Micro-Incubator Protocol for Testing a CO₂ Sensor for Early Warning of Spontaneous Combustion

Abstract

A protocol for detecting the potential occurrence of spontaneous combustion (SC) in stored cottonseeds and peanuts using a micro-incubator is described. The protocol indicates how to quantify CO₂ production rates and final CO₂ levels in wet versus dry cottonseed and peanut samples, which can provide crucial data for the early detection of SC risk in storage facilities. The experimental design utilizes a micro-incubator to simulate conditions found in large bulk crop storage. Parameters monitored include CO₂ concentration, temperature, and relative humidity. The protocol includes preparation methods, experimental procedures for both control (dry) and wet seed tests, and test termination criteria that allow for safe experimentation of likely pathogenic fungi. The protocol has three replicates for wet and dry conditions. The protocol is intended to facilitate future experimental studies and ultimately contribute to the development of a consistently reliable early warning fire detection system for SC in cottonseed and peanut warehouse facilities. A consistently reliable fire detection system would address a critical need in the cotton and peanut industry for improved fire risk management and insurability of storage facilities.

1. Introduction

Cottonseed and peanut warehouses are subject to costly destructive fires. One of the common causes of fires and most difficult to guard against is spontaneous combustion (SC), which can develop because of initial heating due to microbial growth that can then progress to SC because of the high oil content in the stored commodity. A cottonseed warehouse fire in Whitton, Australia, was reported to have caused over 1 million AUD in damages to just the structure alone, not counting the loss of the stored crop [1]. In the US, more and more insurance underwriters are refusing to renew insurance policies for cottonseed storage facilities because of the regularity of fires. Loss of insurance poses an economic hardship to facility owners that they are unable to overcome and thus results not only in the loss of cottonseed storage facilities but many other agribusinesses that rely on storage facilities. In conversations with insurance agencies, they reported that a primary reason for their reluctance to underwrite insurance policies is due to the lack of consistently reliable fire detection, which would translate into a reduction in the number of claims the insurance companies would have to pay. Studies have concluded that the best approach to reduce fire losses is through the use of preventive measures [2]. Hence, there is a critical need for a consistently reliable early warning SC detection system.

Even with a considerable volume of work on fire detection, there still is not a consistently reliable method for using these systems in cottonseed and peanut warehouses [3,4,5,6,7,8,9,10,11,12]. From previous work, a primary indicator of developing SC is elevated regions of temperature within the stored commodity. Using the information about elevated temperature, multiple temperature sensors have been deployed in cottonseed warehouses by suspending them on cables from the rafters down into the stored cottonseed. The problem with the multiple temperature sensor method is that as the amount of stored cottonseed increases, it exerts excessive force on the cables, causing the rafters to collapse, which renders the method impractical. Given the failures of fire detection systems in the past, a new approach to fire detection in cottonseed and peanut warehouses is needed.

To begin a new approach, an understanding of how SC develops in stored cottonseed and peanuts is required. SC develops as a multistep or multistage process where each stage creates the prerequisite conditions for the subsequent stage [13,14,15,16,17,18,19]. Detection in the early stages of SC is critical to limit fire damage as well as to preserve the cottonseed because, during the SC process, the oil in cottonseeds degrades in both quality and quantity with a reduction in oil from 40% to as little as 20%. A report by Gilman and Barron [20] examined self-heating in grain from three commonly observed fungi, Aspergillus flavus (A. flavus), A. niger, and A. fumigatus. Marin et al., 2024 [21] created a model that shows that at aw = 0.84 and temperatures <20 °C, there is zero probability of growth of A. flavus, but with aw = 0.88, even at 20 °C A. flavus will grow after a 7–8 day lag. From these and other reports, it is apparent that as water activity aw increases, the metabolic activity also increases. The work [20] shows that at aw levels supporting any of these three fungal species, that heat production from respiration, will support self-heating and likely lead to eventual SC.

