Calibrating and validating an agrohydrological model to simulate sugarbeet water use under mediterranean conditions

Abstract A calibration of the SWAP agrohydrological model was carried out under Mediterranean conditions using the data obtained from irrigation experiments conducted in Valladolid, Spain from 1992 to 1995. The temperature sums corresponding to each crop development stage were determined from the recorded data. The leaf area indices and root lengths were also estimated as functions of the sugarbeet development stage, using the data obtained from experiments. Furthermore, the Kc and Ky coefficients, i.e. those commonly used to relate the reference evapotranspiration to crop evapotranspiration and to relate the relative yields to the relative crop transpiration, were also estimated at different crop stages. The temperature sums for anthesis and maturity are much higher than those originally included in SWAP, determined for Northern Europe. However, the Mediterranean root lengths and, in particular, leaf area indices follow crop development stage patterns that are similar to those reported for sugarbeet in SWAP calibrations for Northern Europe. A non-linear sugarbeet root-depth distribution was assumed, since most of the roots are commonly found in top soil. The Kc coefficients as functions of the development stages were also changed. Furthermore, Ky coefficients during crop development were not considered as constants, as in the original SWAP. A maximum yield of 89 t/ha seems correct for the SWAP simulations of the highest relative yields. The calibrated parameters obtained were experimentally validated in the same region during the 2005 sugarbeet crop season, considering typical irrigation management. The simulated evapotranspirations follow the same temporal behaviour as that estimated by water-balance from soil water content measurements. The correlation coefficient between simulated and actual sugarbeet evapotranspirations was 0.75, which is statistically significant, as well as the regression line between the said variables. A t-test comparison of means and a Kolmogorov–Smirnov test for probability distribution comparison gave no significant statistical differences between the estimated water-balance and the SWAP-simulated evapotranspirations. Mean root square errors were relatively high, although this may be due to ignoring the drainage term while estimating ETP by water balance. The parameters obtained can be used in future SWAP assessments, particularly in those aimed at reducing water loss by percolation.