Quantitative estimation of soil salinity by means of different modeling methods and visible-near infrared (VIS–NIR) spectroscopy, Ebinur Lake Wetland, Northwest China

Wang, Jingzhe; Ding, Jianli; Abulimiti, Aerzuna; LiangHong, Cai

doi:10.7717/peerj.4703

Cited by 82 publications

(75 citation statements)

References 63 publications

(94 reference statements)

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Wang et al used the PLSR algorithm and the random forest (RF) algorithm to measure soil salinity. According to the validation accuracies, the RF models outperformed the PLSR models [34]. These studies were conducted in different data and different environments, and the results obtained cannot be put together for comprehensive evaluation of these models.…”

Section: Introductionmentioning

confidence: 99%

“…The regularization can reduce the influence of collinearity by adding offset to the optimization function [30]. With the deep research of machine learning methods, the results are uneven and many studies in which machine learning methods are used to estimate soil salinity have been reported [4,21,24,[32][33][34]. Jiang et al monitored soil salinity by integrating multiple biophysical indicators with support vector machine (SVM) and artificial neural network (ANN) regression algorithms.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Performance Comparison of Machine Learning Algorithms for Estimating the Soil Salinity of Salt-Affected Soil Using Field Spectral Data

et al. 2019

View full text Add to dashboard Cite

Salt-affected soil is a prominent ecological and environmental problem in dry farming areas throughout the world. China has nearly 9.9 million km2 of salt-affected land. The identification, monitoring, and utilization of soil salinization have become important research topics for promoting sustainable progress. In this paper, using field-measured spectral data and soil salinity parameter data, through analysis and transformation of spectral data, five machine learning models, namely, random forest regression (RFR), support vector regression (SVR), gradient-boosted regression tree (GBRT), multilayer perceptron regression (MLPR), and least angle regression (Lars) are compared. The following performance measures of each model were evaluated: the collinear problems, handling data noise, stability, and the accuracy. In terms of these four aspects, the performance of each model on estimating soil salinity is evaluated. The results demonstrate that among the five models, RFR has the best performance in dealing with collinearity, RFR and MLPR have the best performance in dealing with data noise, and the SVR model is the most stable. The Lars model has the highest accuracy, with a determination coefficient (R2) of 0.87, ratio of performance to deviation (RPD) of 2.67, root mean square error (RMSE) of 0.18, and mean absolute percentage error (MAPE) of 0.11. Then, the comprehensive comparison and analysis of the five models are carried out, and it is found that the comprehensive performance of RFR model is the best; hence, this method is most suitable for estimating soil salinity using hyperspectral data. This study can provide a reference for the selection of regression methods in subsequent studies on estimating soil salinity using hyperspectral data.

show abstract

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Performance Comparison of Machine Learning Algorithms for Estimating the Soil Salinity of Salt-Affected Soil Using Field Spectral Data

et al. 2019

View full text Add to dashboard Cite

show abstract

“…Random forest (RF) is a machine learning method widely used in the field of classification and regression in recent years [25]. Multiple soil properties have been linked to the Vis-NIR spectrum through this algorithm, such as soil organic carbon (SOC) [26], soil cadmium (Cd) [27], forage phosphorus (P) [28] and soil pH [29]. Numerous research results indicate that RF provides better prediction results than the classical partial least-squares regression (PLSR) [29][30][31].…”

Section: Introductionmentioning

confidence: 99%

“…Multiple soil properties have been linked to the Vis-NIR spectrum through this algorithm, such as soil organic carbon (SOC) [26], soil cadmium (Cd) [27], forage phosphorus (P) [28] and soil pH [29]. Numerous research results indicate that RF provides better prediction results than the classical partial least-squares regression (PLSR) [29][30][31]. Furthermore, compared to other machine learning technologies such as artificial neural networks (ANN) and support vector machine (SVM), the RF model can provide the relative importance of each predictor, which makes the results of the model highly interpretable [27,32].…”

Section: Introductionmentioning

confidence: 99%

Assessing Soil Organic Matter Content in a Coal Mining Area through Spectral Variables of Different Numbers of Dimensions

