Science Service System

Summary of Proposal LAN0918

TitleSeasonly changes on classification of tropical and temperate forests by polarimetric analysis of SAR data
Investigator López Hernández, Juan Ygnacio - University of Freiburg, FELIS
Team Member
Aranguren, Anairamiz - ICAE. Complejo Universitario La Hechicera, Faculdtad de Ciencias
Pacheco, Carlos - Faculdtad de Ciencias Forestales y Ambientales, Laboratorio de Fotogrametría y Sensores Remotos
Ing. For. PhD Gutierrez, Julian - Facultad de Ciencias Forestales y Ambientales, Laboratorio de Fotogrametría y Sensores Remotos
Gutierrez, Néstor - Facultad de Ciencias Forestales y Ambientales, Insituto de Silvicultura, Escuela de Ingenieria Forestales
SummaryObjectives Evaluate seasonal changes in vegetation. Mapping the boundaries of forest cover classes Deciduous trees and evergreen cover for tropical forest. Conifers and broad leaves forests in temperate forest.Evaluate the usefulness of combining optical, TSX and Radarsat2 data. Evaluate the effect of topography on the accuracy of classification of forest lands. Search on the relationship between LAI and the digital values obtained by the set of SAR and optical sensors. Method Landsat scenes fromdifferent seasons will be selected for every area under study. By using orthorectifyed scenes from the tree different places in both temperate and tropical forest the application of both automatic classification and object oriented classification will produce a set of forest cover maps. Polarimetric analysis of TerraSarX and Radarsat 2 scenes for determination of forest boundaries and composition. Five different main procedures will be used to produce the maps: The sources of data for producing maps will be as follows: a) The combination of Landsat TM and TerraSar X data. b) The combination of Landsat TM and Radarsat 2 data. c) The combination of Landsat TM, TerraSar X and Radarsat 2 data d) The use of single data source. Reference points and areas will be selected in every forest. For the German areas high resolution ortho-photographies will be used to determine the composition type of forest. Plots will be set in Black Forest in order to use as a reference points. For the Venezuelan areas a set of permanent plots and field trips will be combined to gather information of the leaf area index (LAI) and composition of theforest in terms of evergreen and deciduous species composition.These information will be used as ground truth data for validating the maps produced by every one of the four methods. Evaluation of the consistency in forest boundaries determination, area covered,producer accuracy and user accuracies for every method will be produced. After the classification an analysis of variable selection will be carried out in order to recommend the best set of input data required to produce the same mapby using approach of statistical model selection. First looking forthe best set of sensors and after looking for the best set of input layer in an automatic classification schema. An statistical analysis on LAI regression will be applied in every forest cover class. Data Requirements Three quad-polarized scenes, every three months from TerraSar-X (ScanSAR quad-pol) and RadarSat 2 (fine quad-pol mode) sensors in Germany and Venezuela.(From DLR) Landsat TM scenes covering three sites selected and taken during similar seasons. (From NASA servers) LAI assessment in situ with LICOR LAI 2000. (FELIS and ULA-Mérida) Deliverables The first six months andafter 12 months will be produced the following maps: 3 Digital maps of forest cover and types for every season. 3 Digital maps of season change in forest during the period of the study. Report wirh the processing steps, relationship between values measured in the field and DN in scenes.

Back to list of proposals

© DLR 2004-2016