Manuel David Vargas Russo Federico Dukatz Graciela Ana Canziani
Transcription
Manuel David Vargas Russo Federico Dukatz Graciela Ana Canziani
Manuel David Vargas Russo Federico Dukatz Graciela Ana Canziani Introduction •Explanation of a methodology for Wetland Environmental Analysis, based on Satelital Images. •Models to optimize the classification and the calculus of vegetation indexes. •Model to calculate tendencies, with the capacity of monitoring the temporal evolution of vegetation. •Finally, combination of different techniques to detect possible anomalies and variations in the study zone. REGWET Vargas Russo, M. - Dukatz, F.- Canziani, G. Description of the Methodology Satelital Images Geometric Correction Separation of Area of Interest Method for Vegetation Indexes TDI Transformation Unsupervised Classification Detection of Anomalies Analysis of Anomalies REGWET Vargas Russo, M. - Dukatz, F.- Canziani, G. Study Site • Satelital data constitutes the base data layer for the landcover map. They were provided by CONAE, from SAC-C within the AM constellation. • The Geometric Correction procedure involves the selection of ground control points. Then they are assigned to a determined geographic referencing system (UTM - zone 21). •Path 226 Row 78. Obtaining Satelital images Geometric Correction Separation of Area of Interest Method for Vegetation Indexes TDI Transformation This sector includes the region studied, bounded by the Paraguay River (West), the Tebicuary River (South) and the 100m contour line (Northeast). Unsupervised Classification Detection of Anomalies Analysis of Anomalies REGWET Vargas Russo, M. - Dukatz, F.- Canziani, G. Separation of Area of Interest In order to make the processing more efficient and more precise, the study area or AOI (Area of Interest) is defined, cut and separated. Obtaining Satelital images Geometric Correction Separation of Area of Interest Method for Vegetation Indexes TDI Transformation Unsupervised Classification Detection of Anomalies Analysis of Anomalies REGWET Vargas Russo, M. - Dukatz, F.- Canziani, G. NDVI (Normalised Difference Vegetation Index) 24/4/2003 10/5/2003 Obtaining Satelital images 26/5/2003 Geometric Correction Separation of Area of Interest Method for Vegetation Indexes TDI Transformation Unsupervised Classification Detection of Anomalies Analysis of Anomalies REGWET -1 Greeness leeves Vargas Russo, M. - Dukatz, F.- Canziani, G. 1 Tasseled Cap Transformation Obtaining Satelital images Geometric Correction Separation of Area of Interest Method for Vegetation Indexes TDI Transformation Unsupervised Classification Detection of Anomalies Analysis of Anomalies REGWET + Vegetation - + Wetness - + Soil Brightness - Vargas Russo, M. - Dukatz, F.- Canziani, G. Tasseled Cap Transformation SBI GVI Obtaining Satelital images Geometric Correction Wetness Separation of Area of Interest Method for Vegetation Indexes TDI Transformation Unsupervised Classification Detection of Anomalies Analysis of Anomalies REGWET Vargas Russo, M. - Dukatz, F.- Canziani, G. TDI Transformation GVIs TDI 10/07/02 27/08/02 18/09/02 23/03/03 26/04/03 10/06/03 Obtaining Satelital images Geometric Correction Separation of Area of Interest Method for Vegetation Indexes TDI Transformation • Temporal Difference Index (TDI). Defined by means of a linear regression. • This method is performed simultaneously over vegetation index data from several different images. Unsupervised Classification Detection of Anomalies Analysis of Anomalies REGWET Vargas Russo, M. - Dukatz, F.- Canziani, G. Vegetation Intensity TDI Transformation α Day 10/07/02 27/08/02 18/09/02 23/03/03 26/04/03 Tg α== Obtaining Satelital images Geometric Correction Separation of Area of Interest Method for Vegetation Indexes TDI Transformation Unsupervised Classification Detection of Anomalies Analysis of Anomalies REGWET Vargas Russo, M. - Dukatz, F.- Canziani, G. 10/06/03 TDI Transformation Increase Tg α== Minimal change Tg α== Decrease Tg α== Obtaining Satelital images Geometric Correction Separation of Area of Interest Method for Vegetation Indexes TDI Transformation Unsupervised Classification Detection of Anomalies Analysis of Anomalies REGWET Vargas Russo, M. - Dukatz, F.