Contents lists available at ScienceDirect Soil & Tillage Research journal homepage: www.elsevier.com/locate/still Multiple AI model integration strategy—Application to saturated hydraulic conductivity prediction from easily available soil properties Mahsa H. Kashani a, *, Mohammad Ali Ghorbani b,c , Mahmood Shahabi d , Sujay Raghavendra Naganna e , Lamine Diop f a Department of Water Engineering, Faculty of Agriculture and Natural Resources, University of Ardabil, Ardabil, Iran b Department of Water Engineering, Faculty of Agriculture, University of Tabriz, Tabriz, Iran c Department of Civil Engineering, Near East University, P.O. Box: 99138, Nicosia, North Cyprus, Mersin 10, Turkey d Department of Soil Science, Faculty of Agriculture, University of Ardabil, Ardabil, Iran e Department of Civil Engineering, Shri Madhwa Vadiraja Institute of Technology and Management, Bantakal - 574115, Udupi. Affiliated to Visvesvaraya Technological University, Belagavi, 590018, India f Université Gaston Berger, Senegal ARTICLE INFO Keywords: Saturated hydraulic conductivity Extreme learning machine Multiple model strategy Multivariate adaptive regression splines M5Tree Support vector machine Prediction ABSTRACT A multiple model integration scheme driven by artificial neural network (ANN) (MM-ANN) was developed and tested to improve the prediction accuracy of soil hydraulic conductivity (Ks) in Tabriz plain, an arid region of Iran. The soil parameters such as silt, clay, organic matter (OM), bulk density (BD), pH and electrical con- ductivity (EC) were used as model inputs to predict soil Ks. Standalone models including multivariate adaptive regression splines (MARS), M5 model tree (M5Tree), support vector machine (SVM) and extreme learning machine (ELM) were also implemented for comparative evaluation with MM-ANN model predictions. Based on several performance indicators such as Nash SutcliffeEfficiency (NSE), results showed that the calibrated MM- ANN model involving the predictions of MARS, M5Tree, SVM and ELM models by considering all the soil parameters used in this study as inputs provided superior soil Ks estimates. The proposed hybrid model (MM- ANN) emerged as a reliable intelligence model for the assessment of soil hydraulic conductivity with an NSE = 0.939 & 0.917 during training and testing, respectively. Accurate prediction of field-scale soil hydraulic conductivity is crucial from the view point of agricultural sustainability and management prospects. 1. Introduction The saturated hydraulic conductivity (Ks) is the infiltration rate into soil at steady state condition (Naganna et al., 2017). The Ks often in- fluences the amount of surface runoff, groundwater recharge and con- taminant plume within soil mass, thereby affecting the stream water quality. Additionally, it is necessary for estimating the primary con- solidation settlement of a foundation, drainage filter design, and as- sessing seepage loss from earthen dams (Naganna et al., 2017). The precise estimates of Ks aid to simulate hydrologically-driven processes, such as groundwater flow and solute transport in both saturated and unsaturated media. Furthermore, the specific influence of Ks con- tributes to analyze the soil water dynamics of lateral and vertical drainage flow (Frind and Pinder, 1973). The soil Ks can be measured through field and laboratory experiments (Fetter, 2000). According to Bagarello et al. (2000), the constant head and falling head permeameters and the grain size analysis are the commonly used la- boratory methods, and the pressure and the tension infiltrometers are the in-situ methods for estimating soil Ks. For in-situ or field measurements, Wu et al. (1999) proposed a single ring infiltrometer method for estimating the saturated hydraulic con- ductivity of top-soil layers. Bagarello et al. (2009) applied the two- ponding-depth infiltration method for a sandy-loam soil and obtained more reasonable Ks measurements than the single-ring infiltrometer through various infiltration tests. Aiello et al. (2014) applied the Beerkan Estimation of Soil Transfer parameters (BEST) procedure and compared it with the method of Wu et al. (1999). Reynolds et al. (2000) conducted field infiltration measurements and laboratory soil core methods to estimate Ks of undisturbed soils. Through laboratory mea- surements, Ks models were validated using statistical methods in- corporating regression lines fitted to different soil data. Comegna et al. (2000) found that there was a strong correlation between the estimated https://doi.org/10.1016/j.still.2019.104449 Received 31 January 2019; Received in revised form 5 August 2019; Accepted 8 October 2019 ⁎ Corresponding author. E-mail address: m.hkashani@uma.ac.ir (M. H. Kashani). Soil & Tillage Research 196 (2020) 104449 0167-1987/ © 2019 Elsevier B.V. All rights reserved. T