• breeding;
  • drought;
  • environment characterization;
  • mega-environment;
  • modelling;
  • water deficit;
  • wheat


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information
  • Plant response to drought is complex, so that traits adapted to a specific drought type can confer disadvantage in another drought type. Understanding which type(s) of drought to target is of prime importance for crop improvement.
  • Modelling was used to quantify seasonal drought patterns for a check variety across the Australian wheatbelt, using 123 yr of weather data for representative locations and managements. Two other genotypes were used to simulate the impact of maturity on drought pattern.
  • Four major environment types summarized the variability in drought pattern over time and space. Severe stress beginning before flowering was common (44% of occurrences), with (24%) or without (20%) relief during grain filling. High variability occurred from year to year, differing with geographical region. With few exceptions, all four environment types occurred in most seasons, for each location, management system and genotype.
  • Applications of such environment characterization are proposed to assist breeding and research to focus on germplasm, traits and genes of interest for target environments. The method was applied at a continental scale to highly variable environments and could be extended to other crops, to other drought-prone regions around the world, and to quantify potential changes in drought patterns under future climates.


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

While growth in population and urban/industrial water demands are both beginning to limit water supply for agricultural production, improving crop yield remains a key strategy globally to meet projected demand in developed and developing countries (Borlaug & Dowswell, 2005). To complicate the task, recent progress in plant improvement has been impeded by climatic changes in crops such as wheat (e.g. Brisson et al., 2010; Richards et al., 2010). Predictions of climate change forecast further increases in rainfall variability and in the occurrence of high-temperature events (IPCC, 2007; Battisti & Naylor, 2009; Coumou & Rahmstorf, 2012; Zheng et al., 2012). Given this context, insights from new technologies are sought. Recent development in genomics has been exploited by major breeding institutions to assist the selection of germplasm that are tested in field trials and to identify genes of interest by systematic screening (Nelson et al., 2007). However, despite the multitude of studies proclaiming the identification of new genes or quantitative trait loci (QTLs) for improved abiotic stress tolerance, a significant contribution in released varieties is yet to be realized (e.g. Chapman et al., 2002; Snape, 2004; Collins et al., 2008; Fleury et al., 2010; Richards et al., 2010). Phenomics platforms are being developed to provide new insights into the understanding of gene function and environmental responses (e.g. Granier et al., 2006; Berger et al., 2010; Furbank & Tester, 2011). However, the number of genotype × environment combinations to be tested is beyond the capacity of any phenotyping platform or breeding programme, even when considering only a limited numbers of genes and environments; for example, 10 bi-allelic genes tested in five environments would correspond to c. 300 million combinations.

Depending on the timing and the intensity of the stress, drought affects different genes and physiological processes, and to different extents (e.g. Slafer, 2003; Hammer et al., 2006; Fischer, 2011). Genes and traits beneficial to yield in some environments can have negative effects in others (e.g. Chenu et al., 2009; Tardieu, 2012), highlighting the importance of working in the target environments when possible. Within a limited geographical area, spatial and temporal variations in rainfall combined with the diversity of soil types give rise to an almost unlimited number of drought patterns that crops might experience. There is thus a need to characterize the environment of the ‘target population of environments’ (TPE; Comstock, 1977) to identify the main stress pattern and enable research and breeding to focus on environments of interest (e.g. Chapman et al., 2000a).

Different approaches have been proposed to classify the environments. Breeders typically sample the TPE using multi-environment trials (METs) often conducted in several locations over several years, and characterize these trials qualitatively (e.g. presence/absence of a disease) or quantitatively (measuring attributes such as rainfall, trial mean yield or check-variety (‘probe genotype’) performance; e.g. Cooper & Fox, 1996; Basford & Cooper, 1998; Brancourt-Hulmel, 1999). While this assists interpretation of genotype × environment interactions, it may misrepresent the TPE, in particular given the unpredictability of seasonal climate. Large-scale environment characterization has been proposed based on the similarity of stresses (both biotic and abiotic) or environment factors (e.g. photoperiod or temperature; e.g. Braun et al., 1996; Löffler et al., 2005; Hodson & White, 2007). However, in the case of complex abiotic stress(es) such as drought, such characterization is difficult given that (1) crops are sensitive to a water stress during most of their cycle, with different processes being involved at different stages, which means that the stress should be characterized over time; (2) the stress is influenced by the crop itself (i.e. by plant growth and transpiration, and by any factors that affect them), by the soil characteristics and by the water supply (rainfall and irrigation), which makes the stress pattern difficult to quantify (timing and intensity) even at a local scale; and (3) both soil characteristics (e.g. available soil water) and rainfall patterns vary spatially and over time, which makes characterization challenging over long periods and large geographical areas.