- Canziani, G. Unsupervised Classification •It is a process for sorting pixels into a finite number of individual classes or categories in the image. •If a pixel satisfies a certain set of criteria, then the pixel is assigned to a predetermined class. Classes Water Vegetation Obtaining Satelital images Geometric Correction Separation of Area of Interest Method for Vegetation Indexes TDI Transformation Unsupervised Classification Algorithm used: ISODATA Identifies statistical patterns in the data. Detection of Anomalies Analysis of Anomalies REGWET Vargas Russo, M. - Dukatz, F.- Canziani, G. Soil Unsupervised Classification Unsupervised Classification on a TDI Trasformation • Defines categories following the intensity of the variation. Unsupervised Classification Obtaining Satelital images Geometric Correction Separation of Area of Interest Method for Vegetation Indexes TDI Transformation Unsupervised Classification AOI GVIs TDI Detection of Anomalies Analysis of Anomalies REGWET Vargas Russo, M. - Dukatz, F.- Canziani, G. Detection of Anomalies • In this example, the green regions show zones without change, while the shades red indicate that the vegetation has suffered a major variation, in this case, an increase in the vegetation. Obtaining Satelital images Geometric Correction Separation of Area of Interest Method for Vegetation Indexes TDI Transformation Unsupervised Classification Temporal Difference Index Detection of Anomalies Analysis of Anomalies REGWET Vargas Russo, M. - Dukatz, F.- Canziani, G. Detection of Anomalies 28-9-2002 10-7-2002 27-8-2002 11-8-2002 26-5-2003 10-5-2003 24-4-2003 TM Bands, 4(NIR), 3(red), and 2(blue). Obtaining Satelital images Geometric Correction Separation of Area of Interest Method for Vegetation Indexes TDI Transformation TDI (7 Images) Unsupervised Classification Detection of Anomalies Analysis of Anomalies REGWET Vargas Russo, M. - Dukatz, F.- Canziani, G. Detection of Anomalies 10-7-2002 11-8-2002 27-8-2002 28-9-2002 24-4-2003 Obtaining Satelital images Geometric Correction Separation of Area of Interest Method for Vegetation Indexes TDI Transformation Unsupervised Classification Detection of Anomalies Images Layers Red: band 4 (NIR) Green: band 3 (R) Blue: band 2 (G) 10-5-2003 26-5-2003 Analysis of Anomalies REGWET Vargas Russo, M. - Dukatz, F.- Canziani, G. Analysis of Anomalies There are reasons to believe that this zone was devastated by fire. NBR transformation was Tasseled applied to Cap two(26-5-2003 images,) one before and the other after the observed change. Obtaining Satelital images Geometric Correction Separation of Area of Interest Method for Vegetation Indexes TDI Transformation Unsupervised Classification Detection of Anomalies Analysis of Anomalies REGWET UC on TDI 3 images (24-04-2003) (10-05-2003) (26-05-2003) - 0 + Vargas Russo, M. - Dukatz, F.- Canziani, G. Analysis of Anomalies NBR (Normalized Burn Ratio) 10/05/03 Obtaining Satelital images Geometric Correction Separation of Area of Interest Method for Vegetation Indexes TDI Transformation Unsupervised Classification Detection of Anomalies Analysis of Anomalies REGWET 26/05/03 TM bands 4 and 7 exhibited the greatest reflectance change in response to fire. The hypothesis is that the ratio of those bands would be most discriminating for burn effects. Difference Vargas Russo, M. - Dukatz, F.- Canziani, G. Analysis of Anomalies Unsupervised Classification (ten classes) only over the limited NBR difference AOI. Obtaining Satelital images Geometric Correction Separation of Area of Interest Method for Vegetation Indexes TDI Transformation Unsupervised Classification Detection of Anomalies Analysis of Anomalies REGWET Vargas Russo, M. - Dukatz, F.- Canziani, G. Conclusion • Wetland zones are very difficult to study because of very particular features. • The satelital image processing is an appropriate tool for detecting temporal changes in vegetation, temperature and anomalies that occur in wetlands, which are isolated, extended and difficult to access. • Transformations performed over a temporal sequence of images generates an advantageous method for problem detection through remote sensing. REGWET Vargas Russo, M. - Dukatz, F.- Canziani, G. Acknowledgements We thank CONAE for providing the images used in these projects, and Raul Colomb and Ida Nöllmann for their unlimited collaboration. Florencia Castets (Personal de Apoyo CICPBA) helped design this presentation. While the European Commission financed the projects "The Sustainable Management of Wetland Resources in Mercosur" (Contract IC18 CT98 0262) and "Regional Aspects of the SustainableManagement of Wetland Resources " (Contract A4-DEV-ICFP-599A4 AM01). REGWET Vargas Russo, M. - Dukatz, F.- Canziani, G.