Individuals were placed on a 1 × 1 grid with a uniform, random, or clustered spatial distribution (

Figure 1). We varied the number of individuals initially established exponentially from 25 to 225. This initial variation in the number of individuals is responsible for the ascending limb of the richness-biomass relationship in our model. We generated a new independent spatial distribution of individuals for each run of the model. The uniform spatial distribution was constructed by spacing each individual a fixed interval apart on the grid. The random spatial distribution was constructed by assigning the coordinates of individuals using a random uniform distribution. The clustered spatial distribution was generated by distributing the centers of the clusters according to a random uniform distribution and then normally distributing the individuals around each of the clusters. We arbitrarily set the number of cluster centers equal to 5% of the total number of individuals to achieve a high degree of clustering in the communities. The standard deviation of the normal distribution was set to the initial diameter of the individuals. This clustering process is similar but not identical to the popular Neyman-Scott point-process [

14]. It is not a true Neyman-Scott point-process because individuals that were slated to become established outside of the 1 × 1 grid were reassigned to a new spatial cluster; additionally, the number of clusters was not a Poisson variable because it was always equal to 5% of the total number of individuals. Due to our boundary restrictions the clusters located in the center of the landscape had a slightly higher number of individuals.

At stand establishment all individuals were assigned the same starting diameter. The relative abundance distribution (RAD) of the species pool was set to uniform such that each species was equally likely to be drawn with replacement from the pool. Although a uniform RAD is not common in nature, more realistic RADs that contain more rare species (e.g., log normal and geometric distributions) simply compressed the richness-biomass relationship along the y-axis by decreasing the mean number of species in the community and did not influence the overall qualitative shape of the relationship (results not shown). Species were drawn with replacement randomly from the species pool.

**Figure 1.**
A top-down view of simulated communities at three points in time: 1, 15, and 30. Each column of the panel depicts the results for a different initial spatial distribution: clustered, random, or uniform. Each circle represents an individual, the color of the circle indicates the species identity, and the area of the circle is equal to the area that the individual occupied in the 1 × 1 window. The same number of individuals was in each community initially.

**Figure 1.**
A top-down view of simulated communities at three points in time: 1, 15, and 30. Each column of the panel depicts the results for a different initial spatial distribution: clustered, random, or uniform. Each circle represents an individual, the color of the circle indicates the species identity, and the area of the circle is equal to the area that the individual occupied in the 1 × 1 window. The same number of individuals was in each community initially.