The first stage leading to SC begins when stored cottonseed has greater than 14% moisture content with ideal conditions, leading to SC at between 25% and 50% moisture content [16,17]. Elevated moisture in the stored cottonseed or peanuts provides the environment for microbes already present on the surfaces of the cottonseeds or peanuts to grow and reproduce rapidly. The metabolic processes of the growing microbes generate heat, and as they reproduce and increase in number, the amount of heat generated also increases. The self-heating process is first mesophilic 20–40 °C and aerobic or anaerobic and then thermophilic 40–70 °C and aerobic or anaerobic, with different species of microbes being active depending on temperature range and availability of oxygen. When the temperature becomes too intense, 70–80 °C, to sustain microbial activity, the final stage in the SC cascade is reached. The final stage in the SC process is the abiotic oxidation of the oil in the commodity, which subsequently rapidly progresses to elevated temperatures, leading to smoldering or open-flame fires, depending upon the availability of oxygen, Figure 1, a figure derived from data presented in [22]. Hence, avoiding excess moisture is critically important to prevent SC; however, in the cases when high levels of moisture are present in the cottonseeds or peanuts, leading to high aw, the need exists for early detection well before significant self-heating and the eventual oxidation stage begins. Chemical reaction rates accelerate with increasing temperature, leading to rapid onset of fire once the cascade hits the oxidation stage, which leaves little time to react to mitigate the situation. Further, by that time, significant damage to the stored product has already occurred. Thankfully, significant CO₂ production, relative to dry seed CO₂ production rates, occurs shortly after the introduction of water, at temperatures >22 °C [22], and detailed later in this report for cottonseed where even at aw = 0.81 significant CO₂ production was observed. Of note, A. flavus, A. fumigatus, and penicillium are all biological safety level-two BSL-2 organisms, and care and expertise are required when growing and experimenting with them. For engineers not trained in mycology, great care is warranted, especially as they likely do not have the required biological safety hoods and ability to lock out all but participating key personnel from the lab so that it qualifies as a BSL-2 level safety laboratory, while the experiment is on-going. Hence, it is suggested to study safer BSL-1 organisms such as A. niger and A. oryzae (which are used in the production of rice wines such as sake). A. oryzae is very close genetically to A. flavus as such, exhibits many of the same traits. Note: unless seeds are thoroughly sterilized, just adding water to seeds from the seed warehouse will likely result in significant growth of A. flavus and/or A. fumigatus at elevated aw levels, as discussed earlier. One method that has been found to work well for sterilization is to soak the seeds in 1% sodium hypochlorite solution for 30 min, followed by a triple rinse with reverse-osmosis (R.O.) water, and then 30 min soak in 70% ethanol, followed by a triple rinse in R.O. water and then submerging the seeds in water and placing them in a 60 °C water bath for 30 min and then if necessary gently driying them at 40 °C. Of note is that A. flavus can also be found inside the seed coats, so this is not always 100% effective, which leads some researchers to use gamma irradiation for seed sterilization. If it is desirable to bring all seeds to the same uniform aw, the use of a water–glycerin mixture can set up a stable and adjustable aw level. A volume fraction of glycerol, Vg of 0.45, to water provides an aw = 0.80, and Vg = 0.25 provides aw = 0.91. An experimental temperature of 30 °C will minimize the microbial lag time, yielding shorter experiments.

An outline of microbial metabolic stages and related chemical reactions in the SC process. The sequence of stages can vary depending on the availability of oxygen:

• Mesophilic (20–40 °C; 68 °F TO 105 °F) Aerobic:

  ○

  Initially, mesophilic bacteria in seeds, typically Aspergillus flavus (A. flavus) and fungi, break down organic material for cottonseed; this is predominantly the lint on the surface of the stored fuzzy cottonseed. This stage produces CO₂ and heat with the primary process of:

  ○

  ${\mathrm{C}}_6{\mathrm{H}}_{12}{\mathrm{O}}_6+6{\mathrm{O}}_2\to 6{\mathrm{C}\mathrm{O}}_2+6{\mathrm{H}}_2\mathrm{O}+Energy\left(ATP\right)$

  ○

  Oxygen Availability: Necessary for aerobic microbial activity and chemical oxidation processes.

  ○

  Heat Production: Significant amounts of heat are produced via microbial activity in this stage

  ○

  O₂ is converted to CO₂ on a 1:1 basis, so pressure build-up in a closed system is predominantly limited to an increase in temperature.

• Mesophilic (20–40 °C; 68 °F TO 105 °F) Anaerobic:

  ○

  Occurs because of the compaction and high moisture content, which leads to anaerobic zones deep within the stored cottonseed.

  ○

  Anaerobic metabolism produces significantly less heat than aerobic processes. Further, while CO₂ is still produced, it is produced at a reduced rate of 1:3 (6 CO₂/glucose unit in aerobic versus 2 CO₂/glucose unit in anaerobic metabolism).

  ○

  ${\mathrm{C}}_6{\mathrm{H}}_{12}{\mathrm{O}}_6\to 2\mathrm{C}{\mathrm{O}}_2+2{\mathrm{C}}_2{\mathrm{H}}_5\mathrm{O}\mathrm{H}$ (ethanol).

The mesophilic microbes create the elevated temperatures necessary for the second stage, where thermophilic microbes take over.

Thermophilic Aerobic Phase (45–80 °C):

• Sufficient O₂ to support aerobic microbial production; predominantly A. fumigatus.
• As the temperature rises, thermophilic microbes takeover, continuing aerobic respiration and further increasing the temperature.

