Introduction

Thermocouple probes are a standard piece of fire science technology. Thermocouples function on the Seebeck effect: the thermoelectric signals created along temperature gradients in a wire. Thermocouples place wires of two different metal types into contact, and differences in each materials’ rate of change in thermoelectric signals as the temperature of each wire changes scales in a predictable manner (Shannon and Butler 2003; Pavlasek et al. 2015). This self-generated voltage difference is detectable by electronic dataloggers.

Thermocouples have been used in the context of wildland fire since at least the 1940s (Vaartaja 1949; Fons 1946). Wildland fire scientists use thermocouples to measure two variables related to fire behaviorflame temperature and rate of spreadwhich are both useful in determining the safety of fire suppression efforts and understanding ecological responses to fire (Rothermel 1972, 1983; Engle et al. 1989; Iverson et al. 2004). Granted, flame temperature alone is a poor indicator of fire intensity, which is the the actual amount of energy released through combustion (Kremens, Smith, and Dickinson 2010; Smith, Cowan, and Fitzgerald 2016). But in the lab, thermocouples are useful in measuring components of flammability (Simpson et al. 2016; Grootemaat et al. 2017; Gao and Schwilk 2018). And in the field, with enough spatial replication, thermocouples can be used to measure rate of spread (e.g., Davies et al. 2009), which is an important variable in calculating fire intensity (Rothermel 1972; Byram 1959).

But heretofore, replicating thermocouple sampling is challenged by the relatively high cost of commercial datalogging systems. Electronic technology has advanced over recent decades, and several aspects of dataloggers for environmental monitoring have improved substantially. For example, today’s dataloggers can record data at millisecond scales to decimals of degrees Celsius, whereas in 1978 Trollope needed multiple “electronic temperature recorders” for different temperature ranges that only logged in 20oC increments (Trollope 1978). A decade later, Engle et al. (Engle et al. 1989) were only logging flame temperature once every two seconds onto a magnetic tape recorder. But while the equipment has improved, the “cost per channel” of dataloggers hasn’t fallen low enough such that most researchers can afford to operate the number of sensors necessary to adequately sample inherently variable processes like wildland fire.

Of course, the prohibitive cost of technology is not a problem unique to fire scientists. Many low-cost solutions have been developed from do-it-yourself (DIY) platforms such as Arduino (www.arduino.cc), an open-source system of microcontrollers and sensors that have been widely applied in environmental research (e.g., Barnard, Findley, and Csavina 2014; Greenspan et al. 2016; Shipley et al. 2017; McGranahan, Geaumont, and Spiess 2018). The computational accuracy of Arduino microprocessors is reported to be sufficient for use in psychological and neurophysiology experiments (D’Ausilio 2012). But the challenges of the wildland fire environment are unique, and the ability of DIY microcontrollers to record as accurately, quickly, and consistently as commercially-available systems. Not only must dataloggers be comparable to industrial-grade thermocouples capable of withstanding extreme heat and flame exposure, they must also be small enough to survive such exposure themselves in the field and portable enough but with sufficient battery life to deploy ahead of fire events in remote locations. Large expensive dataloggers require expensive and/or labor-intensive solutions to protect from these conditions: For example, Jacoby et al. (Jacoby, Ansley, and Trevino 1992) and Butler et al. (Butler et al. 2010) both spent approximately US$500 to build fire-proof boxes to enable their Campbell Scientific dataloggers to record data during wildland fires, and Wally et al. (Wally, Menges, and Weekley 2006) buried theirs in the soil.

We describe here an open-source, DIY datalogging system that senses and records the temperature of standard industrial K-type thermocouples to a degree comparable to a top-of-the-line proprietary commercial system. We compare the performance of the two systems under two different environmental conditionsthe constant environment of a drying oven, and the variable environment of flame exposure. In each comparison, the null hypothesis is that there is no difference between the systems when recording the same sensors in the same environment, and we do not expect to find sufficient statistical evidence to reject this null hypothesis.

Methods

Systems compared

Both the proprietary, commercial system and the open-source, do-it-yourself (DIY) system were compared with the same metal overbraided ceramic fiber insulated thermocouples and fiberglass-coated thermocouple extension wire constructed of Omega Engineering’s chromel and alumel alloys for K-type thermocouple sensors, manufactured within standard limits of error (Omega Engineering Inc, Norwalk, CT). This study focused on K-type thermocouples because they are by far the most-frequently used in wildland fire science on account of their wide range of operating temperatures (McGranahan 2020). Both systems used thermocouple leads constructed from the extension wire and quick-disconnect chromel-alumel couplers (Fig. fig:systems) to connect to metal overbraided thermocouple probes with exposed welded beads. Both systems sampled and recorded thermocouple temperatures at 2 Hz frequencies.

