Details

Autor(en) / Beteiligte

• Saha, Ajoy Kumar
• Hasan, Md. Moudud
• Khan, Md. Redwan Ahmad
• Al-Amin, Md

Titel

Prediction of spatial saturated hydraulic conductivity at the upper soil layer using soil class and terrain attributes

Ist Teil von

• Modeling earth systems and environment, 2022-09, Vol.8 (3), p.3591-3605

Ort / Verlag

Cham: Springer International Publishing

Erscheinungsjahr

2022

Quelle

Alma/SFX Local Collection

Beschreibungen/Notizen

• The knowledge of soil hydraulic properties is indispensable to solve many soil and water management problems related to agriculture, ecology, and environmental issues in water. This research aimed to generate the saturated hydraulic conductivity ( K s ) information of topsoil from soil class and terrain attributes (elevation, roughness, slope, terrain ruggedness index (TRI), topographic position index (TPI), and flow direction) using multiple linear regression (MLR) and support vector regression (SVR) techniques to develop a spatial distribution map of K s for an agricultural landscape. Sixty-five cores of soil sample (diameter 5 cm and length 10 cm) were collected from the topsoil layer (0–10 cm) from the agricultural field in different locations (Upazilas) of the Sylhet region (3452 km 2 ) in Bangladesh and conducted laboratory tests following Darcy’s constant head method. The topsoil was clay or silt clay, having very low K s values. The mean of K s for the agricultural soils is 1.70 × 10 –06  cm/s and varied significantly ( p  < 0.01) among different Upazilas (sub-district). To generate the topsoil layer’s K s values, the developed MLR model (with R 2  = 0.598 and RMSE = 1.12 × 10 –06  cm/s) and SVR model (with training R train 2 = 0.64 , RMSE = 1.22 × 10 –06  cm/s and NSE = 0.58 and testing R test 2 = 0.73 , RMSE = 8.19 × 10 –07  cm/s, NSE = 0.71) were found suitable compared to simple interpolation methods. The SVR model performed better in the modeling process than the MLR based on the goodness of fits parameter. However, the SVR model underestimates the higher K s values in both training and testing stages. In contrast, the MLR model was found to be more balanced. Finally, the spatial variability map of K s for the topsoil layer can be generated from soil texture information and terrain attributes for facilitating agro-hydrological model development in a data-limited area.