The global trade in perishable agricultural products relies heavily on sophisticated cold chain systems to maintain product quality and extend shelf life during extended transportation periods. This is particularly critical for citrus fruits, which represent a significant portion of the international fresh produce trade. South Africa, as the world’s second-largest fresh citrus exporter, contributes substantially to this trade, exporting 2.6 million tonnes in 2022 and generating over $3.3 billion in export revenue (Citrus Growers’ Association 2025; Fresh Produce Exporters’ Forum 2025).

The effectiveness of cold storage facilities in maintaining optimal temperatures throughout the handling and storage process is fundamental to preserving fruit quality and meeting international market standards. However, cold storage operations are energy intensive, creating pressure on facility operators to optimise energy use while maintaining temperature protocols.

Recent studies have highlighted the substantial environmental impact of cold storage operations within freshly produced supply chains. Du Plessis, Van Eeden and Goedhals-Gerber (2024) determined that cold storage contributes 8% of total carbon emissions in the citrus export chain from South Africa to international markets, representing 0.87 kg CO₂e per kg of oranges. Du Plessis, Van Eeden and Goedhals-Gerber (2022) also quantified the energy requirements for cold storage operations at 7.62 kWh per pallet per day, equivalent to 7.52 kg CO₂e daily emissions per pallet.

These energy considerations have led some cold storage facilities to implement energy-saving measures that may compromise temperature control effectiveness. One common practice involves selectively cooling specific areas within the facility while leaving others, such as loading bays and transition areas, without active refrigeration.

This approach raises critical questions about the trade-off between energy conservation and temperature maintenance effectiveness.

Oranges were selected as the model fruit because they are the dominant Western Cape citrus export by volume, have a well-established 5 °C protocol storage temperature, and are commercially handled in large volumes by both participating facilities, allowing matched comparisons. Their relatively thick rind also makes them less susceptible to mechanical damage from instrumentation, so observed effects can be attributed to temperature rather than handling artefacts. Although the focus is on oranges, the airlock and refrigeration findings are expected to be broadly applicable to other citrus varieties that share the same protocol temperature.

To the author’s knowledge, this is the first published controlled comparison of two operational South African citrus cold stores using contrasting refrigeration and airlock technologies under matched protocol conditions. The study therefore provides a quantitative basis for the energy-versus-compliance trade-off that has previously been discussed only qualitatively in the literature.

Formally, the following hypotheses were tested with respect to the impact of temperature-controlled (TC) versus non-temperature-controlled (NTC) airlocks on cooling performance:

This study addresses this knowledge gap by comparing the cooling performance of two different cold storage facility designs: One employing advanced refrigeration technology with TC airlock systems and another using conventional cooling technology with NTC loading areas. The research examines the effectiveness of these different approaches in achieving and maintaining protocol temperatures for exported oranges, providing insights into the relationship between refrigeration technology investment and temperature control performance.

Temperature trials were conducted with oranges in two cold stores in the vicinity of Cape Town during the winter months of July and August. July and August were selected because they coincide with the peak Western Cape Valencia orange export season, when both facilities are operating under representative commercial throughput.

Conducting trials in winter also provides the most rigorous comparison of the two airlock designs, because the smaller temperature gradient between ambient and protocol temperature is the condition under which an NTC airlock is most likely to perform acceptably; observing significant differences under winter conditions, therefore, represents a conservative test. Both facilities specialise in the cold storage of fruit exported from South Africa. Cold store 2 is a facility with TC airlock offloading and loading bays, while CS1 is equipped with NTC airlock loading bays whose loading bays rely on cold air released from the FAC tunnels.

Invasive methods are required to measure fruit pulp temperature, resulting in the loss of fruit, while ambient temperatures can be recorded in non-invasive ways. Therefore, only ambient temperatures were recorded by placing iButtons^® inside the cartons between the oranges. Although fruit pulp temperature is the most direct indicator of physiological state, invasive pulp probing damages the fruit and renders it unsalable, which was not acceptable to the participating commercial facilities. Ambient temperature recorded between fruit inside a closed carton has been shown to track pulp temperature closely once FAC has been completed (typically within ~0.5 °C to 1 °C), and is the standard measurement used in commercial cold chain monitoring (Goedhals-Gerber & Khumalo 2020; Khumalo et al. 2023a). The ambient temperature approach is therefore acknowledged as a limitation for the very early cool-down phase but is not expected to bias the between-facility comparison because the same instrumentation and placement protocol was applied identically to both cold stores. An iButton^® is a small, coin-sized electronic temperature sensor that records and stores temperature readings, with a precision of ± 0.5 °C from −10 °C to 65 °C (Analogue Devices 2026; Khumalo et al. 2023b).

