Study of high SAR backscattering caused by an increase of soil moisture over a sparsely vegetated area: Implications for characteristics of backscattering

We used interferometric methods on a pair of repeat-pass ERS-1 synthetic aperture radar (SAR) images to study soil moisture changes over sparsely vegetated targets. The intensity of the SAR image acquired at one time was higher than that of an image acquired at an earlier time. We used a correlation image computed from the SAR image pair to study the cause of the observed changes in SAR intensity. Because a reduction of correlation over areas with intensity changes was not observed, we interpreted the intensity changes as not being caused by changes in roughness/structure, but by a change in soil moisture owing to rainfall. An increase in soil moisture ranging from 5% to 20% is the most likely explanation for the increase of intensity. These analyses imply that both intensity and phase information should be used in SAR change detection applications. 1. Background Soil moisture is an important environmental parameter that plays a key role in the hydrological cycle, aVecting applications in meteorology, climatology, ecology and agriculture. Studies indicate that soil moisture plays a signifciant role in nearsurface atmospheric variability and is an important component of water and energy transfer between the surface and the atmosphere (Shukla and Mintz 1982, Delworth and Manabe 1989). For ecosystem and agricultural studies, soil moisture is an important factor in modelling ecosystem dynamics and crop yields (Choudhury et al. 1995). Many studies have explored the use of passive microwave systems to infer soil moisture from airborne and spaceborne instruments (Wei 1995, Njoku and Entekhabi 1996). However, these instruments are severely limited by their coarse spatial resolution, usually of the order of tens of kilometres per resolution cell. Synthetic aperture radar (SAR), having high spatial resolution, all-weather acquisition capabilities, and high sensitivity to soil moisture, provides an opportunity to overcome the limitations of the other methods if certain limitations on its use can be addressed. †e-mail: [email protected] ‡[email protected] Internationa l Journal of Remote Sensing ISSN 0143-1161 print/ISSN 1366-590 1 online © 2002 Taylor & Francis Ltd http://www.tandf.co.uk/journals DOI: 10.1080/01431160110040035 D ow nl oa de d by [ U SG S L ib ra ri es P ro gr am ] at 1 6: 26 2 9 A pr il 20 13 Z. L u and D. J. Meyer 1064 Increases in soil moisture result in an increase in radar backscatter approaching 11 dB (Dobson and Ulaby 1986). This is because a change in soil moisture changes the dielectric constant, a primary factor controlling radar backscatter. In theory, the change in soil moisture can be inferred from changes in radar backscatter. Although much progress has been made in the use of SAR to measure soil moisture (Dobson and Ulaby 1986, Oh et al. 1992, Engman 1995), many diYculties remain owing to the eVects of soil and terrain characteristics other than moisture—primarily surface roughness and local terrain slope (Engman 1995, Islam and Engman 1996). In most natural settings, the eVect of roughness is generally equal to or greater than the eVect of soil moisture on radar backscatter. Therefore, to derive changes in soil moisture from single-band, multi-temporal SAR data, one must deal with the eVects of surface roughness and local slope on the radar signal. Repeat-pass SAR images acquired with almost identical viewing geometry can be used for interferometric data analysis. The interferometric phase is a measure of change in distance from SAR sensor positions to the ground target, and can be used to map surface topography or change of topography (see, for example, Massonnet and Feigl (1998) and Lu et al. (2000)). The interferometric coherence or correlation coeYcient (see §2.1) is a measure of the variance of the diVerence of phase values in the two SAR images. The correlation coeYcient of repeat-pass SAR images is primarily controlled by changes in the backscattering phase, and mainly depends on parameters related to land surface (Wegnuller and Werner 1997, Lu and Freymuller 1998). This paper discusses how we studied a feature in southeastern New Mexico that had an anomalously high SAR re ectance, possibly caused by an increase in soil moisture resulting from a recent rainstorm. We used multi-temporal repeat-pass SAR images in an attempt to explain these SAR returns by exploring both intensity and phase information in the radar signal. To determine whether the high return was caused by changes in surface roughness (owing to changes in surface texture) or by the change in moisture content, we studied interferometric coherence using the repeat-pass image pair. After determining that the change was most likely caused by soil moisture rather than surface roughness, we used a geometric optical SAR backscattering model to quantify the amount of soil moisture increase required to cause the observed change in backscatter coeYcient. Finally, an interferometric coherence map was compared to a Landsat 5 Thematic Mapper (TM) image to show that the putative moisture signature occurred mostly in an area devoid of vegetation, consistent with the high coherence maintained between the two dates used in the repeat-pass image pair. 2. Methods 2.1. Correlation coeYcient or interferometric coherence Radar backscattering represents the amount of energy re ected back to the sensor from each resolution element of the imaging surface and is the coherent sum of backscattered energy from individual scatterers. A radar backscatter has a size of the order of a radar wavelength, and therefore the radar backscattering at a pixel level (the size of several metres to tens of metres) consists of contributions from hundreds of thousands of individual scatterers generally considered to be independent. The backscattering of a pixel in SAR imagery can be simpliŽ ed as a complex value of |s0 |e Õ jQ , where |s0 | and Q represent amplitude and phase of the backscattering respectively. D ow nl oa de d by [ U SG S L ib ra ri es P ro gr am ] at 1 6: 26 2 9 A pr il 20 13 Change in soil moisture by interferometric data analysis 1065 The phase Q, at each point in a radar image, is the sum of scattering phase s , a statistical variable, and the propagation phase p , a deterministic variable. The propagation phase Q p is equal to ­ (4p/l)r, where r is the apparent range distance from the antenna to the imaged point. The apparent range distance is the sum of the absolute distance from the antenna to the target and a range delay caused primarily by water vapour in the atmosphere. The backscattering phase, on the other hand, is primarily controlled by the macrostructure and texture of the surface. For simplicity, let us consider SAR backscattering over sparsely vegetated or bare soil to minimize volume scattering as a random in uence on Q s . Both phase and intensity can be altered by surface roughness and dielectric constants, which are primarily controlled by soil moisture content. Surface roughness represents the macrostructure of the scatterers. Changes of surface roughness would modify the relative positions of the scatterers within each pixel and therefore aVect both the intensity and the phase information of the backscattering signal. In this scenario, both the statistical backscattering phase and the deterministic propagation phase will change owing to changes in roughness. On the other hand, a change in soil moisture content may only alter the intensity of the backscattering and the propagation phase (owing to the possible swelling of soil ) and have no eVect on the backscattering phase. Therefore, if we can measure quantitatively the changes of the scattering phase, we can determine whether the change in radar backscattering is caused by surface roughness or water content. Because both amplitude and phase information can be retrieved from SAR imagery, we propose the following equations to quantify the degree of changes of backscattering phase and amplitude between two repeat-pass SAR images.