Around the world, the effects of changing plant phenology are evident in many ways: from earlier and longer growing seasons to altering the relationships between plants and their natural pollinators. Plant phenology is often monitored using satellite images and parametric methods. Parametric methods assume that ecosystems have unimodal phenologies and that the phenology model is invariant through space and time. In evergreen ecosystems such as mangrove forests, these assumptions may not hold true. Here we present a novel, data-driven approach to extract plant phenology from Landsat imagery using Generalized Additive Models (GAMs). Using GAMs, we created models for six different mangrove forests across Australia. In contrast to parametric methods, GAMs let the data define the shape of the phenological curve, hence showing the unique characteristics of each study site. We found that the Enhanced Vegetation Index (EVI) model is related to leaf production rate (from in situ data), leaf gain and net leaf production (from the published literature). We also found that EVI does not respond immediately to leaf gain in most cases, but has a two- to three-month lag. We also identified the start of season and peak growing season dates at our field site. The former occurs between September and October and the latter May and July. The GAMs allowed us to identify dual phenology events in our study sites, indicated by two instances of high EVI and two instances of low EVI values throughout the year. We contribute to a better understanding of mangrove phenology by presenting a data-driven method that allows us to link physical changes of mangrove forests with satellite imagery. In the future, we will use GAMs to (1) relate phenology to environmental variables (e.g., temperature and rainfall) and (2) predict phenological changes.
This is an open access article distributed under the Creative Commons Attribution License
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited