Category: Individual-based Ecology

Guest blog post by Dominik Lammers

Models help us ask a deceptively simple question: if the world works in a particular way, what should we expect to happen?

For honey bee colonies, answering that question is difficult. A colony’s development depends on food, weather, disease, the surrounding landscape, and the behaviour of thousands of individual bees. These influences do not act separately. They affect one another, sometimes in ways that are hard to isolate even in carefully designed experiments.

In a recent study with Fabrice Requier, Andreas Focks, and Jürgen Groeneveld, we explored one of these connections using BEEHAVE, a computer model of a honey bee colony. We developed an exploratory extension called BEEHAVE-PPE, asking whether seasonal colony development could arise from links between pollen, brood pheromones, egg laying, and temperature, rather than following a seasonal egg-laying pattern specified in advance.

The result is not a finished replacement for the original model. It is a hypothesis about processes inside the hive: one that can produce plausible colony dynamics, while also making clear where scientific understanding remains incomplete.

A starting point that made BEEHAVE useful

BEEHAVE has been used to investigate how food availability, weather, parasites, pesticides, and beekeeping practices can affect honey bee colonies. It links processes inside the hive with conditions in the surrounding landscape, allowing researchers to explore combinations of stressors that would be difficult to study directly in real colonies.

To give a simulated colony a realistic seasonal trajectory, the original BEEHAVE uses an annual curve that determines how many eggs the queen lays on each day of the year. This was an effective modelling choice. It allowed the colony to develop in a broadly realistic way under typical Central European conditions and made it possible to investigate many other questions.

At the same time, this relationship shapes much of the simulated colony’s development. The queen follows a known seasonal pattern, while real colonies are likely to adjust reproduction in response to conditions inside and outside the hive.

I was interested in what would happen if that central pattern was no longer specified in advance. That became less like removing a single line from a model and more like beginning a journey. Each change exposed a new problem: without the fixed curve, what would initiate colony growth? What would prevent it from continuing indefinitely? What processes might connect the colony’s nutritional state to its reproduction?

For me, this was both a creative and an intellectual task. It involved imagining possible biological explanations, searching the literature for evidence that could support them, and translating those ideas into code. BEEHAVE-PPE emerged from that process.

A feedback loop inside the hive

The resulting model is built around three linked ideas.

First, the amount of pollen stored in the colony influences the queen’s egg laying. Pollen is the main protein source for feeding brood, so a colony with more available pollen can plausibly support more reproduction.

Second, larvae produce brood pheromones: chemical signals that can influence the behaviour of worker bees.

Third, brood pheromones can encourage workers to collect more pollen rather than nectar. This helps replenish pollen stores and can support further egg laying.

Together, these processes create a feedback loop:

pollen availability -> egg laying -> brood pheromones -> pollen collection -> pollen availability

Temperature affects this loop. In BEEHAVE-PPE, warmer conditions increase the assumed degradation of brood pheromone. This weakens the signal encouraging pollen collection and can slow colony growth.

In one sense, temperature fulfills a role similar to the original seasonal egg-laying curve: it helps shape when growth slows and when a colony reaches its annual peak. But it does so differently. Rather than instructing the queen to lay a certain number of eggs on a particular date, it represents an environmental condition that can differ between places and years. This opens the possibility that the same underlying model could respond differently under different temperature regimes, provided that its temperature relationships can eventually be tested and calibrated.

Letting seasonal dynamics emerge

Under the average weather conditions used in this study, BEEHAVE-PPE produced plausible seasonal colony dynamics. The simulated colonies grew in spring, reached a population peak in early summer, and declined afterwards. The number of adult bees in the simulated colonies followed the broad timing and shape of the French monitoring data used to calibrate the new module.

The model did not prove that real colonies work through precisely this mechanism. It cannot do that. A model can show that a proposed explanation is capable of generating an observed pattern; it cannot establish, by itself, that the explanation is the only or exact one used in nature.

