The manifestation of VIS-NIRS spectroscopy data to predict and mapping soil texture in the Triffa plain (Morocco)‎

الملخص EN

The use of standard laboratory methods to estimate the soil texture is complicated, expensive, time-consuming, and need considerable effort.

The reflectance spectroscopy represents an alternative method for predicting a large range of soil physical properties and provide an inexpensive, rapid, and reproducible analytical method.

This study aimed to assess the feasibility of Visible (VIS: 350-700 nm) and Near-Infrared and Short-Wave-Infrared (NIRS: 701-2500 nm) spectroscopy for predicting and mapping the clay, silt, and sand fractions of the soil of Triffa plain (northeast of Morocco).

A total of 100 soil samples were collected from the non-root zone of soil (0-20cm) and then analyzed for texture using the VIS-NIRS spectroscopy and the traditional laboratory method.

The partial least squares regression (PLSR) technique was used to assess the ability of spectral data to predict soil texture.

The results of prediction models showed excellent performance for the VIS-NIRS spectroscopy to predict the sand fraction with a coefficient of determination R2=0.93 and Root Mean Squares Error (RMSE)=3.72 and good prediction for the silt fraction (R2=0.87; RMSE= 4.55), and acceptable prediction for the clay fraction (R2= 0.53; RMSE=3.72).

Moreover, the range situated between 2150-2450 nm is the most significant for predicted the sand and silt fractions, while the spectral range between 2200 and 2440 nm is optimal to predict the clay fraction.

However, the maps of predicted and measured soil texture showed an excellent spatial similarity for the sand fraction, a certain difference in the variability of clay fraction, while the maps of silt fraction show a lower difference.