Ecosystem

A food web written in space. Plants renew, herbivores and predators meet or miss one another, and each animal’s energy decides whether it reproduces or dies. Abundance is not commanded by a global population equation; booms, lags, crashes and recovery emerge from countless local encounters under limited resources.

The word ecosystem entered ecology through Arthur Tansley’s 1935 argument that organisms and the physical factors of their environment should be treated as components of a combined system. This matters because “three kinds of agents on one screen” would not, by itself, justify the title. The AL.Floys world moves somewhat closer to Tansley’s system-level idea: organisms are embedded in spatial resource conditions, plant renewal can be patchy, optional fertile zones change plant opportunity, seasons modulate primary production, and animal mortality can create temporary local nutrient advantage.

Those mechanisms are deliberately schematic. There is no soil chemistry, water cycle, weather, photosynthetic energy flux or explicit nutrient conservation. The title is justified by the structure of reciprocal dependence, not by ecological completeness.

The model’s central relation is trophic feedback. A simplified causal chain is

plant availability → herbivore feeding and reproduction → predator feeding and reproduction → herbivore pressure → plant availability.

Every arrow contains delay. Plants must first appear in accessible places. Herbivores must encounter and consume them. Increased herbivore success changes predator opportunity only after population and spatial effects accumulate. Predators that overexploit prey may later lose their own food supply. This delayed feedback can create oscillation-like patterns even though no oscillator has been explicitly inserted.

The classic mathematical neighbour is the Lotka–Volterra predator–prey system:

Compact form: (dx)/(dt)=\alpha x-\beta xy, \qquad (dy)/(dt)=\delta xy-\gamma y.

Here, prey grow in the absence of predators; predator–prey encounters reduce prey and support predator growth; predators decline without prey. Alfred Lotka developed related two-species equations before Vito Volterra’s independent 1926 treatment, and Volterra publicly recognised Lotka’s priority for the equations in 1927.

These equations are useful conceptual background, but they are not what ecospheres.ai computes. The engine never updates three global density variables by integrating population differential equations. Each animal has a position, velocity, energy store, species and inherited trait multipliers. A herbivore searches its local neighbourhood for plants and predators; a predator searches for prey. Feeding happens only when the corresponding individuals meet within a small distance. Geometry therefore affects the realised encounter rate.

This distinction is scientifically consequential. In a well-mixed population equation, abundance largely determines the opportunity for encounter. In a spatial individual-based model, the same total population can produce different outcomes depending on clustering, separation, local depletion, refuges and movement. A hundred predators concentrated in one region do not exert the same immediate pressure as a hundred predators evenly distributed across the world.

The engine’s energy variable links behaviour to demography. Moving animals lose energy through a metabolic cost. Herbivores gain it by consuming plants; predators gain it by consuming herbivores. Reaching zero causes death. Exceeding a species-specific reproduction threshold permits an offspring, with parental energy divided in the process.

This variable should be interpreted as an abstract biological resource or energetic state, not as joules. The engine does not calibrate caloric values, body masses, assimilation efficiencies or thermodynamic losses. The conceptual claim is that successful feeding funds persistence and reproduction; the implementation is not an energy-flow budget.

That caveat also separates the sphere from Raymond Lindeman’s trophic-dynamic tradition. Lindeman’s 1942 work made transfers of energy and matter across trophic levels central to ecosystem analysis. AL.Floys borrows the idea that higher trophic persistence depends on lower trophic production, but it does not calculate a conserved or measured energy flux across an ecosystem.

A second feedback introduces limited environmental memory. When a prey or predator dies and nutrient recycling is enabled, the engine deposits a temporary local nutrient patch. While the patch remains, plant spawn cost is reduced nearby, making future plant growth more likely around that location. Mortality can therefore alter later primary-production opportunity in approximately the same region.

The visual gold bloom is not decomposition itself. There are no microbes, detritivores or nutrient atoms. The scientific idea is narrower and still worthwhile: past biological events can modify the future environment, creating a local memory outside any organism.

The resource field adds a different form of spatial heterogeneity. Hidden fertile centres bias plant generation toward some regions; patch sharpness and richness determine how localised and influential those centres are. Animals subsequently reorganise around a landscape that they did not create. This allows the sphere to distinguish endogenous structure—patterns produced by the organisms—from exogenous heterogeneity—patterns imposed by the resource substrate.

Seasonality adds temporal heterogeneity by modulating plant growth. Because the seasonal field affects primary production before it affects consumers, it can generate phase lags across trophic levels. A green field may begin to recover while predators are still declining from an earlier herbivore shortage. The visible populations are therefore records of past conditions, not instantaneous readouts of the current plant-growth multiplier.

The model also contains a bounded inheritance mechanism. Animal offspring inherit multipliers for speed, vision and size; mutation can perturb these values within fixed bounds. If a trait combination improves feeding, escape or reproduction under the current world, its descendants may become more common. This is enough to support limited selection and trait drift.

It is not open-ended evolution. The trait list, mutation rule and permissible range are designed in advance. There is no genome, recombination, sexual reproduction, developmental process or invention of new traits. The term evolving in the subtitle should therefore mean changing inherited trait distributions under differential reproductive success, not unrestricted evolutionary innovation.

Individual-based models are especially useful for this kind of heterogeneity. DeAngelis and Mooij emphasise that such models can explicitly represent spatial, phenotypic, cognitive, ontogenetic and genetic variation that population-average equations usually compress. Ecosystem uses a small subset of that advantage: individuals differ in location, state and inherited behavioural capacity.

The world is maintained far from equilibrium. Plants continuously appear; animals consume, move, reproduce and die; nutrient patches decay; seasonal and resource effects alter opportunity; inherited distributions can change. A visually stable population balance is therefore not a static equilibrium. It is an ongoing throughput of events.

Finite populations introduce another important difference from continuous equations. A density variable can approach a small positive number indefinitely; a finite population can reach exactly zero. In the current default and all shipped public presets, automatic repopulation is disabled. Extinction is therefore genuine within a run unless the visitor manually adds organisms or restarts. Recovery is possible when a stressed population remains above zero; resurrection after extinction is an external intervention.

This distinction should shape the public language. Say local extinction, collapse, or recovery from low abundance when that is what the engine shows. Do not imply that a vanished species spontaneously re-evolves.

Four strands meet in this sphere.

The ecosystem concept. In 1935, Arthur Tansley argued against treating plant communities as quasi-organisms isolated from physical conditions. He proposed the ecosystem as the whole system formed by organisms and environmental factors. That move made feedback between living and non-living components conceptually central.

Mathematical population dynamics. Alfred Lotka and Vito Volterra showed in the 1920s that simple trophic coupling could produce nontrivial population trajectories. Volterra’s 1926 Nature paper responded to questions about interacting species and fisheries; the subsequent correspondence established Lotka’s earlier derivation of the two-species equations. Their work made delayed mutual dependence mathematically explicit.

Trophic and energetic systems ecology. Raymond Lindeman’s 1942 paper organised ecological systems around trophic relations and transformations of energy and matter. It helped shift attention from isolated population counts to flows through whole ecological systems.

Individual-based modelling. Later computational ecology reintroduced explicit organisms with heterogeneous state, location and behaviour. Rather than assuming that population averages are sufficient, individual-based models ask how population-level patterns arise from the histories and interactions of individuals.

AL.Floys is best located at the convergence of those traditions. It is not a canonical implementation of Tansley, Lotka–Volterra or Lindeman. It is a visual spatial artificial ecology that borrows their central questions—system boundary, trophic feedback, resource dependence and population consequence—while using individual agents as the computational unit.

Aspect	Current implementation
Entities / field	Discrete plants, herbivores and predators; temporary nutrient patches; optional fixed resource-patch centres.
Animal state	Position, velocity, energy, species, bounded speed/vision/size multipliers and generation number.
Plant state	Position and species identity; plants do not move.
Update mechanism	Animals wander and mildly flock; herbivores seek plants and flee predators; predators seek herbivores; close encounters permit feeding; metabolism, death and reproduction update populations.
Neighbourhood / interaction range	Spatial-hash queries with species- and trait-dependent perception distances; feeding occurs at shorter range.
Boundary conditions	Default toroidal wrap; an advanced closed world makes animals bounce and uses ordinary Euclidean boundary distances.
Plant renewal	A growth accumulator spawns discrete plants up to a cap, biased by patchiness, resource richness and nutrient boosts.
Inheritance	Animal offspring inherit speed, vision and size multipliers with bounded stochastic mutation.
Environmental feedback	Animal death can deposit decaying nutrient patches that make nearby plant spawning cheaper.
Randomness	Seeded initial placement, movement perturbation, offspring placement, plant spawning and mutation.
Numerical form	Real-time individual updates; no integration of global population ODEs.
Rendering interpretation	Terrain tint, resource overlay, nutrient blooms, event rings, optional vision rings, trails, season veil and population telemetry. Several are visual aids rather than model state.
Public simple controls	Plant growth, Plant cap, Initial herbivores, Initial predators and Herbivore vision.
Public tools	Spawn, Force, Brush and Erase.

