2.5 Application of remote sensing in hydrology .1 Experiences in semi-arid regions
2.5.4 Determination of soil moisture
Considering soil moisture, mainly the upper 10 to 50 cm of the soil are considered. This moisture is used by plants like barley or corn. This is far less than the area used by trees like eucalyptus or olive trees. In a semi-arid rangeland site in South-eastern Arizona soil moisture measurements showed that the recharge in normal winter season with about 30%
of the whole annual rain of 350 mm/a reaches a depth of at least 0.5 m depth in winter rains under grass and a sand loamy soil. This is also the case because the vegetation is senescent and does not use water (Scott et al. 2000).
There are different remote sensing techniques that allow soil moisture to be measured (Table 2.13). Belz (2000) used a visible/near-infrared technique, the procedure of principal component analysis, the Tasseled Cap Transformation: The digital values of the different bands are multiplied by a fixed matrix. This transformation is based on the recognition of a triangularshaped scatter diagrams of Landsat MSS data between band 4 and 2 on scenes of agricultural scenes of Midwestern USA (Kauth & Thomas 1976). The first component of the linear transformation leads to the soil brightness. The axis orthogonal to the soil line is the greenness value. The third component shows the degree of maturing of the crop, called
“yellow stuff” or indicator of wetness. The transformation can be applied in the same way to other regions such as semi-arid regions. However not all components of the image are
TSAVI a NIR aRED b
represented in the transformation, e.g. the non-green vegetation in the semi-arid region.
(Crist & Cicone 1984).
Table 2.13 Remote sensing techniques used to quantify soil moisture (Engman 2000)
Wavelength region Property observed Advantages Disadvantages
Other techniques for investigating the soil moisture by remote sensing are the use of the thermal infrared data (channel 6) of the Landsat TM sensor. Shih & Jordan (1993) differentiated soil moisture status for an area in South-western Florida according to the thermal sensor response and overlaid these results with a landuse classification. This allowed the classification of different moisture statuses within various landuse categories.
Courault et al. (1993) investigated the influence of properties of the bare soil surface on the response to spectral measurements. Using ground measurements like Munsell color, soil water content and surface roughness, soil water was the most influencing factor. They monitored the surface stages of slaking of soil due to simulated rainfall On wet soils the
58 2.5 Application of remote sensing in hydrology
reflectance of the red and near-infrared spectral band of the SPOT satellite simulation radiometer (red band: 600 to 690 nm, near-infrared band: 790 to 900 nm) was reduced. The slope of the soil lines of wet or dry soil are different due to water adsorption in the near-infrared.
Nowadays microwave techniques are used to determine the soil moisture. These techniques are based on the different dielectric constants of dry soil and water. The dielectric constant of dry soil is between 1 and 10 whereas that of water is 81. Therefore an increase in soil water implies an increase in the dielectric constant. For a loamy soil the increase in soil moisture from 01 to 0.3 signifies a fourfold increase of the dielectric constant (5, 19) (Engman 2000). Neusch (2000) investigated the active microwave techniques. A functional relationship between dielectric constant and volumetric soil water content could be not found, only empirical models exist (Neusch 2000).The microwaves are sensitive to surface roughness. The penetration depth of the microwaves depends on the frequency and the soil moisture (0.2 to 0.008 for soil moisture of 0.05 to 0.4 g/cm-3 at a frequency of 1.3 Ghz) Engman 2000, Ulaby et al. 1986). Empirical or semi-empirical models exist to explain the backscatter of the radar signal by soil roughness and soil moisture. The prediction of the soil moisture by these models varies widely (up to 12%) (Neusch 2000). Also vegetation attenuates the microwave signal (Engman 2000). Neusch (2000) studied sensitivities of models for prediction of soil moisture from L-Band to surface roughness and standing vegetation in agricultural fields in Southwestern Germany.
Semi-empirical or empirical models are strongly site dependent: Adaption of the empirically algorithm coefficients is needed if using these models in areas with different terrain characteristics (Neusch 2000). Therefore to exactly determine soil moisture the effects of surface roughness, vegetation and terrain (especially active systems like SAR are sensitive to it) have to be considered. Nevertheless it is difficult to exactly quantify the contribution of soil moisture to the backscattering of the radar signal in distinction to other biophysical parameters such as vegetation or soil texture distribution (Neusch 2000).
Although the current software packages nowadays include image processing modules for radar images a precise terrain model which is needed is frequently unavailable. In addition, standing non-green vegetation often found in semi-arid rangelands interfere with the backscatter of SAR (Moran et al. 2000). Therefore new approaches are tested nowadays like measurement of soil moisture based on radar combined with optical measurements such as based on the temperature measurements in the mid infrared channel of Landsat 5 or 7 (Engman 2000).
Chapter 3
Methodology to determine water harvesting sites in the research area 3.1 Main problems and open questions
Since reliable data on water resources are scarce or even missing in the project area more reliable data resources have to be found. The literature review has shown that new technologies allow this gap to be filled. The tool ‘Geographical Information System’
finally combines the fields of data acquisition, data processing and the decision process.
First of all data on rainfall within the research area have to be analyzed to consider the problem if there is enough rainfall to guarantee the success of water harvesting techniques.
Then the influence of the various parameters presented in the previous chapter should be analyzed and quantified by means of remotely sensed data. Other parameters influencing the choice of water harvesting have to be considered. All these factors that have an impact on the choice of potential water harvesting sites should be incorporated in a decision process model. This model should be flexible enough to adapt it easily to other conditions.
Mainly the natural resources were considered in the current study. Social issues have to be the scope of other studies. The result is a map presenting potential sites for water harvesting. This map could be a planning tool for implementation of water harvesting in the research area.