Development of pedotransfer functions for predicting saturated hydraulic conductivity in a Himalayan catchment: Sikkim, India

Saturated hydraulic conductivity (Ks) plays a vital role in irrigation and drainage system design. Generally, Ks is estimated in the laboratory; however, it is expensive and tedious, especially in the Himalayan ranges where soil sampling is challenging due to topographical constraints. Therefore, in this study, pedotransfer functions were generated using multiple linear regression (MLR) models for the predictability of Ks in a Himalayan catchment in India. Fifty soil samples were collected and divided into two groups at a 70:30 ratio. Different soil attributes derived from 70% of samples were used for MLR generation, and attributes of the remaining 30% of samples were used for model validation. Six different MLR models constituting different independent soil attributes were generated and compared statistically. The results indicate that the MLR model comprising soil texture, bulk density, particle density, soil moisture content (MC), organic carbon content and porosity results in the highest adjusted coefficient of determination (R2; 0.93 and 0.89 during model generation and validation, respectively). Additionally, it was found that the weight basis MC ranged from 14% to 29% with a median value of 24%. These results demonstrate that simple MLR models can be used as an alternative to laborious experimental setups for Ks estimation. These findings can be used as guidelines for proper irrigation planning and design in mountainous catchments. La conductivite hydraulique saturee (Ks) joue un role essentiel dans la conception des systemes d'irrigation et de drainage. Generalement, le Ks est estime en laboratoire; Cependant, il est couteux et fastidieux, en particulier dans les chaines de l'himalaya ou l'echantillonnage du sol est difficile en raison de contraintes topographiques. Par consequent, dans cette etude, des fonctions de pedotransfert (PTF) ont ete generees a l'aide de modeles de regression lineaire multiple (MLR) pour la previsibilite du Ks dans un bassin versant de l'himalaya en Inde. Cinquante echantillons de sol ont ete preleves et repartis en deux groupes selon un rapport de 70:30. Differents attributs du sol provenant de 70% des echantillons ont ete utilises pour la production de MLP, et les attributs des 30% restants ont ete utilises pour la validation des modeles. Six differents modeles MLR constituant differents attributs independants du sol ont ete generes et compares statistiquement. Les resultats indiquent que le modele MLP comprenant la texture du sol, la densite apparente (BD), la densite des particules (PD), la teneur en humidite du sol (MC), la teneur en carbone organique (OC) et la porosite donne le R2 ajuste le plus eleve (0,93 et 0,89 au cours de la generation et de la validation du modele, respectivement). De plus, il a ete constate que la teneur en humidite du sol (MC) sur base ponderale variait entre 14% et 29% avec une valeur mediane de 24%. Ces resultats demontrent que des modeles MLR simples peuvent etre utilises comme une alternative a des configurations experimentales laborieuses pour l'estimation du Ks. Ces resultats peuvent servir de lignes directrices pour une planification et une conception adequates de l'irrigation dans les bassins versants montagneux.