Evaluation of Machine Learning Approaches to Predict Soil Organic Matter and pH Using vis-NIR Spectra

Yang, Mingyong; Xu, Danfeng; Chen, Songchao; Li, Hongyi; Shi, Zhou

doi:10.3390/s19020263

Cited by 104 publications

(55 citation statements)

References 43 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To represent the spatial characteristics of soil properties in complex geographical landscapes, a purposive sampling method was applied in this study [31]. Based on the pedogenesis of SOC and pH of the study area, elevation, MAT, MAP, and NDVI were the main environmental variables considered in collecting samples [2,6,14,23]. The specific location of each sampling point was recorded by a handheld global positioning system (GPS).…”

Section: Soil Sampling and Analysismentioning

confidence: 99%

“…Soil organic carbon (SOC) and pH are important soil properties [1]. They are important indicators to measure soil fertility and soil environmental quality [2,3]. Their changes directly affect the whole soil environment and farmland production [4].…”

Section: Introductionmentioning

confidence: 99%

“…Previous studies have mainly focused on acidic soils. Some researchers have found that carbon mineralization in acidic soils increases after lime is used, and decreases after soil acidification [2,6]. However, the effect of lime on carbon mineralization lasts for a short time.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Spatial-Temporal Changes in Soil Organic Carbon and pH in the Liaoning Province of China: A Modeling Analysis Based on Observational Data

Wang

Zhuang

et al. 2019

Sustainability

View full text Add to dashboard Cite

Quantification of soil organic carbon (SOC) and pH, and their spatial variations at regional scales, is a foundation to adequately assess agriculture, pollution control, or environmental health and ecosystem functioning, so as to establish better practices for land use and land management. In this study, we used the random forest (RF) model to map the distribution of SOC and pH in the topsoil (0–20 cm) and estimate SOC and pH changes from 1982 to 2012 in Liaoning Province, Northeast China. A total of 10 covariates (elevation, slope gradient, topographic wetness index (TWI), mean annual temperature (MAT), mean annual precipitation (MAP), visible-red band 3 (B3), near-infrared band 4 (B4), short-wave infrared band 5 (B5), normalized difference vegetation index (NDVI), and land-use data) and a set of 806 (in 1982) and 973 (in 2012) soil samples were selected. Cross-validation technology was used to test the performance and uncertainty of the RF model. We found that the prediction R2 of SOC and pH was 0.69 and 0.54 for 1982, and 0.63 and 0.48 for 2012, respectively. Elevation, NDVI, and land use are the main environmental variables affecting the spatial variability of SOC in both periods. Correspondingly, the topographic wetness index and mean annual precipitation were the two most critical environmental variables affecting the spatial variation of pH. The mean SOC and pH decreased from 18.6 to 16.9 kg−1 and 6.9 to 6.6, respectively, over a 30-year period. SOC distribution generated using the RF model showed a decreasing SOC trend from east to west across the city in the two periods. In contrast, the spatial distribution of pH showed an opposite trend in both periods. This study provided important information of spatial variations in SOC and pH to agencies and communities in this region, to evaluate soil quality and make decisions on remediation and prevention of soil acidification and salinization.

show abstract

Section: Soil Sampling and Analysismentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Spatial-Temporal Changes in Soil Organic Carbon and pH in the Liaoning Province of China: A Modeling Analysis Based on Observational Data

Wang

Zhuang

et al. 2019

Sustainability

View full text Add to dashboard Cite

show abstract

“…These components are constructed so that they account for most of the variance in the measured spectral data (X) and the SOC content (Y ), and at the same time they maximize the correlation between X and Y . In other words, PLSR leads to the covariance between X and Y being maximized (Bjørsvik and Martens, 2008;Wehrens, 2011). In order to receive a robust model, it is important not to include too many components in model building as this will lead to overfitting (Hastie et al, 2009;Kuhn and Johnson, 2013).…”

