INDITEK ‐2.0: A Bayesian inverse modelling framework for mechanistic evolutionary processes

Abstract Mechanistic eco‐evolutionary models are powerful tools for understanding how life diversified over geological timescales, yet estimating their parameters remains a key challenge. The high computational requirements of these models make probabilistic inversion methods unfeasible. We introduce INDITEK‐2.0, a novel Bayesian inverse modelling framework built upon the cost‐efficient and spatially explicit INDITEK diversification engine. To validate this framework, we conducted a proof‐of‐concept study. We first generated synthetic global distributions of current biodiversity of marine invertebrates using INDITEK with known parameter values. To mimic field data, some realistic random Gaussian errors were added. We then used our Bayesian inverse modelling framework to recover these original (true) parameter values from the dataset of current biodiversity distributions after 500 million years of diversification. We also conducted a suite of sensitivity analyses by disabling specific mechanisms of the model. Our solution is a probability distribution on the parameter space, and we show that the method successfully recovered the target parameter values with a relatively low degree of uncertainty. INDITEK‐2.0's ability to probabilistically infer underlying evolutionary processes provides a new, spatially explicit modelling tool for reconstructing the evolutionary history of biodiversity across Earth's ecosystems and taxonomic groups.