384 IEEE GEOSCIENCE AND REMOTE SENSING LETTERS, VOL. 11, NO. 2, FEBRUARY 2014 C-Band SAR Data for Mapping Crops Dominated by Surface or Volume Scattering Giuseppe Satalino, Anna Balenzano, Francesco Mattia, Senior Member, IEEE, and Malcolm W. J. Davidson Abstract—In this letter, a C-band SAR classification algorithm mapping agricultural crops dominated by surface or volume scattering is derived and assessed. The algorithm is an adap- tive thresholding method based on the iterative solution of the Kittler-Illingworth method applied to exploit temporal series of cross-polarized SAR data. The performances of the classification algorithm have been assessed on ENVISAT ASAR data acquired over Görmin (Germany) during the AgriSAR’06 campaign and on RADARSAT-2 data acquired over Flevoland (The Netherlands) and Indian Head (Canada) during the ESA AgriSAR’09 cam- paign. The results indicate that the classification method improves the accuracy with respect to the one obtained by the threshold method based on a constant value, unless the data distributions are mono-modal. The algorithm is fast and robust versus changes of site location and it is expected to achieve an average overall accuracy better than 80%. Index Terms—Crop classification, cross-polarized synthetic aperture radar (SAR) data, Sentinel-1 (S-1), synthetic aperture radar (SAR). I. I NTRODUCTION T HE forthcoming European Space Agency (ESA) Sentinel- 1 (S-1) C-band SAR constellation will provide continuous all-weather day/night global coverage, with six days exact repetition time (near daily coverage over Europe and Canada) and with radar data delivery within 3 hours [1]. These features open new possibilities for the monitoring of the soil moisture content (m v ) at moderate spatial resolution (100–1000 m) and high temporal resolution [2], [3]. However, for vegetated surfaces, C-band is also sensitive to the characteristics of the above-ground vegetation layer, which can dominate the radar return and obscure the underlying soil moisture information. As a consequence, effective m v retrieval algorithms require a preliminary delineation of areas over which the signal is sensitive to soil moisture conditions. This is achieved when radar scattering from the soil surface is the dominant scattering mechanism with respect to canopy volume scattering [2]. Manuscript received February 13, 2013; revised May 7, 2013; accepted May 8, 2013. Date of publication June 11, 2013; date of current version November 25, 2013. This work was supported by the European Space Agency (ESA) under the contract “GMES Sentinel-1 Soil Moisture Algorithm Develop- ment” funded by the European Union. European Space Agency (ESA) contract “GMES Sentinel-1 Soil Moisture Algorithm Development” European Union. G. Satalino, A. Balenzano, and F. Mattia are with Consiglio Nazionale delle Ricerche (CNR), Istituto di Studi sui Sistemi Intelligenti per l’Automazione (ISSIA), Bari 70126, Italy (e-mail: satalino@ba.issia.cnr.it; balenzano@ba. issia.cnr.it; mattia@ba.issia.cnr.it). M. W. J. Davidson is with the European Space Agency (ESA), ESTEC, 2201 Noordwijk, The Netherlands (e-mail: Malcolm.Davidson@esa.int). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/LGRS.2013.2263034 The objective of this letter is to present and assess a clas- sification procedure for identifying agricultural or sparsely vegetated areas dominated by attenuated surface scattering at C-band. The task is formulated as a two class problem of attenuated surface or volume dominated categories which are detected through an adaptive thresholding of their cross- polarized signatures. As the study application context requires a moderate spatial resolution, the analysed SAR images are characterized by a large k-number of looks (i.e., k ≫ 10) and, therefore, the backscatter probability density function (pdf) of the two classes can be well approximated by a mix- ture of two Gaussian pdfs. Under these circumstances, the Kittler-Illingworth (KI) method [4] is well suited to find the optimal threshold separating the two scattering classes and its iterative version [5] was chosen both to minimize computa- tion time and make use of robust threshold estimates derived from available data. The proposed thresholding algorithm has been evaluated on a large SAR data set acquired during the ESA AgriSAR’06 and AgriSAR’09 campaigns. In particular, the AgriSAR’09 campaign provided a data set consisting of RADARSAT-2 images acquired roughly every week from April to September 2009, providing an excellent experimental basis for simulating the S-1 future performances. In the next section, the experimental data set is illustrated, then the data analysis and the classification algorithm are detailed and its classification accuracy assessed. II. TEST SITES AND SAR DATA The ESA AgriSAR’06 campaign data set [6] consists of satellite SAR observations acquired by the ENVISAT ASAR system over the Görmin agricultural site north-east Germany. It includes 10 ASAR cross-polarized images acquired at swaths I1-I4 (from 19 ◦ to 34 ◦ mean incidence angle), along descending passes, acquired from late-February (DoY 52) to mid-August (DoY 227), covering the complete development period of crops. It is worth mentioning that out of the 10 images at HV polar- izations 6 were acquired at swath I1 at steep incidences. They have been included in the analysis in order to extend the time series, though SAR data at such a low incidence angle may often be influenced from the coherent scattering component and may show backscatter levels significantly higher than those observed at slightly higher incidence angles (e.g., 23 ◦ –30 ◦ ). The Görmin study area is characterized by very large land parcels (on average 80 ha) mainly cultivated with winter wheat, barely, winter rape, and sugar beet. The second data set includes frequent RADARSAT-2 acqui- sitions from April to September 2009 over the two agricultural 1545-598X © 2013 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.