Using multicriterion optimization to classify ecosystems from multitemporal imagery

Mark A. Ridgley, Jeroen C.J.H. Aerts

Research output: Contribution to JournalArticleAcademicpeer-review

Abstract

One element of research on climate change is modeling of the non-fossil carbon cycle. Non-fossil carbon models such as IMAGE-2 use ecosystem maps as inputs for calculating the production of carbon and the levels of carbon in the atmosphere. The NFOSEUR project aims to classify European ecosystems from remote sensing (RS) data. The data consist of monthly NDVI values for each year, where NDVI is a measure of the photosynthetic activity of the ecosystems. Each ecosystem has a characteristic NDVI curve throughout the year, yielding a characteristic “fingerprint” for each ecosystem. Until now, classification of RS data has been based on various statistical procedures. This paper discusses the use of multicriterion optimization to determine the “best” classification. Goal programming and compromise goal programming models yield classifications that compare closely with those produced by conventional statistical approaches. Case studies include a classification of ecosystems for Germany and southern France. Various potential improvements to our models are discussed, including new formulations and the prospects for fuzzy linear and goal programming models.

Original languageEnglish
Pages (from-to)282-293
Number of pages12
JournalPhysical Geography
Volume17
Issue number3
DOIs
Publication statusPublished - 1996

Keywords

  • Classification
  • Ecosystems
  • Multiple-objective optimization
  • Remote sensing

Fingerprint Dive into the research topics of 'Using multicriterion optimization to classify ecosystems from multitemporal imagery'. Together they form a unique fingerprint.

Cite this