Evaluation of DRAINMOD using saturated hydraulic conductivity estimated by a pedotransfer function model

Direct measurement of soil saturated hydraulic conductivity (K-s) is time-consuming and therefore costly. The ROSETTA pedotransfer function model is able to estimate K-s from soil textural data, bulk density and one or two water retention points. This study evaluated the feasibility of running the DRAINMOD field-scale hydrological model with K-s input produced using ROSETTA. A hierarchical approach was adopted to estimate K-s using ROSETTA, with four limited-more extended sets of soil information used as inputs: USDA textural class (H1); texture (H2); texture and bulk density (H3); texture, bulk density, water retention at -33 kPa (theta(33 kPa)) and -1500 kPa (theta(1500 kPa)) (H4). ROSETTA-estimated K-s values from these four groups (H1-H4) were used in DRAINMOD to simulate drain outflows during a 4-year period from a conventional drainage plot (CD) and two controlled drainage plots (CWT1 and CWT2) located in south-east Sweden. The DRAINMOD results using ROSETTA-estimated K-s values were compared with observed values and with model results using laboratory-measured K-s values (H0). Deviations in simulated drainage outflow (D), infiltration (F) and evapotranspiration (ET) resulting from the use of ROSETTA-estimated rather than laboratory-measured K-s values were evaluated. During the study period, statistical comparisons showed good agreement on a monthly basis between observed and DRAINMOD-simulated drainage rates using five soil datasets (H0, H1, H2, H3 and H4). The monthly mean absolute error (MAE) ranged from 0.57 to 0.82 cm for CD, 0.38 to 0.41 cm for CWT1, and 0.15 to 0.22 cm for CWT2. On a monthly basis, the modified coefficient efficiency (E') values were in the range of 0.62 to 0.74 for CD, 0.72 to 0.74 for CWT1, and 0.79 to 0.86 for CWT2. The modified index of agreement (d') for monthly predictions ranged from 0.80 to 0.86 cm for CD, 0.87 to 0.88 cm for CWT1, and 0.89 to 0.93 cm for CWT2. The absolute values of the percent-normalised error (NE) on an overall basis when using ROSETTA-estimated rather than laboratory-measured K-s values were less than 3% in E, less than 1% in F, and less than 15% in D. The results suggest that ROSETTA-estimated K-s values can be used in DRAINMOD to simulate drainage outflows as accurately as laboratory-measured K-s values (H0) in coarse-textured soils. (C) 2008 Elsevier B.V. All rights reserved.