The iButtons^® were inserted before the fruit went into the FAC tunnels and were removed just before the pallets were loaded into the reefer containers.

The Coldstream software was used to calibrate the iButtons^® to record the temperature every 15 min. A 15-min sampling interval is consistent with previous citrus cold chain monitoring studies (Conradie et al. 2022; Goedhals-Gerber & Khumalo 2020; Khumalo et al. 2023a) and is the standard interval used by commercial Coldstream loggers. It is sufficient to capture the loading-event temperature excursions of interest, which typically last from several minutes to several hours, while keeping logger memory and battery use within practical limits for multi-day trials. A pilot test at finer (5 min) resolution did not change the proportion of readings classified as above 5 °C and is therefore not reported separately. The iButtons^® were spread over five pallets for each trial, with 39 iButtons^® used in CS1 and 37 iButtons^® used in CS2 in total. The iButtons^® were inserted into open-top cartons containing oranges, which had to be kept at a temperature of below 5 °C, between the oranges in different layers of the pallets, with their locations recorded for identification.

Nine iButtons^® in pallet 3 of CS1 and one iButton^® in pallet 1 of CS1 malfunctioned, reducing the usable CS1 sample to 29 loggers (10 892 readings) versus 37 loggers (10 471 readings) in CS2. The total number of readings is therefore very similar between facilities, and statistical power for the nonparametric tests used here is not materially affected. To control for the residual difference in N, between-facility comparisons were performed on proportions of readings (e.g. percentage of readings ≤ 5 °C) rather than on absolute counts. As an additional sensitivity check, the analysis was re-run after randomly down-sampling CS2 to match the number of CS1 loggers; the conclusions were unchanged.

Descriptive and nonparametric statistical analyses were used to compare the two datasets. The data were cross-tabulated to determine the number of temperature readings above 5 °C and below or equal to 5 °C for each trial, and percentages were used to allow comparison, given the slightly different sample sizes. The cross-tabulation analysis was supplemented with a maximum likelihood chi-square test to determine whether the frequency of temperature readings above and below or equal to 5 °C differs between CS1 and CS2.

Because the temperature data are recorded over time and decrease as cooling progresses, they cannot reasonably be assumed to be normally distributed, so the nonparametric Mann–Whitney U test was used as the equivalent of the two-sample t test (Keller & Warrack 2000). The Mann–Whitney U test was preferred over the Kolmogorov–Smirnov (K-S) two-sample test because the research hypothesis concerns a difference in central tendency (i.e. whether one facility runs systematically warmer than the other), rather than differences in distributional shape per se. Mann–Whitney U is the more powerful test for stochastic-dominance-type alternatives such as that posed by H₁, and is also the standard nonparametric counterpart of the t test reported in cold chain studies, allowing direct comparison with prior literature. In addition to the significance tests, two effect-size statistics were computed: Cohen’s d, calculated from the means and pooled standard deviation, and the rank-biserial correlation derived from the Mann–Whitney U statistic, which is appropriate for nonparametric two-sample comparisons.

Data from 29 iButtons^® in CS1 and all 37 iButtons^® in CS2 were successfully retrieved, yielding 10 892 and 10 471 valid temperature readings, respectively (see ‘Research methods and design’ section for the data-loss handling protocol).

Figure 2 shows the average and the 20th and 80th percentile temperature profiles for CS1 and CS2. The X-axis has been re-labelled to show elapsed time in hours since the start of the trial, rather than the count of 15-min intervals, to make cooling rates directly comparable between the two facilities. The CS1 trial is longer than the CS2 trial because the fruit had to remain in the cold store waiting for the reefer container – this reflects operational dwell time rather than a difference in cooling capacity, and the proportional metrics in Table 2 are not biased by this difference. The weather was very similar in July and August, with maximum temperatures of 16 °C – 18 °C and is therefore not considered to have an impact on the results.

