Examining The Prospect of Remote Sensing Imagery to Detect Sand Rivers with Water Extraction Potential
Mpala, Sibonakaliso1; Gagnon, Alexandre1; Mansell, Martin1; Hussey, Stephen2
1University of the West of Scotland, UNITED KINGDOM; 2Dabane Trust, ZIMBABWE

This paper concerns sand rivers, also called dry riverbeds, luggas or wadis, which are ephemeral (seasonal) watercourses containing sand that are flooded with rainwater run-off from higher elevated catchment areas once or a few times in a year (Hussey 1997; Herbert 1998; Nissen-Petersen 1998; Hussey 2003). Even though the riverbed is dry for most of the year, there is perennial groundwater flow in the sand (Herbert 1998). However, this resource is not fully exploited in many areas where it could provide an essential water supply and it is not clear what the size of the resource is in comparison with traditional surface and groundwater resources.

Caption: Varying water levels in a sand river

The objective of this paper is to develop and implement remote sensing techniques that can be used to identify alluvial rivers with potential for a year-long sustainable water supply. Such techniques would help cutting down survey costs for water supply developers and NGOs investigating water resources as most of the initial survey work would be done in the office with ground verification being done only for sites that have shown signs of potential.

Being hidden below the surface has meant that groundwater has been the last component of the hydrological cycle to realise the benefits of remote sensing (Becker, 2006). Rango (1994) showed the use of remote sensing to detect most components of the hydrologic cycle, except for groundwater. In the 1960s, the most popular method for groundwater remote sensing was photogeology (Meijerink, 1996), and this exploratory technique was followed by physiographical and geomorphological interpretations. Detection of underground water has been shown to be possible through longwave radar such as carried by Shuttle Imaging Radar (SIR) systems. The detection was only up to a few metres; for instance, at a 23.5-cm wavelength, 2 m of penetration through sand was documented with up to 6 m being possible under ideal conditions (Elachi et al., 1984, Ford, 1989). Thermal remote sensing might also be used to estimate water table depth due to saturated soils having a greater heat capacity than dry soils (Becker, 2006). Heilman and Moore (1982) found good agreement between cool areas in night time thermal images and areas of known shallow groundwater. Also possible is measuring the effects of changes in ground water levels on surface elevation. Depletion or recharge of an aquifer may lead to significant inflation or compaction in unconsolidated sediments, which may in turn be observed as a surface elevation change using interferometric synthetic aperture radar (InSAR) (Galloway et al., 1998).

Zimbabwe is a tropical region and has seasonal rainfall concentrated between November and March, which is also the period where most sand rivers are fully recharged. After the rains the sand-rivers gradually discharge and it time-period between discharge and subsequent recharge that is being investigated. During the rainy season the water level is above the surface of the sand, and this can clearly be seen using optical satellite images. Predictive analysis is required in the dry season, which is the period at which the surface water gradually recedes and the sand river slowly transforms into an alluvial aquifer, and the water is now hidden from view. The interest here is to be able to see the water below the sand using any form of satellite data. Thereafter, the satellite data are compared with water level in the river, which in turn is linked to climatic conditions, with particular emphasis on rainfall patterns. For this purpose hydrological data are being collected, analysed and compared with information derived from remotely sensed satellite data. Data on water level in the river sand have been collected daily since November 2009, as well as daily temperature and rainfall data from weather stations in proximity to the three studied catchments. These data are compared with information interpreted from satellite imagery with some success.

ENVISAT ASAR images were requested for the period October 2009 – June 2012 and five datasets were available for download covering the period January 2012 – April 2012. The team compared the ASAR images with optical images from the Landsat ETM+ images taken at about the same time in the year, although there were no exact matches in the image acquisition dates. The 30m spatial resolution of the ASAR was a challenge as the river channel was in most cases covered by only one or two pixels which meant that edge detection was affected and in some cases the river channel was indiscernible. In all the images the sand river appears darker, with the trees on the river bank being much brighter. This suggested that the surface of the sand river was smooth, suggesting a likely presence of surface water, which tends to specularly reflect the incident microwave radiation. This is consistent with the time at which the images were taken; the middle of the rainy season. The microwave radiation undergoes diffused reflection in the trees and part of the radar energy is scattered back to the radar sensor. Because sand rivers are located in very dry regions, it is common to have a band of rich vegetation 20 – 100m wide on either side of the river and this band can be seen as a continuous bright band on the radar image. Thereafter, the rest of the image is predominantly darker, with lots of bright spots which can be attributed to speckle noise.

Caption: Installation of a water level logger in a sand river

A comparison was also made of the rainfall data with river water level data. Of the 12 days in which rains were received four days showed a corresponding rise in water level in the river. The difference for the remaining eight days can be attributed to presence of a large dam upstream the sand river, which for some days collects all the water into storage without releasing any water downstream. This processing and analysis of the satellite images and their linkages with the in situ hydrological and climatic observations remain to be done for the remaining months of the year as the river progressively gets drier on the surface, just below the surface, and finally until a stage were no moisture is detected at all.

References:
BECKER, M. W. 2006. Potential for Satellite Remote Sensing Of Ground Water.
ELACHI, C., ROTH, L. E. & SCHABER, G. G. 1984. Spaceborne Radar Subsurface Imaging in Hyperarid Regions.
FORD, J. P. B., R. G.; CRISP, J. A.; ELACHI, CHARLES; FARR, T. G.; SAUNDERS, R. 1989. Spaceborne Radar Observations: A Guide for Magellan Radar-Image Analysis.
HERBERT, R. 1998. Water from Sand Rivers in Botswana. HUSSEY, S. W. 2003. The Role of Sand Rivers in Rural Water Supply in Zimbabwe. Loughborough.
NISSEN-PETERSEN, E. 1998. Water from sand-rivers.
RANGO, A. 1994. Application of Remote-Sensing Methods to Hydrology and Water-Resources.