  ○

  ${\mathrm{C}}_6{\mathrm{H}}_{12}{\mathrm{O}}_6+6{\mathrm{O}}_2\to 6{\mathrm{C}\mathrm{O}}_2+6{\mathrm{H}}_2\mathrm{O}+Energy\left(ATP\right)$

  ○

  Heat continues to be produced during this stage till the heat eventually reaches a point where oil oxidation takes over with a rapid increase in temperature, leading to SC. During the oxidation phase, the thermophilic microbes die off because of excessively high temperatures caused by abiotic oxidation processes, which stop CO₂ from being produced by microbial activity.

Thermophilic Anaerobic Phase (45–80 °C):

• Insufficient O₂ to support aerobic microbial production
• As the temperature rises, thermophilic microbes takeover with anaerobic respiration and further increase temperature.

  ○

  ${\mathrm{C}}_6{\mathrm{H}}_{12}{\mathrm{O}}_6\to 2\mathrm{C}{\mathrm{O}}_2+2{\mathrm{C}}_2{\mathrm{H}}_5\mathrm{O}\mathrm{H}$ (ethanol).

  ○

  ${\mathrm{C}\mathrm{O}}_2+4{\mathrm{H}}_2\to {\mathrm{C}\mathrm{H}}_4+2{\mathrm{H}}_2\mathrm{O}$ (methanogenesis).

  ○

  Heat continues to be produced during this stage but at a reduced rate (in comparison to the aerobic thermophilic process).

  ○

  CO₂ is produced during the ethanol production process or consumed if the process follows methanogenesis

  ○

  A lower likelihood of SC in a low O₂ atmosphere significantly reduces lipid oxidation, thereby limiting the necessary third phase; however, in the transition zone between very wet and dry regions, which are likely aerobic, there is strong potential for aerobic processes to lead to SC. The production of methane in the reduced zone can then contribute combustible gas to the SC.

The outline above of the metabolic activity of microbes indicates that increasing CO₂ levels could be an alternative to increasing temperature as an indicator of developing conditions for SC. An important note is that the above chemical equations omit the oxidation step that occurs by the fungi-produced cellulase enzymatic conversion of cellulose to glucose, which can be 2–3 times the direct respiration energy. As such, there is likely 4–5 kJ/mol of energy for each respired CO₂. Hence, CO2 is a sound surrogate for monitoring fungi metabolic activity and, at higher aw, heating of substrate. The observation of CO₂ levels as an alternative to temperature increase is supported by research on cereal grains [23,24,25,26,27,28]. The protocol to be presented for testing microbial production of CO₂ is intended for production in a closed micro-incubator that has previously been described

Factors occurring in a closed incubator that affect microbial CO₂ production and detection, as well as other factors that need to be considered, are:

Microbial CO₂ production depends on the amount of moisture present, so moisture levels must be managed to 25% or greater but less than moisture content that interferes with O₂ diffusion, which would lead to a change to anaerobic processes

• A target moisture content of 25% to produce an equilibrium moisture content of 85% relative humidity, RH, at 25 °C [29]
• Microbial activity on the cottonseeds will initially start as mesophilic aerobic (20–40 °C) processes
• Then, change to thermophilic aerobic (40–80 °C) processes as the temperature rises from microbial respiration and activity
• Comparison of CO₂ production rates between low moisture conditions, 8%, which is a safe storage condition, and at-risk condition, 18% or greater, for SC.
• CO₂ sensors are typically limited in operation to no hotter than 50–60 °C. Since detection is intended for early warning, the sensor operation range should be sufficient for the tests. Sensors should not be located within the wet active microbial zone. Sensors are anticipated to be functional across the full range of mesophilic temperatures and most of the range of thermophilic temperatures. In the range of the sensors, microbial activity should produce significant amounts of CO₂, as reported in studies on grains [30,31,32,33]. The greatest range of CO₂ sensors is 20,000 ppm, which occurs when the atmosphere inside the incubator is 2% CO₂ and conditions are still aerobic and not oxygen-limited.

Anaerobic ethanol production, which produces excessive amounts of CO₂ that must be vented to avoid rupture of the incubator, must be avoided. To avoid potential pressure buildup in the incubator, internal pressure should be measured with a mems barometric pressure sensor, which is capable of detecting a maximum of 1200 hPa and provides the most accurate way to quantify the internal pressure of a closed system.

Overall Objective:

• Obtain experimental evidence that cottonseeds exhibit the same mesophilic and thermophilic microbial processes that have been observed with grains.

  ○

  The CO₂ quantified production rates determined experimentally will be used in computational fluid dynamics and CFD models to estimate CO₂ production rates in storage facilities so early warning detection systems for SC can be designed.

