Winter 2020-2021 Monitoring
Do Bees Heat the Inside of the Hive in Winter?
Published in the Winter 2020 BCHPA BeeScene
To insulate or not to insulate? Not really a question most beekeepers in Canada have to ask themselves. We do have a few areas in BC, Southern Ontario where the bees will likely do fine without that extra protection if well stored in honey, the bees are healthy in fall and they are protected from wind and rain.
If you review the CAPA website for our annual winter losses and the reasons given for losses, you will see that nothing has changed in the 14 years since CAPA started doing these surveys (see figure 1). Pests/diseases, starvation/weak colony/weather and queens make up 99% of the reasons for winter losses.
I also received a spreadsheet of similar results for a double box wooden hive that was conducted in 2016-2017. I will use these results to illustrate some of the property differences between wooden (non-insulated) hives and hives that are insulated.
Figure 4: Photo showing polystyrene hive setup on solid base with an open air gap below each hives screened bottom board. All seams and gaps between hives are taped.
There are several ways to calculate heat loss in “enclosures”. The simplest is to use a standard building calculation approach using Fourier’s Law, which states that for steady state conditions (hourly snapshots for me) heat loss via conduction can be calculated using H=-kAdT/dx, where k is the thermal conductivity of the material, A is the exposed surface, dT is the difference between the inside and outside temperatures and dx is the thickness of the “wall”. Engineers have further simplified this calculation to incorporate common thermal resistance values (R-Values) so that it becomes H=A/RSI(Ti-To), where RSI is the metric version of the R-Value, Ti is the inside temperature, and To is the outside temperature. This calculation is completed for each wall of the hive and the top cover. For my purposes I added an extra 15% heat loss via the top as recommended when calculating heat loss for a building.
Newton's 1st law of thermodynamics states that energy cannot be “destroyed”, it can only be transformed, and therefore must balance out. So, the heat that a hive loses must equal the heat the hive generates. Figure 5 shows you an example of the heat loss of the hive enclosure, but does not currently include losses via the lower entrance (bottom of the cluster).
Honey stores can be considered a thermal mass, which adds more complexity to the problem, but it is a good problem. Honey in a wintering hive acts as food energy, but also as a thermal buffer, thus reducing internal temperature fluctuations due to rapidly decreasing outside temperatures. On warmer winter days, in an insulated hive where the heat loss from the hive is less than the heat generated by the bees (from resting and shivering metabolic rates), the excess heat is absorbed into these stores. When looking at internal temperature vs. external temperature changes, you can clearly see a temperature lag at peak minimums and maximums. This lag is in part due to the absorption and release of this excess energy, as well as from the cluster expanding and contracting.
Figure 5: Simplified heat loss calculation for a single brood hive in my current configuration. More information is available at www.engineeringtoolbox.com under the 'building heat loss' section.
Several numerical analysis methods could be used to increase the accuracy of this calculation. For now, I have chosen to keep it simple.
Next, I have set up my spreadsheet to allow me to visually see the internal temperature profiles and added a simple macro (a small program you write in excel to automate steps) to allow me to increment the calculations in one-hour intervals. This allows me to observe cluster movement, calculate various hourly measurements (heat loss, equivalent honey consumed, water generated, cluster expansion/contraction). The non-calculated temperatures in between the measured values are fit using a non-linear quadratic fit (thank you to Theo Hartmann for helping me here, and with the grid approach). Adding more sensors to the grid would increase the accuracy of the temperature distribution.
Early Results and Discussion
One of the common assumptions used in most of the available studies on honey bee winter cluster behaviour is that ambient temperature is equal to interior temperature (the temperature inside the hive that the cluster is exposed to). In this exercise I wanted to measure the relationship between outside and inside temperatures as a function of thermal resistance (R-value). In the available studies, the cluster/bee metabolic rates and the cluster properties (size, density, thermals) are related back to the outside temperature, which make these studies less relevant to my northern conditions. Why is this important?
From the CAPA survey, 4 of the main causes for winter losses are partly related to the properties/configuration of the hive enclosure: starvation, weather, weak colonies; and Nosema is indirectly related due to the increasing risk of infection as the winter progresses without any cleansing flights. Bees in non-insulated hives will consume more honey per CDH8 (cold intensity units - is a measurement designed to quantify the demand for energy needed to heat the cluster. It is the number of degrees that hourly average temperature is below 8 Celsius, which is the temperature below which bees will be hive bound and clustered), thus risking the bees defecating inside the hive, which could increase the risk of Nosema spreading throughout the colony. In my past trials, I have shown that weaker hives (due to lack of time to grow into stronger hives) can be overwintered successfully in smaller, insulated volumes.
I would also hypothesize that colonies with higher mite counts may have a better chance of surviving cold winters in a well insulated hive, as Varroa mites are known to disrupt the bees' ability to thermoregulate/osmoregulate (regulate the water balance in the hive and within their bodies) due to mites feeding on the fat bodies (damaging this critical organ). These known fat body functions and the other critical functions are likely the driving factors to colony death because of mite damage in winter. Therefore, a well insulated colony would likely reduce the impact of cold stress, thus potentially allowing the colony to survive.
