Feral Honey Bees in Urban Environments

 

Above: Exterior view showing the window shutter from outside.
Below: Interior view of the window shutter box with the access panel removed.

Introduction to the Citizen Science Article Earlier this year, I uploaded a photo of a local Atlanta bee nest entrance—found in a cut-off sycamore limb—to the Free Living Bees website. Not long after, I came across a fascinating citizen science journal article featured on the same site. As someone obsessed with swarms, I found it absolutely captivating.

Urban Population Density The research focuses on the city of Belgrade, Serbia—a large and densely populated urban environment. Residents reported both honey bee nests and swarms, often seeking help with their removal. Interestingly, the number of reports increased exponentially with population density (a relationship shown clearly in the paper’s data).  To verify and expand on these reports, the authors conducted phone interviews with the residents and followed up with on-site field studies.

Reported Swarms vs. Nests At first, I assumed that reports of swarms and nests would occur in roughly equal numbers—but not so. The study found 261 swarms compared to only 90 nests, meaning swarms were reported nearly three times more often.  Most swarms (about 78%) were seen at heights between 1–9 meters, remarkably similar to what Seeley describes in Honeybee Democracy (p. 52). Reported nest heights, on the other hand, were concentrated between 3–15 meters (about 74% of cases).

Interior–Exterior Shutter System Jovana Bila Dubaić kindly sent me photos to help clarify the distinction between (1) the wooden window shutter and (2) the wooden window shutter box. Many buildings in Belgrade feature a combined interior–exterior shutter system.
When closed, pedestrians see the wooden window shutter covering the exterior. When retracted, however, the shutter rolls upward into a wooden box located inside the building, above the window. Astonishingly, 30.7% of reported nests were found inside these shutter boxes.

Exploring with Google Street View Curious, I used Google Street View to explore some of Belgrade’s most populated neighborhoods—imagining myself as a scout bee. What I “saw” was extraordinary: street after street lined with windows that offer easy access to those wooden shutter boxes. From a bee’s perspective, it’s a smorgasbord of indoor nesting opportunities.

In summary Don’t be intimidated by the lengthy title: “Unprecedented Density and Persistence of Feral Honey Bees in Urban Environments of a Large SE-European City.” The paper is well worth the read, and if you reach out with questions, you’ll likely find that the lead author, Jovana Bila Dubaić, is remarkably generous with her responses.

hive preparation for winter

 

Refer to Derek’s 2024 paper in the Journal of Thermal Biology, “Are man-made hives valid thermal surrogates for natural honey bee nests?” Derek reports that “bee space above combs increases heat loss by up to ∼70%; hives, compared to tree nests, require at least 150% the density of honey bees to arrest convection across the brood area.”

In a hive, brood heat typically circulates by (1) rising between frames, (2) striking the inner cover, (3) spreading sideways, and (4) descending through the cold bee space.

To improve insulation, I added an external Bee Cozy (R-value 8) and took internal measures to limit brood heat circulation by blocking the bee space above the combs. I used letter-sized plastic sheets (the type made for overhead projectors), placing one spliced and trimmed sheet directly on the top bars beneath each hive’s inner cover.

Moving the BroodMinder sensor

Graph Interpretation
This hive is built from three stacked 5-frame deep boxes—what I call “the tower.” By late July, my BroodMinder temperature graph became nearly impossible to interpret. Frustration set in: back in May, this same 2025 swarm showed far steadier brood-temperature regulation. Why, I wondered, does the BroodMinder sensor not come with clearer instructions? I should have known better than to confuse gadget guidance with actual beekeeping wisdom.
Temperature Events (red pins)
The BroodMinder “temperature events” logic drops red pins everywhere, but that wasn’t my real concern. The bigger story was that this hive swarmed in early June. That event lowered brood temperature by nearly 10°F. Gradually, though, the colony recovered and brood temperatures crept back into a stable thermoregulation range by late June.
Sensor Location
At $40, the hive gets just one sensor from me. This colony began as a single five-frame deep in mid-April, so I initially placed the BroodMinder at the top of that box, above the brood frame, as the BroodMinder instructions suggested. Since then, I’ve twice bottom-supered, so the hive now stands as three stacked deeps. I’m open to moving the sensor—but only with a theory in hand. Random changes don’t sit well with me. My best reasoning so far: (1) the brood nest shifted downward as I added boxes underneath, and (2) bees prioritize brood temperature regulation over regulating honey storage temperatures.
Conclusion
In early August, I finally moved the sensor from the top of the third box down to the top of the second. Coincidentally, Atlanta cooled off that same week, with outdoor highs never breaking 70°F. As the graph shows, this new placement hit the “brood sweet spot”: readings stayed firmly in the Brood Zone Band (92–98°F) despite the cooler weather. For once, life is good—and I’ll just have to find another beekeeping puzzle to obsess over.
Looking back, there was never much at stake. I changed one variable: the location in the stack where I measured temperature. Still, I can’t resist narrating the drama.