Modelling tools open up opportunities to resolve this challenge. First, they can be applied to multi-site long-term characterization, thus leading to more comprehensive environmental sampling than conventional field studies (e.g. Hammer & Jordan, 2007). Secondly, they allow the simulation of the stress per se by accounting for the interactions between the plants and their environments (e.g. Chelle, 2005; Chenu et al., 2007, 2008). By capturing feedbacks between plant growth and soil water depletion, crop models have been shown to characterize water deficits better than standard indices based on climatic data (Muchow et al., 1996). A few studies have used this approach in multiple locations to characterize the water-deficit patterns experienced by a crop at a regional level (Chapman et al., 2000a,b; Chenu et al., 2011).

The Australian wheatbelt extends over 14 million ha and produces an annual c. 22 million t of grain (data from 2010; source: Australian Bureau of Statistics). The wheatbelt soils range from shallow sandy to deep clay soils. Its climate includes temperate and mediterranean (with winter-dominant rainfall) to subtropical (with summer-dominant rainfall) types, and is characterized by large variations in inter-annual rainfall (e.g. Potgieter et al., 2002; Williams et al., 2002). This paper aims to analyse the drought patterns of wheat crops across the large and contrasting Australian wheatbelt TPE over 123 yr. The goals were (1) to analyse the variability of drought patterns (timing, duration and intensity); (2) to identify the nature of the main water deficits; and (3) to analyse spatial and seasonal variations in the frequency of occurrence of these main water-deficit patterns, as well as the impact of genotypes and management practices.

Materials and Methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information


To represent the Australian wheat cropping system, the major production areas (West, South, South-east and East) were divided into 22 regions (Table 1, Fig. 1). For each region, one to eight locations were chosen in such a way that each location represented between 130 000 and 230 000 ha of planted wheat (averaged data from 1975 to 2000, 2005 and 2006; source: Australian Bureau of Statistics). In total, 60 locations were characterized by their climate (historical records from 1889 to 2011; Fig. 2) and by a soil typical of their region (chosen in consultation with local agronomists).

Table 1. Cropping areas, regions, locations and soils chosen to represent the Australian wheatbelt
Cropping areaRegionSiteNo.StateLat.Long.Soil classificationSoil PAWC (mm)Sowing DateDensity (plants m−2)PAW at sowing (mm)Nitrogen (kg ha−1)
  1. The main cropping areas (West, South, South-east and East) were subdivided into 22 regions, each comprising one to eight locations (numbered 1–60; Fig. 1). Plant available water capacity (PAWC) is indicated for each soil, as well at the five levels of initial soil water used in the simulations (each level corresponded to 20% of the sowing plant available water (PAW) estimated from an initial round of simulations; see text). Initial and applied nitrogen (N) is indicated by ‘x-y-z-a’: x, initial N present in the soil at sowing; y, N applied at sowing as urea for the East and as nitrate in the rest of the wheatbelt; z and a, N applied as nitrate at the stages ‘ beginning of stem elongation’ and ‘mid-stem elongation’, respectively. State abbreviations: QLD, Queensland; NSW, New South Wales; SA, South Australia; WA, Western Australia.

  2. a

    Soil PAWC > 60mm.

  3. b

    > 80 mm of rainfall from sowing to the stage ‘beginning of stem elongation’.

  4. c

    Soil PAWC > 60% of maximum.

  5. d

    > 100 mm of rainfall from sowing to the stage ‘beginning of stem elongation’.

  6. e

    Soil PAWC between 50 and 75% of maximum.

