Summary
Food borne diseases associated to fresh produce continue to cause serious difficulties for public health in the United States. To offset this burden, the produce wash stage has received much attention as a critical control point. However, recent studies indicate a limited understanding of the dynamics of sanitizer control during washing. One problem is that the relationships between water quality constituents and sanitizer levels have only been described via experimental/correlative approaches or by risk models that are difficult to parameterize accurately. Accordingly, there is an urgent need to mathematically describe the fundamental dynamics that generate the observed relationships between sanitizer levels and water quality parameters. Based on such formulations, our long-term goal is to develop optimal sanitizer strategies that are easily automatable and adjustable to specific commodities and washing practices. The primary objective of this proposal is to develop data-informed modeling tools which quantitatively link easily measurable water quality parameters (e.g. turbidity/total dissolved solids) to commodity specific organic load and free chlorine consumption during recirculated wash conditions. Based on USDA experimental data and our recent modeling results, we hypothesize that by using our modeling tools, the industry can obtain reliable predictive capabilities that are not possible with correlations alone.
Technical Abstract
As globalization has broadened the fresh produce supply chain and increased its complexity, more sophisticated methods of surveillance are needed to ensure the safety of fresh products. In particular, the processing juncture is a critical control point that has received much attention. While many wash systems are being used in good faith and according to standard protocols, it is clear that a greater understanding of in-practice dose measurement in relation to water quality constituents and commodity-specific parameters is needed. Part of the problem is that the relationships between sanitizer levels and water quality parameters have mainly been described through experimental and correlative approaches or by risk models that are difficult to parameterize accurately. While these results are important, they alone cannot be used to make precise predictions, or for taking real-time corrective measures in dynamic circumstances. Accordingly, there is an urgent need to mathematically describe the fundamental dynamics that generate the observed relationships between sanitizer levels and water quality parameters. Not meeting this need represents a serious problem because without clear scientific foundations informing sanitizer control, the increased demand for fresh produce translates into increased potential for widespread food borne disease outbreaks. From this perspective, our long-term goal is to develop optimal sanitizer strategies that update in real-time and are easily automated/adjusted to produce type, cut operation and washing practice. Building toward this goal, our main objective for this proposal is to construct data- based modeling tools which quantitatively link easy to measure water quality parameters to commodity specific organic load and free chlorine (FC) levels during recirculated wash conditions. Based on USDA experimental data and our previous modeling results, we hypothesize that a combination of turbidity and total dissolved solids (TDS) can be used to predict organic load and subsequently FC decay kinetics relative to produce/cut type, produce to water ratio and incoming produce rate. The above objective will be realized via the following two tasks for particular cut types of green cabbage and carrots, respectively.
Task 1: Combining correlative techniques and dynamical systems theory, we will build commodity specific mathematical models that quantitatively link TDS and turbidity to FC consumption kinetics.
Task 2: Test and validate model predictions against lab and pilot-plant scale data. This will be accomplished with experimental results from Dr. Yaguang Luo (USDA- ARS) and our own experiments at Cleveland State University. Using sensitivity and uncertainty analysis coupled with this data, we will clearly quantify the range of processing conditions within which our model predictions hold. In terms of immediate industrial impact, our modeling tools can be used to help minimize FC variability under certain processing conditions. In addition to providing a first step for informing the next generation of online chlorine controllers, our model results will provide guidelines for further experimentation, streamlining the scope and frequency of expensive pilot-plant/commercial scale data acquisition.
Research Objectives
The main research objective is to construct data-informed models that quantitatively link easyto-measure water quality parameters to commodity-specific organic load and free chlorine (FC) decay kinetics during recirculated wash conditions. The above objective will be realized via the following tasks for green cabbage and carrots (various cut types for each produce type):
Task 1: Build mathematical models that take into account wash water chemistry, commodity specific aspects.
Task 2: Validate model predictions against lab and pilot-plant scale data.
Findings & Recommendations
1. Because the FC decay predictions hold at multiple scales, the models developed from this project illustrate fundamental chlorine decay dynamics that occur during fresh-cut carrot/cabbage and iceberg lettuce washing. In particular, this gives validity to performing future lab-scale experiments to quantify FC decay associated with different produce/cut types as well as experiments aimed at understanding the impact of continuous FC dosing on FC dynamics during produce washing.
2. The CSU team found that turbidity and TDS measurements are not reliable in predicting FC levels, as there is no consistent, observable relationship linking the increase in organic load (in terms of COD) from cut carrots/cabbage entering the wash tank (3–100 L) and the corresponding increase in turbidity or TDS. Similar results were observed for cut/shredded iceberg lettuce across various scales (3 L, 100 L, and 3200 L). In particular, the model’s predictive success across scales and produce types provides a strong case that COD information is much more reliable than that of turbidity or TDS.
3. The results from this research demonstrate the utility of using mathematical models as tools to elucidate fundamental mechanisms, like FC reaction rates associated with various produce cut types. To minimize expensive experiments at a commercial scale, these models and key lab-scale experiments can aid in planning the logistics of how and what should be measured during commercial-scale experimentation. For instance, for models, like those developed in this project, to be effective tools in validating FC control at the industrial scale, FC data collected at the commercial scale must include quantifiable FC input and water replenishment information.