The motivation behind the work stems from an underappreciated vulnerability in kimchi production. Kimchi seasoning, the potent blend of garlic, ginger, chili pepper, scallions and other aromatics that gives the dish its signature bite, is prepared from raw agricultural ingredients that can harbor contaminating microorganisms. Unlike the finished, fully fermented product, seasoning mixtures are not yet dominated by the protective lactic acid bacteria that suppress competitors during fermentation. If pathogenic bacteria hitch a ride on contaminated seasoning ingredients, they can be transmitted into the fermenting cabbage batch, where their fate depends on a tug-of-war between harsh acidity and favorable storage temperatures. Previous surveys of kimchi and its ingredients have documented microbiological contamination, and outbreaks elsewhere have linked Yersinia to ready-to-eat salads and fresh produce, underscoring that raw plant materials can serve as vehicles for human pathogens.

Yersinia enterocolitica is a psychrotrophic bacterium, meaning it grows happily at refrigeration temperatures, an unsettling trait for a food industry that relies on cold storage as its primary safety net. Infection with the bacterium, known as yersiniosis, typically produces abdominal pain, diarrhea and fever, and cases are frequently reported across Europe and other regions in zoonoses monitoring. Because the organism thrives in the cold, simply chilling kimchi seasoning is not enough to guarantee safety, and producers need to know exactly how fast the pathogen declines, under what combinations of temperature, time and acidity, and whether those declines follow predictable mathematical patterns. That is precisely the knowledge gap the Korean team set out to fill with quantitative modeling rather than one-off experiments.

To capture the combined effects of multiple environmental factors, the researchers turned to response surface methodology, a statistical framework widely used in engineering and food science for optimizing and predicting processes. At the heart of their experimental design was a central composite design, an arrangement of experimental runs placed systematically across a multidimensional factor space that allows a second-order polynomial equation to be fitted to the observed responses. In practical terms, the team varied storage temperature and storage time while holding the initial pH of the seasoning at two distinct levels, pH 4.5 and pH 5.5, and measured how many Y. enterocolitica cells survived in each condition. The fitted equations then generate a response surface, essentially a topographic map of inactivation, that lets modelers read off expected pathogen reduction for any combination of variables within the tested range, including combinations never directly measured in the laboratory.

The statistical backbone of the study came from analysis of variance, which confirmed that every factor built into the models exerted a statistically significant influence on Y. enterocolitica survival. Storage temperature and storage time emerged as the most powerful drivers of inactivation, dominating the behavior of the pathogen across the experimental space. Critically, the analysis also revealed a significant interaction between temperature and time, meaning the effect of one factor depended on the level of the other; cold storage over long durations does not simply add up in a straightforward way, and the models capture that non-additivity. The pH term was likewise significant, but its contribution turned out to be far more interesting than a simple main effect, because acidity did not merely speed up or slow down killing uniformly. Instead, it fundamentally changed the shape of the survival curve.

That shape difference is the conceptual heart of the study. At the milder pH 5.5 condition, inactivation followed relatively linear reduction kinetics, the kind of steady, predictable decline that classical first-order models describe well. At the harsher pH 4.5, however, the survival data took on a distinctly non-linear character that the team characterized as Weibull-type behavior, named after the flexible probability distribution commonly used to describe microbial death curves that bend. Weibull models can represent curvilinear survival patterns, including curves with shoulders or tails, and their shape parameter carries biological meaning: deviations from a straight line are often interpreted as evidence that cells in a population differ in their resistance, or that the environment itself shifts the stress landscape as cells accumulate damage. In this case, the non-linear response under strongly acidic conditions pointed toward enhanced microbial stress responses, a phenomenon in which bacteria exposed to sublethal acid stress activate protective machinery, including acid resistance systems known from related pathogens, that can make the population harder to kill in ways a straight-line model would miss.

For food microbiologists, the practical consequence is that a single generic inactivation equation cannot be safely applied across kimchi seasoning batches whose acidity varies naturally. Kimchi ingredients bring their own buffering capacity and pH to the mixture, and fermentation progresses differently depending on recipe, temperature and microbial community, so the initial pH of the seasoning matrix can plausibly sit anywhere near the two levels the team tested. By modeling pH as an explicit variable, the researchers have produced a framework in which a producer or risk assessor can select the equations appropriate to their product’s acidity and obtain defensible predictions of pathogen decline. The distinction between linear and Weibull-type kinetics also matters for regulatory calculations, because assuming linear death when the true curve flattens out could dramatically overestimate the safety margin of a given storage regime.

The study sits within the broader and rapidly growing field of predictive microbiology, which aims to convert empirical observations of microbial behavior into mathematical models capable of forecasting food safety outcomes. Decades of work in this discipline have established that microbial responses to temperature, pH, water activity and other hurdles can be modeled hierarchically, with primary models describing the shape of growth or death curves and secondary models describing how those shapes change with environmental conditions. The kimchi seasoning models exemplify this approach applied to a food matrix and pathogen combination that had received little quantitative attention. The authors note that quantitative information about Y. enterocolitica behavior during kimchi seasoning storage has been limited, and their response surface models are intended to close that gap by providing usable predictions for contamination risk assessment and the design of storage strategies.

The implications ripple outward to both industry and consumers. For manufacturers of kimchi and related minimally processed vegetable products, the models offer a way to evaluate whether a proposed cold-chain temperature, distribution timeline and recipe acidity collectively deliver an adequate pathogen reduction, before a contaminated batch ever reaches the market. For regulators and risk assessors, they supply the kind of parameterized equations needed for quantitative microbial risk assessment, the formal discipline of estimating illness risk from farm to fork. And for the global community of kimchi lovers, the research quietly reinforces a lesson that food safety scientists have long emphasized: fermentation and refrigeration are powerful tools, but their effectiveness against specific pathogens depends on measurable variables that can now be calculated rather than guessed. As predictive microbiology matures, the days of treating traditional fermented foods as too complex to model are drawing to a close.

The work, funded by the Ministry of Science and ICT of the Republic of Korea through the World Institute of Kimchi, also highlights a larger trend in food safety science: the integration of designed experiments, statistical optimization and mathematical modeling into everyday decision-making for traditional foods. What began as a classically empirical craft, passed down through generations of Korean home cooks, is increasingly described in the language of polynomial equations, Weibull parameters and response surfaces. The Korean team’s demonstration that acidity reshapes the very kinetics of pathogen death in kimchi seasoning is a vivid example of how modern quantitative tools can illuminate even the most ancient of foods, and it offers a template that other producers of fermented and minimally processed vegetables around the world are likely to follow as they confront the same cold-loving pathogens in their own products.

Subject of Research: Predictive modeling of Yersinia enterocolitica inactivation in kimchi seasoning under varying pH and storage conditions

Article Title: Predictive modeling of Yersinia enterocolitica reduction in kimchi seasoning as affected by the initial pH, using response surface methodology

Article References: Song, H., Kim, S.-J., Ha, S., Song, J., & Ha, J.-H. (2026). Predictive modeling of Yersinia enterocolitica reduction in kimchi seasoning as affected by the initial pH, using response surface methodology. Food Science and Biotechnology. https://doi.org/10.1007/s10068-026-02316-4

Image Credits: AI Generated

DOI: 10.1007/s10068-026-02316-4

Keywords: Yersinia enterocolitica, kimchi seasoning, predictive microbiology, response surface methodology, food safety, microbial inactivation kinetics, Weibull model, central composite design, pH, cold storage, foodborne pathogens, fermented foods