sure whether AM effect differ in articles published in Chinese and English language, we conducted a preliminary investigation on the effect of publication language on AM fungal effect, with its two levels being Chinese and English.
As NH3 was reported to be converted into N2O (Gon &Bleeker 2005) so that we collected data related to AM effect on NH3 volatilization. Whether AM effect on NH3 volatilization was different from N2O emission, however, was unknown. Then we set up gas form with its levels being N2O and NH3.
Soil N2O emission was estimated by two indicators: N2O flux and N2O potential. N2O flux was estimated with gas samples collected directly from the original experimental setup. Soil N2O potential was investigated with gas samples collected from incubators where nitrate and/or organic carbon were added. We investigated the possible impact of N2O indicator type on soil N2O emission and its two levels were N2O flux and N2O potential.
Soil N2O emission data were present in different forms: in some articles, N2O flux was present while in others, only the cumulated N2O emission (calculated by multiplying N2O flux with time) was available. To test whether this divergence impact on our conclusion, we considered data type and the two levels of this variable were flux and cumulation.
2.5. Moderator variables
To investigate how biotic and abiotic variables modulate AM effect on N2O, focused on five groups of variables: experiment-, plant-, AMF-, nutrient-, and soil-related variables.
We set three experiment-related variables: scale, intervention and measurement times.
Scale had two levels: lab and field. Studies with experiments taking place under controlled environmental conditions, such as greenhouses, were grouped into lab. Those were carried out in the field conditions were denoted field.
Intervention had three levels: inoculation, control and mutation. Inoculation denoted trials where plants were inoculated or not with AM fungal inoculum to setup AM treatment and control. At control level, the native AM fungi were killed by fungicide or there was no access for AM fungal hyphae. Mutation indicated studies where AM defective mutants were used to setup the control.
Measurement times had two levels: one and multiple. One denoted studies where soil N2O emission was measured once while multiple indicated that soil N2O emission was measured multiple times. In cases where both individual soil N2O emission at some time points and accumulated results during a given period were provided, only accumulated results were extracted.
We defined four plant-related variables to make clear in which case our results apply. They are plant diversity, plant type, plant life cycle and carbon fixation.
Plant diversity had two levels: one and more. Experiments conducted with one plant species were allocated to one. Those involving multiple plant species were denoted more.
Plant type had two levels: crop and non-crop. Crop denoted studies with crop plants, such as maize, rice, ect. Those studies carried out with non-crop plants, such as grass and umbrella plants were allocated to non-crop.
Plant life cycle had two levels: annual and perennial. Plants finishing their life circle in one year or multiple years were allocated to annual and perennial, respectively.
Plant carbon fixation had two levels: C3 and C4. Rice, tomato, wheat, and ryegrass were grouped in C3 while maize, corn, reed, umbrella plants and Brachiaria decumbens in C4.
There were two AMF-related variables: AM colonization difference, AM fungi species and other microbes.
AM colonization difference had two levels: big and small. Big denoted that the difference in AM colonization indicated by the percentage of root length colonization or spore density or the density of the extraradical hyphae were greater than 2% or statistically significant. Small meant that those were smaller than 2% or not statistically significant or not reported. In our study, small only indicated that the difference in AM colonization is not reported.
Studies like (Gui et al. 2021) used 20 –µm pore size nylon mesh to allow or prevent the access of AM fungal hyphae in microcosms. In this case, we treated it as big even though only AM colonization was reported for the AM treatment, not for the control.
AM fungi species had two levels: one and more. One and more denoted studies where in the AM fungal treatment one and more AM fungi species were present, respectively.
Other microbes except AM fungi had two levels: presence and absence. Presence and absence denoted studies where other soil microorganisms except AM fungi were present and absent, respectively. In our dataset, all the trials were grouped into presence.
There were four nutrient-related variables: N addition, P addition, C addition and Micronutrient (M) addition.
N (P and C) addition had two levels: no and yes. Addition of any forms of N (P and C) formed yes. In studies involving forest, farmland and grassland, typical forms of N, P and C were nitrate, phosphate and organic fertilizer respectively. In studies focusing on wetland and bioretention systems, N (P and C) were added via the inflow simulating polluted water or stormwater runoff.
M addition had two levels: no and yes. Added micronutrients were magnesium, zinc, sulfur, calcium, boron, iron and manganese. Studies with any one of these micronutrients added formed the level yes.
Soil-related variables included soil texture, organic matter, sterilization, pH and plant available N and P (short for AN and AP, respectively).
Soil texture had two levels: sandy and not sandy. The data were classified following the USDA Natural Resources Conservation Service soil taxonomy (http://soils.usda.gov). All soils with a sand content > 50% (sand, loamy sand, sandy loam, sandy clay loam and sandy clay) were allocated to the levels sandy, soils with other textures formed the level not sandy.
Soil organic matter (SOM) had three levels low (SOM: <1.38%), medium (SOM: 1.38–2.59%) and high (SOM: >2.59%), which was based on our previous meta-analysis (Zhang et al. 2018). Where the content of organic matter was not available, we estimated it by converting organic carbon via the van Bemmelen coefficient (organic matter = organic carbon * 1.724) (Perie &Ouimet 2008).
Substrate sterilization had two levels: yes and no. When in studies the soil was pretreated to eliminate indigenous AM fungi, for example by fumigation, autoclaving or pasteurization, they were grouped in level yes.
Soil pH had three levels: acidic (≤ 6.5), neutral (6.6–7.3) and alkaline (≥ 7.4), based on the USDA criteria (http://soils.usda.gov).
Soil AN and AP and had two levels: low and high. When the value is less than the median of the data we collected, they were grouped in level low, otherwise, were grouped in level high.