Zhu

Zhang

Wang

et al. 2020

Sensors

Self Cite

View full text Add to dashboard Cite

Soil organic matter (SOM) is a crucial indicator for evaluating soil quality and an important component of soil carbon pools, which play a vital role in terrestrial ecosystems. Rapid, non-destructive and accurate monitoring of SOM content is of great significance for the environmental management and ecological restoration of mining areas. Visible-near-infrared (Vis-NIR) spectroscopy has proven its applicability in estimating SOM over the years. In this study, 168 soil samples were collected from the Zhundong coal field of Xinjiang Province, Northwest China. The SOM content (g kg−1) was determined by the potassium dichromate external heating method and the soil reflectance spectra were measured by the spectrometer. Two spectral feature extraction strategies, namely, principal component analysis (PCA) and the optimal band combination algorithm, were introduced to choose spectral variables. Linear models and random forests (RF) were used for predictive models. The coefficient of determination (R2), root mean square error (RMSE), and the ratio of the performance to the interquartile distance (RPIQ) were used to evaluate the predictive performance of the model. The results indicated that the variables (2DI and 3DI) derived from the optimal band combination algorithm outperformed the PCA variables (1DV) regardless of whether linear or RF models were used. An inherent gap exists between 2DI and 3DI, and the performance of 2DI is significantly poorer than that of 3DI. The accuracy of the prediction model increases with the increasing number of spectral variable dimensions (in the following order: 1DV < 2DI < 3DI). This study proves that the 3DI is the first choice for the optimal band combination algorithm to derive sensitive parameters related to SOM in the coal mining area. Furthermore, the optimal band combination algorithm can be applied to hyperspectral or multispectral images and to convert the spectral response into image pixels, which may be helpful for a soil property spatial distribution map.

show abstract

“…However, the hyperspectral inversion models established for available potassium are mainly constructed based on 1-order or 2-order derivative for spectral reflectance, reciprocal, and logarithm. However, related research points out that traditional integer-order differential transformation ignores the gradual fractional differential information [18,19], especially for high-dimensional data sources such as hyperspectral images with massive information, which may cause some information to be lost or be difficult to extract, and restrict the modeling accuracy.Fractional calculus theory is a mathematical problem for studying the properties of differential and integral operators of any order and its application. First proposed in 1695, its development is almost in synchronization with the theory of integer-order calculus.…”

mentioning

confidence: 99%

Impact of Fractional Calculus on Correlation Coefficient between Available Potassium and Spectrum Data in Ground Hyperspectral and Landsat 8 Image

Gan

Yuan

et al. 2019

Mathematics

View full text Add to dashboard Cite

As the level of potassium can interfere with the normal circulation process of biosphere materials, the available potassium is an important index to measure the ability of soil to supply potassium to crops. There are rarely studies on the inversion of available potassium content using ground hyperspectral remote sensing and Landsat 8 multispectral satellite data. Pretreatment of saline soil field hyperspectral data based on fractional differential has rarely been reported, and the corresponding relationship between spectrum and available potassium content has not yet been reported. Because traditional integer-order differential preprocessing methods ignore important spectral information at fractional-order, it is easy to reduce the accuracy of inversion model. This paper explores spectral preprocessing effect based on Grünwald-Letnikov fractional differential (order interval is 0.2) between zero-order and second-order. Field spectra of saline soil were collected in Fukang City of Xinjiang. The maximum absolute of correlation coefficient between ground hyperspectral reflectance and available potassium content for five mathematical transformations appears in the fractional-order. We also studied the tendency of correlation coefficient under different fractional-order based on seven bands corresponding to the Landsat 8 image. We found that fractional derivative can significantly improve the correlation, and the maximum absolute of correlation coefficient under five spectral transformations is in Band 2, which is 0.715766 for the band at 467 nm. This study deeply mined the potential information of spectra and made up for the gap of fractional differential for field hyperspectral data, providing a new perspective for field hyperspectral technology to monitor the content of soil available potassium.Mathematics 2019, 7, 488 2 of 15 of convenience, speed, and high precision [8][9][10]. Visible, near-infrared and mid-infrared spectroscopy technologies have been widely applied in soil science.In recent years, domestic and foreign scholars have conducted extensive research on soil salinity, moisture, organic matter, and total nitrogen in different types of ecosystems, such as wetlands, forests, grasslands, and farmlands in arid and semi-arid regions [11][12][13][14][15]. There is less research on available potassium content [16]. Liu et al. [17] adopted visible/short-wave near-infrared spectroscopy to measure soil available nitrogen and available potassium. They introduced first-order differential algorithm for spectral pretreatment, and their simulation showed that the model built by least squares support vector machine (LS-SVM) combined with 1-order differential has higher precision. However, the hyperspectral inversion models established for available potassium are mainly constructed based on 1-order or 2-order derivative for spectral reflectance, reciprocal, and logarithm. However, related research points out that traditional integer-order differential transformation ignores the gradual fractional differential informat...

show abstract

Quantitative estimation of soil salinity by means of different modeling methods and visible-near infrared (VIS–NIR) spectroscopy, Ebinur Lake Wetland, Northwest China

Cited by 82 publications

References 63 publications

Performance Comparison of Machine Learning Algorithms for Estimating the Soil Salinity of Salt-Affected Soil Using Field Spectral Data

Performance Comparison of Machine Learning Algorithms for Estimating the Soil Salinity of Salt-Affected Soil Using Field Spectral Data

Assessing Soil Organic Matter Content in a Coal Mining Area through Spectral Variables of Different Numbers of Dimensions

Impact of Fractional Calculus on Correlation Coefficient between Available Potassium and Spectrum Data in Ground Hyperspectral and Landsat 8 Image

Contact Info

Product

Resources

About