Configs.md

@@ -5,12 +5,48 @@ Configs
 
 # 0. Complete configs
  ### 0.1 Pantropical Diversity, Hagen and Skeels et al 2021 
- [config template](https://gitlab.ethz.ch/ites-le/gen3sis/gen3sis_wiki/-/blob/master/configs/Pantropical_config_template.R)
+
+Description:
+Paleoenvironment was reconstructed for the entire globe for the last 110 Myr at a temporal resolution of \textasciitilde170 kyr and a spatial resolution of 2$^{\circ}$ and was characterized by approximate air surface temperature (related to MAT) and an aridity index (related to MAP and PET), following \cite{hagen2019}. Air surface temperature was reconstructed by combining (i) paleotopography, estimated from paleoelevation models \cite{ScoteseWright2018}, with (ii) reconstructions of paleo-Köppen climatic zones based on the geographic distribution of lithologic indicators of climate \cite{Boucot2013, Scotese2015}, modified using the current temperature lapse rate for each Köppen zone based on the current elevation and MAT downloaded from WorldClim2 \cite{FickHijmans2017}. The aridity index was reconstructed from the paleo-Köppen bands and given a value of one for the arid Köppen regions and zero for all the other bands. We additionally modified the input in five ways for use in the landscape modification experiment: we reduced the temperature heterogeneity associated with orogenesis of the Andes region from 110 Ma; we held temperatures of the Indomalayan region at a constant value; we removed the cost associated with crossing water in the Southeast Asian Archipelago; and we changed arid cells in the Afrotropics to non-arid cells from 110 Ma and 23 Ma (further details on the reconstruction in SI Appendix; Fig S11).
+
+We implemented the spatial model of diversification using the general engine for eco-evolutionary simulations, \textit{gen3sis} \cite{gen3sis}. Each simulation followed the diversification of a clade from a single ancestral species distributed broadly throughout non-arid sites within 25$^{\circ}$ of the equator 110 Ma. We also tested the sensitivity of the pantropical diversity patterns to different initial ancestral ranges, including an exclusively extra-tropical ancestor, and found the Afrotropics had lower diversity under these alternative starting conditions (further details in appendix SI; Fig S12). Each simulation follows a clade's radiation from the initial species throughout 110 Myr of reconstructed paleoenvironmental changes across 2$^{\circ}$ grid sites on the global landscape, considering four major processes: dispersal, environmental filtering, niche evolution, and speciation.
+
+\textit{Dispersal}. At each time-step (\textasciitilde170 kyr, 660 time-steps in total), each population could disperse into surrounding grid sites from a dispersal kernel drawn from a Weibull distribution centred on 2$^{\circ}$ (approximately 222 km of latitude at the equator) with shape ${\phi}$.
+
+\textit{Environmental filtering}. The presence of species \textit{i} in site \textit{s} was determined by a match between the species temperature niche width (${\omega_i}$) and the local temperature value ${T_s}$. Each species could be present in a site if $|{T_i} \pm {\omega}|>T_{s}$, where ${T_i}$ is the temperature niche centre. One of the major constraints for the distribution of tropical moist forest taxa is water availability \cite{Bjorholm}. In this study, the paleo-aridity data is derived from a binary layer, thus limiting modelling an aridity niche in a way comparable to that of the temperature niche. Therefore, to implement environmental filtering based on water availability, we place a hard constraint on species entering arid grid cells. This constraint prevents species from entering arid grid cells and the assumption is supported by empirical evidence that biome shifts between forest and arid biomes are exceptionally rare \cite{Crisp}. Extinction occurred when a species no longer occupied any grid cells as a result of mismatches between the species environmental niche and the environment.
+
+\textit{Niche evolution}. Evolution of the temperature niche trait \textit{z} followed a Brownian motion model of trait evolution, where the value of $\bar{z_i}$ at increasing time intervals of $\Delta \textit{t}$ is equal to the value of $\bar{z_i}$ at time \textit{t}, plus a value drawn from a normal distribution with a mean of 0 and standard deviation of $\sigma$.
+
+\textit{Speciation}. Speciation followed the biological species concept \cite{coyne2004speciation} in which species are considered reproductively isolated populations. Populations of a species that became geographically isolated from each other diverged genetically at each time-step, and once divergence had crossed a speciation threshold (\textit{S}) the populations became new distinct species. This equates to a Bateson–Dobzhansky–Muller model of genetic incompatibility \cite{coyne2004speciation}.
+
+We ran 500 simulations over a variable range of the four main model parameters, determined where possible based on empirical data, and subsequently based on a preliminary exploration of the parameter space (see SI Appendix): $\omega$=[0.04, 0.1], corresponding to a niche width of \textasciitilde2.6$^{\circ}$C to 7$^{\circ}$C; $\phi$=[2, 15], corresponding to a dispersal kernel with a right skew to include more long-distance dispersal values for low values of $\phi$ or a dispersal kernel with values centred more closely at 2$^{\circ}$ for higher values of $\phi$; $\sigma$=[0.001, 0.02], corresponding to a range of the standard deviation of the Brownian function from 1/6$^{\circ}$C to 1.3$^{\circ}$C for each 170 kyr time-step, which is within the range found by \cite{Liu2020}; $S$=[1.5, 3], corresponding to a time interval of \textasciitilde2.5 Myr to \textasciitilde6 Myr, which is based on estimated times for reproductive isolation to establish \cite{curnoe}. This range is at the upper limit of the empirically estimated timing for speciation, however \cite{etiene}, and this was also due to the constraints of computational feasibility – when the parameter value was low, reflecting very short speciation times, the number of species increases drastically, preventing the completion of the simulations (Fig. S13; details in SI Appendix). Due to the computational cost of running simulations, we sampled model parameters using Sobol sequences, a quasi-random number generator that samples parameters evenly across the parameter space. 
+
+[config template](https://gitlab.ethz.ch/ites-le/gen3sis/gen3sis_wiki/-/blob/master/configs/Pantropical_config_template.R)
  
