Drs. Rudolf Winter*, Manfred Keil
DLR Oberpfaffenhofen,
DFD, D-82230 Wessling
Germany



* now: JRC Ispra/Varese, Italy
Co-Investigators:
U. Ammer, B. Foerster, A. Haas, Univ. of Munich,
Institute of Landscape Planning and
Nature Conservation
W. Kuehbauch, M. Davidson, Univ. of Bonn, Institut fuer Pflanzenbau
W. Mauser, M. Rombach, Univ. of Munich, Inst. f. Geography
B. Koch, Univ. of Freiburg, Abt.
Luftbildmessung und Fernerkundung
Dr. S. Mohan, RSA, Ahmedabad, Gujarat, India
Information Extraction from Shuttle Radar Images for Forest and Agricultural Applications


OBJECTIVES
The experiment will attempt to extract suitable polarimetric information for objectives of forest mapping and vegetation monitoring as well as for agricultural purposes.

In forest areas in Bavaria (e.g., near the Oberpfaffenhofen supersite and in the Bavarian Forest) and in the Harz Mountains, shuttle data are to be evaluated for different forest formations, age classes and other forest parameters and for the forest situation, e.g., concerning storm damages.

In the rainforest study areas in Brazil, especially in the states of Acre and Rondonia, SIR-C and X-SAR data are to be analyzed for differentiation of land cover classes like primary rainforest, initial and intermediate regrowth, clean pastures and overgrown pastures. By comparison of April and October data, information is to be gained on the dynamics of deforestation by clearing and burning and on other land use changes.

In the supersite Oberpfaffenhofen, the potential for crop classification is to be investigated, in comparison to other test sites like Montespertoli and Matera. The effect on polarimetric backscatter responses is to be studied in relation to biomass and biophysical parameters of vegetation, especially for agricultural purposes. The assessment of soil moisture conditions is another aspect under study in the Oberpfaffenhofen/Weilheim area and in a test site in Gujarat/India.



PROGRESS
Forest investigations in the test site Oberpfaffenhofen

Four data takes of April and October 94 have been investigated for the multipolarimetric and multifrequency information content of different land use classes, especially
for forest areas and storm damage areas near the lake "Ammersee" and a region NE of Oberpfaffenhofen. Signature analysis of different polarization states has been performed in preparation of land use classification. In the region West of Ammersee a land cover classification has been performed using combined SIR-C/X-SAR data.

The X-SAR data takes have been studied for time variation effects, e.g., for different moisture conditions during the first and the second mission. Also interferometric datasets have been studied from the October mission.


Forest investigations in the test sites of Harz Mountains and Bavarian Forest

First investigations have been performed on the SIR-C/X-SAR data in the mountainous terrain of the Harz Mountains and the Bavarian Forest. Both forest areas are partly severely affected by forest damages, e.g., by storm damages and bark beetle calamities. In order to be able to combine the data with geocoded reference data, e.g., in a GIS environment, geocoded terrain corrected SIR-C data are necessary (besides available GTC products of the X-SAR component) in these areas of high relief. Terrain corrected multiband/ multiseasonal data are planned to be available in May/June 1996, within an initiative of DLR and DARA.


Investigations for rainforest monitoring in Acre and Rondonia, Brazil

In the area of Sena Madureira and Rio Branco in the state of Acre, Brazil, two SIR-C/X-SAR data takes of April 1994 and one of October have been studied for rainforest monitoring tasks, in cooperation with the National Brasilian Space Research Institute. Objectives are the differentiation within the rainforest and within the deforested areas (different states of pasture land and regeneration areas). By comparing X-SAR datasets of the April and the October mission, the dynamics in deforestation have been shown in an area NW of Rio Branco. Based on the field work in June/July 1994, analysis of backscatter responses in L-, C- and X-Band have been performed for forest, pasture and regrowth areas. In a region of 40 km by 40 km NW of Rio Branco, a landcover classification has been performed using the textural classifier EBIS, mainly based on different polarizations of L-Band data.

In the state of Rondonia, a region near the town Ji Parana has been investigated by several SIR-C / X-SAR datasets of April and October 1994. These datasets have been registered within the Amazon calibration experiment and as part of the "Pantanal" data take. In November 1995, a first ground campaign has been performed in cooperation with INPE and other Brasilian institutions. Different states of land use are under study, showing different land use intensities as well as different degradation and regeneration states.


Investigation of agricultural areas in the study sites of Oberpfaffenhofen, Montespertoli and Matera

In the Oberpfaffenhofen supersite, investigations for early crop identification have been performed by sets of April data, based on ground truth assessment during the overflights. Methodological studies show the different information content of different polarization/ band combinations. On a field-based approach (field limits coming from other sources), crop classifications have been performed by methods of polarimetric contrast and maximum likelihood classification.

The potential for the assessment of different growth states of winter wheat is under study by comparing the backscatter responses of winter wheat in the different climatic regions of Oberpfaffenhofen, Montespertoli and Matera.

