Spatial Prediction of Soil Organic Matter Using a Hybrid Geostatistical Model of an Extreme Learning Machine and Ordinary Kriging

Song, Yingqiang; Lian-an, Yang; Li, Bo; Hu, Yueming; Wang, An-Le; Zhou, Weizheng; Cui, Xuesen; Liu, Yi-Lun

doi:10.3390/su9050754

Cited by 40 publications

(27 citation statements)

References 40 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…These methods tried to combine the advantages of both worlds, deterministic and stochastic, achieving improved results. Currently, more innovative implementations of the abovementioned hybrid methods are increasingly used; they introduce machine learning (ML) as the deterministic part, along with kriging of the ML residuals as the stochastic part [8][9][10][11][12]. These methods are widely used in environmental sciences mainly for two reasons: their improved prediction accuracy and the elimination of most of the restrictions (e.g., statistical assumptions) that regression, kriging, and their variations (RK, KED, etc.)…”

Section: Introductionmentioning

confidence: 99%

Comparing Machine Learning Models and Hybrid Geostatistical Methods Using Environmental and Soil Covariates for Soil pH Prediction

Tziachris¹,

Aschonitis²,

Chatzistathis³

et al. 2020

IJGI

View full text Add to dashboard Cite

In the current paper we assess different machine learning (ML) models and hybrid geostatistical methods in the prediction of soil pH using digital elevation model derivates (environmental covariates) and co-located soil parameters (soil covariates). The study was located in the area of Grevena, Greece, where 266 disturbed soil samples were collected from randomly selected locations and analyzed in the laboratory of the Soil and Water Resources Institute. The different models that were assessed were random forests (RF), random forests kriging (RFK), gradient boosting (GB), gradient boosting kriging (GBK), neural networks (NN), and neural networks kriging (NNK) and finally, multiple linear regression (MLR), ordinary kriging (OK), and regression kriging (RK) that although they are not ML models, they were used for comparison reasons. Both the GB and RF models presented the best results in the study, with NN a close second. The introduction of OK to the ML models’ residuals did not have a major impact. Classical geostatistical or hybrid geostatistical methods without ML (OK, MLR, and RK) exhibited worse prediction accuracy compared to the models that included ML. Furthermore, different implementations (methods and packages) of the same ML models were also assessed. Regarding RF and GB, the different implementations that were applied (ranger-ranger, randomForest-rf, xgboost-xgbTree, xgboost-xgbDART) led to similar results, whereas in NN, the differences between the implementations used (nnet-nnet and nnet-avNNet) were more distinct. Finally, ML models tuned through a random search optimization method were compared with the same ML models with their default values. The results showed that the predictions were improved by the optimization process only where the ML algorithms demanded a large number of hyperparameters that needed tuning and there was a significant difference between the default values and the optimized ones, like in the case of GB and NN, but not in RF. In general, the current study concluded that although RF and GB presented approximately the same prediction accuracy, RF had more consistent results, regardless of different packages, different hyperparameter selection methods, or even the inclusion of OK in the ML models’ residuals.

show abstract

Section: Introductionmentioning

confidence: 99%

Comparing Machine Learning Models and Hybrid Geostatistical Methods Using Environmental and Soil Covariates for Soil pH Prediction

Tziachris¹,

Aschonitis²,

Chatzistathis³

et al. 2020

IJGI

View full text Add to dashboard Cite

show abstract

“…However, additionally considering the spatial autocorrelation of SOC residuals, ANNK apparently outperformed SVR, in spite of a close performance of ANN and SVR. Obviously, two-step hybrid approaches performed better than the corresponding one-step models, which was also revealed by comparisons made in Kumar (2015), Zeng et al (2016) and Guo et al (2017) about GWRK and GWR [48,86,87], and in Dai et al (2014) and Song et al (2017) about ANNK and ANN [72,88]. Dai et al (2014) used ANNK and ANN for mapping soil organic matter content with auxiliary variables, including climate data, relief data and spectral indices of the remote sensing data, and obtained a higher accuracy than that of this study [88].…”

Section: Comparison Among Geostatistical ML and Hybrid Techniquesmentioning

confidence: 58%

“…The validation set (n = 132) was used to test and compare the performances of OK, GWR, SVR, ANN, GWRK and ANNK models based on the root mean squared error (RMSE, Equation (8)), mean error (Equation (9)) and coefficient of determination (R 2 , Equation (10)) [66,72].…”

Section: Model Testing and Comparisonmentioning

confidence: 99%

A Comparative Assessment of Geostatistical, Machine Learning, and Hybrid Approaches for Mapping Topsoil Organic Carbon Content

