Should laser diffraction become the new standard for soil particle size analysis ?

 This is the title of one of the presentations of the different methods used to determine  the clay content in soils. You can download the presentation here.

The recording is also available with the other presentations:

Improving the data downloaded from the ASD spectra in USGS

 I continue downloading ASD spectra from thre USGS library,  in order to have a data set of several clay minerals and as soon as you look to the spectra I get more ideas to improve the data set. 

The header in the spectrum gives details such as the sample number, formula, type and spectral purity, so iI will use these data as variables that will improve the work.

I show you this spectra of the sample Kaolinite CM5 as an example:

USGS Reference Materials (Illite)

We can get spectra from the USGS library and with some work we can import them to R or any other chemometric software to work with them. In this case I search for "illite" ( a type of clay mineral) and I found these 12 samples to work in coming posts with them.

       

We can see the spectra samples overplotted:

The USGS Library give a lot of details about each sample, such as the origin, purity, particle size, ......, and even the picture of the sample:

Illite GDS4.2 Marblehead ASDNGb AREF

How much kaolinite could contain my soil sample?

Now that we have our reference kaolinite near infrared spectrum stored  and studied, we can acquire a we spectrum (unknown) and compare it with the reference material trying to have a clear idea od how much kaolinite could contain. Soils high or very high in kaolinite have disadvantages for agriculture, so looking to the comparison we can take some actions.

We are going to use one spectrum (the first one) from the soilspec package, where we have a data set called:  "rutherglenNIR" which is a subset of soil spectra samples from the Rutherglen Area.

First, we compare the kaolinite reference material spetrum with the spectrum of the unknown sample:

And we have a look to the kaolinite featured spectra region, so we can trimm the spectrum and start to make some comparissons after the normalization processes:

Normalization Step 1 :

Normalization Step 2:

Now we can compare the areas, and  get the conclussion that the unknown soil sample has a relative low content of kaolinite compared to the reference spactrum.

Of course we can confirm this with numbers such as Area, Relative Depth,....

Reference Materials (Kaolinite)

 In the spectra of reference materials there are two kaolinite spectra, one is called kaolinite_114 and the other kaolinite_113. We can take them apart and overplot them in the same scale to see the differences:

One of the characteristics of the kaolinite spectra is the double is the doublet spectra feature in the 2200 nm region, so we can trim this region, take it apart and keep it as a diagnostic spectral feature to compare with our soil samples to determine if our soil samples contain kaolinite and in what proportion.

We can trim the zone of interest for the kaolinite and make calculations for the Area, Depth, Slope, etc. After we calculate the same values for our soil sample and compare the results to estimate  how much of the reference material (in this case kaolinite) could contain our soil sample.

In the next figures we see how the spectrum is trimmed, normalized, and the area is calculated:

Reference Materials for Soil in soilspec package.

 

How "spectra2colour" function from "soilspec" get the colour predicted values

 In the soil spectrum the colour regions are defined, and the mean reflectance value obtained for each. With the three values (RGB) we obtain a colour prediction when proyected on the RGB cube space.

 

Colour predictions with soilspec package

 The soilspec package comes with a data set called specSoilCol, which has 12 spectra of four types (A, B, C and D) of soils at three deferent depthss (Horizons 1, 2 and 3). 

The spectra is a Reflectance scale:

Now we use the function spectra2colour in the soilspec package and we obtain the predicted colours for the 12 samples that we have to compare with the real colours of the samples.

Normally the soil in the Horizon 1 is darker due to a higher quantity of organic matter.

Colour data available in soilspec package

 Once we have installed the soilspec package we need some data (spectra) to work with, so we can import our own data, or to use the data available in the package (datsoilspc).

The data set has 391 spectra samples (range from 350 to 2500 nm in 1 nm steps) acquired with an AgriSpec instrument using a probe. All the information of this dataset is given in the description of this file in the poject.

Once we have the data in the workspace we can plot it as a first and quick look, so we realize that we must treat these spectra in order to have a better idea of the characteristics of the samples.

The spectra come with four constituent values: clay, sand, silt and Total Carbon.

The package come with another data set with spectra of a profile with factor variables for the Horizon ("H1", "H2", "H3"), Soil ("A", "B" , "C" and "D") and the reflectance spectra scanned every 10cm down to the profile (see the file description in the package).  

Before to work with the spectra in the coming posts, I show some videos to understand what a profile is, and how the colour can talk about the characteristics of the soil.

In the next video  professor Ray Weil study the colour in a profile.

Can we use spectroscopy to predict a colour (based on the spectrum) and to compare the predicted value with the original colour?. I think this a good exercise that the author of the soil package propose in the book "Soil Spectral Inference with R" which is the companion book for this package.

soilspec R package (installation)

 I want to test during the coming days the soilspec package. It is a package not available yet from CRAN, but you can get it from the author author`s GitHub . I use in this case the tar.gz file, and after I install from the folder where I download it.

Take care during the installation, because you can receive some warnings and errors telling you that you don´t have some dependencies, as in my case the "MatrixStats" and "clhs" packages. Once I have installed these packages, I have the soilspec package ready to use and in the coming posts we will work with it.

Veris® - Soil Mapping

It is interesting to be updated of the new infrared instruments applied to agricultural vehicles in order to measure soil parameters like Nitrogen, Organic Matter, pH and others to create a Soil Mapping and extract the maximum  benefits from the land, so in this aspect is  important to follow the Glosolan Spectroscopy Webinars where the experts will show in the presentation slides some of the companies which are in the front line in this matter.

GLOSOLAN Soil Spectroscopy Webinar #1 2024

GLOSOLAN continua con su labor de mantenernos informados en como las técnicas espectrales de Infrarrojo contribuyen al análisis del suelo. Durante este año 2024 se desarrollaran varios Webinarios muy interesantes, mostrando en este post el primero.

Podéis descargaros la presentación aquí.

Foro Romano y Palatino

He tenido la ocasión de estar en Roma este mes y la verdad es que me ha encantado la ciudad, y por supuesto la comida.

He tomado bastantes fotos y con las del foro Romano he preparado este corto video.