In the test site Weilheim, Southwest of Oberpfaffenhofen, parallel registered airborne E-SAR data are investigated in comparison to SIR-C/X-SAR data. In this area, the study of soil moisture conditions is the main objective.

Investigation of soil moisture and vegetation cover in Gujarat/India

By the Indian Co-Investigator, several SIR-C and X-SAR datasets, mainly from April 1994, are under study for soil moisture and vegetation cover in a region of Gujarat in India.


SIGNIFICANT PRESENT RESULTS
Forest investigations in the test site Oberpfaffenhofen

In the forest area West of the lake Ammersee, ten landcover classes including five forest types could be classified with an overall classification accuracy of about 63 %, in an area which has a quite inhomogeneous distribution of land cover. Storm damaged areas could be separated by about 70% accuracy, the forest / nonforest separation accuracy is around 95 %. C-Band HV polarization seems to be important for structure identification between deciduous and coniferous forest. L-Band HV has a high correlation to natural forest age classes and biomass. Interesting is the result of using a three-date multitemporal complex X-SAR data set of the October mission. By an interferometric processing, the interferometric coherence of a one-day repeat cycle can be successfully be used to improve the classification of five main land cover classes, in addition to the three-date multitemporal X-SAR data set.

Investigations for rainforest monitoring in Acre and Rondonia, Brazil

Investigations of different polarimetric states of L- and C-band showed a better forest / non-forest separation in the L-band data, especially in the regeneration states. L-band HH-Polarization allowed the separation of alluvial rainforest areas along rivers. Integration of C-band and X-band data delivered a better differentiation of pasture land. For land use classification, the EBIS texture tool was successful which allows the integration of texture parameters for classification besides the evaluation of local histograms.


Investigation of agricultural areas in the study sites of Oberpfaffenhofen

Astonishing positive results have been reached in the Oberpfaffenhofen test site for early crop identification, using a five-band combination of all three frequencies. L-HV has proven to be an important layer also for this mapping, based on field-wise classification.


FUTURE PLANS
Forest Monitoring Investigations in the Harz Mountains and in the Bavarian Forest are planned to be performed, based on terrain-corrected, combined SIR-C/X-SAR products from the April and October missions. Storm damages and insect calamities are a heavy problem in those areas, e.g., the bark beetle damages in the area of the Bavarian Forest National Park.

Rainforest and Deforestation Monitoring Investigations will be continued, the results of the study sites of Acre (Sena Madureira) and Rondonia (Pantanal data takes) are to be compared with results from other projects in the Manaus test site. Different approaches are planned to get information for biomass estimates especially for forest regrowth, in cooperation with INPE, Brazil.

Investigations for crop identification will be continued, including October data takes.


PUBLICATIONS
Davidson, M. W., R. Steingiesser, W. Kuehbauch: Exploiting multifrequency multi- polarization radar images for mapping crop types early during the growing season, Int. Symposium "Retrieval of bio- and geophysical parameters from SAR data for land applications," Toulouse, Oct. 10-14, 1995, in press.

Davidson, M. W., R. Steingiesser, W. Kuehbauch: Assessing Agricultural Land Use early During the Growing Season Using Multi-frequency and Multi-Polarisation SIR-C Backscatter Features, EUSAR '96, Koenigswinter, Germany, 1996, in press.

Dos Santos, J. R., M. Keil, H. Kux, M. S. P. Lacruz, D. Scales: Interactive Analysis of polarimetric SIR-C and Landsat TM data for the spectral-textural characterization of the land cover in SW Amazon, Brazil. Paper for presentation at ISPRSS '96 conference, July 1996, Vienna/Austria, planned.

Haas, A., B. Foerster, U. Ammer, The suitability of multifrequency, multipolarimetric and multitemporal radar data for forest monitoring, IGARSS '96 , May 1996, Nebraska, planned.

Keil, M., D. Scales, R. Winter: Investigation of Forest Areas in Germany and Brazil using SAR Data of the SIR-C/X-SAR and other SAR Missions, IGARSS '95 , 10-14 July 1995, Vol. II, p. 997 - 999, Florence.

Keil, M., M. Schmidt, D. Scales, H. Kux, J. R. dos Santos: Investigation of polarimetric SIR-C / X-SAR data for characterization of land cover in Acre and Rondonia, Southwest Amazon, Brazil.

Paper for presentation at PIERS-96 Conference, July 1996, Innsbruck/Austria, planned.

Stolz, R., W. Mauser: First Evaluations of Shuttle SIR-C and X-SAR Data for Landcover Classification. IGARSS '95 , 10-14 July 1995, Vol. II, p. 1058-1060, Florence.

Table of Contents

Converted to HTML by Alvin Wong, al.wong@jpl.nasa.gov

The Jet Propulsion Laboratory
4800 Oak Grove Drive
Pasadena, Cailfornia 91109