Chen

Ren

et al. 2019

IJGI

View full text Add to dashboard Cite

Accurate digital soil mapping (DSM) of soil organic carbon (SOC) is still a challenging subject because of its spatial variability and dependency. This study is aimed at comparing six typical methods in three types of DSM techniques for SOC mapping in an area surrounding Changchun in Northeast China. The methods include ordinary kriging (OK) and geographically weighted regression (GWR) from geostatistics, support vector machines for regression (SVR) and artificial neural networks (ANN) from machine learning, and geographically weighted regression kriging (GWRK) and artificial neural networks kriging (ANNK) from hybrid approaches. The hybrid approaches, in particular, integrated the GWR from geostatistics and ANN from machine learning with the estimation of residuals by ordinary kriging, respectively. Environmental variables, including soil properties, climatic, topographic, and remote sensing data, were used for modeling. The mapping results of SOC content from different models were validated by independent testing data based on values of the mean error, root mean squared error and coefficient of determination. The prediction maps depicted spatial variation and patterns of SOC content of the study area. The results showed the accuracy ranking of the compared methods in decreasing order was ANNK, SVR, ANN, GWRK, OK, and GWR. Two-step hybrid approaches performed better than the corresponding individual models, and non-linear models performed better than the linear models. When considering the uncertainty and efficiency, ML and two-step approach are more suitable than geostatistics in regional landscapes with the high heterogeneity. The study concludes that ANNK is a promising approach for mapping SOC content at a local scale.

show abstract

“…Although the IDW and OK spatial interpolation algorithms fully consider the spatial distance relationship, the accuracy of the RF algorithm is slightly higher than IDW and OK when considering the ME, RMSE, and r. Compared with SVM regression, the accuracy of the RF algorithm is much higher. Theoretically, OK and IDW consider the spatial autocorrelation of soil samples, whereas RF considers the spatial heterogeneity of environmental variables [45]. In other words, when the validation data are closer to the measured data in the soil sampling locations, the performance of OK and IDW is likely better than that of RF.…”

Section: Comparison Of the Spatial Prediction Methodsmentioning

confidence: 99%

“…However, the spatial differences in wetlands SOCc are large and easily affected by environmental variables [5,6]. Therefore, interpolation based only on the autocorrelation of the sample data will result in slightly larger errors [45].…”

Section: Comparison Of the Spatial Prediction Methodsmentioning

confidence: 99%

Investigating Spatial and Vertical Patterns of Wetland Soil Organic Carbon Concentrations in China’s Western Songnen Plain by Comparing Different Algorithms

Ren

Mao

et al. 2020

Sustainability

View full text Add to dashboard Cite

Investigating the spatial and vertical patterns of wetland soil organic carbon concentration (SOCc) is important for understanding the regional carbon cycle and managing the wetland ecosystem. By integrating 160 wetland soil profile samples and environmental variables from climatic, topographical, and remote sensing data, we spatially predicted the SOCc of wetlands in China’s Western Songnen Plain by comparing four algorithms: random forest (RF), support vector machine (SVM) for regression, inverse distance weighted (IDW), and ordinary kriging (OK). The predicted results of the SOCc from the different algorithms were validated against independent testing samples according to the mean error, root mean squared error, and correlation coefficient. The results show that the measured SOCc values at depths of 0–30, 30–60, and 60–100 cm were 15.28, 7.57, and 5.22 g·kg−1, respectively. An assessment revealed that the RF algorithm was most accurate for predicting SOCc; its correlation coefficients at the different depths were 0.82, 0.59, and 0.51, respectively. The attribute importance from the RF indicates that environmental variables have various effects on the SOCc at different depths. The land surface temperature and land surface water index had a stronger influence on the spatial distribution of SOCc at the depths of 0–30 and 30–60 cm, whereas topographic factors, such as altitude, had a stronger influence within 60–100 cm. The predicted SOCc of each vertical depth increased gradually from south to north in the study area. This research provides an important case study for predicting SOCc, including selecting factors and algorithms, and helps with understanding the carbon cycles of regional wetlands.

show abstract

Spatial Prediction of Soil Organic Matter Using a Hybrid Geostatistical Model of an Extreme Learning Machine and Ordinary Kriging

Cited by 40 publications

References 40 publications

Comparing Machine Learning Models and Hybrid Geostatistical Methods Using Environmental and Soil Covariates for Soil pH Prediction

Comparing Machine Learning Models and Hybrid Geostatistical Methods Using Environmental and Soil Covariates for Soil pH Prediction

A Comparative Assessment of Geostatistical, Machine Learning, and Hybrid Approaches for Mapping Topsoil Organic Carbon Content

Investigating Spatial and Vertical Patterns of Wetland Soil Organic Carbon Concentrations in China’s Western Songnen Plain by Comparing Different Algorithms

Contact Info

Product

Resources

About