The monitoring campaign ran from October 2024 to September 2025 in Liangcun Town, Xingguo County, in the hilly subtropical monsoon region of southeastern China. The greenhouse had approximately twelve years of continuous vegetable cultivation behind it, and both sites followed local conventional production practices. During the study, celery was grown in the greenhouse and radish in the open field from October through March, and pepper was cultivated in both environments from April through September. The researchers collected soil once per month from the 0–20 centimeter cultivated layer using an S-shaped sampling design, mixing five subsamples into each of three independent composite samples per environment—72 samples in total. Sampling was deliberately scheduled after at least three rain-free days whenever practicable, to avoid transient moisture spikes that could mask the slower seasonal signals the team was hunting for.

The microclimate measurements told an unambiguous story. Hourly air temperature differences between the greenhouse and the open field ranged from −6.1 to +6.1 degrees Celsius across the year, with the largest positive differences concentrated between 10:00 and 18:00. Daytime warming inside the greenhouse peaked in early spring, from February to April, when the temperature difference generally exceeded 4 degrees and reached its maximum of +6.1 degrees. At night the greenhouse still retained heat, but the effect was weaker than daytime warming. Relative humidity told a parallel story: the greenhouse was more humid than the open field during most hours of the year, with differences ranging from −0.6 to +43.0 percent. The humidity contrast was greatest in winter, smallest in summer, and intermediate in spring and autumn—likely modulated by seasonal management practices such as ventilation, shading, and heat conservation.

Within the greenhouse soil itself, temperature followed a seasonal pattern similar to air temperature but varied more smoothly, while soil volumetric water content stayed within roughly 15–20 percent all year. In spring, summer, and autumn, greenhouse air temperature ran above greenhouse soil temperature, which in turn ran above open-field air temperature; in winter, the ordering shifted, with greenhouse soil temperature warmest of all at 10.5 degrees Celsius while greenhouse and open-field air temperatures fell to 7.5 and 7.0 degrees respectively. Interestingly, winter combined high atmospheric humidity inside the greenhouse with comparatively low soil water content, indicating that humidity accumulation and soil water status were not fully synchronized—potentially reflecting the interplay of irrigation, ventilation, and heat-conservation practices.

Soil bacterial community composition differed significantly between the two environments. At the phylum level, Pseudomonadota, Acidobacteriota, Actinobacteriota, Chloroflexi, and Bacteroidota together accounted for more than 90 percent of total bacterial relative abundance. Linear discriminant analysis effect size identified differentially abundant taxa: Pseudomonadota, Bacteroidota, and Chloroflexi were relatively more abundant in greenhouse soil, while Acidobacteriota and Actinobacteriota were more abundant in the open field. At the lowest reliable taxonomic level, Sphingomonas, Vicinamibacterales, and Vicinamibacteraceae were relatively more abundant in greenhouse soil, whereas Pyrinomonadaceae and Pseudarthrobacter were more abundant in open-field soil. Principal coordinate analysis based on Bray-Curtis dissimilarities showed clear separation of the two communities, with the first two axes explaining 35.79 and 21.02 percent of community variation respectively.

PERMANOVA testing confirmed that cultivation environment was significantly associated with bacterial community composition, with a pseudo-F of 7.32, an R-squared of 0.577, and p < 0.001. But the temporal dimension proved just as striking. Month-resolved analyses revealed significant differences among the twelve sampling months (pseudo-F = 1.44, R-squared = 0.597, p < 0.001), and both greenhouse and open-field communities followed continuous trajectories through the year when monthly centroids were connected in chronological order. This month-resolved approach retained temporal information that seasonal grouping alone could not fully capture, demonstrating that bacterial communities in both environments are in constant month-to-month flux rather than settling into four static seasonal states.

Alpha diversity added another layer of nuance. The Shannon index was significantly higher in greenhouse soil in spring, autumn, and winter but significantly lower in summer; the Simpson index showed the opposite pattern. Environmental differences were strongest in autumn (p < 0.01), while differences in the remaining seasons were significant at p < 0.05. In other words, the direction of the greenhouse effect on bacterial diversity was strongly season dependent rather than consistently positive or negative throughout the year. This finding echoes comparable temporal variation reported in other soil microbial studies, including subtropical systems where bacterial communities and co-occurrence networks differed between wet and dry seasons, and warming experiments showing that bacterial responses can depend on crop type.