Experimental Objective:

• Obtain the CO₂ production rates between cottonseed that are within safe storage conditions, less than 12% moisture content, and cottonseed that has elevated moisture, 25% nominal moisture.
• Determine if CO₂ production rates and sensing technology successfully predict the early stages of SC development.

2. Materials and Methods

CO₂ sensor specifications, limitations, and confounding factors (typical specifications as provided by various CO₂ sensor manufacturers such as www.vasalia.com and www.sensirion.com):

• Accuracy ±35–50 ppm CO₂ + 3–5% of reading (dependent upon sensor)
• Maximum CO₂ level 10,000–20,000 ppm (depending upon sensor)
• Temperature limits of operation to 50 °C to 60 °C (depending upon sensor)
• As reported by the manufacturer on one high-end scientific grade CO₂ sensor [30], CO₂ readings depend upon the following:

  ○

  Temperature

  ○

  Pressure

  ○

  Oxygen concentration

  ○

  Nitrogen concentration

  ○

  H₂O vapor pressure

The above table, from Vasalia, notes the dominant impacts on sensor readings. While their technical note [28] also mentions sensor deviations can also occur from other process variables, such as water vapor pressure and nitrogen concentration, as N2 in the air is unchanging, that is unlikely to be of concern; however, water vapor could be as the purpose is to detect the difference between CO2 produced by very wet seeds versus dry seeds; any effect of water vapor should be included implicitly in the experimental results.

After a review of the manufacturer’s datasheets and application notes, key notes of interest for infrared absorption spectroscopy CO₂ gas sensors:

• For an isobaric system, the CO₂ sensor reading exhibits a strong dependence on temperature and must, therefore, be corrected to obtain accurate readings,
• For an adiabatic system, no heat gain or loss in the system; the system will evolve towards increasing temperature as the microbial production increases, and along with temperature rise, it will, in turn, cause the pressure to increase. Using the standard gas law provides an estimate that 1000 ppm CO₂ at 25 °C and 1013 hPa will when internally heated to 50 °C, result in an increase in pressure to 1100 hPa. In examining Table 1, we see that for this adiabatic closed cylinder the CO₂ reading will not change. So, in a commercial setting with an uncontrolled system, the dependence of CO₂ reading on pressure and temperature, as well as O₂ and N₂, will need to be corrected. In a closed adiabatic system, temperature and pressure correction can be ignored with minimal impact on the accuracy of sensor readings.
• Dependence of CO₂ reading on concentrations of O₂ and N₂. Advanced CO₂ sensors use IR gas sensing, and the absorption bands in the IR region do not overlap for these specific gases; therefore, the dependence is most likely limited to changes in pressure caused by fluctuations in these additional gases. A similar situation will arise with moisture vapor pressure in the system as well.

  ○

  For an adiabatic system undergoing a 1:1 exchange between O₂ and CO₂, pressure should not change so that no correction would be required.

  ○

  Water added to the system from a chemical reaction of microbial activity would be about 1%, so it should not affect vapor pressure measurably.

• In an adiabatic closed system, as in the micro-incubators, once the samples are sealed inside the incubator, the ambient barometric pressure will remain fixed, assuming the container has rigid walls. Thus, the barometric pressure over the course of the test should remain fixed at the same ambient level as when the system was sealed; therefore, for a given experimental setup, all incubators should be sealed under the same barometric conditions. Monitoring barometric conditions in the incubators during an experiment is very desirable. The barometric pressure at the beginning of an experiment when the incubators are sealed should be recorded so CO₂ can be back-corrected to standard temperature and pressure for comparison to other experimental tests.

  ○

  If opting for a simpler apparatus that allows for pressure to equalize via a water-trap air-lock, then pressure and temperature will both have to be accounted for with corrections sent to CO₂ sensors to be applied to the reading or corrected offline via correction equations; however, this approach is not recommended as CO₂ gas will escape utilizing this approach which will confound the targeted objective to measure CO₂ production rate and final CO₂ concentration level and the time to achieve final level.

• Relative humidity should be monitored as this can impact the internal pressure and is required to provide results that can be compared between experimental tests.

Test Sensor Summary:

○

In a closed system, test sensors should be utilized to measure:

▪

CO₂

▪

Temperature

▪

Pressure (barometric pressure provides the most accurate approach)

▪

Relative Humidity

▪

Time (to quantify CO₂ production rate)

▪

Suggested sensor: SCD31 NDIR, infrared gas detector, manufactured by http://www.sensirion.com (accessed 10 June 2024). It measures CO₂, temperature, and relative humidity.

