Florian Hartig

Understanding the assembly of communities and biotic interaction networks

Why do some species coexist while others exclude each other, and how do networks of competition, pollination and other interactions arise? We combine data from forest plots, interaction networks and eDNA surveys with models to infer and test mechanisms such as density dependence and trait matching.

Mechanisms behind biodiversity patterns in macroecology and macroevolution

Why are there more species in the tropics, and how did today’s distribution of species richness arise? We link ecological and evolutionary processes such as speciation, dispersal, extinction and adaptation to large-scale patterns, using eco-evolutionary simulations such as gen3sis and phylogenetic models of diversification.

Systems ecology and process-based simulation models

How do forests, vegetation and populations respond to climate and land use? We develop and apply process-based models, from forest growth and dynamic vegetation models to population models, to understand and project the dynamics of ecological systems. R interfaces such as r3PG and rLPJGUESS make these models accessible.

Model-data integration, simulation-based inference and Bayesian statistics

How can complex simulation models be fitted to data, and how certain are their predictions? We develop methods to calibrate process-based and stochastic models, from approximate Bayesian computation to Bayesian calibration and uncertainty quantification, and software such as BayesianTools that makes these methods usable.

Statistics, machine learning and deep learning in ecology

How can statistics and machine learning deliver robust inference, not only predictions, from ecological data? We develop and test such methods; our R packages, such as DHARMa for residual diagnostics, sjSDM for joint species distribution models and cito for deep learning, bring them into everyday practice, and we apply them to predict and monitor biodiversity, for example in the BaySenseAI and bAImo projects.