The default autoRepopulate value is false, and the shipped presets inherit that value. A population reaching zero remains extinct until the visitor intervenes or restarts.

Trophic lag — causes arrive at different levels at different times

A reduction in plant growth first changes plant abundance or accessibility. Herbivore energy and reproduction respond later. Predator decline may arrive later still. The lag is the visible signature of dependency propagating through a causal chain.

Boom and bust — success can undermine its own conditions

Abundant plants can support rapid herbivore growth. High herbivore numbers then deplete plants; predators may also rise in response to prey abundance. The favourable state creates pressures that later destroy it, illustrating delayed negative feedback.

Spatial refuge — abundance is not the whole story

A prey population may survive because individuals occupy regions predators have not reached, or because plant patches separate feeding from hunting pressure. Spatial organisation changes encounter rates even when total counts remain the same.

Local extinction — finite populations can cross an irreversible boundary

A trophic group can disappear entirely. Once the last individual is gone, ordinary internal dynamics cannot recover it. The event distinguishes finite agent models from continuous abundance variables.

Resource concentration — the substrate organises consumers

Where fertile resource patches or strongly patchy plant growth are present, herbivores cluster around renewed food and predators follow. Environmental heterogeneity becomes a generator of biological spatial pattern.

Nutrient memory — death changes what can grow later

Gold nutrient blooms appear near animal mortality, then decay while locally lowering plant spawn cost. The field records a past event outside the surviving organisms.

Trait drift — ecological success changes inherited averages

In Evolution Drift, lineages vary in speed, vision and size. Mean values and trait diversity can move as some variants reproduce more effectively. The visible ecology and the inherited population state become coupled.

First 30 seconds

1. Load Balanced Web.
2. Identify the three trophic types before opening any controls: stationary plants, mobile herbivores and mobile predators.
3. Follow one herbivore. Notice the alternation between foraging, local wandering and flight from a predator.
4. Lower Plant growth sharply while leaving all other settings unchanged.
5. Wait. The lesson is not the immediate plant response but the delayed herbivore and predator consequences.
6. Restore the baseline or reload Balanced Web, then compare Predator Pressure.

Three experiments

Experiment	Question	Do	Watch for	Why it matters
Bottom-up limitation	How does reduced primary production propagate upward?	Balanced Web → reduce Plant growth substantially → wait without touching other controls.	Plant scarcity, later herbivore stress, then delayed predator decline.	Makes trophic lag and dependency visible.
History from initial conditions	Can the initial predator burden decide whether prey survive?	Increase Initial predators → use Restart so the reset-bound control takes effect.	Early prey loss, possible prey extinction, followed by predator starvation.	Shows path dependence and finite-population thresholds.
From sensing to demography	Can one individual behavioural capacity alter population-level outcomes?	Balanced Web → vary Herbivore vision only, first low then high.	Search efficiency, escape timing, energy state and later abundance.	Connects perception radius to trophic dynamics.

For a fourth, advanced experiment, load Evolution Drift and compare trait telemetry over several generations. Treat it as bounded inheritance under selection, not as open-ended evolution.

Parameter	What it really controls	Increase it	Decrease it	What to watch
Plant growth	Rate at which discrete primary resources are renewed.	More bottom-up support and faster recovery.	Greater starvation pressure.	Delayed effects across trophic levels.
Plant cap	Maximum number of living plant agents.	Higher standing-resource ceiling.	Tighter carrying conditions for herbivores.	Long-run consumer support, not just green coverage.
Initial herbivores	Starting grazing pressure; reset-bound.	Faster initial consumption and more predator food.	Gentler plant pressure and smaller prey base.	Early transients and later overshoot.
Initial predators	Starting top-down pressure; reset-bound.	More immediate hunting.	Greater initial herbivore release.	Extinction cascades and predator starvation.
Herbivore vision	Radius for detecting both food and predators, modified by inherited traits.	More effective search and earlier threat response.	More local, less informed movement.	Feeding success versus escape.
Predator vision (advanced)	Radius for locating prey.	More efficient search.	More spatial refuge for prey.	Hunting fronts and prey persistence.
Mutation rate / strength (advanced)	Frequency and magnitude of bounded inherited trait changes.	Faster divergence and greater variation.	More stable trait distributions.	Mean traits and diversity across generations.
Resource richness (advanced)	Strength of a fixed fertile substrate.	Stronger spatial heterogeneity.	A more homogeneous world.	Persistent resource-centred clusters.
Nutrient recycling (advanced)	Whether animal death creates local plant-growth advantage.	Enables mortality-to-resource feedback.	Removes that environmental memory.	Regrowth around past deaths.
Season cycle / strength (advanced)	Periodic modulation of plant renewal.	Stronger external forcing.	Flatter primary production.	Phase lags between plants, herbivores and predators.
Resource overlay (visual)	Display of the hidden fertile field.	More legible substrate.	Less visual guidance.	It reveals opportunity; it does not itself change it.

