Monthly Archives: December 2016

Looking Back in Time with SAR Satellite Imagery: Tracing the Path of a Dead River

In a recent blog post, I had explained how the low-frequency SAR signal can penetrate dry soil and give us sub-surface imaging capability. Building on that, I want to highlight our recently published paper on an application of the penetration property of SAR images, through which we detected a buried paleochannel in the Cholistan desert area in Eastern Pakistan. A “palaeochannel” is a dried up old river bed or stream bed that has been either filled or buried by younger sediment. Paleochannels either change their courses due to past seismic or flooding activities or cease to exist due to various climatological factors. The Hakra paleochannel in the Cholistan desert is well-renowned in the region, especially with its connection to the old Indian Saraswati river.


The Cholistan desert and main network of irrigation canals in Punjab, Pakistan. Figure from Islam et al. (2016).

In our paper published in the SPIE Journal of Applied Remote Sensing, we used both optical and SAR remote sensing imagery to identify and delineate the Hakra river paleochannel. The dried river channel is buried under sand and not visible from the surface in optical / IR wavelengths, but SAR signals can penetrate dry sand (see earlier blog post)! The detailed methodology is given in the paper. To summarise the methodology, we utilized a 3-band false color combination of bands 3, 5, and 7 from Landsat 8 reflectance data and merged it with pre-processed Envisat ASAR imagery through data fusion to generate one image product for analysis. Data fusion was done through the Principal Component (PC) fusion method, in which the 3-band false color composite is transformed into principal components, the first component is replaced with the SAR data, and the resulting new merged 3-band composite in the PC feature space is transformed back into regular feature space.


3-band multisensor fused image generated from Principal Component image fusion of Landsat 8 reflectance data false color composite (bands 357) and Envisat ASAR calibrated sigma-nought image. The Hakra palaeochannel signature is visible as linear green segments extending toward southwestren direction from the visible portion of Hakra. Figure and more details in Islam et al. (2016).

Ideally, we would have liked to use L-band SAR data for this study, as it penetrates more into dry sand, however ALOS PALSAR L-band data was not available for this study. We settled therefore for the next best frequency, i.e. C-band, and utilised data from Envisat ASAR satellite. Sentinel-1 data is also C-band, however we needed a long-term time series to choose the best data for analysis, and Sentinel-1 being a recently launched satellite, does not provide that advantage. Furthermore, the Envisat ASAR datasets selected for this study were acquired in the hottest / driest part of the seasons, so as to capture maximum subsurface signal.

The remote sensing results were validated with in-situ geophysical surveys for groundwater, i.e. electrical resistivity and conductivity. The presence of high apparent electrical resistivity with corresponding low soil water conductivity values intersects well with the paleochannels identified from the remote sensing data. We also utilized ancillary data and historical evidences like locations of old wells and forts for validation.


Point locations of old forts and water sources (at which water conductivity readings were taken) in the regions overlaid on the detected Hakra palaeochannel from the Landsat 8 and Envisat ASAR fused imagery. Figure and more details in Islam et al. (2016).

I had presented the initial results of this work during my TEDxIslamabad 2014 talk. This paper is the result of collaborative research between research groups at GREL-IST and IGIS-NUST. We also thank officials from Pakistan Council for Research in Water Resources (PCRWR) for guidance and support during this research.

See the paper here:

Islam Z., Iqbal J., Khan J., Qazi W. A. (2016). Paleochannel delineation using Landsat 8 OLI and Envisat ASAR image fusion techniques in Cholistan desert, Pakistan. J. Appl. Remote Sens. 0001;10(4):046001.  doi:10.1117/1.JRS.10.046001