How to restore tropical forest cover to degraded landscapes is a challenge at the forefront of environmental policy. Achieving international commitments to bring hundreds of millions of hectares of land under restoration will require strategic planning, including identifying sites where rates of forest succession are sufficient to restore tree cover with minimal intervention. However, our capacity to produce ecological forecasts of forest succession at landscape scales remains limited. In a new paper, we address this gap with models for forest succession that can be parameterized with satellite remote sensing data. While Landsat data has long been used to monitor deforestation rates, monitoring the continuous changes that occur during succession is more challenging, in part because satellite data contains many sources of noise unrelated to forest structural change. We present a state-space modeling framework that explicitly disentangles measurement error in Landsat-derived spectral reflectance from successional changes related to forest structure.
I’m particularly excited about the modeling efforts in this paper. State-space models are widely used in wildlife, epidemiology, and other applications with “messy” data, but have not yet been widely used for satellite imagery. We demonstrate that state-space approaches are a feasible technique for disentangling process and measurement error from time series of imagery. All of our code is available at: https://doi.org/10.5281/zenodo.3873639
Citation: Caughlin, T.T., Alvarez‐Buylla, C.B., Asner, G.P., Glenn, N.F., Bohlman, S.A., Wilson, C.H., In Press. Monitoring tropical forest succession at landscape scales despite uncertainty in Landsat time series. Ecological Applications, e2208. https://doi.org/10.1002/eap.2208https://esajournals.onlinelibrary.wiley.com/doi/10.1002/eap.2208