Date of Award


Level of Access

Campus-Only Thesis

Degree Name

Master of Science (MS)


Forest Resources


Steven A. Sader

Second Committee Member

Mary Kate Beard-Tisdale

Third Committee Member

William Glanz


Coffee culture is extremely important in Central America as well in other producer countries. Coffee is one of the major cash crops and important component of the Central American economy. The Mesoamerican Biological Corridor Program in Central America has promoted sustainable cultural activities, including shade coffee systems as an important conservation strategy. Accurate detection and mapping of coffee crops has been problematic using any type of remotely sensed data because spectral reflectance patterns of coffee plants are similar to secondary forest types and other woody crops. Surprisingly little research has been published on this topic. The thesis research is focused on testing remote sensing methods to discriminate coffee plantations from other crops and forest types in two study areas in Central America. For both study areas, Landsat ETM + images were subset, geometrically corrected, haze corrected, and clouds and water masked prior to further processing. Other spatial data sets were acquired and combined with Landsat to support the analysis. Chapter one, (El Salvador study area) is an exploratory analysis of spectral reflectance characteristics of coffee and other cover types. Six different Landsat waveband combinations were tested employing four spectral channels: red, near infrared, mid infrared and thermal; two vegetation indices: normalized difference vegetation index (NDVI), normalized difference moisture index (NDMI); and a texture layer. The band combinations were evaluated through signature separability analysis (Transformed Divergence). The results indicated that the spectral separability between coffee and other cover types was enhanced when the visible red, near infrared, mid infrared reflective wavebands were combined with the thermal band and the normalized difference vegetation index (NDVI) layer. In chapter two, the objective was to implement a supervised land cover classification and conduct an accuracy assessment. The research analyzed non-shade and shade coffee spectral characteristics in a Costa Rica study area where better data existed on the location and ground characteristics of coffee plantations. A geographic information system (GIS) distance analysis was conducted to determine physical characteristics of known coffee plantations. Elevation data from the Shuttle Radar Topographic Mission (SRTM), not previously available for Central America, was acquired from NASA Marshall Spce Flight Center, and used for analysis. A coffee environmental stratification model (CESM) based on elevation and precipitation information was tested as a post-classification method. To minimize topographic influence, effective incident angle layer (cos(I)) per pixel was calculated from the SRTM data and added to the band combinations to evaluate its contribution to the accuracy assessment. Supervised land cover classifications were performed with three different band combinations with and without the CESM. Accuracy assessment, Kappa Analysis and "Z-tests" were performed. Results indicated that BC 34567-NDVI-cos(I) produced the highest accuracy to discriminate coffee plantations. However there was no significant difference between datasets that included or excluded CESM. Although accuracy results were higher than reported in previous research in the highland tropical forest, the Kappa analysis indicated only moderate agreement with reference data. Future research is recommended to employ spectral analysis and test other satellite platforms for detecting and mapping coffee crops and other land cover in tropical Central America.

Files over 10MB may be slow to open. For best results, right-click and select "save as..."