deconstructing counting board debris

Green contours surround 2 varroa mites

 
My beekeeping journal
I have several notebooks to journal my hobbies. There is one for the backyard chickens and one for the worm farm - you get the idea, I’m a mess! What’s going on in these journals? Usually, it's just my pencil or ink block letter writings, but when I get excited about a topic, out comes the straight edge and I draw with a mechanical pencil, colored pencil, and highlighter.
The big picture
I usually capture photographs of my counting board and hive inspections for ease of use when I later write in my beekeeping journal or need to capture time and date for my blog. Here, I’m focusing on my counting board debris for a small spring swarm trap capture and using a chatbot interface (for example, ChatGPT). In the spring, the debris begins as beautiful, clear wax mirrors; over time, I see grainy lumps of pollen; then darker wax cappings appear.

In the summer, I find small hive beetles and female varroa mites in my debris observations. For spotting tiny varroa mites, I use a magnifying glass and methodically raster scan (line by line style) the counting board. The mites are approximately 2mm in width (left-to-right) and are nearly all found resting on their feet (back facing upwards). If you look closely you can see their feet sticking out on the mouth side of the insect. I asked the chatbot 1) what is the size of the varroa mite, and 2) is a varroa mite an insect? The chatbot supported this observation as a female mite and goes on to say that a male mite is much smaller and rarely seen outside the brood cell. Here’s a table that chatbot produced. The table values look reasonable, but if you type “add sources” to the chatbot reply, you might be more confident with the results.

Trait	Insects	Arachnids (like Varroa mites)
Body segments	3 (head, thorax, abdomen)	2 (cephalothorax, abdomen)
Legs	6	8
Antennae	Yes	No

Taking the iPhone photo
Here are my steps to photograph the debris board.

• Place the debris board in bright indirect sunlight
• Touch the iPhone screen to focus
• Adjust the brightness slider (icon looks like the sun) until the debris-free region matches the expected white color

I took an additional up-close photo that captured a small portion of the counting board containing varroa mites. Why? This is necessary if you want to see the varroa mite legs with the pixel resolution of my iPhone
Computer vision prototyping
I typed “how to get started with OpenCV and Python” into the chatbot prompt and followed the numbered steps. I made incremental progress by replying to the chatbot with write code using OpenCV and Python to: 1) read my JPG file, 2) define the mite color by clicking on the image, 3) calculate a mask using color distance, 4) dilate the mask, 5) threshold the mask, 6) find external and internal contours, 7) filter contours with area limits, 8) filter contours with approximate convexity limits, 9) filter contours with approximate circularity limits, 10) draw the filtered contours on the my JPG using a heavy green line, and 11) write this result to an image file. This seems like lots of work, but it's a factor of 10 easier than writing my own code from scratch. Sometimes the chatbot would suggest that I upload the file to execute all the steps for me. FYI, there is a “+” icon near the prompt area to browse your computer and upload your JPG file.The mites are reddish-brown and have a shiny spot when their backs are facing the iPhone. In photography, the shiny spot is often referred to as a "specular highlight", and this requires infilling with the processing step #4, dilation.
Computer Vision Conclusion
To detect all the mites I had to trial-and-error adjust my program filters. The computer vision project was interesting, but I quickly reached the pixel limit of my iPhone and my hobby lighting skills. The best news, the filtered contours drew my attention to a few mites that I missed without the program.

BroodMInder Temperature Event

In the beginning
On April 13, 2025, a swarm selected my swarm trap, I transferred these swarm trap frames into a five-frame deep box and added a BroodMinder temperature sensor to the top bar of a brood frame. Hourly the BroodMinder stores the temperature °F (purple line), and when I am within bluetooth range, the temperature data is transferred to the cloud using my iPhone.   The outdoor temperatures °F (red dashed line) are provided by a web provider (Weather Source). 
Flash forward and the graph interpretation
Flash forward to June 2025, and the hive has been twice bottom-supered into something we call ‘the tower’ - a stack of 3x5-frame deep boxes. In an ideal hive, the brood temperature (purple line) is centered in the gray band (92-98 °F) with very little fluctuation over time. However this is not such an “ideal” hive, bursting with overwintered bees. The brood temperature of ‘the tower’ fluctuates in sync with the outdoor temperature and on a good day achieves temperatures in the gray band. I have not discussed the red pins, but hang on, I'm getting there.
The red pins
If the brood temperature is above 92°F, then the sensor logic compares previous temperature measurements with the current measurement. Large temperature increases over a short period of time are flagged as a ‘Temperature Event’ or a swarm detection. “No way,” I said.  “Experimental feature,” I said. On June 3, 2025, I was out skating and my spouse called to say, ‘the tower is swarming, the bees are pouring out of the entrance.”  When I returned to the house, things looked normal-ish…well, the flight activity was less vigorous. After transferring the brood temperature data to the cloud, the graph shows that the spouse’s phone call and the red pin align in time, 11:30 AM. I guess the ‘Temperature Event’ detection worked - it’s not an experimental software gimmick after all. 
Adjusted expectations
The BroodMinder does not inspect, manage, or automate my beekeeping; I put those fantastical expectations aside. Apparently, I created more beekeeping tasks for myself like replacing its battery annually, weekly transferring the brood temperature data to the cloud, and the biggest task of all - interpreting the graph. After the temperature event, “the tower” brood temperature continued to fluctuate in sync with the outdoor temperature. After 16 days, the brood temperature reappeared in the gray band. A re-queening intervention was not attempted and, in hindsight, not required. For a small first-year hive, the broodminder is an interesting data logging tool, and requires lots of graph interpretation.  If graph interpretation is your thing, here’s a link to the BroodlMinder. 