Redundancy analysis was used to evaluate statistical associations between the measured atmospheric microclimate variables and bacterial community structure. Air temperature and relative humidity jointly explained 15.59 percent of the variation in community structure—RDA1 and RDA2 accounted for 9.03 and 6.56 percent respectively. The air-temperature vector was oriented mainly toward negative RDA1 and positive RDA2, whereas the relative-humidity vector extended primarily along positive RDA1, indicating different directions of statistical association with community structure. Samples showed partial separation by environment and season in the ordination, consistent with temporal covariation between bacterial composition and atmospheric conditions. But the authors are careful: most variation remained unexplained, and because soil pH, electrical conductivity, organic matter, salinity, available nutrients, and other edaphic properties were not measured concurrently, the analysis describes statistical associations rather than identifying principal drivers or establishing causality.

Potential co-occurrence networks, constructed separately for greenhouse and open-field soils using Spearman correlations with false discovery rate adjustment, also differed markedly. The greenhouse network contained 242 OTU nodes and 481 edges, while the open-field network contained 280 nodes and 827 edges. The open-field network had a higher average degree (5.907 versus 3.975), higher average clustering coefficient (0.381 versus 0.324), higher network density (0.021 versus 0.016), and a shorter average path length (5.367 versus 5.749), indicating a denser pattern of potential associations among open-field bacterial OTUs. Modularity, however, was slightly higher in the greenhouse network (0.699 versus 0.683), illustrating that connectivity and modular organization do not necessarily change in parallel—and that no single topological metric should be used to declare one network intrinsically more complex or stable than another.

The balance of positive and negative correlations also shifted between environments. Positive edges accounted for 75.68 percent of greenhouse network edges and 66.99 percent of open-field edges, with negative edges representing 24.32 and 33.01 percent respectively. The authors stress that these correlations describe statistical co-variation among OTU abundances and should not be equated with direct ecological interactions: positive correlations do not necessarily indicate facilitation or cooperation, and negative correlations do not necessarily indicate competition or inhibition. Eight highly connected OTUs were identified in each network, and their composition differed between environments—though the researchers describe them conservatively as highly connected OTUs rather than keystone taxa, since high node degree alone does not establish a keystone ecological function.

The study has clear limitations that the authors acknowledge directly. Soil temperature and volumetric water content were monitored only within the greenhouse, so the team could not determine whether soil thermal-moisture regimes differed between environments or quantitatively attribute bacterial community differentiation to soil temperature or moisture. Soil physicochemical properties and quantitative water and fertilizer inputs were not monitored concurrently, and the study involved a single paired greenhouse-open-field production system, restricting inference to this system rather than greenhouse cultivation as a whole. Future reciprocal soil-transplant experiments—in which greenhouse and open-field soils are exposed to both environments—could help separate current microclimate effects from soil legacy and management-history effects.

Despite these constraints, the combination of continuous one-year microclimate monitoring, month-resolved community analysis, and co-occurrence network comparison provides an unusually detailed temporal description of bacterial communities in a subtropical protected-agriculture system. The results document year-round differences in microclimate and soil bacterial community dynamics between greenhouse and open-field cultivation, while showing that measured atmospheric variables alone are statistically detectable but insufficient to account for bacterial community differentiation. For the rapidly expanding world of protected agriculture, the message is that greenhouse soils are not simply warmer versions of open-field soils—they are distinct microbial ecosystems whose composition, diversity, and association structures shift month by month. Understanding and managing those shifts will require integrating synchronous soil-environment monitoring, physicochemical measurements, and quantitative management records to resolve the environmental mechanisms underlying bacterial community variation in protected cultivation systems.

Subject of Research: Year-round microclimate monitoring and monthly 16S rRNA sequencing reveal greenhouse and open-field differences in soil bacterial communities and co-occurrence networks in subtropical China.

Article Title: Microclimate characteristics, temporal dynamics of soil bacterial communities, and potential co-occurrence network differences between greenhouse and open-field cultivation environments

Article References: Xue, L., Chen, Y., Zhong, Y., & Wan, C. (2026). Microclimate characteristics, temporal dynamics of soil bacterial communities, and potential co-occurrence network differences between greenhouse and open-field cultivation environments. Frontiers in Microbiology, 17, Article 1988234. https://doi.org/10.3389/fmicb.2026.1988234

Image Credits: AI Generated

DOI: 10.3389/fmicb.2026.1988234

Keywords: greenhouse agriculture, soil microbiome, bacterial communities, microclimate, 16S rRNA sequencing, co-occurrence networks, protected cultivation, subtropical red soil, temporal dynamics, Frontiers in Microbiology, characteristics, temporal