Frank Linton, Anna Stumme, Brett Padula, Gail Ifshin, Gregory Behrmann 2020-09-17 00:52:17
abstRact
Adding sensors to beehives may reduce the need for inspections while providing beekeepers with early warnings of issues that need attention. Some beekeepers have been using some sensors in their hives for years, though on the whole, the practice is virtually non-existent. These days, smartphones, IOT (Internet of Things), and precision agriculture are making it possible to provide sensing technology economically and the value of sensors in beehives is beginning to be explored. In this article, the first of a three-part series, we show how a temperature sensor grid can be used to precisely monitor the size of the winter cluster, the transition to brood rearing, the state of the brood, and sometimes, swarm events.
intRoduction
I have been using temperature sensors in my beehives for many years. With one or more temperature sensors inside the hive, sensors that report their readings to my cell phone or to my computer via the internet, I can tell a lot about what is going on inside my hives without opening them. In this article I will report preliminary results of using a new temperature sensing device, one with many times the spatial resolution of the units now on the market.
Typically, the companies that provide beehive sensor units (see https:// ColonyMonitoring.com) offer one temperature sensor per hive, or at most one per box. The data they provide is useful. For example, they can indicate the presence of the winter cluster (50oF – 90oF), the presence of brood (93 F – 95 F), and so on.
Eventually I realized, however, that while it is nice knowing that the winter cluster exists or that brood is present, it would be much more useful if I knew the size and location of the winter cluster, the amount of brood and how it was changing. Obtaining temperature data in much greater detail could help me either to reduce inspections or to intervene earlier should it become necessary.
hypotheses
Imagine an economical bee-friendly sensor grid placed atop every box in each of your hives, reporting the temperatures, or better yet, reporting conclusions directly to your cell phone or computer, informing you daily of the health and productivity of your colonies.
A suitable arrangement of temperature sensors might be able to tell you about the location, size, and movement of the winter cluster. This article will explore the extent to which one experimental temperature sensor grid supports that hypothesis. In parts 2 & 3, we will explore other hypotheses as well as actual results of monitoring details of brood rearing, queen health, flight activity, and honey production.
design of the sensoR gRid
Some decisions must be made when constructing an actual sensor grid. Where would the sensors best be located, how many sensors are optimal, and with what frequency should they be sampled? The answers to these questions depend on the colony behavior one intends to monitor, its size, its movement, and its rate of change. For example, to monitor the winter cluster, the sensors might be located within frames, between frames, between boxes, or under the inner cover. The number of sensors could range, conceptually at least, from one per hive to one per cubic inch. And the frequency of taking readings could range, again conceptually, from once per minute to once per day.
For the winter cluster, for example, I would like to know its size and its position within the hive, and to get some indication of the cluster’s ability to reach up and out to its honey stores. I would also like to know when the cluster begins raising brood. And finally, I would like to see how external conditions affect temperatures inside the hive.
Through Toni Burnham, a wellknown Washington, DC beekeeper, I made contact with Anna Stumme and Brett Padula, two engineering graduate students at the Catholic University of America who were looking for an engineering project relating to honey bees, and their professor, Gregory Behrmann. I suggested a temperature sensor grid and they accepted the challenge.
Based on my colony monitoring experience, their engineering know-how, and real-world constraints of time and money, we made an educated guess at a starting point and decided to build a temperature sensor grid using a modified inner cover. We modified it in two ways. First we put slots in the cover directly over the frames, so that if the cover were to be placed between boxes, the bees could move freely from box to box. In a preliminary trial, the bees moved through this modified inner cover unimpeded.
Second, we placed the sensors in four cross-sections of nine sensors, located at equal intervals along the length of the hive. Each sensor was between two frames (or between a frame and the side of the box) of an 8-frame hive body. This arrangement placed one sensor in every six square inches of space over the box (1.33” x 4.5”). See Figure 1. This placement captured heat rising or radiating from below, but may also have allowed for some heat dispersal, possibly influencing adjacent sensors.