○

In a test incubator that is open to the atmosphere or for future studies where there is a potential for saturated wet regions and test samples that might become anaerobic, additional sensors should be added to help quantify the state of microbial activity:

▪

O₂

• To notify change from aerobic to anaerobic conditions
• Corrections to CO₂ sensor readings

▪

NH₄

• To notify change from aerobic to anaerobic conditions
• Corrections to CO₂ sensor readings

▪

CH₄

• Track potential changes to methanogenesis

Experimental Progression: Aerobic microbial activity is initially mesophilic, then transitioning to thermophilic with an endpoint at 50 °C or CO₂ in excess of 10,000 ppm in a closed adiabatic system design (water jacket designed to eliminate heat transfer out or into a test incubator).

Experimental Results:

• Final CO₂ levels of microbial activity on wet versus dry cottonseeds
• CO₂ production rates throughout the experiment which is a function of test temperature and initial moisture content.
• The ratio of CO₂ production rates and final CO₂ levels between microbial activity on wet cottonseeds versus dry cottonseeds. Feasibility of this as a usable metric for identifying the development of SC in warehouses.
• CO₂ production rates to allow for the development of CFD simulation models to assess the practicality of the approach for commercial application. CFD modeling will help to identify the dilution effects that will occur in commercial facilities where the microbe-produced CO₂ is mixed with ambient air in the headspace. CFD modeling can assist in the design of pulsed ventilation schemes that will allow for both proper ventilation of the product as well as allowing for a build-up of the CO₂ so that sensors can detect developing SC within the stored cottonseed.

2.1. Experimental Micro-Incubator Protocol

A micro-incubator has the advantages of reducing the amount of a commodity volume required for testing as well as reducing the amount of labor and expense and can potentially expedite the results. Lashermes and coworkers reported how to control the external surface temperature of a micro-incubator [30,31]. Their design uses a water jacket with a control system that monitors the internal temperature of the test commodity inside the incubator and adjusts the temperature of the water jacket to maintain the exterior surface of the incubator to be 0.5 °C lower than the internal temperature of the tests commodity inside the incubator. The water jacket limits heat gain and loss, so it achieves adiabatic conditions that can accurately simulate conditions found in bulk storage of a commodity in a warehouse. The use of a temperature-controlled water jacket offers many advantages over incubators where the exterior temperature of the incubator is not controlled, allowing heat to be lost from the microbial activity occurring in the test commodity inside the incubator; however, the simpler, less costly micro-incubator could be useful for pre-study tests of protocols and sensors [32,34]. Bulk storage results, however, cannot be simulated in a micro-incubator without the use of a temperature-controlled water jacket. The following is the description of an experimental trial of the proposed method, where the test is designed to test out the concept, leaving a more detailed test for a full research project that can be run once this protocol has been trialed, assessed, and potentially adjusted as necessary to obtain accurate results.

Micro-incubator Apparatus: The micro-incubator is a steel cylindrical 2.0-litre pressure pot, by Vevor (available from http://www.amazon.com). The pressure pot has a nonstick coating on the inside to prevent the plastic bucket from sticking. A water jacket could be placed around the outside of the pressure pot for precise temperature control. For our experiment, for ease of use, we omitted the water jacket and instead ran the test inside a temperature-controlled laboratory.

Due to the likely pathogenic nature of the fungi growing on the seeds, the system was designed to isolate the user from the fungi completely. This was accomplished by placing dry seed samples inside a specially made 50 mL centrifuge sample jar. The jar was modified with a 0.3-micron PTFE gas exchange filter in the lid to allow CO₂ to escape while retaining the moisture and dangerous fungi safely inside the sample jar. As the wet-up protocol was simply to add an estimated amount of water to each jar. The target moisture was estimated by running a gravimetric oven test on samples of the seeds, which were found to be 7.8% moisture, and then requisite water was added to bring the moisture up to a target of 25% moisture on a wet basis. As fungi are driven by water activity, not moisture content, this key metric was obtained for each jar’s contents using a micro PCB module with a relative humidity sensor that was incorporated into the lids of each sample jar. To obtain measurements from the sensor, the wires were run out through the lid and sealed to the lid. This allowed for later monitoring of the water activity after the contents came to equilibrium moisture over the next several days. Into each micro-incubator, 3 sample jars were placed. This method allowed for safely setting up the experiment with dry seeds that are pathogen-free, and thereafter, as the pathogenic fungi grow, they are safely contained throughout the experiment and can be safely disposed of by trained laboratory personnel in a bio-safety level 2 laboratory after the conclusion of the experiment.

The temperature sensor from each SCD31 sensor located in each micro-incubator was used to monitor temperature as a biotic indicator of the progress of SC.

2.2. Experimental Design

A total of 3 replicate tests, cottonseeds, were run. Dry control tests, 3 replicates, were also run to establish the CO2 production rate of nominal dry cottonseeds, where dry seeds represent typical safe storage moisture contents, 8%, of each material, with wet tests being moisture contents sufficient to achieve equilibrium moisture contents >80–85% RH, that we have estimated to likely occur at nominally 25% moisture content, wet basis, for cottonseed at 25 °C based on [28].

