as per the above answer please let us know you want to measure area or the depth of the sand? i.e. horizontal extension or vertical profile of the desert.

Using remote sensing tools in desert sand dunes is not an easy task. Ta have an idea about this subject I suggest to read this paper: Hugenholtz et al. 2012. Remote sensing and spatial analysis of aeolian sand dunes: A review and outlook. Earth-science Review. Vol:11. Pp: 319:334.0

You get a free accepted pending revision from: http://people.uleth.ca/~chris.hugenholtz/web/EARTH1742.pdf

If soil has no significant organic matter content or iron oxide, extremely dry you can predict percent of sand. The best combination then will be indices of blue, SWIR and MIR bands. The challenge that you will face is the atmospheric correction for the blue band.

Thank you all very much, these are my comments for what has been mentioned so far:

@Kenneth M Towe

YES Kenneth, we want to quantify the amount of sand over an area of 2 million Km2, for now we are very open to any crazy idea! the depth could be 100 m max,

@Shrikant Mukate,

we are interested for 3D sand availability?

@Hicham Elbelrhiti

Thank you very much for the formidable paper, I had a look and it helped me somehow but i am still looking for top latest technology and methods,

@Jawad T Al-Bakri

thanks jawad for your idea, could you share any paper or link about possible indices and bands combination?

------

Dear all,

considering the above question and my comments, I would like to add that we are interested to know RS and GIS data & techniques to create a "predictive models" to quantify the 3D sand availability,

we would like to know all the possible ideas, challenges, what has been done before...and what are the input that can be incorporated into the model?

please share any crazy idea, link, paper that might help all of us to develop a better understanding and explore the advantages of RS in depth

Salim :)

The same question on LinkedIn as well and you are kindly invited to post your constructive comments:

https://www.linkedin.com/groupItem?view=&gid=76109&type=member&item=5971778902756990980&commentID=5972461471223812096&report.success=8ULbKyXO6NDvmoK7o030UNOYGZKrvdhBhypZ_w8EpQrrQI-BBjkmxwkEOwBjLE28YyDIxcyEO7_TA_giuRN#commentID_5972461471223812096

To estimate the volume of barchans you can have a look on this paper: Elbelrhiti et al. 2008. Barchan dune corridors: Field characterization and investigation of control parameters. JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 113, F02S15, doi:10.1029/2007JF000767,

Dear Salim,

First of all you need the courage to work on such huge area (2 million Km2). I suppose it is the Sahara desert for that I propose to have a look on this paper:

Ballantine et al. Mapping North African landforms using continental scale unmixing of MODIS imagery. Remote sensing of environment. Vol. 97 (4). Pp:470-483.

In other hand, I think that using remote sensing tools to quantify the 3D sand availability present several challenges. First sand is a detrical sedimentary rock, which means that the diameter of the grain is its principal physical characteristics. In other word, for the sand the chemistry don’t present any sense and for that reason (among others) sand dunes could have different color because it is originated from different kinds of rock. For example the black sand is composed mainly by the magnetite while the yellowish is probably rich in quartz. So, if you want to use an automatic method to identify dunes field (for example band ratio indices) you need to use and run all the possible indices that could identify dune fields.

Second, to cover this huge area you need to use lot of satellite images and you must choose between coarse resolution images or images like landsat or Aster. Anyway, it is difficult to use high spatial resolution images.

Finally I think it is not easy but it is possible. I don’t know if you are familiar with French. If it is the case you can have a look on this paper to avoid some elementary problems in your project:

Imad Lekouch, Pierre Rognon, Belkacem Kabbachi, Mohammed Benssaou, Abdelkarim Ezaidi. 2010. Essai de morphométrie sur un système dunaire du Sud-Ouest marocain à partir d’une image Landsat. Volume 21, Numéro 1, janvier-février-mars 2010 p. 54-62

This is a commentary that we did to the former paper:

Elbelrhiti et al. 2010. COMMENTAIRE SUR L’ARTICLE INTITULÉ

“ESSAI DE MORPHOMÉTRIE SUR UN SYSTÈME DUNAIRE DU SUD-OUEST MAROCAIN À PARTIR D’UNE IMAGE LANDSAT” D’IMAD LEKOUCH ET AL.

[SÉCHERESSE 2010 ; 21(1) :54-62]. Revue Télédétection, 2010, vol. 9, n° 2, p. 151-155

@Hicham Elbelrhiti

many thanks for your detailed answer, I am from Algeria so i speak FR of course!

I will go through the data you mentioned and we will be in touch in the near future to investigate al the options one by one,

thank you again and again Hisham

with best regards

Salim :)

Wow, a very large task that you are taking on, but a nice challenge and remote sensing should surely be one of the tools that you will want to use.  Some good comments and suggestions already given and I'll add some possible data sets to consider:

* For such a large area you will need satellite images with large footprints with the ones that come to mine being Landsat TM and MODIS.  My guess is that for this scale of interest that a way to get a 1st order calculation will be about as good as can be done with the data sets that are available.  I am not sure if it is best to try to do automatic classification into 'sand types' or do an interactive visual analysis and interpretation to map the areas.