 [config generator](https://gitlab.ethz.ch/ites-le/gen3sis/gen3sis_wiki/-/blob/master/configs/ESH_config_generator.R)
  
 ### 0.2 Evolutionary Speed Hypothesis, Skeels et al 2021
   
+Description: 
+We implemented multiple models of diversification using the spatially-explicit general engine for eco-evolutionary simulations, gen3sis, version 1.348. The simulations follow the diversification of a clade from a single ancestral species at the beginning of the Cenozoic (Fig. S1). The simulations track populations as they disperse and diversify throughout 65 myr of reconstructed temperature and aridity changes across a gridded global landscape at 220 km × 220 km resolution50. Biodiversity patterns emerge from simulated phylogenetic trees, spatial distributions of species, and functional trait evolution.
+At each time step (~170 kyr), each population can disperse into surrounding sites from a dispersal kernel drawn from a Weibull distribution with a fixed shape parameter (Ψ = 2.5) and variable scale parameter (Φ; Fig. S2). The size (N) of population i in site j is determined by a Gaussian function of resource use efficiency – the distance between the temperature value in the site (Tj) and the population’s temperature optimum (Ti) – following101,102 (Fig. S3); 
+Nij = K * exp(−(Ti − Tj/ω)2)		equation 1
+where ω is a parameter that determines the strength of environmental filtering, with small values leading to a sharper decline in abundance as the species temperature niche optimum (Ti) becomes more different from the temperature of the site (Tj). Nij equals K in the absence of competitors if population i is perfectly adapted to the site. The carrying capacity for each site (K) is independent of temperature but decreases exponentially with the aridity index in each site (Aj), according to the function:
+K = Kc * exp(−1*Aj)			equation 2
+where Kc is a constant [30,000] (Fig. S4). We model a zero-sum game where sites have finite resources available, which places ecological limits on the maximum number of individuals in a site across populations of all species present (Nj). Therefore, if the total community abundance across all species (Nj) is ≥ K, the addition of any new species decreases the resources available to all present species. In saturated communities (Nj ≥ K), the realised abundance of each species (N̂ij) is apportioned according to the resource use efficiency of each species, following 103: 
+N̂ij = Nij ∗ min(Nj, K)/Nj			equation 3
+Local extinction occurs deterministically if N̂ij = 0 or stochastically as a sigmoidal function of N̂ij:
+1/(1+exp(−μd * (μt − N̂ij)))		equation 4
+where μt is the population size threshold below which extirpation in site j becomes more likely and μd is the rate of decay of the function (Fig. S5). Extinction of a species occurs when it no longer occupies any sites.
+Evolution of the temperature niche trait (Ti) and body mass (Bi) for each independently evolving population approximates a Brownian motion model of trait evolution. These traits are each scaled to be between 0 and 1. The value of the trait at increasing time intervals of δt is equal to the value of the trait at time t plus a value drawn from a normal distribution with a mean of 0 and standard deviation of σ. We model separate rates for temperature evolution (σT) and body-size evolution (σB; Fig. S6). 
