Estimating Number of Pteridophyte and Melastomataceae Species from Satellite Images in Western Amazonian Rain Forests

摘要: The paper studies the variation of rain forests vegetation by means satellite images and data 30 field 9 species understorey pteridophytes (ferns fern allies) Melastomataceae mainly small trees shrubs). Pteridophytes were seen as indicators more general floristic patterns in present study. numbers individuals taxonomically ecologically defined groups estimated using Landsat TM measurements primary lowland Amazonian Ecuador. k-nearest neighbours (k-nn) estimation method was applied. Different spectral features weighting tested. final estimations computed standard deviations bands TM1–TM5 TM7 calculated a window 7x7 pixels weighted optimal weight seeking method. estimates evaluated leave-one-out cross-validation separate test set. A destriping developed. systematic brightness towards eastern edge image corrected. worked well if striping continuous from one to another. results showed that number could be relatively accurately with root mean squared errors (RMSE) for mostly high. pteridophyte accurate than species. Visualisations ecological varied spatially, observed spatial pattern largely explained topography. Success based suggest these is related canopy trees.