What BEEHAVE-PPE does show is that the seasonal development of a colony need not be prescribed as a curve from the outset. A plausible combination of pollen availability, brood signalling, worker behaviour, and temperature can generate it.

That changes the role of the model. Instead of only reproducing a known seasonal pattern, it asks what biological connections could be responsible for that pattern.

Learning from where the model fails

The model also revealed a clear limitation. It produced plausible dynamics under averaged weather conditions, but it was vulnerable to prolonged periods in which bees could not collect pollen.

In the simulation, a long interruption weakens the feedback loop. Less pollen leads to reduced egg laying; fewer larvae produce less brood pheromone; and the weaker pheromone signal reduces the incentive to collect pollen when conditions improve. The colony can become trapped in a low-pollen, low-brood state.

This is unlikely to be the complete story in real colonies. Colonies can buffer difficult periods through stored resources and changes in brood and worker management. BEEHAVE already contains potentially relevant processes, including brood cannibalism and worker self-metabolism, but these are not yet represented in sufficient nutritional detail to support recovery within the new feedback loop.

That result gives the next steps a clearer direction. Rather than simply knowing that the model behaves unrealistically after sustained poor weather, we can identify the chain of events that causes it. This points to promising improvements, including better representation of nutrient reserves, resource recovery through brood cannibalism, and the colony processes that help it restart reproduction after a difficult period.

The data needed to go further

Developing and testing a model depends on data. The French dataset used in this study, covering 250 colonies, was especially valuable because it provided an unusually broad picture of seasonal colony development under comparable climatic conditions. It made it possible to see not just how one colony behaved, but what a larger set of colonies broadly did over a season.

This kind of baseline information is more limited than it may seem. For several important aspects of honey bee colony development, some of the most detailed observations still trace back to research from decades ago. We have valuable knowledge about individual processes, but less information showing how pollen availability, egg laying, brood development, worker behaviour, temperature, and population size change together over time.

New long-term observation approaches, including continuously monitored colonies, could help fill this gap. Targeted measurements of these linked processes would make it possible to test not only whether BEEHAVE-PPE produces realistic patterns, but whether it does so for the right biological reasons.

Why this matters for future stressor research

BEEHAVE is often used to explore how poor forage, adverse weather, parasites, pesticides, and beekeeping practices may affect colony development. In the original model, these pressures act on a colony whose broad reproductive trajectory is already set by the annual egg-laying pattern.

BEEHAVE-PPE changes that relationship. Because egg laying, brood production, pollen collection, and colony strength can influence one another, a stressor can affect more than one isolated part of the model. A shortage of pollen, for example, may not only reduce food available on a particular day. It may also reduce reproduction, alter brood signals, change later foraging behaviour, and affect the colony’s capacity to recover.

The current version may respond too strongly when the feedback loop is interrupted. But if these dynamics can be stabilised and supported by stronger empirical evidence, future versions could give a more complete picture of how stressors interact. They could help investigate when the effects of a stressor are amplified by the colony’s internal state, which combinations are most damaging, and where a colony’s natural buffering mechanisms provide protection.

This is particularly relevant as weather patterns, flowering times, forage availability, and temperature change together. BEEHAVE-PPE is not yet a forecasting tool for climate change or colony risk. It is a first step from an imposed seasonal pattern toward interacting biological mechanisms.

Its main value is that it makes both a plausible explanation and its remaining gaps visible. That is one of the strengths of models: they do not only tell us what we can predict. They show us what we still need to understand before prediction becomes possible.

Original source: 

Lammers D, Requier F, Focks A, Groeneveld J (2026) Food for thought: could the queen’s egg-laying rate in the BEEHAVE honey bee model emerge from the effects of brood pheromones, weather conditions, and pollen availability? Individual-based Ecology 2: e185721. https://doi.org/10.3897/ibe.2.185721