As shown in the next section, the internal temperature (Ti) of an insulated hive is driven by A/RSI (Exposed Surface/Thermal Resistance). On the other hand, with a non-insulated hive (typical wood) we get Ti =~ Tambient (Temperature Inside is comparable to Ambient Temperature) where the clustering behaviour would be the primary survival mechanism.
After just one and half months in my trial, the data clearly shows that internal T min( the lowest temperature that the cluster is exposed to inside the hive) is driven by hive enclosure heat loss (Thermal Resistance A/RSI). The simple X-Y chart R2 which equals the square of correlation shows the data fitting very well with my model (heat loss calculation). Anything above 0.9 shows that the data variance is low. In this case Tambient is related to heat loss value (derived from the insulation value), R2>0.9. When run a similar comparison between Tambient and Tmin we get a very low R2 (poor relation, more data variation) which indicate that something else is driving that relationship.
The opposite relation occurs for the wooden hive where the relationship clearly indicates that Tambient drives the Tinterior low, therefore the assumption Tambient = Ti is acceptable when discussing non-insulated hives. As stated by Mitchell (2016), the clustering behaviour is likely a secondary winter survival mechanism only applicable during very sustained cold periods. Under my current hive configuration, outside temperatures below -10°C to -15°C start showing internal hive temperatures dropping below 10°C. Under non-insulated configurations, any heat released from the cluster (via conduction/convection) is quickly lost in a wooden hive.
Based on the Myerscough (1993) modelling approach, a 19,200 bees “cluster” will produce (at resting metabolism) approximately 10W, at an average box temperature of 18°C. At bee exposure temperatures below 18°C, the bees start shivering to increase heat generation. The challenge is that the heat output of the cluster is not based on one temperature but the local temperature of each bee.
Figure 6 & 7: Charted Linear equations for insulated hives (SI – Single Box; DL - Double Lower Box; DU – Double Upper Box) and typical wood hive (WL – Wooden Lower Box).
Figure 8: (Left Images)Temperature profile of single box hive at T ambient of -32°C (The grid layout is equivalent to 1” x 1” per square.) with its centre cross section. Estimated 20,000 to 30,000 bees in this hive based on a late fall inspection. Right pictures: Using Southwick Metabolic Rate MR chart and using the lower internal hive temperatures, we get a MR of between 8.9 to 13.3 Watts for this one hour snapshot. The picture showing several circles uses the temperature grid and hemisphere geometry to estimate cluster radius and several mantle isotherm ranges.
Now I will take you through the logic of calculating the Honey equivalent Grams/Hr consumed and Water generated based on the calculated heat losses (A/RSI). As you can see, most of these calculations and charts are designed to allow a beekeeper to apply the data to better understand the impact of our beekeeping practices.
Using the calorific value for 1 kg of honey converted to Watts, you can estimate honey consumption and the water generated required to metabolize the honey. 1lb of honey produces about 0.67 lbs of water (0.18 lbs from 18% water content of honey + 0.49lbs produced while metabolizing sugars) (from Linton, F., Bee Culture Jan 1, 2015). About 40% of this water is recycled by the bees to consume the next batch of honey. This leaves 420ml of water per 1 kg of honey consumed to deal with (Oliver R, ABJ Jan 2020). Excess water is one of the main causes of dysentery.
How much energy does 1 kg of honey contain and how can you convert the metabolic rate to honey consumed? (Note that 1.163 W-h/cals is the conversion factor to convert food calories to Watt-hours).
1 kg honey = 3040 calories (nutritional)
1 kg honey = 3040 cals x 1.163 W-h/cals
1 kg honey = 3.5 kW-h
Here is the chart of honey consumed and water generated equivalent to the heat loss. I am currently trying to figure out a way to calculate heat lost via convection/conduction from the lower part of the cluster. The sum of these values should theoretically be equal to the heat generated.
Figure 9: Table showing cumulative honey and water generated based on heat loss via the hive enclosure (excluding lower cluster and lower entrance).
On a very cold day (-32°C) this single box hive was losing approximately 16W per hour, which is equivalent to about 5g of honey/hr and 3ml of water produced. These amounts quickly add up as we have 24 hrs in a day and 7 days in a week. In the above chart you can see that during the coldest week (Week 45), the internal temperatures were still very reasonable with a low 8.6°C internally, with an average just below the shivering (endothermic heat) trigger of 18°C. Above 18°C we call this ectothermic heat generation (resting metabolism). During this period, the hive consumed an equivalent of 1.3 lbs of honey and generated 161ml of water, just to balance upper hive heat losses. In a wooden hive, the cluster would be severely stressed to maintain itself for extended periods of time, so heat conservation is critical.
In future articles, I would like to cover my findings in relation to the slatted racks being monitored. I will likely be using this hive location (the slatted rack) as the lower boundary limit for my lower heat loss calculation. I will also be discussing water generation (water vapour) and the benefits of using insulation, 9 frames per box, and an open screened bottom board to manage condensation. We can clearly see that a non-insulated hive in cold climates (Southern Canada – near the US border) will generate a lot of moisture during sustained periods of cold.