Comparison: Experimental results will compare dry material control CO₂ production rates and final CO₂ levels in test incubators to that of wet material in incubators.

Experimental Termination: The experiment will terminate when either the internal temperature of one of the test incubators is >50 °C or the test duration exceeds once several days of statistically significant CO₂ production rate, compared with the CO₂ production rate of the dry-seeds control group. Note: once the CO₂ level exceeds the maximum sensor capacity, the micro-incubator will be aired out to reset the gas monitoring. While this fresh oxygen may have some effect on the fungi’s growth, it is the best compromise that would allow for near-continuous tracking of CO₂ production rates. We also note that in a real seed pile, oxygen and CO₂ are free to migrate the seed pile, so allowing the CO₂ to build up in a small container is not highly realistic either; so our premise is that periodic ventilation of the chamber is closer to real-world than leaving it undisturbed (which would not allow for measurements past 24 h even at the onset of the test).

Anticipated Products of Research: The research is anticipated to produce a working protocol for the use of off-the-shelf sensors for use in managing cottonseed warehouses in order to prevent spontaneous combustion fires reliably. This is expected to allow the insurance companies to continue to provide insurance policies for operators of cottonseed storage facilities, which are critical for the cotton industry. The research is anticipated to help protect and stabilize the industry from collapse due to the loss of cottonseed storage facilities.

2.3. Biological Safety

A. flavus is classified as a biological safety level 2, BSL-2, organism because of both its ability to cause disease in people with compromised immune systems and asthma, as well as its propensity for producing Aflatoxin, AF. AF has been shown to cause liver cancer and is regulatory limited in the food system to no more than 2 ppb in the EU and no more than 20 ppb in the US. Further danger is presented from airborne spores from A. flavus and especially A. fumigatus, which is a serious infectious agent that causes respiratory disease in human health. For heating studies, A. fumigatus is the primary fungal organism that will take over as temperatures exceed 40 °C. As such, great care is warranted when working with these organisms, and because of the inherent dangers, BSL-2 safety protocol is required by most institutional biosafety committees, IBC, when even a possibility of inadvertently growing them exists in an experimental protocol. It is possible to run a study with BSL-1 organisms by sterilizing the seeds and then inoculating them with BSL-1 fungi. The suggested fungi for low water activity, aw = 0.75, is A. candidus, a BSL-1 fungi. A. candidus was shown by Milner et al., 1947 [22] to produce up to 600 times more CO₂ than dry seeds and is commonly found on most seeds as it, along with A. flavus and A. fumigatus, are all common soil fungi. If an experiment is also designed to track heating, then the selection of A. niger or A. oryzae, as discussed in earlier section, are safer organisms to study than A. flavus or A. fumigatus. Genetically, A. oryzae is nearly identical to A. flavus; however, it is reported to have lower metabolic growth and needs higher aw > 0.9 to thrive. Thus, A. oryzae may not exhibit self-heating at the same production levels as A. flavus. For self-heating studies, A. niger has been shown to run at aw > 0.90 and to foster self-heating, such as A. flavus and A. fumigatus [20,21,22].