+Speciation is based on an allopatric model of speciation, and populations of a species that become geographically isolated from each other diverge genetically at each time step. Under the null model (M0), where population divergence is independent of temperature and body size, the amount of genetic divergence (g) at each time step is drawn from a uniform distribution [0.01, 1]. Diverging populations become distinct species once genetic divergence has crossed threshold S. If populations return to sympatry, they coalesce towards genetic homogeneity at a rate of 1 per time step. We additionally model three alternative scenarios in which rates of population divergence are temperature dependent (M1), body-size dependent (M2), or temperature and body-size dependent (M3). Under M1, the genetic divergence of populations i and k (gi,k) is a function of the sum of the average temperatures (T̂) experienced by diverging populations i and k across all sites within their geographic range, scaled exponentially with the parameter λ (Fig. S7): 
+gi,k = (Σ[T̂i, T̂k]/2)λ			equation 5
+As temperature values are standardised between 0 and 1, the maximum value of g at each time step is equal to 1, and λ determines the rate of exponential decline towards 0 as species inhabit cooler grid cells. Under M2, gi,k is a function of the sum of the average standardised body sizes (B̂) of diverging populations i and k, scaled exponentially with the parameter λ (Fig. S7): 
+gi,k = (Σ[1−B̂i, 1−B̂k]/2)λ			equation 6
+Here, genetic divergence exponentially approaches a value of 1 as body size decreases, at a rate of λ. Finally, under M3, gi,k is a function of the average of 1−B̂ and T̂ (B̂T̂), scaled exponentially with the parameter λ, such that genetic divergence is faster in small-bodied populations in warmer regions and decreases exponentially as species increase in body size and occupy cooler sites (Fig. S7). 
+gi,k = (Σ[B̂T̂i, B̂T̂k]/2)λ			equation 7
+We ran the simulation model 500 times under each of the 4 scenarios, varying 6 key parameters: the divergence threshold (S, parameter range = [2, 10]), the rate scaling factor for the rate of population divergence under M1–M3 (λ, [2, 5]), the strength of environmental filtering for the temperature niche trait (ω, [0.01, 0.035]), the rate of body-size evolution under Brownian motion (σB, [0.001, 0.02]), the rate of temperature niche evolution (σT, [0.001, 0.015]), and the dispersal kernel (Φ,[330, 880]). 
+
 config templates: [M0](https://gitlab.ethz.ch/ites-le/gen3sis/gen3sis_wiki/-/blob/master/configs/ESH_m0_scotese_equalarea_220km_template.R); [M1](https://gitlab.ethz.ch/ites-le/gen3sis/gen3sis_wiki/-/blob/master/configs/ESH_m1_scotese_equalarea_220km_template.R); [M2](https://gitlab.ethz.ch/ites-le/gen3sis/gen3sis_wiki/-/blob/master/configs/ESH_m2_scotese_equalarea_220km_template.R); [M3](https://gitlab.ethz.ch/ites-le/gen3sis/gen3sis_wiki/-/blob/master/configs/ESH_m3_scotese_equalarea_220km_template.R)
 
 [config generator](https://gitlab.ethz.ch/ites-le/gen3sis/gen3sis_wiki/-/blob/master/configs/Pantropical_config_generator.R)