Science Service System
You are here : Home : Proposals_Summary

Summary of Proposal LAN1123

TitleImproving Semi-natural Grassland Administration with TerraSAR-X
Investigator Voormansik, Kaupo - University of Tartu, Space Technologies Workgroup
Team Member
Dr Reinart, Anu - Tartu Observatory, Department of Remote Sensing
Mr. Praks, Jaan - Aalto University, Department of Radio Science and Engineering
SummaryThe Objectives
The main objective of the study is to work out an algorithm and assess the applicability of TerraSAR-X data for wooded meadow mowed area calculation in order to improve semi-natural area administration. There are many wooded meadows especially in Northern Europe. They are formed during centuries in the combination of human and nature interactions and are under protection because wooded meadows are one of the richest ecosystems when looking the number of species per hectare. In order to preserve them government bodies pay subsidies to the farmers who take care of them. The subsidies are paid proportionally to the area that was actually mowed in a particular year. Thus exact measurement of the mowed area is important, but it is challenging because the mowed areas vary from year to year as many of the wooded meadows are located in river floodplains where the water level determines the accessible land every year differently. Hence it is not possible to use last year area measurement results as reference and the measurement procedure needs to be repeated every year. Currently the areas are measured using GPS-based field works, which is quite expensive.
The Method
Backscatter analysis is foreseen as the main method, but interferometric coherence could also be studied. Image pairs with acquisitions before and after the mowing with different polarisations would be studied to find out in which mode and processing technique it is possible to extract the mowed areas in most efficient and accurate way. The areas that are calculated from SAR imagery would be compared against precise in situ GPS measurements for method accuracy assessment. Next ESA SAR Toolbox (NEST), ENVI+IDL and MapInfo software would be used for the geospatial and SAR data processing during the project.
Data Requirements
Total of 14 stripmap and spotlight images with different polarisations would be needed. The acquisitions should be made from May to August 2011.
The outputs of the project would be articles in peer reviewed and popular science magazines. A detailed report about test project results would also be presented to Estonian Agricultural Registers and Information Board for their evaluation.
The Funding
The salaries of team members are covered by the University of Tartu, Aalto University and Tartu Observatory. Mowed area GPS-measurement field works are funded by Estonian Agricultural Registers and Information Board. The purchase of satellite imagery is funded by Tartu Observatory.
Final ReportThe goal of the study was to develop a robust methodology for detecting the areas where grass was cut on a given year using minimum number of SAR images.
As of February 2013 the field works campaign and SAR data processing has been completed. One conference proceeding have been published and is available at (p 29-32 in the PDF). The full scale research article was written from June to October 2012 and submitted to the International Journal of Remote Sensing end of October 2012. The paper received a major revision in January 2013 and the improved revision is now uploaded. The article is expected to be published during 2013.
Using the dual pol. TerraSAR-X HH/VV data all possible known parameters were calculated including HH, VV, HH+VV, HH-VV backscatter, HH/VV coherence magnitude and phase, T12 coherence magnitude and phase, dual pol. entropy/alpha decomposition (H2a) parameters entropy, alpha and alpha dominant. Most useful parameters for cut grass detection appeared to be dual pol. mean alpha angle and especially alpha dominant angle calculated from the 2x2 T-matrix. Using the field works data it was found that when high grass is freshly cut and left lying on ground the dual pol. alpha dominant angle increased from around 10 degrees to over 25 degrees. Still no reliable method was found for distinguishing high grass from low grass using dual pol. X-band HH/VV data.
Besides the dual polarisation HH/VV data also 7 single pol. HH high resolution spotlight images were ordered within the current project LAN1123. After the data processing it seems that it is almost impossible to reliably retrieve information about grassland status using only single channel HH-polarisation SAR data as no observable relations between HH-backscatter and grassland status was found (except slight decrease of about 1-2 dB in HH-backscatter when grass grew higher).
Also the repeat pass interferometric information could not help, because the coherence of the grasslands was almost completely after the 11/22 day period due to temporal decorrelation.

Back to list of proposals

DLR 2004-2016