Detection of Inland Open Water Surfaces Using Dual Polarization L-Band Radar for the Soil Moisture Active Passive Mission

• Published in: IEEE transactions on geoscience and remote sensing Vol. 54; no. 6; pp. 3388 - 3399
• Main Authors: Seung-bum Kim, Ouellette, Jeffrey D., van Zyl, Jakob J., Johnson, Joel T.
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.06.2016; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Algorithms; Classification; Heterogeneity; L-band; Microwave radiometry; Noise levels; Polarization; Radar; Soil moisture; Spaceborne radar; Spatial resolution; Synthetic aperture radar; synthetic aperture radar (SAR); Vegetation mapping; Water bodies; water detection; synthetic aperture radar (SAR); water detection; Classification
• ISSN: 0196-2892; 1558-0644
• DOI: 10.1109/TGRS.2016.2517010

A dual-copolarization algorithm to classify inland open water bodies free of flooded vegetation using an L-band radar is presented and evaluated, with a view to applying the method to the Soil Moisture Active Passive (SMAP) mission for hydrological science and soil moisture retrieval applications. Past radar-based water body detection algorithms have applied a threshold to a single-polarization measurement, with water body detection declared if the observed cross section is less than the specified threshold. However, such methods are subject to ambiguities associated with scene variability and terrain slopes, making a universal threshold value difficult to derive and complicating the global application of such methods. Because SMAP will provide measurements in both HH and VV polarizations, the copolarization ratio is also available for water body detection. A threshold of -3 dB applied to the HH/VV polarization ratio is found effective in detecting water bodies at 40° incidence angle based on analysis of theoretical model predictions and measurements from airborne synthetic aperture radar and the spaceborne Aquarius scatterometer. When the water surface is calm and its radar response is very small (i.e., at the radar thermal noise level), the HH/VV ratio method fails. However, a combination of an HH/VV threshold (at -3 dB) and an HH threshold (at -25 dB) is shown to allow water body classification even in this situation. This proposed "combined" algorithm is assessed in four different geophysical scenarios. The resulting water body detection error is shown to be less than 10% for these cases, which satisfies SMAP requirements to allow accurate soil moisture retrieval, and the corresponding false alarm rate is smaller than 2%. The robustness of the proposed approach to subpixel heterogeneity has been also investigated. The performance of the algorithm remains sensitive to the noise level of the radar observations: for SMAP, a radar noise-equivalent sigma 0 of -28.5 dB or less is required in order to facilitate acceptable performance.