The datalogging systems compared in laboratory trials. Left: The Campbell Scientific CR1000 is a commercial system that represents an industry standard in environmental monitoring. Note that while on-board memory is an option, we used Campbell Scientific’s proprietary software on a Windows laptop computer to monitor thermocouple status and save data in real time. Right: A do-it-yourself (DIY) thermocouple datalogging system based on the open-source Arduino system. Data are saved to an SD card and thermocouple status monitored via the optional OLED screen.

The datalogging systems compared in laboratory trials. Left: The Campbell Scientific CR1000 is a commercial system that represents an industry standard in environmental monitoring. Note that while on-board memory is an option, we used Campbell Scientific’s proprietary software on a Windows laptop computer to monitor thermocouple status and save data in real time. Right: A do-it-yourself (DIY) thermocouple datalogging system based on the open-source Arduino system. Data are saved to an SD card and thermocouple status monitored via the optional OLED screen.

Proprietary commercial system

We used a Campbell Scientific CR1000 measurement and control datalogger, a commercial system widely considered an industry standard for environmental monitoring, including wildland fire behavior (e.g., Butler et al. 2010). The CR1000 is programmed via CRBasic, a proprietary BASIC-like language, and data are accessed with proprietary software such as the free-of-cost PC200W package. The CR1000 features eight differential analog inputs capable of reading K-type thermocouples with no additional hardware; that the sensors are K-type thermocouples is specified in the CRBasic program. Limits of error for K-type thermocouples up to 1250oC are \(\pm\) 2.2oC (Campbell Scientific 2018). The CR1000 was connected to a laptop computer running Windows 7 to monitor thermocouple status and record data in real time (Fig. fig:systems, Left).

Open-source DIY system

For laboratory comparison against the CR1000, we constructed a thermocouple datalogging system based on the open-source Arduino system (Fig. fig:fritzing). Hardware included an Arduino MEGA 2560 board (arduino.cc), which uses the Atmel ATmega2560 microcontroller chip. The Arduino MEGA is programmed via the open-source and freely-available Arduino Independent Development Environment, based on the open-source, object-oriented Processing language (processing.org) that is very familiar to users of C++.

Other hardware included breakout boards from Adafruit Industries (Brooklyn, NY; adafruit.com). Each thermocouple was connected to the MEGA board via MAX31855 thermocouple amplifiers; each connection included a 0.1\(\mu\)F ceramic capacitor for noise reduction. Adafruit reports these boards produce temperature readings for K-type thermocouples within \(\pm\) 26oC. The status of thermocouples was monitored by a 128x32 pixel OLED display.

Data were written to a removable SD card via the SD datalogger shield in human-readable format; files are accessible via any program capable of reading .txt and .csv files. Each observation was stored as a comma-delimited text string. Arduino sketches (code) are available online; see link in Appendix app:links. Full component lists and costs for each system are outlined in Appendix tab:parts.

Schematic diagram of the Arduino MEGA thermocouple datalogging system. Components include: A: Arduino MEGA microcontroller board. B: MAX31855 breakout board for thermocouple signal amplification. Six shown here, allowing space to show connections that are covered up when breakout boards for eight sensors are connected (see Fig. fig:systems). As per adafruit instructions, each board includes a 0.1\muF capacitor for electromagnetic noise reduction. C: Datalogging shield for writing to removable SD memory card. D: 128x32 OLED display. Kit includes HC4050 level shifter and 220\muF capacitor as depicted on the breadboard.

Schematic diagram of the Arduino MEGA thermocouple datalogging system. Components include: A: Arduino MEGA microcontroller board. B: MAX31855 breakout board for thermocouple signal amplification. Six shown here, allowing space to show connections that are covered up when breakout boards for eight sensors are connected (see Fig. fig:systems). As per adafruit instructions, each board includes a 0.1\(\mu\)F capacitor for electromagnetic noise reduction. C: Datalogging shield for writing to removable SD memory card. D: 128x32 OLED display. Kit includes HC4050 level shifter and 220\(\mu\)F capacitor as depicted on the breadboard.

Experimental design

We conducted two experiments to compare the performance of the open-source, DIY system against the proprietary, commercial system. In the first, we compared how closely the two systems recorded thermocouple responses to constant temperatures produced by a drying oven; this experiment was designed to determine whether the systems agreed on the temperature of the environment in which the thermocouples were placed. In the second, we compared how similarly the two systems recorded thermocouple responses to variable heat inputs produced by a Bunsen burner; this experiment was designed to determine whether the systems differed in their capacity to detect changes in the temperature of the environment in which the thermocouples were placed.

Recording constant temperature

To determine whether the systems agreed on the ambient temperature, we routed seven thermocouple probes into a drying oven via the top vent and arranged probe termini such that they were neither in contact with each other nor with any metal surfaces inside the drying oven. Four rounds of trials were conducted at four settings on the drying oven. Although the drying oven did not allow users to specify temperatures, temperatures recorded for the four settings ranged as follows: 48-58oC, 105-120oC, 165-180oC, and 220-240oC.