The sensors were programmed to collect data approximately every 30 minutes. Data was stored on micro SD cards mounted on the circuit boards that controlled data collection. The micro SD cards and the batteries that powered the system were swapped out periodically. Eventually, data transfer will be automated. Figure 3 displays Anna’s image of the sensor board and descriptions of its components.
monitoRing the WinteR clusteR Remotely
Honey bees engage in three major activities within the hive: overwintering, raising brood, and processing nectar into honey. It seems plausible that each of these activities has its own heat signature, with variations indicating the level of activity, details of its performance, and the bees’ ability to carry it out. If the sensor grid we are piloting works out as expected, and reveals that a colony is doing well, there will be less need for disruptive inspections, and if it reveals that the colony is struggling, the beekeeper can investigate as soon as the need becomes apparent.
In this area, near Washington, DC, a colony in a deep box with some stores can get through the winter with a medium super of honey. Syrup or fondant is easily fed if needed. Typically, a colony would work its way up through the stores over the course of the winter, approaching the inner cover as spring appears. Usually the rate of honey consumption will increase as more and more brood is being raised in anticipation of the spring nectar flow.
With these facts in mind, I chose to place the grid over the honey super rather than between the hive body and the honey super, in the fall. This placement will necessarily result in a less accurate measure of the cluster’s size and position early in the winter when the cluster is well below the top of the hive, but should give me increasingly precise readings toward spring as the colony moves up into the honey super and begins to raise brood.
About noon on November 9, 2019, we put the sensor grid in place over the honey super to observe the position, size, and movement of the expected winter cluster. Ten days later, on November 19, we replaced the batteries and swapped out the SD cards so we could analyze this first set of data. There were several surprises. See Figure 4.
WinteR clusteR: the findings
First, there are obvious daily periodic changes. When the sun hits the hive in the morning, temperatures shoot up in the entire hive, though none quite as high as the exterior temperature sensor (blue line), which is located with the electronics in a sealed box atop the hive. Also, when the exterior temperature falls below 32 F, the exterior sensor stops recording, and I have replaced those readings with 32 F.
Second, in general, the daily pattern is for temperatures to shoot up when the sun hits the hive full blast, and then gradually cool down all day and night until the morning sun begins to warm the hive again. Furthermore, there is little difference between the coolest in-hive temperatures and the exterior temperatures — at this point in the winter season.
Third, the temperatures in the hive rise and fall in unison. The warmer places are always warmer and the cooler places are always cooler. There are no sites where the temperatures are both stable and high — above 50 F (44 F according to some authors — see Sidebar 1), so there does not appear to be a cluster; but read on.
Fourth, actually there is a cluster. It is just that the cluster is down in the brood box, well below the sensors. Both exterior temperatures and cluster temperatures influence the sensor grid. Of the two, the exterior temperatures are the stronger, but the influence of the cluster is also apparent. The sensors with the higher readings, the cluster-influenced readings, are toward the center of the hive and those with lower readings are toward the edges of the hive, especially in the corners, see Figure 5, which shows the average temperature at each sensor over this data collection period. The presence of the cluster in the hive body is corroborated by an infrared image of the hive, taken a little later. See Figure 6.
A second data collection period of 24 days (Figure 7) reveals the same pattern of daily ups and downs as Figure 4. And again, the six central sensors reported consistently higher temperatures than the surrounding sensors, but not the unchanging higher temperatures that would indicate direct contact with a cluster immediately below it.
Fifth, as a user of this technology, I have re-confirmed that a) it is important to make things fool-proof, and b) prototypes, by their nature, are not fool-proof. Consequently, some data has been lost as I have learned to be very careful when exchanging batteries and SD cards.
Sixth, the lack of direct temperature- based evidence for a winter cluster, even though the infrared image shows one to be present, indicates the potential utility of a temperature sensor grid directly above the brood box. However, even if resources were available for a second sensor board, and even though bees do move freely through this sensor board in summer, it seems to me that a mostly solid sensor board might well interfere with the cohesiveness of the winter cluster and impede its upward movement. The ideal of placing a sensor grid permanently above every box awaits a design that presents a nearly invisible barrier to the bees.
The next data segment, Figure 7, November 19 to December 12, is similar to the previous one. A horizontal line at 43 F indicates where we might expect to see the edge of the cluster, but again, it is not visible in this view.