WestWA zone 1Carnamah1WA−29.69115.89Sand9808-May/22-May/30-May/07-June/17-June15044-64-83-98-9845-20-30-30a
WA zone 2Badgingarra3WA−30.34115.54Gravel5411-May/23-May/02-June/11-June/21-June15026-45-54-54-5445-20-40-40a
Corrigin4WA−32.33117.87Deep sandy duplex7411-May/23-May/02-June/11-June/21-June15026-45-60-74-7445-20-30-0
Cunderdin6WA−31.66117.25Deep sandy duplex7411-May/23-May/02-June/11-June/21-June15022-36-55-74-7445-20-30-0
Wongan Hills7WA−30.84116.73Deep loamy duplex9011-May/23-May/02-June/11-June/21-June15028-47-66-90-9045-20-30-30a
WA zone 3Katanning8WA−33.69117.56Deep sandy duplex7409-May/21-May/29-May/06-June/18-June15036-56-74-74-7445-20-30-30a
Northam10WA−31.64116.67Loamy earth13509-May/21-May/29-May/06-June/18-June15046-73-103-129-13545-20-30-30a
WA zone 4Bencubbin11WA−30.81117.86Sandy earth11211-May/25-May/03-June/11-June/22-June10020-36-52-76-11230-20-20-30a
Kellerberrin13WA−31.62117.72Sandy earth11211-May/25-May/03-June/11-June/22-June10025-41-59-77-11230-20-20-30a
Merredin14WA−31.48118.28Shallow loamy duplex10111-May/25-May/03-June/11-June/22-June1009-22-37-56-9130-20-20-30a
Morawa15WA−29.21116.01Deep loamy duplex9011-May/25-May/03-June/11-June/22-June10025-38-58-85-9030-20-20-30a
Mullewa16WA−28.54115.51Sandy loam3411-May/25-May/03-June/11-June/22-June10014-24-33-34-3420-20-0-0
Southern Cross17WA−31.23119.33Gravel5411-May/25-May/03-June/11-June/22-June10015-27-40-54-5420-20-0-0
Narembeen18WA−32.07118.40Loamy earth13511-May/25-May/03-June/11-June/22-June10022-45-61-89-12230-20-20-30a
 WA zone 5Esperance19WA−33.61121.78Deep sandy duplex7409-May/23-May/01-June/10-June/22-June15038-64-74-74-7445-20-30-30a
  Hyden20WA−32.44118.90Deep sandy duplex7409-May/23-May/01-June/10-June/22-June10016-30-45-62-7430-20-20-30a
  Lake Grace21WA−33.10118.46Shallow sandy duplex5709-May/23-May/01-June/10-June/22-June10018-32-43-57-5730-20-20-30a
  Salmon Gums22WA−32.99121.62Shallow sandy duplex5709-May/23-May/01-June/10-June/22-June10012-27-47-57-5720-20-0-0
  Ravensthorpe23WA−33.58120.05Deep sandy duplex7409-May/23-May/01-June/10-June/22-June10032-52-72-74-7445-20-30-30a
SouthUpper Eyre PeninsulaMinnipa24SA−32.84135.15Red light sandy clay loam9004-May/17-May/26-May/05-June/19-June10012-22-35-51-8450-20-30b-30c
 Ceduna25SA−31.90133.42Grey calcareous sandy loam6404-May/17-May/26-May/05-June/19-June1005-13-27-40-6250-15-0-0
 Lower Eyre PeninsulaRudall26SA−33.69136.27Grey calcareous loamy sand8703-May/16-May/25-May/04-June/19-June10012-25-37-52-8550-20-30b-30c
 Cummins27SA−34.27135.73Dark loamy clay10503-May/16-May/25-May/04-June/19-June15032-52-76-102-10550-40-40d-40c
 Yorke PeninsulaBalaklava28SA−34.14138.42Sandy loam7203-May/14-May/25-May/05-June/19-June15022-42-60-72-7250-20-30b-30c
 Roseworthy29SA−34.53138.69Dark brown cracking clay9403-May/14-May/25-May/05-June/19-June15023-53-75-94-9450-40-40d -40c
 Mid North (SA)Port Pirie30SA−33.17138.01Loamy sand13902-May/15-May/24-May/03-June/18-June10021-40-60-83-12950-20-30b-30c
 Murray MalleeLoxton31SA−34.44140.60Loamy sand13402-May/16-May/26-May/07-June/19-June10012-27-43-59-9150-20-30b-30c