Upon arrival at CS2, the oranges were offloaded into a TC airlock receiving area, which is reflected in the lower initial temperature in Figure 2. The fruit was cooled down much faster in CS2 and then maintained at its ideal temperature of below 5 °C. In CS1, the average temperature hovered slightly above 5 °C for an extended period before finally dropping below 5 °C. Defining cooling time as the elapsed time from the start of the trial until the average pallet temperature first reached the 5 °C protocol threshold, CS2 reached protocol within approximately 1.5 h of arrival, whereas the CS1 average only crossed 5 °C after roughly 25 h and remained marginal for an extended period thereafter. This represents an order-of-magnitude difference in cooling rate between the two facilities.

This comparative study demonstrates significant differences in cooling performance between cold storage facilities employing different refrigeration technologies and airlock systems. The investigation of two cold storage facilities handling export oranges reveals that advanced cooling technology with TC airlock systems substantially outperforms conventional systems with NTC loading areas.

CS2 maintained 85.92% of temperature readings at or below the critical 5 °C threshold, compared to only 57.46% for CS1, with a mean temperature difference of 3.16 °C between facilities. Statistical analysis using the chi-square and Mann–Whitney U tests confirmed highly significant differences (p < 0.001) between the temperature distributions of the two facilities, with large effect sizes (Cohen’s d = 1.04; rank-biserial r = 0.73). The faster cooling rates observed in CS2, particularly evident in the initial cooling phase, indicate that TC airlock systems effectively prevent temperature rises during fruit handling and loading operations.

These findings should be interpreted within the energy–compliance trade-off that motivates NTC airlock designs.

The 15% – 25% energy savings reported by Fikiin and Markov (2014) for selectively cooled airlock systems can be substantial: At 7.62 kWh per pallet per day (Du Plessis et al. 2022), a facility handling 1000 pallets per day might save in the order of 1500 kWh – 2000 kWh daily. However, the present results show that 42.5% of readings in CS1 exceeded the 5 °C protocol, putting fruit value at risk. Given that arrival rejections, claim deductions and reduced shelf life can easily exceed the value of the energy saved, the apparent NTC energy advantage is likely to be partially or wholly offset by quality losses in commercial operation. A formal cost–benefit analysis incorporating fruit value, rejection rates and current South African energy tariffs is identified as an important next step.

Although the study recommends investment in advanced TC airlock systems, the author recognises that a full retrofit may not be financially viable for small-scale operators. Several intermediate measures can narrow the compliance gap at a fraction of the cost: (1) installation of high-speed roll-up doors and air curtains at the loading bay to reduce heat infiltration during pallet movement, (2) scheduling loading operations to coincide with the coldest part of the day, (3) shortening pallet dwell times in the staging area through better synchronisation with reefer container availability, and (4) routine commissioning of existing FAC tunnels to ensure that residual cold air supplied to the loading bay is maximised. These measures preserve much of the energy advantage of NTC systems while substantially reducing the worst temperature excursions observed in this study.

Although the present study did not measure fruit quality directly, the temperature differences observed are well within the range that has been linked to measurable quality losses in citrus. Sustained excursions above 5 °C in the early cold chain have been associated in the literature (Goedhals-Gerber & Khumalo 2020; Khumalo et al. 2023a) with accelerated weight loss, increased decay incidence and elevated rates of market rejection on arrival in European ports. The 3.16 °C mean difference and the threefold difference in time spent above protocol observed here are therefore likely to translate into commercially significant differences in shelf life and arrival quality. Direct measurement of fruit quality outcomes in a follow-up study is recommended to quantify this link.

These findings have important implications for cold storage facility design and operation in the context of energy efficiency versus temperature control effectiveness. The superior performance of the advanced facility indicates that investment in comprehensive refrigeration technology, including TC airlock systems, is justified for maintaining export quality standards. Future research should investigate the long-term economic implications of comprehensive refrigeration system investment versus energy-saving operational practices, including analysis of fruit quality retention, shelf-life extension and market acceptance rates.