Global changes in landscapes between 1992 and 2015

This article is originally published at https://nowosad.github.io/

Poster: https://zenodo.org/record/1323780

Land-cover change is a pervasive phenomenon caused by climate change, and, in recent decades, by the rapid population growth and accelerated industrialization. Therefore, the assessment of land cover changes is of prime importance for planing and management of natural resources. It provides necessary information for making decisions on trade-off between development and conversation. We present the first comprehensive map and GIS-based database of global land cover during the 1992–2015 period. The map is derived using a post-classification change detection algorithm applied to the new European Space Agency (ESA) global time series of land cover maps at 300m resolution (CCI-LC). To smooth possible errors stemming from incorrect category assignments at individual pixels and obtain distinct changes on relevant geographical scale, we tessellate the entire landmass into local landscapes – 9km × 9km tiles of 900 CCI-LC pixels. Change is detected by comparing the 2015 pattern of land cover categories within a tile with its 1992 pattern. Only 22% of tiles registered a meaningful change in their patterns. The collection of all changed tiles constitute a SQL-searchable spatial database to be used for analyzing land cover transitions and global mapping of change trajectories.
A global map of change trajectories provides a visualization of spatial distribution of all major changes and serves as a guide to a more focused use of the database. Globally dominant CCI-LC transitions during the 1992-2015 period was forest → agriculture (19%). The vegetation type that experiences the largest net loss was the trees at -559,124 km2 globally. We concluded that post-classification change detection is well-suited for a fairly accurate estimation of the global forest area and a depiction of a geographical distribution of forest losses/gains, but, in comparison with estimates stemming from a forest-dedicated change detection method using high resolution images, it provides a low estimation of forest loss and a high estimation of forest gain. For other vegetation types estimations of losses and gains are expected to be more accurate due to more homogeneous definitions of non-forested CCI-LC categories.

Thanks for visiting r-craft.org
This article is originally published at https://nowosad.github.io/
Please visit source website for post related comments.