 Waikerie34SA−34.18139.98Hypercalcic calcarosol8203-May/15-May/25-May/07-June/20-June10021-41-59-80-8250-20-30b-30c
 MalleePinnaroo33SA−35.26140.91Loamy sand7903-May/15-May/25-May/07-June/20-June10021-41-62-79-7950-20-30b-30c
 Walpeup32SA−35.12142.00Loamy sand13402-May/16-May/26-May/07-June/19-June10020-41-59-92-13450-20-30b-30c
South-eastSouth-western NSWUrana35NSW−35.33146.03Clay loam17204-May/17-May/30-May/10-June/24-June10059-78-109-136-17250-40-40d-40c
  Yanco36NSW−34.61146.42Brown sodosol19104-May/17-May/30-May/10-June/24-June10060-72-104-152-19150-40-40d-40c
 South-eastern NSWWagga Wagga37NSW−35.16147.46Red sodosol18104-May/18-May/26-May/06-June/20-June15095-131-168-181-181100-50-50-40e
  Lake Bolac40VIC−33.10118.46Sandy clay loam15101-May/12-May/20-May/31-May/16-June15064-106-147-151-151100-50-50-40e
 South MalleeHopetoun41VIC−35.73142.37Loamy sand15303-May/14-May/23-May/06-June/19-June10019-40-63-90-12650-40-40d-40c
  Birchip43VIC−35.98142.92Clay loam9903-May/14-May/23-May/06-June/19-June10016-34-51-83-9950-20-30-30c
EastWestern NSWMerriwagga44NSW−33.92145.52Sandy loam16504-May/17-May/28-May/09-June/24-June10058-76-90-126-16550-40-40d-40c
  Parkes45NSW−33.14148.16Sandy clay loam19604-May/17-May/28-May/09-June/24-June150102-135-190-197-197100-50-50-40e
  Gilgandra46NSW−31.71148.66Brown dermosol12804-May/17-May/28-May/09-June/24-June10083-128-128-128-12850-50-50-0d
  Condobolin47NSW−33.07147.23Sandy loam13204-May/17-May/28-May/09-June/24-June10085-98-115-132-13250-60-60-0d
  Dubbo48NSW−32.24148.61Red dermosol14204-May/17-May/28-May/09-June/24-June10084-124-142-142-14250-50-50-0d
  Nyngan49NSW−31.55147.20Sandy clay loam21904-May/17-May/28-May/09-June/24-June100107-154-186-219-21950-60-60-0d
 Eastern NSWGunnedah50NSW−30.98150.25Black vertosol27206-May/19-May/29-May/09-June/23-June150119-186-261-272-27250-70-60b-0
  Wellington51NSW−32.80148.80Sandy clay loam10106-May/19-May/29-May/09-June/23-June15091-101-101-101-10150-50-50-0d
 Northern NSWNarrabri52NSW−30.34149.76Grey vertosol23304-May/18-May/29-May/09-June/24-June100117-174-210-218-21830-130-0-0
  Moree53NSW−29.48149.84Grey vertosol19404-May/18-May/29-May/09-June/24-June100106-152-186-194-19430-80-0-0
  Walgett54NSW−30.04148.12Grey vertosol19404-May/18-May/29-May/09-June/24-June10092-115-148-186-19430-80-0-0
  Coonamble55NSW−30.98148.38Sandy clay18104-May/18-May/29-May/09-June/24-June10084-130-169-181-18150-70b-60-0
 South-western QLDRoma56QLD−26.57148.79Brown vertosol11904-May/18-May/27-May/08-June/25-June100103-115-119-119-11930-50-0-0
 Western Darling DownsMeandarra57QLD−27.33149.88Grey vertosol20102-May/17-May/27-May/07-June/23-June10098-145-189-200-20130-80-0-0
  Goondiwindi58QLD−28.55150.31Red brown15902-May/17-May/27-May/07-June/23-June100119-147-157-159-15930-80-0-0
 Eastern Darling DownsDalby59QLD−27.18151.26Grey vertosol20605-May/18-May/30-May/08-June/24-June100145-198-206-206-20630-130-0-0
 Central QLDEmerald60QLD−23.53148.16Black vertosol13416-Apr/28-Apr/14-May/27-May/05-June100120-129-134-134-13430-50-0-0