The simple public tier is intentionally narrow. That is appropriate for first contact: Plant growth is the strongest single pedagogical control because it initiates a causal cascade rather than merely changing appearance.

Balanced Web

The reference artificial ecology. Moderate resource heterogeneity makes plant patches visible without overwhelming the trophic interaction. Use it for one-variable experiments.

Predator Pressure

Sharper, faster predators make hunting and prey flight easier to observe. Compare it with Balanced Web to examine stronger top-down pressure, but remember that the preset changes several behavioural values simultaneously.

Boom-Bust Cycle

Higher plant renewal, altered reproduction thresholds, seasonal forcing and a visible population graph amplify temporal fluctuation. It is the best preset for recognising lag, but a visible cycle should not be labelled “Lotka–Volterra” without mechanism-specific evidence.

Evolution Drift

Increased mutation makes bounded inherited variation perceptible within an ordinary visit. It is the best entry to eco-evolutionary feedback: ecological success alters which trait values persist.

Sparse Field / Low Power

Primarily a performance and legibility regime. Lower counts and larger organisms make individual encounters easier to follow, so it can still be pedagogically useful even though its first purpose is computational accessibility.

Recommended learning order: Balanced Web → Predator Pressure → Boom-Bust Cycle → Evolution Drift.

• Spawn adds the selected trophic type within the gesture radius. It is an external population intervention, not reproduction.
• Brush also adds organisms in this engine; it does not repaint or convert existing individuals. The target species determines what is introduced.
• Force alters animal velocities but does not move plants. It can create a temporary refuge, hunting disruption or local crowding without changing population counts directly.
• Erase probabilistically removes organisms within the radius. When nutrient recycling is active, erased prey or predators can deposit nutrient patches, so erasure is not necessarily environmentally neutral.

The most informative intervention is often a small one: add a few predators near a dense herbivore patch, or use Force to separate a predator from prey, then observe whether the perturbation remains local or changes later population dynamics.

This is not a realistic natural ecosystem. It omits species diversity, age and stage structure, mating systems, disease, explicit decomposers, hydrology, climate physics, soil chemistry, migration between habitats and conservation-based nutrient or energy budgets.

The energy state is an abstract life-history currency, not joules. Nutrient patches are a heuristic feedback, not mechanistic decomposition. Seasons are a periodic multiplier, not meteorology. Resource centres are designed spatial opportunity, not measured geology or soil fertility.

The inherited traits support limited adaptive change, but there is no genome, sexual recombination, developmental biology or open-ended innovation. Use evolution only with that qualification.

Finally, an oscillating graph does not prove a Lotka–Volterra mechanism. The same visual rhythm can arise here from spatial encounters, finite population noise, plant renewal, energetic thresholds, resource patchiness, seasonality or their combination.

Ecosystem adds dependency, scarcity and delayed feedback across unlike populations. Flock’s agents mainly coordinate with peers; here the success of one class changes the survival conditions of another. The sphere introduces birth, death, extinction, resource renewal, inheritance and environmental memory—the first world in the atlas where order is inseparable from what is consumed and what is lost.

Foundational and primary sources

1. A. G. Tansley, “The Use and Abuse of Vegetational Concepts and Terms,” Ecology 16, 284–307 (1935). DOI: 10.2307/1930070.
2. Alfred J. Lotka, Elements of Physical Biology (Williams & Wilkins, 1925).
3. Vito Volterra, “Fluctuations in the Abundance of a Species considered Mathematically,” Nature 118, 558–560 (1926). DOI: 10.1038/118558a0.
4. Alfred J. Lotka and Vito Volterra, 1927 correspondence in Nature on priority for the two-species equations: DOI 10.1038/119012a0 and 10.1038/119012b0.
5. Raymond L. Lindeman, “The Trophic-Dynamic Aspect of Ecology,” Ecology 23, 399–417 (1942). DOI: 10.2307/1930126.

Modern scientific context

6. Donald L. DeAngelis and Wolf M. Mooij, “Individual-based modeling of ecological and evolutionary processes,” Annual Review of Ecology, Evolution, and Systematics 36, 147–168 (2005). DOI: 10.1146/annurev.ecolsys.36.102003.152644.