Ali. M. Al-Turki Hesham M. Ibrahim Prediction of water content at different potentials from soil property data in Jazan region Ali. M. Al-Turki Hesham M. Ibrahim Department of Soil Science King Saud University This research was funded by the National Plan for Science, Technology and Innovation (MAARIFAH), King Abdulaziz City for Science and Technology, Kingdom of Saudi Arabia, Award Number 12-AGR2575-02

Introduction In dry regions, effective irrigation management is crucial to maintain crop production and sustain limited water resources. Effective irrigation requires good knowledge of soil water content in the root zone. Measurements of SWRC are required to determine water availability to plants and to simulate the transport of water and solutes in the soil environment Direct measurement of SWRC is laborious, time consuming, and expensive Pedotransfer functions (PTFs) are widely used to determine SWRC from basic soil properties Multiple linear regression (MLR) and artificial neural networks (ANN) are the two most common methods used to develop PTFs

Jazan region Jazan region is about 13500 km2 (about 0.6% of the total area of KSA), and its population is approximately 1400000 (about 5% of the total population of KSA). The region has two main watersheds: Baysh and Jazan

Study Location Watershed delineation was carried out in WMS 9.3 Jazan Watershed delineation was carried out in WMS 9.3 using 10 m DEM Jazan watershed: Total area 612.5 km2 Average temp 21-40 oC Average rainfall 230 mm Based on a 10 m DEM for the region, the delineation for the Jazan watershed was carried out using the Watershed Modeling System package. The main channel and streams were checked against a topographic map for the region to ensure high accuracy for the delineation process.

Mountains in the East

Jazan Dam

Doum Palm trees

Wadi Jazan, central region low flooding

Wadi Jazan, west region low flooding

Plain fields

Sorghum fields

Mango farm

Typical soil profiles West Middle East Soil profiles F101: Loamy sand, light yellowish brown color, very week structure, medium drainage. F114: Loamy sand, dark grayish brown color, week structure, good drainage. F110: very gravel loamy sand, dark yellowish brown color, massive structure, good drainage.

Landforms and soil texture distribution Wadi and alluvial plain are the main dominant landforms, with small areas of sabakhat and costal plain in the west, and lava and alluvial fan in the east. The dominant soil texture is sand and loamy sand, with very small areas of clay and silt loam textures.

Objectives To predict water content at variable potentials (0, -10, -33, -60, -100, -300, -500, -800, -1000, and -1500 kPa) using three hierarchical approaches based on the Rosetta model: Soil texture class (STC) Percent of sand, silt, and clay (SSC) Bulk density, percent of sand, silt, and clay, and water content at -33 and -1500 kPa (SSC+WC) To determine the ability of the three approaches to predict available water content (AWC)

Sampling locations Outlet Jazan watershed: 43 measurement locations

Measurements and analysis The following soil properties were measured in the collected soil samples: Particle size distribution (PSD) (Sand, 2-0.05; silt, 0.05-0.002; clay, <0.002 mm) Bulk density (BD) Total Calcium carbonate (CaCO3) Saturation percent (SP) Electrical conductivity (EC) Organic carbon (OC)

Measurements and analysis In each soil sample, the water content at matric potentials of -10, -33, -60, -100, -300, -500, -800, -1000, and -1500 kPa was determined using pressure plate extractor. The SWRC was determined by fitting the retention data to the equation of van Genuchten (1980): Where is the volumetric water content (cm3 cm-3), and are the residual and saturated water content (cm3 cm-3), (cm-1) and n (-) are shape parameters of the SWRC

Measurements and analysis The following statistical indices were used to evaluate the accuracy between measured and predicted WC: Root mean square error (RMSE): Mean relative error (MRE):

Measurements and analysis D-index: Nash-Sutcliffe coefficient of efficiency (NSCE): Where and are the measured and predicted values is the average measured value, n is the total observations

Results SWRC The van Genuchten equation adequately described the SWRC with an average R2 of 0.95. The best fit was obtained with the third approach

Results Root mean square error Model Water potential (kPa)   -10 -33 -60 -100 -300 -500 -800 -1000 -1500 RMSE STC 0.133 0.103 0.073 0.062 0.055 0.043 0.042 SSC 0.129 0.075 0.063 0.041 0.040 SSC+WC 0.119 0.053 0.039 0.032 0.026 0.021 0.024 0.025 The largest (0.119-0.133 cm3 cm-3) RMSE was observed close to saturation The third approach showed lower RMSE at all potentials

Results Mean relative error Model Water potential (kPa)   -10 -33 -60 -100 -300 -500 -800 -1000 -1500 MRE STC 12.52 6.61 4.33 3.57 3.12 1.86 1.45 1.11 0.89 SSC 12.24 7.24 4.89 4.04 3.51 2.13 1.68 1.31 1.29 1.06 SSC+WC 10.89 2.42 0.64 0.56 0.61 -0.07 -0.38 -0.69 -0.91 The first and second approaches always overestimated WC The third approach provided lower MRE, and showed slight underestimation of WC at water potentials <=300 kPa

Results D-index Smaller D-index was observed close to saturation Model Water potential (kPa)   -10 -33 -60 -100 -300 -500 -800 -1000 -1500 D-index STC 0.355 0.685 0.780 0.773 0.750 0.722 0.683 0.662 0.651 0.625 SSC 0.369 0.673 0.777 0.763 0.751 0.719 0.698 0.686 0.654 SSC+WC 0.362 0.802 0.893 0.911 0.927 0.931 0.913 0.895 0.889 0.875 Smaller D-index was observed close to saturation Larger D-index with the third approach

Results NSCE NSCE values are negative at water potentials >33 kPa Larger NSCE with the third approach

Results Available water content Maximum prediction accuracy with the third approach (R2=0.8)

Conclusions The van Genuchten equation adequately described the SWRC in the Jazan region with an average R2 of 0.95 The three approaches failed to describe water content accurately at saturation conditions (>-10kPa) The third approach gave the best prediction of WC as indicated by an average NSCE value of 0.75 as compared to 0.16 and 0.18 for the first and second approaches, respectively The ability to predict the amount of available water in the soil profile will facilitate the accurate estimate of irrigation requirements and achieve effective irrigation scheduling

Thank You!