Spatial Modeling of Soil Behavior the Integration of Geographical System and Advanced Mathematical Analysis
Al-Noor Journal of Engineering Management and Computer Science · 2026
DOI: 10.71229/4v311a92 ↗Kurz gefasst
Fachartikel in der Zeitschrift Al-Noor Journal of Engineering Management and Computer Science, 2026. Verfasst von Abeer Sabri Bshara und 1 weiteren Autor:innen. Die Arbeit nennt 21 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-SA 4.0).
Abstract (Originalsprache)
The land is covered with soil, except somewhere of the ice or bare rock, and its present varies in their continuously. Accurate determination of soil properties plays a vital role in the safe and inexpensive construction of various structures. This work presented an integrated Geostatistic-Artificial Neural Network (ANN) approach to characterise and predict the behaviour of the soil as a proof of concept target variable, taking electrical conductivity (EC) in a virtual landscape containing four classes of soil (N = 100) as the target variable. The four hypothetical soil classes were: (Saline-Gypsiferous, Silty Clay Loam Alluvial, Sandy Loam and Calcareous) simulated in a spatially 100 x 100 grid domain, with using physicochemical parameters taken from literature. The spatial dependency of EC was tested using experimental semivariogram analysis and Moran's' I, and an interpolated salinity surface was created using a distance-weighted estimator; EC spatial data coordinates and eight physicochemical data descriptors were used for training a Multi-Layer Perceptron with two hidden layers (16 and 8 neurons respectively). The geostatistical analysis showed high spatial structure with a very low nugget to sill ratio (6.4%), a high positive Moran's I value (0.278), and a range of 169.82 grid units which gave a smooth looking, physically sound interpolated surface that correctly identified and plotted the high salinity zone. The overall predictability of the ANN was high (R-squared = 0.68) but low (R-squared = 0.214) for the held-out test set with class-stratified analysis indicating that the only significant contribution to predictability came from the ability of the ANN to separate soil classes rather than from the ability to express within-soil class variance. This reflects that the limited sample size compared to the complexity of this part of the ANN was a major limiting constraint on the generalizations. These results suggest that geostatistical and machine-learning techniques are complementary and not mutually exclusive: variogram-based techniques provide valuable and powerful, “data-efficient” spatial description, but ANNs have the potential to include multivariate covariates if sufficient training data are present. The extension of this pipeline to larger sizes or field measured datasets, complete kriging and hybrid regression-kriging implementations is recommended for future research to enable operational soil salinity mapping in real arid-semi-arid alluvial environments.
Zitieren
APA
Bshara, A. S., & Mohammed, F. A. (2026). Spatial Modeling of Soil Behavior the Integration of Geographical System and Advanced Mathematical Analysis. Al-Noor Journal of Engineering Management and Computer Science. https://doi.org/10.71229/4v311a92
Vancouver
Bshara AS, Mohammed FA. Spatial Modeling of Soil Behavior the Integration of Geographical System and Advanced Mathematical Analysis. Al-Noor Journal of Engineering Management and Computer Science. 2026. doi:10.71229/4v311a92
BibTeX
@article{Bshara2026,
author = {Bshara, Abeer Sabri and Mohammed, Fatimah Abdulrazzaq},
title = {Spatial Modeling of Soil Behavior the Integration of Geographical System and Advanced Mathematical Analysis},
journal = {Al-Noor Journal of Engineering Management and Computer Science},
year = {2026},
doi = {10.71229/4v311a92}
}
EUVIMED