Environmental risk analysis of farming systems in a semi-arid environment: effect of rotations and management practices on deep drainage C.G.H. Dı ´az-Ambrona a,b , G.J. O’Leary a,c, * , V.O. Sadras d , M.G. O’Connell e , D.J. Connor a a Joint Centre for Crop Innovation, School of Agriculture and Food Systems, The University of Melbourne, Vic., Australia b Universidad Polite ´cnica de Madrid, Avda Complutense, s/n 28040 Madrid, Spain c CSIRO Land and Water, Mallee Research Station, Walpeup, Vic. 3507, Australia d CSIRO Land and Water, APSRU, PMB 2, Glen Osmond, SA 5064, Australia e Department of Primary Industries, Tatura, Vic. 3616, Australia Received 18 June 2004; received in revised form 19 January 2005; accepted 22 January 2005 Abstract We investigated the impact of crop rotations and management practices on the water balance of farming systems in a semi- arid region of south-eastern Australia, where drainage beyond the root zone and rising water tables contribute to salinisation of soils and water streams. The CropSyst model was locally tested and used with long-term weather data to analyse the risk of deep drainage associated with alternative farming practices; water balances were estimated using both cascade and finite difference approaches. Model estimates were compared with (a) yield, biomass, phenology, soil water content and water use measured in a 5-year experiment involving two rotations, i.e. FWP (fallow-wheat–pea) and MWP (mustard–wheat–pea) and (b) yield measured in grower-managed wheat crops encompassing three seasons and broad ranges of weather, rotations, management practices and soils. In the rotation experiment, the simulations of yield were close to measured yield for wheat (observed range from 0 to 2.4 t ha 1 ; R 2 = 0.78; RMSE = 0.42 t ha 1 ), mustard (range from 0 to 0.6 t ha 1 ; R 2 = 0.57; RMSE = 0.20 t ha 1 ) and field pea (range from 0 to 1.4 t ha 1 ; R 2 = 0.72; RMSE = 0.29 t ha 1 ). Simulated wheat yield also compared well with on- farm measurements (observed range from 0 to 3.5 t ha 1 ; R 2 = 0.72, RMSE = 0.21 t ha 1 ). The model provided reasonable estimates of crop biomass, phenological development, soil water content and water use. Water use tended to be overestimated in some cases using the finite difference approach, but soil water content was well simulated in the presence of crop residues. The use of a finite-difference approach highlights the need to account for upward movement of water in the soil and its consequences on deep drainage. Median annual drainage simulated at 1.5 m was +9.1 mm for the fallow-wheat rotation, 5.4 mm for fallow-wheat–pea, 8.4 mm for mustard–wheat–pea, and +0.6 mm for wheat–wheat. The 3-year rotations (FWP and MWP) had the lowest median drainage and lowest probability of large drainage events in wet years. In the fallow-wheat rotation, median annual drainage was 24.8 mm with stubble retention and zero tillage, compared to 2.8 mm for stubble burnt and conventional tillage. Whilst the www.elsevier.com/locate/fcr Field Crops Research 94 (2005) 257–271 * Corresponding author. Tel.: +61 3 5091 7200; fax: +61 3 5091 7210. E-mail address: garry.o’leary@csiro.au (G.J. O’Leary). 0378-4290/$ – see front matter. Crown Copyright # 2005 Published by Elsevier B.V. All rights reserved. doi:10.1016/j.fcr.2005.01.008