Figure 1. The 22 regions (coloured and named in each box) and 60 sites (1–60; see Table 1) used to represent four cropping areas of the Australian wheatbelt: the ‘West’ area (green colours); ‘South’ (blue); ‘South-east’ (purple); ‘East’ (orange). State abbreviations: QLD, Queensland; NSW, New South Wales; SA, South Australia; WA, Western Australia.

Download figure to PowerPoint


Figure 2. Map of monthly cumulative rainfall (bars) and monthly average temperature (solid line) in the 22 regions of the Australian wheatbelt for the period 1889–2011. Data are averaged across locations within each region (Table 1 and Fig. 1). Scales are given in the shaded box at the top of the figure.

Download figure to PowerPoint

Simulations were performed for the 60 sites over 123 yr of historical climatic data, with the commercial variety ‘Hartog’ (Triticum aestivium L.), using the Agricultural Production Systems Simulator (APSIM) crop model (Wang et al., 2002; Keating et al., 2003). A first set of simulations was run to identify representative sowing dates over long periods and soil water content at sowing for each site, according to farmer local practices, soil characteristics and preceding rainfall. Based on these conditions at planting, a second set of simulations was run to characterize the drought patterns that wheat crops experience at these sites.

Finally, a third set of simulations was performed for an early, a standard and a late genotype to determine the impact of maturity on drought pattern (with the same initial conditions as in the second set of simulations, including sowing dates).

Determination of sowing opportunities and associated soil water content for each site – first set of simulations

To identify representative sowing dates and initial soil moisture at each site, an initial set of simulations was performed over 123 yr using the APSIM crop model. Historical daily climate data (solar radiation, maximum temperature, minimum temperature and rainfall) for the period of 1889–2011 were gathered for 60 sites across the wheatbelt, from the SILO patched point data set (; Jeffrey et al., 2001). Soil characteristics were collected from a database generated by local experts from the Agricultural Production Systems Research Unit (APSRU; and from Daniells et al. (2002).

For each location and each year, a summer fallow was assumed to begin from 1 November with a soil containing 20% of its potential available soil water capacity (PAWC). For each simulation, the timing of possible planting events for wheat, as well as the soil moisture at each planting event, was recorded (Table 1). Planting events were defined to occur when a set of region-specific conditions (chosen with the input of local agronomists) were met: (1) the sowing had to occur between 1 May and 1 July in all regions except Central Queensland, where it ranged from 15 April to 15 June; (2) a sowing event could only occur if rainfall over the 10 preceding days was at least 10 mm in the East, or at least 5 mm in the South and the South-east; and in the West a linear decrease over time was applied for the minimum rainfall required (from a minimum of 20 mm over 3 d for 1 May, to a minimum of 5 mm over 3 d for 1 July); and (3) a minimum soil water availability was required for the soils of regions with summer-dominant rainfall: 50 mm for Coonamble, Dubbo, Gilgandra, Merriwagga, Urana, Wellington and Yanco; 80 mm for Condobolin, Dalby, Emerald, Goondiwindi, Gunnedah, Meandarra, Moree, Narrabri, Nyngan, Parkes, Roma, Wagga Wagga and Walgett. In the rest of the wheatbelt, sowing could occur without any requirement on the soil moisture. A maximum of three sowing opportunities were considered per season in each site.

From this initial set of simulations, five sowing dates each representing 20% of sowing opportunities were identified for each region (Table 1). As soil water content at sowing was mainly uniform across sowing dates, five unique initial soil water conditions were used for the second set of simulations (i.e. without distinction across sowing dates; Table 1). The levels of initial soil water were chosen to each represent 20% of the conditions encountered during the planting window for the considered site, over 123 yr.

Characterization and classification of seasonal drought patterns (‘environment types’) in the Australian wheatbelt – second set of simulations

Based on information obtained from the initial simulations (date and soil water content at sowing), a second round of simulations was run to characterize the seasonal drought patterns occurring at the 60 sites across the Australian wheatbelt for 1889–2011. In total, 184 500 simulations were generated (60 locations × 5 sowing dates × 5 initial soil water × 123 yr) for a standard check variety, ‘Hartog’ (mid-maturity genotype), with conventional tillage, no biotic limitations and nitrogen fertilization as described in Table 1.

Apart from various crop traits such as grain yield and plant biomass, the APSIM model generated a water-deficit index (‘water supply/demand ratio’; Chapman et al., 1993. This index indicates the degree to which the soil water extractable by the roots (‘water supply’) is able to match the potential transpiration (‘water demand’). The water demand (Wd, in mm) corresponds to the amount of water the crop would have transpired in the absence of soil water constraint and is estimated daily based on the amount of crop growth on that day (CGR, in g mm–2), and the atmospheric saturation vapour pressure deficit (VPD in kPa):

  • display math

(TEcrop, the transpiration efficiency coefficient for above-ground biomass, which is assumed constant for a species (Tanner and Sinclair, 1983) and is set at 0.006 g m−2 mm−1 kPa in APSIM-wheat; αCO2, a coefficient adjusting the transpiration efficiency for CO2 concentration, increasing linearly from 1 to 1.37 when CO2 concentration increases from 350 to 700 ppm (Reyenga et al., 1999). Crop water demand is capped at 1.5 times the potential evapotranspiration to limit transpiration on days with high VPD.

