Prediction of spatial soil property information from ancillary sensor data using ordinary linear regression: Model derivations, residual assumptions and model validation tests

Lesch, Scott M.; Corwin, Dennis L.

doi:10.1016/j.geoderma.2008.09.014

Cited by 49 publications

(47 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Interpolated maps of HMs spatial distribution were generated using Surfer software version 9. In order to check the validity of performed estimations, the jack-knifing technique, the common technique of assessing the predictive capability of different regression models (Lesch and Corwin 2008), was used. Consequently, interpolated and actual (observed) values were compared using the error measurement of standardized root-meansquare error (RMSE %), which is calculated as follows (Hengl et al 2004):…”

Section: Geostatistical Studiesmentioning

confidence: 99%

Spatial distribution of heavy metals in soils and plants in Zinc Town, northwest Iran

Delavar

Safari

2015

Int. J. Environ. Sci. Technol.

View full text Add to dashboard Cite

Identification of heavy metals spatial variability in soil and plants may provide useful information on how to manage the polluted sites. The main objective of this study was to determine the spatial distribution of selected heavy metals in soils and natural plants of Zanjan city, northwest Iran. A total of 184 composite topsoil samples (0-10 cm) and 98 natural plant samples were systematically taken from an area of about 4000 ha located around an industrial complex, covering rangeland and agricultural and industrial land uses. All samples were analyzed for their total concentration of Zn, Pb and Cd. The results showed that the average concentrations of Zn, Pb and Cd in the soil samples were up to 294.2, 152.8 and 5.6 mg kg -1 , respectively, whereas in the plant samples, these values decreased to 131.4, 113.2 and 2.5 mg kg -1 , respectively. These contents are much higher than the normal range in soil and plant communities, leading to classify the studied area as a polluted site. Variography analyses revealed a similar spatial structure for the studied heavy metals in the soil and plant samples. Based on interpolated maps, the highest concentrations of the selected heavy metals in the soil and plant samples were found in the vicinity of industrial complex. These findings clearly highlight the role of industrial activities in simplifying the entrance of dangerous trace elements to the human food chain. Application of ordinary kriging technique to predict the heavy metals spatial variability in the plant community resulted in logical estimations with acceptable error values.

show abstract

Section: Geostatistical Studiesmentioning

confidence: 99%

Spatial distribution of heavy metals in soils and plants in Zinc Town, northwest Iran

Delavar

Safari

2015

Int. J. Environ. Sci. Technol.

View full text Add to dashboard Cite

show abstract

“…where the expectation and variance of the test statistic were computed using the formulas given in Lesch and Corwin (2008). Additionally, the normality assumption was assessed using standard residual quantilequantile plots and the Shapiro-Wilk test (Myers, 1986;Shapiro and Wilk, 1965).…”

Section: Model Specifi Cation and Assumptionsmentioning

confidence: 99%

Electromagnetic Induction Sensor Data to Identify Areas of Manure Accumulation on a Feedlot Surface

Woodbury

Lesch

Eigenberg

et al. 2009

Soil Science Soc of Amer J

View full text Add to dashboard Cite

A study was initiated to test the validity of using electromagnetic induction (EMI) survey data, a prediction‐based sampling strategy, and ordinary linear regression modeling to predict spatially variable feedlot surface manure accumulation. A 30‐ by 60‐m feedlot pen with a central mound was selected for this study. A Dualem‐1S EMI meter (Dualem Inc., Milton, ON, Canada) pulled on 2‐m spacing was used to collect feedlot surface apparent electrical conductivity (ECa) data. Meter data were combined with global positioning system coordinates at a rate of five readings per second. Two 20‐site sampling approaches were used to determine the validity of using EMI data for prediction‐based sampling. Soil samples were analyzed for volatile solids (VS), total N (TN), total P (TP), and Cl− A stratified random sampling (SRS) approach (n = 20) was used as an independent set to test models estimated from the prediction‐based (n = 20) response surface sample design (RSSD). The RSSD sampling plan demonstrated better design optimality criteria than the SRS approach. Excellent correlations between the EMI data and the ln(Cl−), TN, TP, and VS soil properties suggest that it can be used to map spatially variable manure accumulations. Each model was capable of explaining >90% of the constituent sample variations. Fitted models were used to estimate average manure accumulation and predict spatial variations. The corresponding prediction maps show a pronounced pen design effect on manure accumulation. This technique enables researchers to develop precision practices to mitigate environmental contamination from beef feedlots.

show abstract

“…The selection of the model was based on the MLM approach. The statistical and geostatistical models listed in Table 2.4 are special cases derived from the MLM following a set of assumptions (Lesch and Corwin, 2008). The MLM formulation is (Hengl, 2009):…”

Section: Statistical Modelingmentioning

confidence: 99%

“…They have used and sometimes compared several statistical methods such as simple linear regression (LR), multiple linear regression (MLR), variance analysis (ANOVA), covariance analysis (ANOCOVA) and kriging techniques, such as ordinary kriging (OK), cokriging, universal kriging (UK) and kriging with external drift (KED). These statistical and geostatistical techniques are derived from the general universal model of spatial variation as defined by Matheron (as cited in Hengl, 2009), which is also called the geostatistical mixed linear model (MLM) by Lesch and Corwin (2008). For any given dataset, the selection of an adequate prediction model derived from the MLM approach cannot be exclusively based on the model performance and goodness of fit.…”

Section: Introductionmentioning

confidence: 99%

“…The selection must be primarily made by the verification of the model assumptions. In this study we applied the guidelines of Lesch and Corwin (2008), which explained the connection between the LR model and the MLM, reviewed the underlying assumptions and presented statistical diagnostic tools. We also followed the methodology of Hengl (2009) who developed a decision chart, based on the sequential verification of the model assumptions, to guide in selecting an appropriate prediction model.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Integration of hydrogeophysics and remote sensing with coupled hydrological models

Francés¹

View full text Add to dashboard Cite

Prediction of spatial soil property information from ancillary sensor data using ordinary linear regression: Model derivations, residual assumptions and model validation tests

Cited by 49 publications

References 30 publications

Spatial distribution of heavy metals in soils and plants in Zinc Town, northwest Iran

Spatial distribution of heavy metals in soils and plants in Zinc Town, northwest Iran

Electromagnetic Induction Sensor Data to Identify Areas of Manure Accumulation on a Feedlot Surface

Integration of hydrogeophysics and remote sensing with coupled hydrological models

Contact Info

Product

Resources

About