Individuals entered the system at any point in the three-dimensional environment. Species identities of inflowing propagules were chosen at random from a uniform distribution. Individuals were assigned a unique ID and a randomly chosen cell quota between 0.5 and 1.0, which translated to starting cell diameters of 1 to 1.25 μm.

Consumption increased an individual's cell quota and decreased the size of the resource that an individual had encountered. Growth was simulated as an increase in cell quota that corresponded to a proportional increase in biomass. That is, we assumed that cell quota (Q) and cell volume (V) were proportional and hence, that Q was related to a cell's radius through a geometric relationship: Q ∝ V = ⁴/₃ π r³. Estimating cell size was important as the ability to consume a resource particle relied on whether a cell and a resource particle were in physical contact. The energetic cost of growth was proportional to the product of growth rate and individual cell quota, such that larger individuals required more energy to increase by a fraction of their body mass.

Individual cell quotas diminished according to a randomly chosen species-specific maintenance cost, i.e., a numeric value subtracted from the cell quota. This cost was the implied energetic cost of maintaining basal metabolism and cellular maintenance. When active individuals transitioned to a dormant state, the cell maintenance cost decreased by a species-specific factor that was likewise chosen at random. This cell maintenance cost was reversed when individuals resuscitated to a metabolically active state.

Reproduction was clonal and occurred via binary fission and without mutation. Individuals reproduced with a probability directly determined by the ratio of cell quota to maximum cell quota (i.e., 1000). Hence, the probability of reproducing when randomly sampled was between 0.0 and 1.0. The daughter cell was given a unique individual ID along with the species ID and geographical coordinates of its mother cell.

An individual sampled at random died if its cell quota was less than the species-specific maintenance energy, whether in an active or dormant state. Depending on the model, the remains of dead individuals could be scavenged and consumed by other cells. Otherwise, dead individuals were effectively lost to the system, as in most community ecology models.

We simulated levels of spatial complexity in terms of the spatial structure of the environment and modes of individual dispersal (Figures 1, 2). The first level of spatial complexity in the environment was a well-mixed model in which the locations of individual organisms and resource molecules changed at each time step in an uncorrelated way. Hence, every organism and resource particle had the same chance of moving to any location within the environment at each time step in the model. While this is unrealistic in highly structured systems of relatively large size, the microscale spatial dynamics of more fluid systems can be highly complex, rapidly mixed, and unpredictable (Rusconi et al., 2014). This model created a well-mixed environment and allowed for passive dispersal. In the second level of spatial complexity, resource molecules only changed their location via Brownian motion, resulting in a system that lacked mixing and where, through the breakdown of large resource particles, smaller resource particles became aggregated. We refer to this second level as “spatially structured.”

The first mode of individual dispersal excluded active dispersal and was entirely passive. While passive dispersal has the benefit of not requiring an energetic investment, we expected this mode of dispersal to only have a benefit under well-mixed conditions. The second mode of dispersal allowed active movement, but did not allow chemoreception. This “run and tumble” strategy allowed individuals to move along a straight line in a randomly chosen direction and, once hitting the edge of the environment, turned and moved in a new, randomly chosen direction. By chance alone, individuals would be able to encounter resource particles. The third mode of dispersal simulated chemotaxis, i.e., the ability to sense the location of resources and actively move toward them. Because chemotaxis involves sense perception, we encoded it as a more energetically expensive mode of dispersal. In this way, one could envision a potential trade-off between the greater energetic cost of chemotaxis and the efficiency associated with more rapid encounter of resources.

Our simulations captured important features of resource complexity that we hypothesized would affect encounter rates. First, we considered resource diversity, which refers to the different types of resources that are supplied to a system. In the “monoculture” simulation, only one type of resource was supplied. Second, we considered a “polyculture” scenario, where three different types of resources were supplied and each could only be used by specialist consumers (Figure 1). Third, we considered that some resource particles are more difficult to break down, or recalcitrant, than others. To simulate recalcitrance, we imposed a “lock and key” scenario that required individuals to invest time and energy to break down complex molecules for consumption. Importantly, a given model could include this “lock and key” feature while being a monoculture or polyculture.

