A framework for quantifying water extraction and water stress responses of perennial lucerne Hamish E. Brown A,C , Derrick J. Moot B , Andrew L. Fletcher A , and Peter D. Jamieson A A New Zealand Institute for Plant & Food Research Limited, Private Bag 4704, Christchurch, New Zealand. B Faculty of Agriculture and Life Sciences, PO Box 7647, Lincoln University, Canterbury, New Zealand. C Corresponding author. Email: brownH@crop.cri.nz Abstract. A generic framework was developed and validated for predicting the water extraction and water stress responses of perennial lucerne (Medicago sativa) to improve existing crop models. Perennial forages have roots established throughout a soil profile so require a different approach to quantify water extraction patterns than annual crops. Two years of experimental data from two fields in New Zealand, each containing dryland and irrigated lucerne crops, were analysed to develop the theory of the water extraction framework. This showed that the temporal pattern of water extraction was consistent and each year commenced in the shallowest layer and progressed downward. Water extraction from each soil layer was quantified as the minimum of soil water supply and crop demand for that layer. For each soil layer, water demand was represented by transpiration demand (the product of potential evapotranspiration and crop cover) minus the sum of water extraction in overlying layers. This approach gave accurate descriptions of water extraction patterns over a range of rainfall and irrigation situations. Water supply from each soil layer (l) was quantified as the product of plant-available water and an extraction rate constant (kl l ). The kl l of lucerne could not be calculated using the traditional curve-fitting procedure so kl l was calculated by integrating the water extraction framework described above with a soil water balance and fitting kl l to minimise residuals for water extraction predictions in each soil layer. This gave kl l values that decreased from 0.035/day in the 0–0.2 m layer of soil to 0.01/day in the deepest layer measured (1.8–2.3 m). The water extraction framework was validated against another 3 years of dryland and irrigated lucerne data and gave accurate predictions of water extraction patterns throughout the soil profile. Water stress was quantified from actual transpiration relative to transpiration demand (T/T D ). The most sensitive variable was leaf area expansion, which decreased from an optimum at T/T D = 1 to zero at T/T D = 0.2, followed by radiation-use efficiency, which decreased from an optimum at T/T D = 1 to zero at a T/T D of zero. The framework for quantifying water extraction and the techniques determined for identifying appropriate parameters to measure and characterise the framework are expected to be generally applicable to perennial forages in a wide range of environments. Additional keywords: alfalfa (syn. lucerne), leaf area index, mechanistic simulation, radiation-use efficiency, soil water content, transpiration, water extraction depth. Introduction An accurate estimation of crop water extraction and its effect on dry matter (DM) yield is a principal outcome of most crop models. A robust quantification of the effect of environment on water extraction and yield is needed to achieve this. Soil water extraction (WE) by the roots is the consequence of transpiration (T ) and can be defined as the minimum of supply (T S ) from the crop’s roots or demand (T D ) from the atmosphere (Monteith 1986): WE ¼ T ¼ minðT S ; T D Þ ð1Þ T D is normally calculated as the product of crop cover (I/I o ) and potential evapotranspiration (P ET ) (French and Legg 1979). T S is the maximum potential water extraction that individual soil profile layers can supply (W Sl ) summed to the extraction depth (ED) of the crop (Monteith 1986): T S ¼ X l ¼ ED l ¼ 1 W Sl ð2Þ where l is the soil layer under consideration, l = 1 is the top soil layer, and higher numbers are sequentially deeper layers. Passioura (1983) developed an approach for quantifying W Sl as a function of the plant-available water (PAW l ) in each layer (Eqn 3), which has been widely used to simulate water extraction for given crop–soil combinations (Chapman et al. 1993; Hochman et al. 2001; Dardanelli et al. 2004): W Sl ¼ PAW l  kl l ð3Þ where kl l is the proportion of PAW l that can be extracted each day. The kl l is normally quantified by fitting an exponential curve to the decline in PAW l over time (Meinke et al. 1993; Robertson et al. 1993a; Dardanelli et al. 2004). However, the Ó CSIRO 2009 10.1071/CP08415 1836-0947/09/080785 CSIRO PUBLISHING www.publish.csiro.au/journals/cp Crop & Pasture Science, 2009, 60, 785–794