Forest dynamics drive large-scale ecological processes, including the terrestrial carbon cycle, forest succession, and biodiversity maintenance at landscape, regional and continental scales. Individual tree vital rates (growth, survival and reproduction) underlie these dynamics, and are primarily estimated from repeated measurements of tree stems in forest inventory plots. However, logistical constraints limit the spatial coverage of these field measurements and extrapolation to larger scales can propagate spatial errors. Remote sensing data could provide a solution to the spatial mismatch between field data and large scale forest dynamics.

I demonstrate a hierarchical Bayesian approach to link remote sensing and forest inventory data to predict forest dynamics across a heterogeneous landscape covering a 3809 km² area in western Panama. I use generalized linear models to predict LiDAR-derived metrics of forest structure, including tree canopy cover and tree height, from 30 x 30 m resolution data from the Landsat archive. I then develop a state-space model with tree canopy cover and tree height as output variables and tree vital rates as input variables. This modelling approach can incorporate data on individual trees from forest inventory plots to inform demographic parameters over large spatial extents. Bayesian methods enable a clear route to propagating uncertainty in parameter estimation throughout the model, including measurement error from remote sensing classification. I apply the model to quantify whether forest succession over a ten year period was limited more by seed arrival or by the growth and survival of established trees. Meeting the demand for large-scale predictions of forest dynamics will require synthesizing site-specific data across a range of empirical studies. Statistical models that integrate multiple data sources to make spatially-explicit forecasts over large areas could support the UN’s target of restoring hundreds of millions of hectares of degraded land within the next 15 years, and help mitigate global climate change.