During each round, one system was connected to the thermocouples and allowed to record temperature for 12 hrs, after which the systems were switched and the other system allowed to record temperature for 12 hrs. Once both systems had recorded temperature for 12 hrs at each setting, the setting of the drying oven was increased and logging began again once the thermostat on the drying oven reached the new equilibrium temperature. Data from this experiment consisted of mean temperature readings for each system from five, randomly-sampled 30-min periods from the 12 hrs of data, each separated by at least 30 min.

Response to variable heat input

To determine whether the systems differed in their capacity to detect temperature changes, we created an array of thermocouples at four positions above a Bunsen burner in a fume hood (Fig. fig:FumeHood). At each position, metal test tube holders held four thermocouple probes in as close proximity as possible, with two leading to each system for a total of eight thermocouples per system (three pairs were discarded prior to analysis due to faulty readings attributable to the thermocouple leads themselves, not the datalogging systems).

Design of the laboratory comparison of datalogger responses to variable heat input. (A) Arrangement of thermocouples relative to Bunsen burner on the test tube stand. Two thermocouples per datalogger were placed in each of the four clamps. (B) An example of the Bunsen burner, thermocouples, and thermocouple leads in the fume hood.

Design of the laboratory comparison of datalogger responses to variable heat input. (A) Arrangement of thermocouples relative to Bunsen burner on the test tube stand. Two thermocouples per datalogger were placed in each of the four clamps. (B) An example of the Bunsen burner, thermocouples, and thermocouple leads in the fume hood.

Five different trials were conducted, in which the thermocouples were exposed to flame from the Bunsen burner for increasingly-longer periods: 10 s, 30 s, 1 min, 2 min, and 4 min. Each trial began with the burner off and the thermocouples all registering ambient air temperature.

Data analysis

To test the correlation between temperature readings among the two systems, the five, 30-min mean temperatures for each oven setting were fit as the response (ArduinoMEGA) and predictor (CR1000) variables in linear mixed-effect regression (LMER) models using the lmer function from the lme4 package in the R statistical environment (Bates et al. 2015; R Core Team 2019). To control for variability among thermocouples, we fit thermocouple as a random intercept effect. Because responses varied significantly among the four temperature trials in a comparison of models with and without the categorical trial term (\(\chi^2\) = 191.42, \(P <\) 0.001), individual LMER models were fit for each oven setting. From each model we extracted \(\beta_1\) regression coefficients and 95% confidence intervals with functions fixef and confint, respectively. The degree to which regression coefficients differed from the expected relationship between systems was assessed by whether or not the 95% CI of \(\beta_1\) included 1.0 (slope of a perfect positive correlation) for each oven setting.

To determine the similarity among systems in response to temperature changes induced by heat input from the Bunsen burner, we fit generalized linear mixed-effect regression models (GLMM) using the glmmabmb function in the glmmADMB package for R (Fournier et al. 2012). Data were fit to logistic growth curves to best describe the curvilinear, asymptotic trends of temperature increase and stabilization. Due to the differences in position with respect to the Bunsen burner, variability among thermocouples was modeled as both fixed and random intercept effects to compare the relative magnitude of variability among thermocouples to that between systems. Thus, \(\beta\) regression coefficients and 95% confidence intervals were extracted for both terms from each GLMM.

Results and Discussion

Systems comparison

Recording constant heat

Both systems tended to agree on ambient temperatures in the drying oven. Differences between systems were within the \(\pm\) 26oC range of error associated with K-type thermocouples, which sacrifice high precision for lower cost and a wide range of operating temperatures. Indeed, there was greater variation among the thermocouples themselves than among the systems within readings for the same thermocouple (Fig. fig:oven_graphs-1).

Relationship between temperature readings of seven thermocouples by two datalogging systemsthe open-source, DIY Arduino MEGA and proprietary, commercial Campbell Scientific CR1000 (Fig. fig:systems) at four settings of a drying oven (see axis ranges for temperature values of each setting). Black diagonal lines represent the expected 1:1 relationship of a positive correlation between temperature readings among systems; values above the line indicate the Arduino MEGA recorded higher temperatures on the same thermocouple, while values below the line indicate the CR1000 recorded higher temperatures on that thermocouple. Colors represent unique thermocouples. Points denote mean values for 5, 30-min periods, and error bars show mean and standard deviation for each thermocouple.

Relationship between temperature readings of seven thermocouples by two datalogging systemsthe open-source, DIY Arduino MEGA and proprietary, commercial Campbell Scientific CR1000 (Fig. fig:systems) at four settings of a drying oven (see axis ranges for temperature values of each setting). Black diagonal lines represent the expected 1:1 relationship of a positive correlation between temperature readings among systems; values above the line indicate the Arduino MEGA recorded higher temperatures on the same thermocouple, while values below the line indicate the CR1000 recorded higher temperatures on that thermocouple. Colors represent unique thermocouples. Points denote mean values for 5, 30-min periods, and error bars show mean and standard deviation for each thermocouple.