The crop water supply (Ws, in mm) is calculated for each layer of the soil where roots are present (Wslayer) and depends on the root growth and soil property of each layer:

  • display math

dul,layer, the water content (in mm mm–1) of the layer at the drained upper limit (i.e. field capacity); θll,layer, the water content (in mm mm–1) of the layer at the lower limit; t, time (in days) since the start of water extraction from that layer (Meinke et al., 1993; Robertson et al., 1993); kl, the water extraction rate (in d–1), which is effectively parameterized as the product of root length density and the hydraulic conductivity of the soil).

The water-deficit index is defined as the ratio between Ws and Wd, and was capped between 1.0 (no water stress) and 0.0 (no water available to the crop). For each environment (defined by a site, year, sowing date and initial soil water), this daily index was centred around flowering and averaged over 100°Cd from emergence to 450°Cd after flowering, after which senescence greatly reduced plant transpiration and can thus lead to an ‘artificial’ increase of the water-deficit stress. Note that APSIM still accounts for water stress in these conditions but, as leaf area has dropped, the impact of stress is mediated through the decrease in biomass accumulation and substantial retranslocations among organs.

The partitioning clustering function (clara) in the R statistical package (R Development Core Team, 2011) was used to cluster the seasonal water-deficit pattern into four environment types (ETs). The method minimized the sum of dissimilarities between the stress-index pattern of each environment of the TPE and the median situation of the ET they related to. An average pattern of water-deficit index was calculated to describe each ET. The occurrence of each ET was interpreted for the different regions with respect to the sowing dates and initial soil moisture situations, and over time.

A Pearson's chi-squared test was applied to identify significant differences between individual seasons and long-term periods (1889–2011) in terms of the occurrence of drought ETs at the national level.

Seasonal drought patterns of genotypes with contrasting maturity - third set of simulations

To evaluate the impact of maturity on the seasonal drought pattern, a third set of simulations was run with the same conditions as the second set but for a quick-maturing, a standard-maturing and a slow-maturing variety (i.e. 184 500 simulations for each genotype). These genotypes all shared the characteristics of ‘Hartog’ but had the phenology of ‘Westonia’ (early), ‘Hartog’ (medium) and ‘Bolac’ (late). For each simulation, their seasonal patterns were classified based on which previously defined ET they were most similar to, that is, based on the minimum sum of squared differences for the considered water-deficit pattern compared with the water-deficit pattern of the previously defined ETs.


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

The Australian wheatbelt has highly diverse rainfall and drought patterns

Monthly rainfall and temperature patterns in the 22 regions across the Australian wheatbelt (Fig. 2) illustrate the great variability that wheat crops experience in Australia. Rainfall is summer-dominant in most of the East and winter-dominant in the West and some parts of the South, while rain is more evenly distributed over the year in the South-east. Cumulative average rainfalls from May to November (i.e. the approximate wheat cropping season) range from < 200 mm (in Murray Mallee) to > 400 mm (in WA zone 3), while average daily temperature during this period ranged from 10.9°C (in Wimmera) to 19.5°C (in Central Queensland).

This variability in climatic patterns, together with a great variability in soil types (Table 1), led to variation in the number and the duration of water-stress events that occurred within the cropping season (Figs 3, S1). In the East, wheat crops were subjected to few but long continuous water-stress events. For instance, crops in Central Queensland had 3.3 stress events on average, each lasting 13.7 d on average (a stress event being defined here as a continuous period when the water-stress index was < 0.7). By contrast, the rest of the wheatbelt (i.e. mainly the area where rainfall is not summer-dominant) tended to have more and shorter stress events (with some exceptions, such as the Upper Eyre Peninsula; Supporting Information Fig. S1). For instance, in the Yorke Peninsula, c. 50% of the crops had six stress events or more, and the average duration of an event was 4.9 continuous days.


Figure 3. Number and duration of stress events per crop cycle for the 22 regions (named in Fig. 1) across the Australian wheatbelt. A stress event was defined as the number of consecutive days when the simulated water-stress index was below 0.7. The frequency of each combination of ‘number of events × duration in number of entire days’ is coloured to represent low (yellow) to high (red) probability of occurrence. Data were simulated for the check variety ‘Hartog’ over 123 yr of historical data. Scales are given in the shaded box at the top of the figure (y-axis limited to 30 d for clarity).