2.4. Protocol Examination of: Fixed Temperature Experimental Test

To explore the practicality of the protocol, a test using a modified version of the aforementioned protocol was conducted. The modifications were made to ensure a long-term test could be conducted while keeping the personnel conducting the test isolated from the growing organism, which might be BSL-2 fungi. As the 2.0 L test tank was estimated to likely fill up with CO₂ within a few days at the start of the test and much faster as the fungal colonies expand across the substrate, it was deemed acceptable and necessary to vent the test tank periodically to avoid saturating the CO₂ sensors in order to allow for monitoring of the CO₂ PRODUCTION-RATE {CO₂_PR = CO₂ mg kg⁻¹ (24 h)⁻¹}. There were serious safety concerns with using seeds placed directly into the test tank, as this exposes the worker ventilating the tank each day to the growing fungi and their spores during the venting operation. To avoid this potential exposure, the following changes to the protocol were made. Instead of placing the seeds directly into the test chamber, the seeds were encapsulated into three modified 50 mL centrifuge sample jars. Each jar lid had two holes drilled out; one for the wires for a relative humidity, RH, sensor and the other to allow CO₂ gas to escape from the sample jar, while trapping in any fungal spores or particulates. By embedding an RH sensor into each jar and bringing out the wires, this allowed workers to periodically monitor the interior water activity aw, inside each sample jar without opening it and exposing themselves to the fungi. To allow the CO₂ to escape in the outer test tank and keep in moisture and spores, the sample jar lid was fitted with a 0.3-micron PTFE (Teflon) gas exchange filter that is also hydrophobic so as to prevent loss of moisture. To seal the wires to the lid, hot PVA (polyvinyl alcohol) glue was used to provide a gas-tight seal to the wires. Noting from [22] that they were able to monitor CO₂ production at 30 °C, coupled with the concept that earliest possible detection is the primary response rate of interest to our research and further noting that this will occur shortly after water is added to the seeds while they are at lower temperatures; the experiment was conducted without the micro-incubator temperature control apparatus at target of 25 °C temperature. This greatly simplified the experimental apparatus and streamlined the study. Using results from [22], the outer tank volume and the mass of seeds were estimated to provide for reasonable CO₂ sensor readings over the course of a day, which was projected to give accurate results and provide for a reset of the outer tank atmosphere back to near ambient levels of CO₂, each day. The test was replicated with three CO₂-monitored outer tanks, where into each tank, three replicate sample jars were placed for a total of nine test sample jars. Seeds were sealed inside each sample jar, with a PTFE gas exchange membrane to allow CO₂ to escape into the outer tank, which was fitted with a CO₂ sensor. Before the test began, the three CO₂ sensors, one used in each outer tank, were compared with each other in a side-by-side test comparison with the test being conducted inside a closed box that was filled with CO₂ gas and allowed to equalize to allow for the gas to evenly diffuse to ensure each sensor was seeing the same gas. The CO₂ concentration was brought up to over 10,000 ppm, and the box was slowly vented over 3 h while monitoring the sensors to ensure they were all tracking each other to within the accuracy tolerances provided by the manufacturer’s datasheet over the course of the test. The start of the test monitored natural air-dried seeds straight from the seed storage house for five days before the introduction of water. To gain an initial estimate of water concentration in the air-dried seeds, three samples from the seed pile were obtained, and replicated gravimetric oven moisture samples were conducted. This provided an estimate of 7.8% moisture content starting point for the seeds. From this initial moisture, it was estimated that 3.3 g of added water to 12.1 g air-dried seeds would provide a target moisture content near 25% moisture (MC wet basis) for a target water activity of 0.75 for cottonseeds. After the initial dry seed monitoring time, 5 days, was over, 3.3 g of water was injected into each sample jar using a 1.0 mL syringe, taking care to distribute the water as evenly as possible. Given the sealed sample jar, it was anticipated that in a few days, the water would fully equilibrate throughout the substrate. Observations of the sample, through the clear sample jar (without opening), on day 1 and day 2 indicate that this might not be the case. The relative humidity sensors in each sample jar indicated water activity, aw, levels ranged from 0.79 to 0.92 with an average aw: {Tank 1: 0.88, Tank 2: 0.81, Tank 3: 0.86} see Table 1 for full details. The mean aw for tank 1 was μ_aw = 0.88, tank 2 μ_aw = 0.81, tank 3 μ_aw = 0.85. The overall mean μ_aw = 0.85 with standard deviation σ_aw = 0.04.

Table 1. Water activity, AW, levels for each sample jar, and its assignment to the outer experimental apparatus tank.

Sample Jar	Tank	AW
1a	1	0.86
1b	1	0.86
1c	1	0.92
2a	2	0.83
2b	2	0.79
2c	2	0.82
3a	3	0.82
3b	3	0.83
3c	3	0.91

Table 1. Water activity, AW, levels for each sample jar, and its assignment to the outer experimental apparatus tank.

Sample Jar	Tank	AW
1a	1	0.86
1b	1	0.86
1c	1	0.92
2a	2	0.83
2b	2	0.79
2c	2	0.82
3a	3	0.82
3b	3	0.83
3c	3	0.91

Given that the three sample jars were placed into each tank with 10.6 g seeds (estimated dry weight), each tank held 31.8 g seeds (dry weight) in a 2 L volume. CO₂ concentration in each tank was monitored continuously, along with relative humidity and temperature.

3. Results

For the cottonseed test, one day after water addition, AWA, monitored over 10 h, the mean CO2_PR = 490 CO₂ mg kg⁻¹ (24 h)⁻¹ with a standard deviation of 144. On day two, AWA, the mean CO₂_PR = 2544 CO₂ mg kg⁻¹ (24 h)⁻¹ with a standard deviation of 665, and on day 4, mean CO₂_PR = 6965 CO₂ mg kg⁻¹ (24 h)⁻¹ with a standard deviation of 616. Figure 2 details a graph of the CO₂ concentration, in the units of mg kg⁻¹ substrate, that was obtained by converting from the sensor’s ppm concentration reading from tank #2 and shows the progression of increasing CO₂ production over 4 days, post AWA (chart created by python code included in Supplementary Materials).