At the lowest temperature setting (48-58oC), the Arudino MEGA registered temperatures 18oC higher than the Campbell Scientific CR1000, which was the only case of 95% confidence intervals not including 1.0 (Fig. fig:OvenCIs). At the third-highest setting (165-180oC) the CR1000 tended to register temperatures  3oC higher than the Arduino MEGA. At the highest setting (220-240oC) the two systems had a very high level of agreement, with slopes no different than 1.0 and a very narrow 95% confidence interval (Fig. fig:OvenCIs).

Deviation from the expected 1:1 relationship (\beta_1 = 1.0) between temperature recordings by two datalogging systemsthe open-source, DIY Arduino MEGA and proprietary, commercial Campbell Scientific CR1000 (Fig. fig:systems)at four settings of a drying oven (see Methods and Fig. fig:oven_graphs-1 for temperature ranges of each setting).

Deviation from the expected 1:1 relationship (\(\beta_1\) = 1.0) between temperature recordings by two datalogging systemsthe open-source, DIY Arduino MEGA and proprietary, commercial Campbell Scientific CR1000 (Fig. fig:systems)at four settings of a drying oven (see Methods and Fig. fig:oven_graphs-1 for temperature ranges of each setting).

Response to variable heat input

Both systems similarly registered the temperature response of thermocouples exposed to the Bunsen burner. As above, there was greater variability among thermocouples than between systems (Fig. fig:BBplots). While the Arduino MEGA tended to record temperatures at an average 5% higher than the CR1000 (\(\beta\) = 0.05, 95% CI 0.020.09), the average variability among thermocouples was an order of magnitude greater (\(\beta\) = 0.22, 95% CI 0.160.27).

Variability in response to five durations of exposure to flame from a Bunsen burner among two datalogger systemsthe open-source, DIY Arduino MEGA and proprietary, commercial Campbell Scientific CR1000 (Fig. fig:systems) and five pairs of thermocouples (for placement see Fig. fig:FumeHood). Broken black lines show the logistic regression curves for each of the five thermocouples included in this analysis fit with the drm function from the drc package in R (Ritz et al. 2015). Solid lines show variability among each thermocouple by system.

Variability in response to five durations of exposure to flame from a Bunsen burner among two datalogger systemsthe open-source, DIY Arduino MEGA and proprietary, commercial Campbell Scientific CR1000 (Fig. fig:systems) and five pairs of thermocouples (for placement see Fig. fig:FumeHood). Broken black lines show the logistic regression curves for each of the five thermocouples included in this analysis fit with the drm function from the drc package in R (Ritz et al. 2015). Solid lines show variability among each thermocouple by system.

On the applicability of the DIY system

We’ve demonstrated that open-source, do-it-yourself solutions to expensive thermocouple dataloggers are not only feasible in terms of cost and durability but also comparable in performance to an industry-standard commercial system. While neither out-of-the-box DIY microcontrollers nor the systems described here can fully substitute for the breadth of functionality provided by commerical systems like the CR1000, we show here that DIY systems can be relied on to collect robust data on fire behavior in the wildland fire environment. Although the Arduino system based on the ATmega microprocessor did not record temperatures exactly like the Campbell Scientific CR1000, the minor discrepancies are within the margins of error associated with using K-type thermocouples in the wildland fire environment. K-type thermocouples are expected to vary by \(\pm\) 26oC; this variability is apparent in results from both of our trials, in which recorded temperatures among thermocouples varied more than between the systems (Figs. fig:oven_graphs-1 & fig:BBplots).

Moreover, responsiveness to changes in the media around the thermocouple probeair, soil, or plant tissue, depending on the in agris contextis often of greater interest to the wildland scientist than precision in temperature readings. The data recorded when the microcontroller samples the electrical signals from the thermocouple probe are in fact a measure of the temperature of the thermocouple wires themselves, not of the surrounding media. As heat is applied or remains constant, thermocouples necessarily underestimate actual temperature (e.g., Iverson et al. 2004); after peak heating, thermocouples necessarily overestimate temperature as the beads release heat through convection (McGranahan 2020). The difference in probe vs. medium temperatureand the rate of changeis a function of the material in the thermocouple, the size of the bead fusing the wires, and the magnitude of the temperature difference (Shannon and Butler 2003; Blevins and Pitts 1999; Lemaire and Menanteau 2017). While better estimates of the temperature of the surrounding medium can be determined by extrapolating from the responses of two coincident probes of different diameters (Walker and Stocks 1968), wildland fire scientists are likely best off using the thinnest possible (i.e., most responsive) probes and a high-frequency microcontroller to detect the arrival of flame fronts. Arduino boards have been shown to function accurately at these high frequencies (D’Ausilio 2012).