Budapest beekeeping startup

Farmers Market; Tánczos selling Acacia and Linden honey

András describes the bumble bee colony 

 

Zoltan led the honey extraction and honey tasting 

Examining brood comb

 

As a young person, my dad, amongst his other farming community chores, kept beehives in Bakonyjákó, Hungary. Visiting this farming community was on my bucket list, and since work sent me to a conference in Vienna, Austria, I could not resist a piggyback visit to Hungary.

In preparation for the trip, we contacted a Hungarian-speaking guide. Also, I had ChatGPT create a list of common beekeeping terms in English, Hungarian, and the Hungarian phonetic pronunciation. I spent hours repeating out loud the bee vocabulary and listening to Google Translate. Warning, others say I can be annoying in my repetitive habits.

Unexpected first contact with a beekeeper
Setting off from Budapest by car to Bakonyjákó, a wrong turn for lunch brought us to a farmers' market. At the first food stall, we met Tánczos selling honey. Without any spoken words required, I was offered a wooden tongue depressor and jumped into honey tasting. The guide helped me understand that the two varietal honeys are acacia and linden. I spoke to the beekeeper using my limited German and discovered my new friend has 30 hives.

An afternoon with a beekeeping startup, Mézerderdő méhészet
Our guide found a beekeeping activity for us. András greeted us at the entrance of the startup wearing a green hoodie with the text, Méhész vagyok, mi a szupererőd? The guide whispered, “I’m a beekeeper - what’s your superpower?” Three years ago, Zoltan and András purchased a plot of land at the edge of Budapest, and they now run a part-time beekeeping startup focused on what appears to include insect education, beekeeping experiences, and honey sales. The plot supports 50 hives, all with vigorous flight activity, and colonies kept in two box styles. Half of the hives were traditional Hungarian style of boxes, all of the same size, and the other half a mix of box heights, including a very tall deep box.

The afternoon was shared with 30 paying visitors who split into two groups. We set off with András and started by looking at their native and commercial bumble bee hives. My ChatGPT vocabulary is only vaguely useful, perhaps confused in translation, but the guide says family when describing a colony. I’ll guess that these guys interchange the words colony and family when speaking to visitors. Their relaxed and playful style of education wins the trust of the crowd. Somehow, the visit and the word family (instead of colony) shifted my perception, and I find myself in a The Matrix movie moment. The beekeeping experience includes passing around honey supers and brood comb, extraction of a few frames, honey tasting, and more honey tasting. These guys are living the life, and it felt like I had run away with the circus.

Saying goodbyes, I had more questions for András and found him smoking a cigarette - yeah, this is beekeeping in Hungary, and smoking is part of the relaxation culture.

tulip poplar windfall, other spring flowers and phenology '25

 

Based on my Atlanta backyard observations and record keeping, I found this year's first Tulip Popular flower windfall a few days early (day 93) - see my table below.  I compared my first tulip poplar windfall observations with other spring surrogates.  The National Phenology Network leaf out model is another way to determine if spring is early, on time, or late.

Spring Leaf Index Anomaly Map uses the first leaf out of cloned lilac and honeysuckle cultivars as these woody plants are among the first to leaf out or bloom and are common across much of the country.  The model also uses temperature and weather events beginning January 1 of each year.  The current spring is compared with a 1991-2020 average. The map shows that the 2025 Georgia Piedmont region's first leaf out day of the year is a few days early compared to the 30-year average.

Other Atlanta nectar sources at this time (not shown):

• Chinese Wysteria (invasive)
  
• White Clover
• Spiderwort
• Amur Honeysuckle (invasive)
• Dandelion

1st Tulip Poplar Windfall	Day of Year
10-Apr-2011	100
27-Mar-2012	87
19-Apr-2013	109
12-Apr-2015	102
15-Apr-2017	105
5-Apr-2018	95
11-Apr-2019	101
7-Apr-2021	97
6-Apr-2022	96
24-Mar-2023	83
6-Apr-2024	97
3-Apr-2025	93
median	97