The code utilized to perform the plots and compute the conversion from ppm to mg per kg-day are included as Supplementary Materials along with the report. Note the units are corrected from CO₂ ppm to transferable units of mg CO₂ per kg substrate (fuzzy gin-run cottonseeds in this case).

Table 2. Replicated Test Results; units: mg CO₂ (kg_seeds 24 h)⁻¹.

Control	Day 1	Day 2	Day 3	Day 4
−5.1	382.0	2077.0	4154.0	6555.0	Tank 1
−45.7	436.0	2225.0	3905.0	6668.0	Tank 2
16.5	653.0	3304.0	5338.0	7673.0	Tank 3
−11.4	490.3	2535.3	4465.7	6965.3	mean
31.6	143.4	669.8	765.7	615.5	Stdev.

shows the results of the CO₂ production rate for the Control, day 1 and day 2. Analyzing with Tukey’s ad-hoc honest-significant-differences, HSD, post-comparison test yields no significant difference between day 1 and control (p = 0.41) but does result in a significant difference between day 2 versus control (p = 0.0001) and between day 2 and day 1, p = 0.0003 with similar results for days 3 and 4 (

Table 3. Tukey’s Multiple Comparison of Means—Tukey HSD.

Group 1	Group 2	Mean Diff.	p-Adjusted	Reject Ho	Label
Control	Day 1	501.8	0.7775	FALSE	a
Control	Day 2	2556.8	0.0012	TRUE	a
Control	Day 3	4477.1	0.0000	TRUE	a
Control	Day 4	6976.8	0.0000	TRUE	a
Day 1	Day 2	2055.0	0.0057	TRUE	b
Day 1	Day 3	3975.3	0.0000	TRUE	b
Day 1	Day 4	6475.0	0.0000	TRUE	b
Day 2	Day 3	1920.3	0.0091	TRUE	c
Day 2	Day 4	4420.0	0.0000	TRUE	c
Day 3	Day 4	2499.7	0.0014	TRUE	c

).

4. Summary

Cottonseed and peanut storage warehouses are subject to costly destruction due to an inherent predilection of the commodities for spontaneous combustion, SC. This report covers the development of an experimental protocol that will be utilized in upcoming research to develop an early warning system. A feasibility study was conducted utilizing a modified version of the protocol that includes safety features to enable an isolated culture of the fungi on seeds that provides a safe method to observe CO₂ production as fungal colonies spread across the substrate without exposing the workers to the fungi and it is potentially toxic spores and mycotoxins. As the test used seeds from a cottonseed storage facility, there was a high probability of the seeds having A. candidus, A. flavus, A. niger, and A. fumigatus. Before the addition of water, CO₂ production was not observable above the noise floor of the sensors. Fungal growth was observed, via CO₂ monitoring, to have a rapid onset, with only a 1–2 days lag, resulting in high CO₂ production levels. On day four, after water addition, the mean production rate of three replicates was CO₂_PR = 6965 CO₂ mg kg⁻¹ (24 h)⁻¹. As the literature clearly shows that the fungi’s metabolic respiration rate is tied to both water activity, aw, and temperature; these values should be utilized as references, not as absolute values as changes in either the dominant fungal organism, temperature or aw, will result in significant deviations from these values. Of interest is that the high metabolic respiration rates required for SC are easily measured as significant increases in CO₂ production rates. As such, early detection utilizing a CO₂ production rate monitor should be feasible for cottonseed storage facilities as fungal growth on fuzzy cottonseed is even more productive than is reported for common cereal grains. The feasibility study used a non-temperature controlled apparatus, leveraging published results from the literature [22] which found CO₂ production could be observed at slightly elevated laboratory room temperatures. The tests were replicated trials and deemed successful. The exploration utilizing an adiabatic test chamber that seeks to mimic seed pile conditions that are commonly observed in cottonseed and peanut storage facilities is left to future research where the full protocol enables observation of the cascading progression of microbial biotic activity as the temperature increases from ambient to >50 °C that will most likely see a transition from A. flavus, or A. niger, to A. fumigatus as the temperature increases above and beyond 40 °C. If the researcher’s needs dictate an exploration of the change-over in fungal dominance that occurs with elevated temperature, the full protocol will provide the necessary conditions to support this exploration. For the case of early detection of wet seeds at risk of progressing to SC, the fixed-temperature protocol tested and described here should be sufficient. In either case, the design of experiments presented herein adapts to the available sensors and apparatus to provide two scientifically sound protocols for conducting transferable trials that will result in data that can be directly transferred for use in computational fluid dynamic models that will be required in commercial developments of early warning systems. To ensure maximum safety while conducting a study, the reader is referred to Section 2.4 fixed-temperature protocol as the sealed sample jar with permeable gas exchange membrane provides the safest approach for inexperienced engineers who need to stretch into mycology to seek answers.