Given that the temperatures move in unison, we can obtain a view of their relative values by taking their averages (Figure 8). As before, the hive is warmer in the center, but all sensors are warmer than the exterior; the automated color coding shows the relative values. Figure 9 is a graph of the values of the Mid-Rear crosssection of the hive. As in Figure 8, the middle three temperatures are the highest, and this view looks like a cluster with an edge, yet neither Figure 7 nor Figure 8 show a sharp temperature drop-off at approximately 50 F. This observation, i.e., the lack of a sharp cluster edge at 44-50 F, is counter to a number of sources reporting a cluster edge temperature (see Sidebar 1: Edge of Winter Cluster: What Scientists Say).
mid-febRuaRy
Next, look at data from ten days in mid-February, Figure 10 and Figure 11, when, if the hive had brood, it was located well below the sensor grid. When we retrieved this data on February 21, we also added a shim — for feeding fondant — between the honey super and the temperature sensor grid.
The last few days of mid-February data in Figure 11 show nearly-continuously high temperatures for a halfdozen or so sensors. Note however that the high temperatures are in the 80s, not the mid-90s. I infer this to mean that there may be brood some distance below these sensors. In general you can see that these higher temperatures tend to vary less over a 24-hour period than the lower ones, indicating that for some reason, most likely brood rearing, bees are keeping these temperatures relatively constant.
One difference to notice between the November data and the February data: At first, in the fall, the lowest temperatures inside the hive nearly matched the temperature outside, but as winter wore on, outside temperatures fell and interior temperatures rose, so that even the lowest interior temperature was well above the outside temperature.
To summarize, a sensor grid placed above the honey super can monitor the winter cluster’s size and position and provide some evidence of colony health. In mid-November, average highs in the center of the hive were in the 50s. By mid-February, they were in the 70s. The rising temperatures appearing in early spring indicate the approach of the colony to the sensor grid and the stable high temperatures at the end of the period may signal the presence of brood. At this point we can conclude that the colony is overwintering well.
in the next installments
In the next installments we shall see the transition to brood rearing, brood volume and location, swarming, and the effects of swarms on brood volume.
RefeRences
Doke, Frazier, and Grozinger. 2015. Overwintering honey bees: biology and management. Current Opinion in Insect Science, 10:185–193
Linton. 2017. Monitoring Colony Activity with Temperature Sensors: a Research Agenda. Apimondia 2017. Istanbul Turkey.
Linton. 2020. https://colonymonitoring.com, a guide to currently available colony monitoring products, including hive scales, temperature sensors, microphones, cameras, radar, & more.
Mangum. Winter Clusters Seen with Colors of Heat. American Bee Journal, October 1, 2016
Ocko, & Mahadevan. 2014. Collective thermoregulation in bee clusters. J. R. Soc. Interface 11: 20131033. http://dx.doi. org/10.1098/rsif.2013.1033
Owens. 1971. The Thermology of Wintering Honey Bee Colonies. Agricultural Engineering Research Division, Agricultural Research Service.
Stabentheiner, Pressl, Papst, Hrassnigg, and Crailsheim. 2003. Endothermic heat production in honeybee winter clusters. The Journal of Experimental Biology 206, 353-358 353. doi:10.1242/jeb.00082.
Stumme and Padula. (2019) An Automated Sensor System for Monitoring Beehive Health. Research Day Poster; Catholic University of America. Social Innovation Startups, 2018-2019.
about the authoRs
Frank Linton, fnlinton@gmail.com, runs the website https://colonymonitoring.com.
Anna Stumme is currently an Electrical Engineer at the U.S. Naval Research Laboratory.
Brett Padula is currently a Mechanical Engineer at the U.S. Army Research Laboratory.
Gail Ifshin, is Co-founder and President of Everybody Grows.
Greg Behrmann is Clinical Associate Professor and Associate Dean, Catholic University of America.
Sidebar 3: BroodMinder’s Flying Carpet
BroodMinder’s Flying Carpet animation on YouTube is a riveting depiction of data gathered over nearly a year using nine temperature sensors placed in three rows of three within a hive. The YouTube video, Figure 12, displays three animated views of this time-series data simultaneously. One sensor is in the center of the hive, the others are at the edges. The temperatures in the spaces between the sensors are filled by interpolation. The changes over time corroborate the time-series data graphs in this document. Repeated viewings will reveal subtleties.
©American Bee Journal. View All Articles.