Abstract
Wetlands have a high biodiversity, and are key regulators of the flow of water and the fluxes of mineral and nutrients from land to sea. Wetlands only cover a few percent of the global land surface, but their soils contain as much carbon as the entire biosphere. Yet the extent, volume and carbon content of the world’s wetlands are not accurately known, in particular, for tropical and sub-tropical regions. Despite their importance, wetlands have historically been regarded as wastelands, and have largely been ignored in studies of climate change. Improved maps of the distribution of global tropical wetlands, their volumes and carbon contents are urgently needed. Because wetlands are characterised by their water, soil and vegetation conditions, they are difficult to identify from satellite images of earth. Few of the existing efforts at mapping the global land surface have attempted to identify wetlands.
Wetlands exclusively occur under certain topographic conditions, and where the soil and water conditions are such that inundation can, and actually does, occur on a regular (annual) basis. Taking this as a starting point, a set of novel indexes relating to wetlands was developed. The first index is a climatic topographic wetness index. Using a global digital elevation model, combined with global climate data, a tropical global map of surface wetness was created. Using global optical satellite images from the moderate resolution imaging spectroradiometer (MODIS) a second wetness index was developed. Compared to previous satellite-based wetness indexes, the index attempts to remove the vegetation influence and focus on the soil surface wetness. From an annual time-series of MODIS images, the inundation cycle of the global tropics was captured. The two wetness indexes are strong candidates for mapping the distribution of global tropical wetlands.
Traditional image classification is based on reference data, and usually attempts to delineate features from a single image. As wetlands are characterised by annual variations in inundation, an approach for classifying wetlands from a chrono-sequence of annual MODIS images was developed. In the chrono-sequence, only locations with similar climatic seasonality, and within spatial proximity were classified based on a reference site. Wetlands, often with vegetation and wetness phenology out of phase compared to adjacent dry lands, can thus be delineated.
Initial results are promising for all approaches developed in this study; however, lack of reference sites and reference data has hitherto prevented the development of a global tropical wetland map.