From a budget standpoint, an effective way to compare datalogger systemsespecially as the capacity for sampling frequency, data storage, and even wireless communications continues to increaseis cost per channel. Among commercial systems, less-expensive dataloggers typically support fewer channels; for example, among HOBO dataloggers (Onset Computer Corporation, Bourne, MA), the single-channel U12 model used frequently in fire ecology (e.g., Strong, Ganguli, and Vermeire 2013; Kral et al. 2015; Russell et al. 2015) costs US$140, while the four-channel UX120-014M costs US$275. Supplies like industrial-grade thermocouple probes, connectors, and lead wires are fixed regardless of the system used, and scale linearly with the number of sensor replicates deployed. Thus, when overbraided, insulated thermocouple probes are included (Table tab:parts), the cost per channel of these systems are approximately US$180 and US$110, respectively. Even with eight channels, the CR1000an industry-standard for environmental monitoringstill costs approximately US$250 per channel. Note that none of these calculations include the cost of protecting these systems from exposure to the extreme conditions of wildland fire.

To extend the functionality of the open-source, DIY thermocouple datalogging system as a solution for measuring flame temperatures during wildland fires in situ–or rather in agris, “in the field”–we also developed a compact, six-channel, mobile Arduino-based thermocouple datalogging system, which is described briefly in Appendix 6 and Table tab:parts. This step was essential because the Arduino MEGA only solves one of the challenges of commercial systemsprohibitive costbut was still fairly large and relied upon a 110V wall outlet for power. Measuring temperatures of wildland fire presents additional challenges, namely a battery-powered device that can itself withstand or be protectable against the extremely high temperatures of wildland fire (Jacoby, Ansley, and Trevino 1992; Butler et al. 2010; Wally, Menges, and Weekley 2006).

DIY thermocouple systems invert the cost-per-channel calculus because their base components are less expensive and fairly independent of the number of channels employed. Even with the additional US$15 per channel to add thermocouple amplifiersmost commercial systems can receive K-type thermocouples directlya single-channel in agris system costs about US$107 before equipment to protect it from exposure to fire. Cost-per-channel for a four-sensor system falls to US$72, and the six-sensor system described in Fig. fig:FeatherFlame cost US$68 (US$80 per channel with all fire shields included). And Arduino-based systems can support many more sensors than used heredepending on the application, microcontroller, and additional sensors, input capacity of Arduino-based systems range from 10 analog inputs or 20 digital input/output pins, to 16 analog inputs and 54 digital input/output pins. However, if high sampling frequencies are required, users must write more technical programs that handle data in a machine-readable format, as processing the text strings required for human-readable file formats can slow the microprocessor.

The combined benefits of cheap, multi-sensor units is a boon for the wildland fire scientist seeking technology capable of addressing the inherent variability of wildland fuelbeds. Patterns of variability in wildland fire behavior have been described at spatial scales \(<\) 10 m-2 (Gimeno-García, Andreu, and Rubio 2004; Hiers et al. 2009). Fine-grain sampling across wildland landscapes requires a spatially-hierarchical approach for which our in agris system is ideally suitedmultiple dataloggers with multiple sensors per datalogger can be deployed throughout the burn area in a manner conducive to geospatial analysis (e.g., Zopfi 2020).

Acknowledgements

We recognize support from the North Dakota State Agricultural Experiment Station and USDA-NIFA Hatch project number ND02393. Dr. Aaron Daigh made valuable contributions to the design of these experiments.

Conflict of interest

The authors declare that they have no conflict of interest. All products used herein were purchased by the authors at standard market prices.

Appendix B. Components and prices for thermocouple datalogging systems

Components for the three thermocouple datalogger systems described in this paper. The CR1000 represents an industry-standard proprietary commercial system, while the ArduinoMEGA and FeatherFlame (App. 6) systems are based on open-source, do-it-yourself technology. Abbreviated sources include ADA = Adafruit Industries (adafruit.com), AMZ = amazon.com. Feather and FeatherWing components are Adafruit products. Quantities and costs in italics are variableat least one is needed for basic functionality, systems as described use values from the table. Ranges in total cost reflect differences between these quantities. Costs listed in US Dollars. \(\dagger\)Components optional for basic functionality but support the specific systems described here.
System Component (source) Cost each Quantity Total cost
Campbell Scientific CR1000
CR1000 datalogger (Campbell Scientific) $1,500 1 $1,500
12V power supply (Campbell Scientific) $275 1 $275
K-type quick-connect thermocouples (omega.com) $35 8 $280
K-type extension wire\(\dagger\) (omega.com) $32 1 $32
Arduino MEGA
Arduino MEGA microcontroller board (arduino.cc) $40 1 $40
SD card data logging shield (ADA) $14 1 $14
128x64 OLED kit\(\dagger\) (ADA) $35 1 $35
16 GB SD card (AMZ) $6 1 $6
9V power supply (ADA) $7 1 $7
Prototyping breadboard (ADA) $6 1 $6
10-pack 0.1\(\mu\)F ceramic capacitors (ADA) $2 1 $2
MAX31855 breakout board (ADA) $15 8 $120
K-type quick-connect thermocouples (omega.com) $35 8 $280
K-type extension wire \(\dagger\) (omega.com) $32 1 $32
Feather Flame complete in agris system
Feather M0 datalogger $20 1 $20
FeatherWing 128x32 OLED \(\dagger\) $14 1 $14
FeatherWing DS3231 precision real-time clock \(\dagger\) $14 1 $14
Battery: 3.7V 400mAh Lithium Ion Polymer (ADA) $7 1 $7
32 GB micro SD card (AMZ) $10 1 $10
10-pack 0.1\(\mu\)F ceramic capacitors (ADA) $2 1 $2
MAX31855 breakout board (ADA) $15 6 $90
Pelican brand 1020 case \(\dagger\) (AMZ) $20 1 $20
HVAC end cap, 26-ga galvanized 10" \(\dagger\) (AMZ) $18 1 $18
Junction box, lid, 2x squeeze connectors \(\dagger\) (AMZ) $20 1 $20
Flexible metal conduit \(\dagger\) (AMZ) $20 1 $20
K-type quick-connect thermocouples (omega.com) $35 6 $210
K-type extension wire \(\dagger\) (omega.com) $32 1 $32

Appendix C. FeatherFlame: A DIY solution for field measurements

The Arduino-based Feather system from Adafruit Industries provides a compact solution for recording thermocouple responses in situ during wildland fires. (A) Three stackable Feather boards (Table tab:parts) replace the MEGA board, datalogging shield, and OLED display in the laboratory system (Fig. fig:fritzing). The board has \muSD removable storage and an ATmega microcontroller. Battery not shown. (B) An example of how the FeatherFlame is deployed in agris. We affix three thermocouple probes to rods that form a 1m equilateral triangle 15cm from the ground, a fourth probe on the soil surface, and a fifth probe in the center of the triangle. (C) The dataloggers are protected from surface flame fronts by first scraping away vegetative matter so the box can be placed on mineral soil, then covering with a galvanized steel HVAC end cap. Dataloggers are placed away from the probe array to minimize disruption to fuels around the probes, which is made possible by leads protected by flexible metal conduit or high-temperature foil HVAC tape. (D) The steel junction box protects the connectors between the overbraided thermocouple probes and the leads from the datalogger.

The Arduino-based Feather system from Adafruit Industries provides a compact solution for recording thermocouple responses in situ during wildland fires. (A) Three stackable Feather boards (Table tab:parts) replace the MEGA board, datalogging shield, and OLED display in the laboratory system (Fig. fig:fritzing). The board has \(\mu\)SD removable storage and an ATmega microcontroller. Battery not shown. (B) An example of how the FeatherFlame is deployed in agris. We affix three thermocouple probes to rods that form a 1m equilateral triangle 15cm from the ground, a fourth probe on the soil surface, and a fifth probe in the center of the triangle. (C) The dataloggers are protected from surface flame fronts by first scraping away vegetative matter so the box can be placed on mineral soil, then covering with a galvanized steel HVAC end cap. Dataloggers are placed away from the probe array to minimize disruption to fuels around the probes, which is made possible by leads protected by flexible metal conduit or high-temperature foil HVAC tape. (D) The steel junction box protects the connectors between the overbraided thermocouple probes and the leads from the datalogger.

Our compact, mobile Arduino-based datalogger relies on the Feather system from Adafruit Industries (Brooklyn, NY), which also uses an ATmega microcontroller, is powered by a compact, rechargeable lithium-ion battery, and stores data on a removable \(\mu\)SD card (Fig. fig:FeatherFlame). The unit fits into a small airtight box (we used the Pelicantm 1020 case), which we shielded from flame and radiant heat with a galvanized steel cap (Fig. fig:FeatherFlame). The reliability of the datalogger system and the fire protection components were proven over the course of 243 individual datalogger deployments in 27 prescribed fires in North Dakota, USA (Zopfi 2020). While the specifics of the FeatherFlame systemincluding the spatial sampling scheme based on the low-cost/high-replication advantages of the DIY system and R scripts used for analysisare described elsewhere (e.g., Zopfi 2020), links to online information about the construction of the FeatherFlame system are provided in Appendix app:links.

References cited

Barnard, H. R., M. C. Findley, and J. Csavina. 2014. “PARduino: A Simple and Inexpensive Device for Logging Photosynthetically Active Radiation.” Tree Physiol. 34 (6):640–45. https://doi.org/10.1093/treephys/tpu044.

