Estimation of The Ability of Hyper Spectral Images In Estimation of The Extent of Soil Contamination By Lead

Any change in the properties of the constituents of the soil is known to change the quality of the soil. Pollutants, including heavy metals, arsenic, cadmium, nickel, lead, chemical pesticides and herbicides, insecticides, fungicides, fertilizers, waste, sewage, and oil might damage the soil. Lead continuous distribution and the cumulative result of industrialization and urbanization are rapidly causing serious problems for the environment and food security (Zhang et al., 2010). Previous studies show heavy metals in the visible and near infrared region have a specific spectrum and feature that can be used to identify and trace elements (Kemper and Sommer, 2003; Choe et al., 2008; Ji et al., 2010 ). Hyperspectral sensors in a continuous, very narrow and several spectrum bands, range in the visible, near infrared, mid-infrared and thermal infrared imaging track. Spectroscopic method with a lower cost is faster and can be used on a wider level. The purpose of this research is to identify the spectral changes associated with heavy metals in soil using spectroscopy, following the determination of distribution of heavy metals using hyper spectral images.
Study area: The study area is located in the south-western city of Khorramabad. According to soil moisture and temperature regimes of the Iran map, is soil moisture regime is xeric and its thermal regime is thermic. A total of 17 samples up to 30 cm depth were taken in areas where there was the risk of further contamination.
In this study, Hyperion data was used in the processing levels LIR.
Laboratory Lead levels were measured using atomic absorption. The lead levels were measured using spectroscopy, which meant using spectrophotometric spectral reflectance of each sample in the range of 200 to 2400 nm being measured in every 10 nm.
Discussion of Results reflected from the sample surface using a spectrometer in the range of 200 to 2500 nm at 10 nm (Figure 1) are observed:The basic assumption in the detection of heavy metals is that soil reflectance decreases with increasing amount of them (Pandit et al, 2010). Graph spectral samples 1, 2, 9 and 10 are also visually compatible with that the assumption and sample 1 with the highest lead levels had lower reflectance and sample No. 10 with the lowest lead is the highest reflection.
Partial least squares regression: In this research lead levels were identified directly using the method of least squares either (using spectra) and indirectly using spectroscopy and other soil elements. In this method estimating the amount of lead was performed in the four different ways: 1 – The lead estimated using spectrum in the range of 200 to 2100 nm 2 – estimation using the clay content 3 – estimation using the amount of organic matter and spectrum 4- estimation using the amount of clay, organic matter and spectrum. Principal component analysis was applied because more than 95% of the variance between the data had raised the estimated clay content using partial least squares models.
Verification: Methods used in these models included a sample for each time which was removed for estimation accuracy and the estimation of accuracy amount. Results of overall accuracy verification and also the calibration show that combined with clay models with the lowest mean square error are the best estimates. Lead Index According to this research, with regard to the relationship between reflection and organic matter, lead levels were calculated using hyper-spectral images. Therefore,according to the first derivative of spectrum with soil properties, the first derivative of spectrum in all areas first was calculated according to the following equation: Eq.1 In this regard, d the first derivative spectra in the wavelength ,  Wavelength,   The distance between the range 1  i  and 1  i  the wavelength is. The correlation between organic matter and lead is a function of the following equation: Pb  0.458OC  0.7243 Eq.2 The coefficient of determination obtained from the mean square error obtained using respectively 0.579 square mean error remaining 1.517 is.
The results of this study indicate that a combination of visible and near-infrared spectral range can be used to detect heavy metals. Interestingly, we found a high lead correlation with the amount of clay and organic matter. High effect of wavelengths 500 to 600 nm in estimating lead was achieved due to the existence of organic material. Therefore, it seems that the estimation of heavy metals in areas with clay that is the ability to have a lot of storing organic matter prove higher accuracy. The use of indexes or relationships to estimate the amount of lead using hyper-spectral images can be successful when the relationship between the elements with other soil properties is identified. The results of the models PLSR, MTMF and Lead indicators suggest that the techniques employed have been able to identify the lead content effects in contamination, except that the outputs are different for each for number, classification and images that can be used according to the purpose.