As you can see from my charts and calculations, I am also trending and measuring temperature isotherms between 10°C and 15°C (Owens 1971) to help me estimate cluster mantle perimeters over time at different external temperatures. One MS Excel feature that I like using is called 'slicers', which allows me to quickly toggle values, ranges and criteria on and off in spreadsheet pivot tables (i.e. Day vs Night, Tambient Ranges, Wind, etc…). This allows you to quickly check relationships and identify potential drives. There are some interesting cluster movements and temperature spikes that can be observed in my time lapse videos of hourly temperature heat map distributions, especially after very cold mornings and rapid warm up periods. Is this due to the bees adjusting heat output (a delayed feedback loop) or to the thermal mass effect? All interesting concepts to look into. The Ta range selector allows me to quickly exclude certain temperature ranges. You can clearly see a tightening of the data point (Ta Vs Est Heat Loss) as the outside temperatures drop. Less variability is good when testing out linear models. Another interesting chart is Ta Vs T Max (warmest temperature in hive). You will notice a floor of about 20C regardless of outside temperature. This tells me the bees have set this as their lowest core temperature. You will also notice that close to 90% of the data points are between 20C and 25C which agrees with existing research for broodless hives. Scatter in data points like those observed in Ta Vs Tmin shows a bit of downwards focus as the outside temperature cools but it is has a lot of variability (data points scattered above and below the mean). In this case, much of this variability is due to the effect of insulation (reduced heat loss). In the data I analyzed for the wooden hive, this same chart (Ta Vs T min) shows a very linear relationships with very little variability (R2>0.9). This tells me that in a wooden hive the interior temperature can be assumed to be close to outside temperature. The implication here is that the bee cluster must therefore adjust to every external temperature fluctuation and deal with every extreme climatic event biologically through clustering with no real help from the enclosure. Add the impact of solar radiation on a cold sunny day where the hive will quickly gain heat for a short period of time while the sun is out, often breaking cluster and then being forced to quickly re-cluster if clouds form or the sun sets. Remember insulation works both ways (also keeps heat out) and it works to ensure a very thermally stable environment. My Day/Night slicer allows me to verify this solar effect.
Figure 10: Example of slicers (left column), associated X-Y charts and charted “cluster” >10C and >15C % Area coverage
Ensuring that you have healthy, well nourished bees and follow best hive management practices is critical to successfully overwinter. The hive enclosure can provide insurance against starvation, smaller wintering clusters, abnormal weather and possibly reduces the risk due to Nosema and Varroa in colder climates. Other future experiments will measure the thermodynamic effect of having a top entrance (heat loss vs. moisture generated) and the effects of a slatted rack in summer.
One reason I write these articles is to force me to think through things and to encourage others to experiment and ask questions. Much of the official research is geared towards commercial beekeeping (as it should be) or pure academia (which I enjoy) but with no clear beekeeper application. A lot of what I do wouldn't be practical for very large commercial operators, but hopefully some of the concepts can help them. My main focus is on the small scale beekeeper who can be significantly impacted by winter losses.
I would like to thank the BCHPA who assisted me with covering the cost of few sensors used in this study. If any of the readers are interested in participating, have ideas or would like to get access to the data, please don’t hesitate to reach out to me at email@example.com
Youtube video of time lapse cluster movement (heat map), calculations: https://youtu.be/yLI6Xges2OE
Owens, C.E. (1971). The thermology of wintering honey bee colonies. USDA Technical Bulletin 1429.
Mitchell, D. (2016). Ratios of colony mass to thermal conductance of tree and manmade nest enclosures of Apis mellifera: implications for survival, clustering, humidity regulation and Varroa destructor. Int J Biometeorol. 60(5): 629-638.
Ocko, S.A. & Mahadevan, L. (2014). Collective thermoregulation in bee clusters. J. R. Soc. Interface 11: 20131033.
Watmaugh, J. & Camazine, S. (1995). Self-organized thermoregulation of honeybee clusters. J. Theor. Biol. 176: 391–402.
Southwick, E.E. (1983). The honey bee cluster as a homeothermic superorganism. Comp. Biochem. Physiol. Vol. 75A(4): 641-645.
Oliver, R. (2017). Understanding Colony Buildup and Decline: Part 13c – The Winter, and Hive Design, ScientificBeekeeping.com, January 2017.
Oliver, R. (2020) The Nosema Problem: Part 7b The Causes Of Dysentery In Honey Bees: Part 2, ScientificBeekeeping.com
Myerscough, M.R. (1993). A Simple Model for Temperature Regulation in Honeybee Swarms. Journal of Theoretical Biology, Volume 162, Issue 3.
Szabo, Tibor I. (1985). The Thermology of Wintering Honeybee Colonies in 4-Colony Packs as Affected by Various Hive Entrances, Journal of Apicultural Research, 24:1, 27-37, DOI: 0.1080/00218839.1985.11100645.