* You may want to consider do some analysis that includes the 2km topo data that exist for the entire earth and attemp to merge it with the information extracted from the satellite images to get a handle on possible 3D information extraction.

* In the 1980s and early 90s there were several SIR (Shuttle Imaging Radar) missions that collected radar image data for different parts of the earth, with the L band mission collecting some very dramatic 'sub surface' features in the deserts of North Africa; actually showed major water drainages covered by tens of meters of sand.  When the sands are very dry the different radar frequencies (c, x, and L bands) will pernetrate thru the sand different amounts.  Perhaps some of these types of data will give you some 'thickness' information.

good luck

Pat

Hi Salim,

In my opinion, Hyperspectral data and imaging spectroscopy technique may be the better option and accurate way to quantify sand in the desert. The field and lab spectrometry will get an accurate result with data mining such as PLSR SVM and MARS methods for mapping sand amount pixel by pixel through the image. Of course, multispectral data can be also good data if there is no covering hyperspectral data. The moisture content measurements should be considered when using hyperspectral data for quantifying sand amount.This is will be a benefit for surface method. For more depth, synthetic aperture radar (SAR) data will be more helpful for dry sand through empirical models.

Best wishes

Hi Salim,

I believe you just need a digital elevation model over the area of interest to calculate volume of topography over a selected reference serface. The best candidate would be SRTM, less so ASTER v2 DEM, (both are free). But the most accurate DEM is WorldDEM by Airbus Defence and Space (not free). If you need to discriminate sand from vegetation simply use one of the NDVI indexes on the Landsat multispectral data (free).

Best Regards,

Kaz

@Pat S. Chavez,

Thank you very much for your very helpful comments about sand classification and SIR  missions,

@Said M.A. Nawar,

thanks said for your input for on this discussion, me too i am very optimistic about HSI but as u said we will face the prob of volume of the area which is huge...sure SAR can do the job,

@Kazimierz Becek,

many thanks for your comment, i do agree that DEM  can contribute to this research and as u said we are counting on the availability of FREE data :)

thank you all very much,

The discussion is still ON

Hyperspectral and high-resolution lidar data would be great, but I think that for the size of the area that is of interest you are not going to find these types of data sets and they are just not available.  You are going to have to make use of what is operationally available for such a large area, which means less that the ideal data sets.

Topgraphy (such as DEMs) will be good over sand dunes, but a large percentage of sand is often in 'sand sheets' and not just in the form of sand dunes, so in general these areas are relatively flat and topo will not be as helpful as over dune type areas.

Pat

Dear Salim,

  Once you talk about the amount of the sand in the desert, it means the volume of the sand (at least from my point of view!). Therefore, knowing the surface or the topology of the desert won't be enough. It is not accurate to assume a certain depth of sand for all the regions in the desert. In order to calculate the exact volume, you will also need to know the depth of the sand at each area as well. From my point of view, this work is somehow futuristic. However, for the time being in my opinion your solution will be a type of ultrasound or deep penetration radar to calculate the depth of the sand layer in different regions. Here, one should distinguish materials at different depths of the desert for different desert regions.

Good luck

Peyman

Hi Salim,

First of all, You should define some sample plots from the desert maching with your RS data spatial resolution. Get a free multispectral data and if posible hyperspectral ata such as CHRIS PROBA or Hyperion. And check the stepwise regression between each waveband, vegindex and measured sand quantify. You may find a relationship. In this method, you should devide your measured plot for modeling and testing. Use modeling plots to find ideal formula and use testing plots for validation.

I hope it would be helpful 

Dear Salim,

In theory you need two parameters to be able to calculate the volume of sand, the surface occupied by dunes and its depth. Using remote sensing it is more or less easy to define the first parameter (the surface) but I think it is not possible to determine the second parameter using RS tools even with active RS like radar which can determine at the best 1 m of sand thickness. However there is an indirect way to calculate (or at least estimate) the volume if we know the surface occupied by sand dunes. There are imperial formulae to calculate the volume if we know the width, high or length of sand dunes and these formulae depends on the types and the morphology of the dunes that you have in your studied area.

I don’t think that the use of free DEM can give a good result, first because DEM’s cannot give an idea about the thickness of sand. Second because the coarse resolution your are not able to take in consideration the small dunes. For example, small barchans has about 20 m of width and you will not be able to detect this kind of dunes with coarse resolution of RS data, nevertheless these dunes has an important role in the redistribution of sand flux inside the dune field.

I suggest you to use the hyperspectral data set and do some regression with the field data of sand composition. Even the multispectral data may give some indication, but needs to be studied/established.

@all,

thank you all very much for your time to keep the discussion progressing, till now we have got several ideas on how the work should be done,

I know this work can not be done using one type of dataset, and combining several data and methods is a must,

please keep posting any suggestion and ideas about the idices that can help discriminate the sand from other land cover types and how the 3d measurements could be done ?

many again to all of to all of you,

best,

Salim