Download figure to PowerPoint

Four drought environment types to describe the main water-limited stress patterns in the wheat-cropping production area of Australia

Despite the variability in the number and duration of individual stress events (Fig. 3), the global stress patterns of the different regions had some similarity (Fig. 4). For each region, a cluster analysis on simulated crops (i.e. 123-yr simulations for one to eight locations depending on the region, five sowing dates, and five levels of soil water content at sowing) allows depiction of the four main drought patterns of the region. While these regional ETs differed across regions, they had some similarity: the first clusters of each region (ETregion1) corresponded to environments with no or light stress(es); most of the second clusters (ETregion2) corresponded to post-flowering mild stresses; the third clusters (ETregion3) were usually characterized by stresses beginning before flowering that were relieved during the grain filling; and the last clusters (ETregion4) typically corresponded to severe stresses that began during the vegetative period and generally lasted until maturity. These general descriptions suit most but not all of the regions; for example, in Upper Eyre Peninsula, a few crops (3%) were affected by an early drought and, in WA zone 1, the abundance in rainfall (Fig. 2) resulted in almost no stress for all the ETs. The frequency of occurrence of these ETs varied greatly across regions (data not shown).


Figure 4. Dynamics of the simulated water-stress index for the main environment types (ETs) when clustered separately within each region. The stress index corresponds to the ratio of soil water supply to crop water demand and is represented as a function of cumulative thermal time relative to flowering, from crop emergence to 450°Cd after flowering. Data were simulated for the check variety ‘Hartog’ over 123 yr of historical data for the 22 regions of the wheatbelt (named in Fig. 1). Scales are given in the shaded box at the top of the figure.

Download figure to PowerPoint

Given the similarity in the regional ‘global’ drought patterns, a cluster analysis was performed at the national level. Four main ETs were identified as representative of drought patterns that wheat crops experience in the Australian wheatbelt (Fig. 5). As c. 185 000 drought patterns were summarized by only four classes, each of these ETs comprised a broad range of water-deficit patterns but they explained 53% of the total variability.

The ETs at the national level could be described in a similar way as at the regional level. The first ET (ET1) represented stress-free and short-term water-deficit environments. The second ET (ET2) was characterized by mild water shortage that mainly occurred during grain filling and was relieved by maturity. ET3 was defined by more severe water stress that occurred earlier, during the vegetative stage, and was relieved during mid-grain filling. Finally, in ET4, plants were constrained by water from the early stage onwards, and experienced a severe water deficit throughout the grain-setting and the grain-filling periods.

Across the wheatbelt, the frequency of occurrence of the environment types ET1:2:3:4 was 23:33:24:20%, respectively.


Figure 5. Simulated water-stress index for four environment types (ETs) identified across the Australian wheatbelt (all regions combined). Inset, coefficient of determination of the clustereing for different number of ETs. The stress index corresponds to the ratio of soil water supply to crop water demand and is represented as a function of cumulative thermal time relative to flowering, from crop emergence to 450°Cd after flowering (after which senescence greatly reduced plant transpiration).

Download figure to PowerPoint

Regional variability in the occurrence of the main drought patterns

Each of the four ETs occurred in all the regions of the wheatbelt, but ET4 was rare in WA zone 1 (Fig. 6). Overall, stresses occurred most frequently around flowering (44% of ET3 and ET4 (ET3-4)) and during grain filling (77% of ET2-3-4), while the generally stress-free conditions were less frequent (23% of ET1). Light to mild stresses (ET1-2) were most common closer to the coast, where rainfall is more abundant during the cropping period (Fig. 2), and ET1 was rare in the north-east. Further inland, severe stresses (ET3-4) were generally dominant, partly as a result of smaller rainfalls over the cropping season. While higher seasonal rainfalls were correlated with a higher frequency of light to mild stresses (ET1-2; r= 0.68), several examples illustrated that rainfall was not the only key driver of drought pattern (Figs 2, 6). For instance, the driest region (Central Queensland, averaging 115 mm of within-season rainfall) had only slightly more long and severe ET4 stresses (30%) than regions where it rains twice as much (e.g. South Mallee; 215 mm; 28% of ET4), and only slightly more ET3 stresses (30%) than one of the wettest regions (South-eastern NSW; 277 mm; 27% of ET3). These differences could mainly be explained by the differences in soil characteristics and cropping management. Globally, cropping soils in the West and South are mainly sandy-loamy soils (potential soil water available to the plants is typically lower < 100 mm) while soils in the East are heavy clay-vertosol that typically contain 150 mm or more extractable water (Table 1). Furthermore, in the East, given the summer dominance of rainfalls and their high inter-seasonal variability, farmers only plant a crop when the soil has a minimum of moisture already. Overall, the greatest occurrence of severe stress (ET3-4) resulted from a combination of low rainfall and poor sandy soils for the South and West, or of low rainfall and shallow vertosol soils for some parts of the East. Severe stresses (ET3-4) varied from 77% of occurrence in South-western Queensland to < 20% in regions with high seasonal rainfalls (Wimmera, WA zone 1 and WA zone 3). Overall, most Australian wheat is produced in regions frequently subjected to severe stress.


