WATER RESOURCES RESEARCH, VOL. 30, NO.2, PAGES 175-179, FEBRUARY 1994 Effect of saturated areason backscattering coefficient of the ERS I synthetic aperture radar: First results P. Merot, A. Crave, and C. Gascuel-Odoux Science du Sol, InstitutNational de la Recherche Agronomique, Rennes, France S. Louhala Geosys, Toulouse, France Abstract. One way to estimate spatial distribution of water content at the soil surface consists of using active microwave remote sensing. It has been theoretically and experimentally demonstrated for unsaturated conditions. Nevertheless, radardataare ambiguous whenponding conditions occur,as in variable source areas, dueto the contradictory influence of the dielectric effectand the specular effecton the backscattering attenuation coefficient, cr 0. A procedure is considered, based on a topographic analysis,to take into accountthe influence of the two effects on radar response. The present results stress the capabilities of the ERS (European Remote Sensing Satellite) radar to surveysaturated areas in time and space. Introduction Mapping areas where an excess of water occurs at the soil surface is of major importance in hydrology. Surface runoff in the widest sense may be generated by two main mecha- nisms. First, Hortonian overland flow ("infiltration excess runoff")occurswhen the rainfall intensity exceedsthe soil infiltration capacity. This is independentof the watershed morphology and of position in the landscape; for homoge- neous soils it involves the whole watershed considered. Second, saturation overland flow andreturn flow, according to the variable source area concept, which was first pro- posed by Cappus [1960] and then developed by different authors [Kirkby and Chorley, 1967;Beven and Wood, 1983; Merot,1988] represent surface runoff generated by rainfall on saturated areas. According to this concept, not all the watershed area will contributeequally to runoff; the runoff- generating zone may be of varying size and represents the groundwater exfiltration areas.Its occurrence depends both ontopographical factors, .specifically convergenceof runoff flow and slope andon soilcharacteristics, specifically per- rneability. The distribution and the extent of these areas controlthe saturation overland flow and return flow. Differ- ent authors [Beven and KiPkby, 1979;Grayson et al., 1992] have developed models based onthissecond concept. The difficulty of mapping therunoff exfiltration zone, due to the rapid variation in space and timeof these areas, has often been pointed out.It is themain limit to validating the variable source area concept and thecorresponding models. One way to estimate water content at the soil surface consists of using active microwave remotesensing. This technique has been theoretically and experimentally demon- strated [Chanzy, 1991; Engman et al., 1989; Mo et al., 1984; Ulaby et al., 1978; Van de Griend andEngman, 1985]. However, when ponding conditions occur, asin variable source areas, radar data are ambiguous due to the contra- Copyright 1994 by the American Geophysical Union. Paper number 93WR02920. 0043-1397/94/93 WR-02920505.00 dictory influenceof the dielectriceffect and the specular effecton the backscattering attenuation coefficient, cr 0. The dielectric effectis caused by the dielectric capacity of soils, which increases with the water content; the backscattering coefficient has a positive correlation with the volumetric water content of soils until a threshold close to field capac- ity. The specular effect is due to the reflection at the saturatedsoil surface or on ponding water that involves a decreasing backscattering coefficientwhen the saturatedsoil surface increases. In a previous paper, Brunet al. [1990] showed that saturatedareas can be detected usingradar data. These data came from a helicopter-borneinstrument, ERASME (Etude par Radar des Sols de la Mer) with the same configuration as the ERS 1 (European Remote Sensing) satellite synthetic aperture radar (SAR); under a threshold backscattering value the measured plots corresponded to saturated areas; over this threshold value they correspondedto soils at field capacity. In another paper [Merot and Chanzy, 1991] an experimentcharacterizing the SAR response in controlled ponding conditions was reported. The extent of surface saturation was measured on two plots with different rough- nesses. The extent varied in time from 50% to 0%. The surface saturation was simultaneously measured by the crane-mounted radar RAMSES with two configurations (5.3 GHz, VV, 23 ø, and 5.3 GHz, HH, 15 ø) and by photography. It was shown, on pixels of 10 m 2size, that the backscattering coefficientdecreases when the percentageof saturatedarea increases. The correlation gradient was the same for the different roughnesses. Because of ambiguousSAR values between measurements in pondingconditionsand measure- ments in drier conditions, the necessity of complementary data was emphasized. In thispaper,asa part of a European Space Agency (ESA) ERS 1 project [Louhala et al., 1993],preliminary results from ERS 1 SAR are presented. The objectiveis to analyze the radar data and to emphasize an interpretation procedure taking into account the two processes of backscattering attenuation when the soil is drying and ponding. 175