Reconstructing Taiwan’s land cover changes between 1904 and 2015 from historical maps and satellite images

Yi Ying Chen*, Wei Huang, Wei Hong Wang, Jehn Yih Juang, Jing Shan Hong, Tomomichi Kato, Sebastiaan Luyssaert

*Corresponding author for this work

Research output: Contribution to JournalArticleAcademicpeer-review


A new reconstruction of changes in Taiwan’s land cover and estimated uncertainty between 1904 and 2015 is presented. The reconstruction is made by integrating geographical information from historical maps and SPOT satellite images, to obtain spatially explicit land cover maps with a resolution of 500 × 500 m and distinguishes six land cover classes: forests, grasslands, agricultural land, inland water, built-up land, and bare soil. The temporal resolution is unbalanced being derived from four historical maps describing the land cover between 1904 and 1994 and five mosaic satellite images describing the land cover between 1995 and 2015. The uncertainty of the historical maps is quantified to show the aggregation error whereas the uncertainty of the satellite images is quantified as classification error. Since 1904, Taiwan, as a developing country, has gone through a not unusual sequence of population growth and subsequent urbanization, a decoupling of the demand for agricultural land from population growth, and a transition from shrinking in forest area to forest expansion. This new land cover reconstruction is expected to contribute to future revisions of global land cover reconstructions as well as to studies of (gross) land cover changes, the carbon budget, regional climate, urban heat islands, and air and water pollution at the national and sub-national level.

Original languageEnglish
Article number3643
Pages (from-to)1-12
Number of pages12
JournalScientific Reports
Issue number1
Publication statusPublished - 6 Mar 2019

Fingerprint Dive into the research topics of 'Reconstructing Taiwan’s land cover changes between 1904 and 2015 from historical maps and satellite images'. Together they form a unique fingerprint.

Cite this