Section: Model Building and Validationmentioning

confidence: 99%

Error propagation in spectrometric functions of soil organic carbon

Ellinger

Merbach

Werban

et al. 2019

SOIL

View full text Add to dashboard Cite

Abstract. Soil organic carbon (SOC) plays a major role concerning chemical, physical, and biological soil properties and functions. To get a better understanding of how soil management affects the SOC content, the precise monitoring of SOC on long-term field experiments (LTFEs) is needed. Visible and near-infrared (Vis–NIR) reflectance spectrometry provides an inexpensive and fast opportunity to complement conventional SOC analysis and has often been used to predict SOC. For this study, 100 soil samples were collected at an LTFE in central Germany by two different sampling designs. SOC values ranged between 1.5 % and 2.9 %. Regression models were built using partial least square regression (PLSR). In order to build robust models, a nested repeated 5-fold group cross-validation (CV) approach was used, which comprised model tuning and evaluation. Various aspects that influence the obtained error measure were analysed and discussed. Four pre-processing methods were compared in order to extract information regarding SOC from the spectra. Finally, the best model performance which did not consider error propagation corresponded to a mean RMSEMV of 0.12 % SOC (R2=0.86). This model performance was impaired by ΔRMSEMV=0.04 % SOC while considering input data uncertainties (ΔR2=0.09), and by ΔRMSEMV=0.12 % SOC (ΔR2=0.17) considering an inappropriate pre-processing. The effect of the sampling design amounted to a ΔRMSEMV of 0.02 % SOC (ΔR2=0.05). Overall, we emphasize the necessity of transparent and precise documentation of the measurement protocol, the model building, and validation procedure in order to assess model performance in a comprehensive way and allow for a comparison between publications. The consideration of uncertainty propagation is essential when applying Vis–NIR spectrometry for soil monitoring.

show abstract

“…In recent years, the development of machine learning algorithms also allowed their application for the prediction of element concentration in soil by the use of hyperspectral imaging (400-2500 nm) [21][22][23]. Compared with support vector machine regression (SVMR), the random forest (RF) model is a more effective machine learning method for developing diagnosis models [23].…”

Section: Introductionmentioning

confidence: 99%

Development of Visible/Near-Infrared Hyperspectral Imaging for the Prediction of Total Arsenic Concentration in Soil

Wei

Zhang

et al. 2020

Applied Sciences

View full text Add to dashboard Cite

Soil total arsenic (TAs) contamination caused by human activities—such as mining, smelting, and agriculture—is a problem of global concern. Visible/near-infrared (VNIR), X-ray fluorescence spectroscopy (XRF), and laser-induced breakdown spectroscopy (LIBS) do not need too much sample preparation and utilization of chemicals to evaluate total arsenic (TAs) concentration in soil. VNIR with hyperspectral imaging has the potential to predict TAs concentration in soil. In this study, 59 soil samples were collected from the Daye City mining area of China, and hyperspectral imaging of the soil samples was undertaken using a visible/near-infrared hyperspectral imaging system (wavelength range 470–900 nm). Spectral preprocessing included standard normal variate (SNV) transformation, multivariate scatter correction (MSC), first derivative (FD) preprocessing, and second derivative (SD) preprocessing. Characteristic bands were then identified based on Spearman’s rank correlation coefficients. Four regression models were used for the modeling prediction: partial least squares regression (PLSR) (R2 = 0.71, RMSE = 0.48), support vector machine regression (SVMR) (R2 = 0.78, RMSE = 0.42), random forest (RF) (R2 = 0.78, RMSE = 0.42), and extremely randomized trees regression (ETR) (R2 = 0.81, RMSE = 0.38). The prediction results were compared with the results of atomic fluorescence spectrometry methods. In the prediction results of the models, the accuracy of ETR using FD preprocessing was the highest. The results confirmed that hyperspectral imaging combined with Spearman’s rank correlation with machine learning models can be used to estimate soil TAs content.

show abstract

Evaluation of Machine Learning Approaches to Predict Soil Organic Matter and pH Using vis-NIR Spectra

Cited by 104 publications

References 43 publications

Spatial-Temporal Changes in Soil Organic Carbon and pH in the Liaoning Province of China: A Modeling Analysis Based on Observational Data

Spatial-Temporal Changes in Soil Organic Carbon and pH in the Liaoning Province of China: A Modeling Analysis Based on Observational Data

Error propagation in spectrometric functions of soil organic carbon

Development of Visible/Near-Infrared Hyperspectral Imaging for the Prediction of Total Arsenic Concentration in Soil

Contact Info

Product

Resources

About