Soil Water Retention Characteristics of Black Soils of India and Pedotransfer Functions Using Different Approaches

Information on soil hydraulic properties is a prerequisite in irrigation management decisions and crop planning. Such information on soils of the black soil region (BSR) occupying $7.7\times {10}^7\mathrm{ha}$ of India is sparse. Soil profile information for 49 representative sites (244 samples) was collected and used for analysis. Ten different functions were evaluated for their efficacy to describe soil water retention characteristics (SWRC) of the BSR soils. Campbell model fitted to measured SWRC data with relatively lower root mean square error ($\mathrm{RMSE}=0.0214{\mathrm{m}}^3·{\mathrm{m}}^{-3}$), higher degree of agreement $d=0.9653$), and lower absolute error on average ($\mathrm{MAE}=0.0165{\mathrm{m}}^3·{\mathrm{m}}^{-3}$). The next best description was by van Genuchten (VG) function with RMSE ($0.0249{\mathrm{m}}^3·{\mathrm{m}}^{-3}$), $d(0.9489)$, and MAE ($0.0868{\mathrm{m}}^3·{\mathrm{m}}^{-3}$). Pedotransfer functions (PTF) were developed to predict field capacity (FC) and permanent wilting point (PWP) using nearest neighbor (kNN) algorithm and artificial neural networks (ANN). Four levels of input information used for point PTF development include (1) textural data (data on sand, silt, and clay fraction-SSC), (2) level 1+bulk density data (SSCBD), (3) level 2+organic matter (SSCBDOM), and (4) level 1+organic matter (SSCOM). The RMSE of predictions by kNN PTFs ranged from 0.0346 to $0.0611{\mathrm{m}}^3·{\mathrm{m}}^{-3}$ with an average of $0.0483{\mathrm{m}}^3·{\mathrm{m}}^{-3}$. The ANN PTFs performed with an average RMSE of $0.0550{\mathrm{m}}^3·{\mathrm{m}}^{-3}$ and a range of 0.0367 to $0.0905{\mathrm{m}}^3·{\mathrm{m}}^{-3}$. Relatively better estimates of $\mathrm{FC}/\mathrm{PWP}$ were obtained using SSCBD-based PTF. Accuracy of FC and PWP estimates obtained by using analytical functions was relatively greater than the estimates by kNN and ANN PTFs. Campbell and VG functions were relatively more accurate. The study demonstrated the efficacy of kNN technique vis-a-vis neural regression with the additional benefit of appending the development data as and when desired. The proposed PTFs could be useful in making irrigation management decisions for BSR soils of India. Identification of the most suitable SWRC function for the study soils will help in crop modeling/water balance studies of the region.