Bates, Douglas, Martin Mächler, Ben Bolker, and Steve Walker. 2015. “Fitting Linear Mixed-Effects Models Using Lme4.” J Stat Softw 67 (1):1–48. https://doi.org/10.18637/jss.v067.i01.

Blevins, Linda G, and William M Pitts. 1999. “Modeling of Bare and Aspirated Thermocouples in Compartment Fires.” Fire Saf. J. 33 (4):239–59.

Butler, B W, D Jimenez, J Forthofer, K Shannon, and P Sopko. 2010. “A Portable System for Characterizing Wildland Fire Behavior.” In Proceedings of the VI International Conference on Forest Fire Research, edited by D. X. Viegas, 13. Coimbra, Portugal: University of Coimbra.

Byram, G.M. 1959. “Combustion of Forest Fuels.” Edited by K.P. Davis. Forest Fire: Control and Use, 61–89.

Campbell Scientific. 2018. “CR1000 Datalogger Operator’s Manual.” Logan, Utah USA.

D’Ausilio, Alessandro. 2012. “Arduino: A Low-Cost Multipurpose Lab Equipment.” Behav. Res. Methods 44 (2):305–13. https://doi.org/10.3758/s13428-011-0163-z.

Davies, Gwilym M., Colin J. Legg, Alexander A. Smith, and Angus J. MacDonald. 2009. “Rate of Spread of Fires in Calluna Vulgaris -Dominated Moorlands.” J. Appl. Ecol. 46 (5):1054–63. https://doi.org/10.1111/j.1365-2664.2009.01681.x.

Engle, DM, TG Bidwell, AL Ewing, and JR Williams. 1989. “A Technique for Quantifying Fire Behavior in Grassland Fire Ecology Studies.” Southwest Nat 34 (1):79–84.

Fons, Wallace L. 1946. “Analysis of Fire Spread in Light Forest Fuels.” J Agr Res 72 (3):93–121.

Fournier, David A, Hans J Skaug, Johnoel Ancheta, James Ianelli, Arni Magnusson, Mark N Maunder, Anders Nielsen, and John Sibert. 2012. “AD Model Builder: Using Automatic Differentiation for Statistical Inference of Highly Parameterized Complex Nonlinear Models.” Optim. Method. Softw. 27 (2). Taylor & Francis:233–49.

Gao, Xiulin, and Dylan Schwilk. 2018. “Grass Canopy Architecture Influences Temperature Exposure at Soil Surface.” Fire 1 (3):35.

Gimeno-García, E, V Andreu, and J.L Rubio. 2004. “Spatial Patterns of Soil Temperatures During Experimental Fires.” Geoderma 118 (1-2):17–38. https://doi.org/10.1016/S0016-7061(03)00167-8.

Greenspan, Sasha E., Wayne Morris, Russell Warburton, Lexie Edwards, Richard Duffy, David A. Pike, Lin Schwarzkopf, and Ross A. Alford. 2016. “Low-Cost Fluctuating-Temperature Chamber for Experimental Ecology.” Methods Ecol. Evol. 7 (12):1567–74. https://doi.org/10.1111/2041-210X.12619.

Grootemaat, Saskia, Ian J. Wright, Peter M. van Bodegom, and Johannes H. C. Cornelissen. 2017. “Scaling up Flammability from Individual Leaves to Fuel Beds.” Oikos 126 (10):1428–38. https://doi.org/10.1111/oik.03886.

Hiers, J. Kevin, Joseph J. O’Brien, R. J. Mitchell, John M. Grego, and E. Louise Loudermilk. 2009. “The Wildland Fuel Cell Concept: An Approach to Characterize Fine-Scale Variation in Fuels and Fire in Frequently Burned Longleaf Pine Forests.” Int J Wildland Fire 18 (3):315–25. https://doi.org/10.1071/WF08084.

Iverson, Louis R, Daniel A Yaussy, Joanne Rebbeck, Todd F Hutchinson, Robert P Long, and Anantha M Prasad. 2004. “A Comparison of Thermocouples and Temperature Paints to Monitor Spatial and Temporal Characteristics of Landscape-Scale Prescribed Fires.” Int J Wildland Fire 13:311–22.

Jacoby, P. W., R. J. Ansley, and B. A. Trevino. 1992. “An Improved Method for Measuring Temperatures During Range Fires.” J Range Manage 45 (2):216–20.

Kral, Katherine C., Ryan F. Limb, Torre J. Hovick, Devan A. McGranahan, Aaron L. Field, and Peter L. O’Brien. 2015. “Simulating Grassland Prescribed Fires Using Experimental Approaches.” Fire Ecol 11 (3):34–44.

Kremens, Robert L., Alistair M.S. Smith, and Matthew B. Dickinson. 2010. “Fire Metrology: Current and Future Directions in Physics-Based Measurements.” Fire Ecol 6 (1):13–35. https://doi.org/10.4996/fireecology.0601013.

