Developing and comparing pedotransfer functions and spectral transfer functions for predicting water retention in some soils of Kurdistan province

The soil water retention curve (SWRC), as an important hydraulic soil property, is used in modeling water flow and solute transport in the unsaturated zone of the soil. Direct measurements of SWRC are difficult, time-consuming and costly. Hence, researchers have proposed indirect methods such as pedotransfer functions to estimate soil water retention curve using readily available soil data. Over the last decades, soil spectral data as a rapid, low cost, and non-destructive method has been widely applied to estimate basic soil properties. Consequently, in this study, the feasibility of using soil spectral information in the visible and near-infrared region, as input variables for transfer functions, and evaluation its performance was investigated compared to basic soil properties in estimating soil water retention curve.

A number of 100 soil samples were collected and their spectral reflectance over 350-2500 nm region were measured using a handheld spectroradiometer apparatus. Some basic soil properties such as particle size distribution, particle density, bulk density, organic carbon content and calcium carbonate equivalent, and soil moisture content at matric potentials of -10, -33, -50, -100, -300, -500, -1000, and -1500 kPa were also determined with pressure plate - membrane apparatus. Spectral reflectance curves of the samples were recorded using RS3 software on a portable computer connected to a spectroradiometer with 5 readings per soil sample. After spectral preprocessing, the correlation coefficient between absorption features of soil in each wavelength with soil moisture content at different matric potentials were investigated. Stepwise multiple linear regression was applied to derive pedo-transfer functions (PTFs) and spectral transfer functions (STFs) that uses basic soil properties and soil spectral reflectance as input, respectively. The accuracy of the proposed functions were assessed by adjusted coefficient of determination (R2adj), normalized root mean square error (NRMSE), mean error (ME), and the ratio of performance to deviation (RPD).

Pedo-transfer functions (PTFs) provided more accurate estimates at the dry-end of the soil moisture curve than the wet-end, due to the high correlation of soil moisture with soil particle size distribution at the dry-end of the soil moisture curve. The results of the statistical parameters showed that the derived PTFs for estimating soil water retention at 10 to 1500 kPa matric suctions have good prediction accuracy. However, STFs also had reasonable but poorer results than the proposed PTFs in estimating the studied characteristics.

Overall, the results of this study revealed that, despite the relatively poorer results of STFs than PTF, due to lower costs, time and field data, soil spectral data can be used as an indirect and novel method for estimating volumetric soil moisture content at different matric potentials.