We simulated three aspects of trophic complexity that we hypothesized would influence resource encounter (Figure 1). First, we simulated a simple heterotrophic “consumer-resource” interaction where all individuals were solely consumers of inflowing resources. The second level of trophic complexity allowed for “scavenging” which is specified in the model as the consumption of resources from dead individuals (e.g., Rozen et al., 2009). Our third level of trophic complexity simulated a situation wherein individuals produced metabolites as a result of breaking down and consuming resources. These metabolites could potentially be consumed by other species that, in turn, produced metabolites that also be consumed. This situation was meant to simulate conditions that are characteristic of cross-feeding or syntrophy (Pande et al., 2015). Additionally, because the metabolites that individuals of particular species produced were chosen at random at the start of the model, it was possible for mutualistic cross-feeding to occur, i.e., where two species produced metabolites that the other could consume.

Encounter rates were most affected by interactions between dispersal mode and the spatial structure of the environment (Figure 3). In well-mixed environments, neither chemotaxis nor “run and tumble” produced greater encounters than the energetically free strategy of passive dispersal. However, in structured environments, chemotaxis produced 10–100 times more encounters while passive dispersal frequently resulted in no encounters. Chemotaxis is an important microbial trait that has strong effects on resource encounter rates (e.g., Datta et al., 2016; Smriga et al., 2016). The energetic cost of this directed movement must have been offset by the energy saved in not encountering resources at random, as with our simulated “run and tumble” mode of dispersal. It is generally assumed that chemotaxis is a trait that is favored in productive habitats and selected against in oligotrophic environments where searching for sparse chemically complex resources could be energetically wasteful (Ottemann and Miller, 1997). This view is supported by genomic evidence suggesting that bacteria from productive environments tend to harbor more motility genes than bacteria from oligotrophic environments (Giovannoni et al., 2005; Lauro et al., 2009). Though our models did not support these observations, the energetic cost of chemotaxis, as modeled, may not have been great enough to substantially tax individual cell quotas. More likely, however, individuals that became energetically depleted through chemotaxis simply went dormant.

Habitat heterogeneity is thought to limit the probability of decomposition and to influence the accumulation of resources in spatially structured environments like soil (Schmidt et al., 2011; Dungait et al., 2012; Lehmann and Kleber, 2015). Our simulations support this view. Except in the case of chemotaxis, we found that encounter rates were lower in spatially structured environments, which resulted in the aggregation and persistence of resources in the environment (Figure 3). This pattern is most clearly seen in the comparisons between mixed and structured environments, where the “run and tumble” strategy or passive dispersal resulted in greatly reduced encounter rates (Figure 3). Our results highlight the benefits of chemotaxis in overcoming the challenges of resource acquisition in spatially structured environments. On the other hand, in well-mixed environments, chemotaxis did not result in greater encounters compared to the energetically less expensive “run and tumble” mode of dispersal and the energetically free passive dispersal strategy. These model-based findings suggest the importance of considering energetic trade-offs related to modes of dispersal and the spatial distribution of resources.

The recalcitrance of resources reduced encounter rates in both mixed and structured environments (Figure 3). In order for microorganisms to consume complex resources, they needed to invest energy in cleaving resource particles into consumable parts, slowing encounter with those breakdown products. This finding confirms how structurally complex or recalcitrant resources are thought to influence the growth and activity of microorganisms within habitats, such as soil (Schimel and Weintraub, 2003; Allison et al., 2011). Along with the spatial distribution of resources, chemical complexity slows the loss of energy from the resource pool by limiting encounter. This “slow release” effect has been hypothesized to affect ecosystem dynamics (e.g., Wetzel, 1999) and its interaction with spatial structure is consistent with emerging views on the controls of organic matter persistence in soil environments (Lehmann and Kleber, 2015). Furthermore, time series of our simulated data suggest that the slow turnover of recalcitrant substrates contributes to a more stable dynamic of resource encounter (Figure 4). Whereas the inclusion of complex resources (i.e., “lock-and-key” models) had little-to-no influence on average encounter in well-mixed models, the presence of recalcitrant resources did impart greater stability in encounter rates over time, regardless of dispersal mode or trophic complexity (Figure 4).