Lemaire, R., and S. Menanteau. 2017. “Assessment of Radiation Correction Methods for Bare Bead Thermocouples in a Combustion Environment.” Int. J. Therm. Sci. 122 (December):186–200. https://doi.org/10.1016/j.ijthermalsci.2017.08.014.

McGranahan, Devan Allen. 2020. “An Inconvenient Truth About Temperature-Time Data from Thermocouples.” Plant Ecol, in press. https://doi.org/10.1007/s11258-020-01064-7.

McGranahan, Devan Allen, Benjamin Geaumont, and Jonathan W Spiess. 2018. “Assessment of a Livestock GPS Collar Based on an Open-Source Datalogger Informs Best Practices for Logging Intensity.” Ecol Evol 8 (11):5649–60.

Pavlasek, P., C. J. Elliott, J. V. Pearce, S. Duris, R. Palencar, M. Koval, and G. Machin. 2015. “Hysteresis Effects and Strain-Induced Homogeneity Effects in Base Metal Thermocouples.” Int. J. Thermophys. 36 (2-3):467–81. https://doi.org/10.1007/s10765-015-1841-3.

R Core Team. 2019. “R: A Language and Environment for Statistical Computing (V. 3.5.3).” Vienna, Austria: R Foundation for Statistical Computing.

Ritz, Christian, Florent Baty, Jens C Streibig, and Daniel Gerhard. 2015. “Dose-Response Analysis Using R.” PLoS One 10 (12). Public Library of Science.

Rothermel, Richard C. 1972. “A Mathematical Model for Predicting Fire Spread in Wildland Fuels.” Research Paper INT-115. Intermountain Forest and Range Experiment Station, Ogden, UT: US Department of Agriculture, Forest Service.

———. 1983. “How to Predict the Spread and Intensity of Forest and Range Fires.” General Technical Report INT-143. Intermountain Forest and Range Experiment Station, Ogden, UT: US Department of Agriculture, Forest Service.

Russell, Morgan L., Lance T. Vermeire, Amy C. Ganguli, and John R. Hendrickson. 2015. “Season of Fire Manipulates Bud Bank Dynamics in Northern Mixed-Grass Prairie.” Plant Ecol 216 (6):835–46. https://doi.org/10.1007/s11258-015-0471-y.

Shannon, K S, and B W Butler. 2003. “A Review of Error Associated with Thermocouple Temperature Measurement in Fire Environments.” In Proceedings of the 2nd Fire Ecology Congress, 3 pp. Orlando, FL: American Meteorological Society.

Shipley, J Ryan, Julian Kapoor, R Andrew Dreelin, and David W Winkler. 2017. “An Open-Source Sensor-Logger for Recording Vertical Movement in Free-Living Organisms.” Methods Ecol. Evol.

Simpson, Kimberley J., Brad S. Ripley, Pascal-Antoine Christin, Claire M. Belcher, Caroline E. R. Lehmann, Gavin H. Thomas, and Colin P. Osborne. 2016. “Determinants of Flammability in Savanna Grass Species.” Edited by Hans Cornelissen. J Ecol 104 (1):138–48. https://doi.org/10.1111/1365-2745.12503.

Smith, Jane E., Ariel D. Cowan, and Stephen A. Fitzgerald. 2016. “Soil Heating During the Complete Combustion of Mega-Logs and Broadcast Burning in Central Oregon USA Pumice Soils.” Int J Wildland Fire 25 (11):1202. https://doi.org/10.1071/WF16016.

Strong, Dustin J., Amy C. Ganguli, and Lance T. Vermeire. 2013. “Fire Effects on Basal Area, Tiller Production, and Mortality of the C4 Bunchgrass, Purple Threeawn.” Fire Ecol 9 (3):89–99. https://doi.org/10.4996/fireecology.0903089.

Trollope, W.S.W. 1978. “Fire Behaviour Preliminary Study.” Proc Grassld Soc Sth Afr 13 (1):123–28. https://doi.org/10.1080/00725560.1978.9648846.

Vaartaja, O. 1949. “High Surface Soil Temperatures on Methods of Investigation, and Thermocouple Observations on a Wooded Heath in the South of Finland.” Oikos 1 (1):6. https://doi.org/10.2307/3565034.

Walker, JD, and BJ Stocks. 1968. “Thermocouple Errors in Forest Fire Research.” Fire Technol. 4 (1):59–62.

Wally, Alaä L., Eric S. Menges, and Carl W. Weekley. 2006. “Comparison of Three Devices for Estimating Fire Temperatures in Ecological Studies.” Appl. Veg. Sci. 9 (1):97. https://doi.org/10.1658/1402-2001(2006)9[97:COTDFE]2.0.CO;2.

Zopfi, Megan Elizabeth. 2020. “Characteristics and Spatial Heterogeneity of Prescribed Fire Behavior in North Dakota Grasslands.” M.S. Thesis, Grand Forks, ND: University of North Dakota.