A genetic algorithm‐based stacking algorithm for predicting soil organic matter from vis–NIR spectral data

Spectral data obtained from optical spaceborne sensors are being recognized as a valuable source of data that show promising results in assessing soil properties on medium and macro scale. Combining this technique with laboratory Visible-Near Infrared (VIS-NIR) spectroscopy methods can be an effective approach to perform robust research on plot scale to determine wildfire impact on soil organic matter (SOM) immediately after the fire. Therefore, the objective of this study was to assess the ability of Sentinel-2 superspectral data in estimating post-fire SOM content and comparison with the results acquired with laboratory VIS-NIR spectroscopy.The study is performed in Mediterranean Croatia (44&#176; 05&#8217; N; 15&#176; 22&#8217; E; 72 m a.s.l.), on approximately 15 ha of fire affected mixed Quercus ssp. and Juniperus ssp. forest on Cambisols. A total of 80 soil samples (0-5 cm depth) were collected and geolocated on August 22nd 2019, two days after a medium to high severity wildfire. The samples were taken to the laboratory where soil organic carbon (SOC) content was determined via dry combustion method with a CHNS analyzer. SOM was subsequently calculated by using a conversion factor of 1.724. Laboratory soil spectral measurements were carried out using a portable spectroradiometer (350-1050 nm) on all collected soil samples. Two Sentinel-2 images were downloaded from ESAs Scientific Open Access Hub according to the closest dates of field sampling, namely August 31st and September 5th 2019, each containing eight VIS-NIR and two SWIR (Short-Wave Infrared) bands which were extracted from bare soil pixels using SNAP software. Partial least squares regression (PLSR) model based on the pre-processed spectral data was used for SOM estimation on both datasets. Spectral reflectance data were used as predictors and SOM content was used as a response variable. The accuracy of the models was determined via Root Mean Squared Error of Prediction (RMSEp) and Ratio of Performance to Deviation (RPD) after full cross-validation of the calibration datasets.The average post-fire SOM content was 9.63%, ranging from 5.46% minimum to 23.89% maximum. Models obtained from both datasets showed low RMSEp (Spectroscopy dataset RMSEp = 1.91; Sentinel-2 dataset RMSEp = 0.99). RPD values indicated very good predictions for both datasets (Spectrospcopy dataset RPD = 2.72; Sentinel-2 dataset RPD = 2.22). Laboratory spectroscopy method with higher spectral resolution provided more accurate results. Nonetheless, spaceborne method also showed promising results in the analysis and monitoring of SOM in post-burn period.Keywords: remote sensing, soil spectroscopy, wildfires, soil organic matterAcknowledgment: This work was supported by the Croatian Science Foundation through the project "Soil erosion and degradation in Croatia" (UIP-2017-05-7834) (SEDCRO). Aleksandra Per&#269;in is acknowledged for her cooperation during the laboratory work.

Download Full-text

Determination of Soil Organic Matter and Carbon Fractions in Forest Top Soils using Spectral Data Acquired from Visible-Near Infrared Hyperspectral Images

Soil Science Society of America Journal ◽

10.2136/sssaj2011.0053 ◽

2012 ◽

Vol 76 (2) ◽

pp. 586-596 ◽

Cited By ~ 15

Author(s):

S. M. O'Rourke ◽

N. M. Holden

Keyword(s):

Organic Matter ◽

Soil Organic Matter ◽

Spectral Data ◽

Near Infrared ◽

Hyperspectral Images ◽

Carbon Fractions

Download Full-text

Hyperspectral Estimation of Soil Organic Matter Content using Different Spectral Preprocessing Techniques and PLSR Method

Remote Sensing ◽

10.3390/rs12071206 ◽

2020 ◽

Vol 12 (7) ◽

pp. 1206

Author(s):

Lanzhi Shen ◽

Maofang Gao ◽

Jingwen Yan ◽

Zhao-Liang Li ◽

Pei Leng ◽

...

Keyword(s):

Organic Matter ◽

Soil Organic Matter ◽

Dimensionality Reduction ◽

Spectral Data ◽

Wavelet Packet ◽

Estimation Model ◽

Data Transformations ◽

First Derivative ◽

Model Based ◽

Reduction Methods

Soil organic matter (SOM) is the main source of soil nutrients, which are essential for the growth and development of agricultural crops. Hyperspectral remote sensing is one of the most efficient ways of estimating the SOM content. Visible, near infrared, and mid-infrared reflectance spectroscopy, combined with the partial least squares regression (PLSR) method is considered to be an effective way of determining soil properties. In this study, we used 54 different spectral pretreatments to preprocess soil spectral data. These spectral pretreatments were composed of three denoising methods, six data transformations, and three dimensionality reduction methods. The three denoising methods included no denoising (ND), Savitzky–Golay denoising (SGD), and wavelet packet denoising (WPD). The six data transformations included original spectral data, R; reciprocal, 1/R; logarithmic, log(R); reciprocal logarithmic, log(1/R); first derivative, R’; and first derivative of reciprocal, (1/R)’. The three dimensionality reduction methods included no dimensionality reduction (NDR), sensitive waveband dimensionality reduction (SWDR), and principal component analysis (PCA) dimensionality reduction (PCADR). The processed spectra were then employed to construct PLSR models for predicting the SOM content. The main results were as follows—(1) the wavelet packet denoising (WPD)-R’ and WPD-(1/R)’ data showed stronger correlations with the SOM content. Furthermore, these methods could effectively limit the correlation between the adjacent bands and, thus, prevent “overfitting”. (2) Of the 54 pretreatments investigated, WPD-(1/R)’-PCADR yielded the model with the highest accuracy and stability. (3) For the same denoising method and spectral transformation data, the accuracy of the SOM content estimation model based on SWDR was higher than that of the model based on NDR. Furthermore, the accuracy in the case of PCADR was higher than that for SWDR. (4) Dimensionality reduction was effective in preventing data overfitting. (5) The quality of the spectral data could be improved and the accuracy of the SOM content estimation model could be enhanced effectively, by using some appropriate preprocessing methods (one combining WPD and PCADR in this study).

Download Full-text