**1. Introduction**

In Australia, the wheat industry is challenged by complex genotype x environment x managemen<sup>t</sup> (GxExM) interactions [1,2], due in part to the high spatial and temporal variability of the Australian climate (e.g., [3]). In the eastern part of the continent, annual variations in temperature and rainfall that are influenced by El Niño–Southern Oscillation (ENSO) [1,4,5] affect frost, heat, and drought stress patterns [5–7], and ultimately, wheat production [1,3,5]. Drought and warmer temperatures, but also greater frost risk due to the clear night sky, are generally associated with the onset of El Niño episodes [5,8,9], and limit grain yield [10–13]. Stronger ENSO climate oscillations are expected in the near future, as climate forecasts project more frequent extreme El-Niño and La-Niña conditions [14,15].

As the major driver of inter-annual climate variability in Eastern Australian [4,5,16], ENSO is a quasiperiodic climate pattern that occurs across the tropical Pacific Ocean every 3–8 years. It is caused by variations in the surface temperature of the tropical eastern Pacific Ocean, and the air surface pressure in the tropical western Pacific [17]. The Southern Oscillation Index (SOI), as measured by surface pressure anomaly difference between Tahiti and Darwin, has been used to investigate ENSO effects on crops. Five SOI phases have been defined through grouping all sequential two-month pairs of the SOI into five clusters, using principal component analysis and a cluster analysis [18]. Hammer et al. [1] found that using the 5-phase SOI classification (based on SOI values for the current and previous month) could significantly increase wheat profits (up to 20%) and decrease failure risk (up to 35% less risk) in Goondiwindi, South-Eastern Queensland, Australia, through adapting wheat cultivars and nitrogen fertiliser.

Strategies for yield improvement include breeding new cultivars and adapting managemen<sup>t</sup> practices to the target population of environments [19]. Climate forecasting offers new opportunities in terms of agricultural planning and operation [4]. In the Australian broad-acre dryland wheat production area, most major decisions occur prior to sowing. Producers can potentially react to early indicators of upcoming rainfall and temperature. Early estimation of SOI phases can thus help farmers adjust managemen<sup>t</sup> practices such as which cultivar to sow, when to sow, and what nitrogen fertilisation to apply [5,20,21].

In Eastern Australia, wheat crops rely heavily on soil-stored plant available water (PAW) [6,22]. An appropriate combination of sowing data, variety maturity, and pre-sowing PAW is crucial to allow flowering and grain filling to occur with minimal stress, in particular frost, heat, and drought stress, and thus, to maximise yield potential [6,7,23–25]. In this context, crop modelling can assist farmers to adapt their practices to specific SOI phases through adequate choice of maturity type and sowing date, in order to ge<sup>t</sup> extra benefit and increased profit [26].

The aims of this paper were to determine the values of (i) fixed adaptation (no distinction between the years) and (ii) adaptations to specific pre-sowing plant available water (PAW) and/or SOI phase. In this study, adaptation strategies were defined in terms of sowing, maturity type, and nitrogen fertilisation, to target the greatest long-term productivity at each site. The APSIM crop model [27], together with a phenology model [28], frost impact module [12] and heat impact module [10], were used to predict flowering time and yield of wheat, and search for the best long-term adaptation strategies.

## **2. Materials and Methods**
