Firstly : For satellite images : You can Use Landsat ( High Resolution ) images or MODIS ( Moderate Resolution ...). After you have to choose the type of classification : 

- If you know the different types of soils in your area and their spectral signature you can use Supervised or engineering classification  : 

You can see these two papers for more informations :

http://gisgeography.com/image-classification-techniques-remote-sensing/

http://www4.hcmut.edu.vn/~cnan/Principles%20of%20geotechnical%20engineering%20(Fifth%20Edition,%20Das)/83-99.PDF

Firstly : For satellite images : You can Use Landsat ( High Resolution ) images or MODIS ( Moderate Resolution ...). After you have to choose the type of classification : 

- If you know the different types of soils in your area and their spectral signature you can use Supervised or engineering classification  : 

You can see these two papers for more informations :

http://gisgeography.com/image-classification-techniques-remote-sensing/

http://www4.hcmut.edu.vn/~cnan/Principles%20of%20geotechnical%20engineering%20(Fifth%20Edition,%20Das)/83-99.PDF

Hi Yacine,

given the dominant presence of sandy deposits, the extent of your research area and the relative stability of soil type over time, I think medium to coarse spatial resolution, low temporal resolution (time passing between subsequent sensor visits), high radiometric and spectral resolutions are needed.

Thus, optical multispectral scanners (Landsat-like, or MODIS) or hyperspectral sensors (EO-1 or Hyperion) could record spectral signatures of different soil properties, but I am not sure this would be enough for classifying a soil type. In particular, optical passive remote sensing sensors do not penetrate more than a few millimiters into the soil (might be helpful for Horizont O, but in your case, I guess this reflectance would be represatitve of weathered regolith). Subsurface soil profiles should be confirmed, as well as spectroradiometer's curve obtained in situ should be used to carefully correct radiometrically remote sensed data.

Moreover, dry soil with almost non-existet organic matter is increasingly reflective with increasing wavelenghts..less complex means less identifiable as spectral signature and more difficult to distinguish from one soil to another.

This doesn't mean it cannot be done, just that it is more complicated than, for example, with vegetation.

An alternative to optical sensors might be active microwave RADAR. I think (active) Microwave RADAR remote sensed data was used in the Sahara region.

Anyway, there are a number of soils features that can be extracted from remote sensing techniques:

- soil texture and moisture content

- organic matter and iron-oxide content

- soil salinity

- surface roughness

Take a look at Jensen's "Remote Sensing of the Environment". It's the book where I found these information.

Good luck Yacine!

Nic

Thank you Mohamed, and Nicolas for your interest in my question and your answer with the reference will help me set up an experiment and refine my research question about Soil of arid regions and remote sensing.

Greetings Yacine,

I found one research paper on your area of concern,

http://www.academicjournals.org/journal/AJAR/article-full-text-pdf/D00C9C238184

Remote sensing technologies can provide spectral signatures of the soil or surface you are studying. Satellite images can provide surface optical properties. Such properties are recorded as pixels or picture elements, they are the dots on an image gathered by a satellite. Computer software maybe written to scan through image, the dots one by one, and such data maybe stored in a database for further analysis.

Algeria has approximately 2.381 million sq. kms. land area mostly arid. Soil analysis and classification is done to study where agricultural crops or plants may be grown.

Here are some practical steps on how it may be done.

1. Acquire a satellite image Landsat or MODIS, or any satellite generated image of your country or area of interest.

2. Get technical information as to how to interpret the map accordingly. The idea is to know how to interpret each point or dot in the map.

3. Use software or computer program to create a database of each pixel of the map. Each pixel has its corresponding longitude and latitude. Make approximation as to the actual land area each of the dots represent. Each dot is an actual place on the Algerian map.

Say for example, you were able to have a database with 100,000 dots, dot no. 1 is on the northern most part of the country near the mediterrenean sea ... and the last dot may be near the border with Niger or Mali.

Sample record in the database

1. dot_no

2. latitude

3. longtitude

4. satellite_color

5. satellite_interpretation

6. actual soil analysis details

7. classification

Satellite images can give you surface readings as acquired by sensors positioned kilometers from the surface using optics, infrared, etc. They give general ized trends or readouts, which must be validated by actual soil analysis and additional research work. For example, the pixels on areas on the northernmost may be colored brown, and there are 3,000 dots colored brown in the entire map. Soil analysis maybe done on a sampling of such dots. I could remember in my undergrad soil science class, we use to drive down a cylinder a foot or so into the ground to get soil samples and bring it to the lab for chemical and physical analysis. If there is a brown dot a kilometer from where you are, it may represent another thousand brown dots, 500 kms away. Do your best to understand and interpret each dot or pixel grouping.

Then, putting your data in a computer database will allow you to do statistical analysis of all sorts, this will allow great flexibility in your research work. You may use SPSS, PSPP or R to do the stat work. You may also you JAVA or Phyton for computer databasing component of your research work.

Then you may go back to your map, and do enhancements on the zoning or classification.

This may be lots of very detailed work, but this is how I will do it if i were in your place. Hope this helps. 

May Allah, swt, guide you in your research work.

These are all great answers. I suggest to start with a simple MLC unsupervised image classification per image or image mosaic of the same date and sensor. It will give a quick overview of the main soil regions. 

 Dr Van Gils, 

what do you mean by MLC? can you elaborate more on it?