Classification of coffee-forest landscapes using landsat TM imagery and spectral mixture analysis

Mikaela Schmitt-Harsh, Sean P. Sweeney, Tom P. Evans

Research output: Contribution to journalArticlepeer-review

9 Scopus citations


This research applies linear spectral mixture analysis (LSMA) to a Landsat TM image, and assesses the value of fraction images (green vegetation, shade, soil) and the thermal band (TM-B6) in discriminating shade-grown coffee systems from forests. Four combinations of TM bands and fraction images were compared, and a maximum likelihood algorithm was used to classify five land cover classes: high-density woodlands, low-density woodlands, coffee agroforests, crop / pasturelands, and urban settlements. The classification accuracy of each model combination was assessed using both Kappa analyses and quality and allocation disagreement parameters. Results indicate improvements to classification accuracies following inclusion of TM-B6 and fraction images as inputs to the classification; however, only the use of TM-B6 led to significant improvements at the 95 percent confidence level. The highest classification accuracy achieved was 86 percent (Kstandard = 0.82), with producer's and user's accuracy of coffee agroforests reaching 89 percent and 90 percent, respectively, an improvement over previous research aimed at spectrally distinguishing coffee from other woody cover types.

Original languageEnglish (US)
Pages (from-to)457-468
Number of pages12
JournalPhotogrammetric Engineering and Remote Sensing
Issue number5
StatePublished - May 2013

ASJC Scopus subject areas

  • Computers in Earth Sciences


Dive into the research topics of 'Classification of coffee-forest landscapes using landsat TM imagery and spectral mixture analysis'. Together they form a unique fingerprint.

Cite this