Our simulations revealed that trophic complexity had only modest effects on encounter rates (Figure 3). Both scavenging and cross-feeding led to slight increases in resource encounter in both well-mixed and structured environments across all modes of dispersal. However, the small effect of trophic complexity on resource encounter should be cautiously interpreted. First, we only considered a few types of trophic interactions and we encoded them in fairly specific forms. In addition to scavenging and cross-feeding, microbial communities engage in a plethora of trophic interactions, which could affect encounter and consumer-resource interaction strength. Second, aspects of our modeling may have dampened the effects of scavenging and cross-feeding on encounter rates. For example, because our models included dormancy which reduces mortality, seed banks may have reduced the importance of scavenging by decreasing available necromass. Finally, it is possible that scavenging and cross-feeding affected resource encounter for a small cross-section of models but that the signal was overshadowed by the larger effects of spatial and resource complexity. We envision two ways that future studies could more closely examine the roles of trophic complexity in driving microbe-resource encounters. First, studies could explore the influence of trophic levels or ask whether trophic interactions that lower the abundances of primary consumers (e.g., predator-prey, parasitism) drive resource encounters. Second, the possibility that scavenging could increase the availability of resources to organisms warrants greater attention to the energetic value derived from necromass.

Greater resource encounters led to greater productivity and greater total abundance, and smaller seed banks (Figure 5). While these relationships can be expected since resource encounters are required to fuel growth, we also observed that modes of dispersal influenced the general forms of these relationships. For example, in spatially structured models, chemotaxis not only led to greater encounters but also some of the largest seed banks. In contrast, the “run and tumble” strategy produced the overall smallest seed banks despite encounter rates that were often very low (Figure 5). We attribute the stark difference between these two strategies to their energetic cost, where chemotaxis was twice as energetically costly as “run and tumble”. Resource complexity and trophic complexity had no effect on relationships of abundance and productivity to rates of encounter regardless of whether the environment was structured or well-mixed.

Consistent with energy- and nutrient limitation as a control on microbial activity, we observed that seed banks were largest in simulations that included recalcitrant resources and passive dispersal (Figure 6). In contrast, the relative size of seed banks was not related to the macroscale properties of total resource supply and bulk resource concentration (Figure 6). Finally, we often found that passively dispersing individuals did not encounter resource particles (Figure 7). On closer inspection, this appeared to be due to the small size and sparse spatial distribution of many resource particles, i.e., “crumbs,” which has been suggested as an alternative to resource recalcitrance as a reason for why large amounts of total resources seem to be resistant to microbial consumption (Arrieta et al., 2015). In soil systems, low concentrations of organic matter have been found to limit its use by microbes (Don et al., 2013), presumably because dilute resources limit the encounter with organisms. Consequently, it may be that natural systems in which we find large seed banks are likely to be those characterized by recalcitrant or chemically complex resources, such as plant structural materials, as well as sparse distributions of relatively small resource particles.

Time series from our IBMs revealed important influences of microbial seed banks across 2000 generations. Active dispersal and a strong capacity for dormancy helped maintain large and stable communities over time (Figures 7, 8). However, while chemotaxis resulted in the highest rates of encounter and greatest abundances, it also produced substantially larger seed banks than models that incorporated the less energetically expensive “run and tumble” strategy (Figure 7). This result suggests that while chemotaxis may be effective for locating resources, its energetic demands can cause organisms to enter dormancy as a consequence of exhausting their cell quotas in pursuit of resources. We found that communities went extinct less often and fluctuated less wildly when there was a high capacity for dormancy, i.e., low maintenance energy and low probability of randomly resuscitating (Figure 8). Under a weak capacity for dormancy, communities of active and passive dispersers often went extinct even when the bulk concentration of resources was high and relatively stable (Figure 8). These findings not only confirm our support the importance of seed banks to microbial communities in a variety of systems, but also indicate that extinction can occur under relatively high resource concentrations, i.e., if individuals fail to encounter and consume resources.