Figure 6. Map of the frequencies of each environment type (ET) (pie charts) across the Australian wheatbelt. Data were simulated for the check variety ‘Hartog’ over 123 yr of historical data for the 22 regions of the wheatbelt (named in Fig. 1). The size of the pie charts is proportional to the wheat-planted area in the associated region. The ETs are presented in Fig. 5. State abbreviations: QLD, Queensland; NSW, New South Wales; SA, South Australia; WA, Western Australia.

Download figure to PowerPoint

The frequency of occurrence of stress events and their duration (Figs 3, S1) tended to be correlated with the frequency of ETs (Fig. 6) (data not shown). However, regions like Central Queensland and Murray Mallee had similar frequencies of ET1:2:3:4 stresses (18:23:30:30% and 12:24:33:32%, respectively) but they had highly different patterns of individual stress events: in Central Queensland, the median crop experienced three stress events of 10 d on average, while Murray Mallee experienced six stress events each lasting 5.5 d on average.

Regional variability in the occurrence of yield could partly be explained by the drought environment type

The simulated grain yield was highly constrained by the water-deficit pattern (Fig. 7). Overall, yield tended to decline from ET1 to ET4. However, the yield distributions of ET1 and ET2 were similar for many regions. The range of simulated yield for each ET remained broad, with some large overlap in yield among ETs. This arose from the fact that (1) each drought ET encapsulates numerous different drought patterns (184 500 simulations clustered into four ETs) and (2) yield is affected by factors other than drought (e.g. temperature and radiation). Note that yield simulated here corresponds to environment-potential yield assuming no heat-shock, frost, pest, disease or lodging damage.


Figure 7. Simulated yield distribution for each environment type (boxplot) across the Australian wheatbelt. Data were simulated for the check variety ‘Hartog’ over 123 yr of historical data, for the 22 regions of the wheatbelt (named in Fig. 1). The lower, middle and upper lines of the boxplot indicate 25, 50 (median) and 75% of yield values, respectively. Scales are given in the shaded box at the top of the figure.

Download figure to PowerPoint

The occurrence of the crop environment type varied between successive years, and severe drought patterns were frequent in recent years

The variability in seasonal rainfall over the years led to a high variation in the occurrence of the ETs over time (Fig. 8). Climates oscillated from periods with low to high occurrence of light to mild, ET1-2 stresses (i.e. high to low severe, ET3-4 stresses; see the pattern depicted by the 5-yr average (solid lines) in Fig. 8). Variations were even greater for individual years, and successive years could change from extremely low to extremely high ET1-2 frequency and vice versa (e.g. from 8.3 to 73% on average for 1914 to 1915 and from 37.5 to 95.1% on average from 1972 to 1973; Fig. 8a). Years with contrasting patterns among areas were also frequent (e.g. in 2010, 79% of ET3-4 in the West, while only 6% elsewhere), highlighting that the spatial variability observed in the Australian climate also varies over time (Fig. 8b).


Figure 8. Frequency of occurrence of the four environment types (ETs) over time, for weather records from 1889 to 2011, for (a) the entire wheatbelt and (b) the 22 regions of the Australian wheatbelt (named in Fig. 1). The solid line (a1 and b) corresponds to the 5-yr moving average of ET1-2 frequencies, with each point being the average of the considered year and the previous 4 yr (the complement to this line corresponds to the frequency of ET3-4). The horizontal bars (a2 and a3) indicate years significantly below (red) and above (blue) the overall average of ET1-2 frequency (= 0.05). In (a1), (a2) and (b), as in the rest of the paper, ETs simulated do not correspond to historical drought occurrence but to the effect of historical within-season climate, as initial conditions in the simulations were calculated over 123 yr and not for each specific year (i.e. the same set of sowing dates and levels of initial soil water were used for each year). In (a3), additional simulations were performed for wheat crops cultivated in consecutive years such that each year's sowing conditions resulted from climatic conditions of the previous year (sowing at the first opportunity, with soil water conditions dependent on the climate from the previous year). Data were simulated for the check variety ‘Hartog’. Vertical dot lines were added every decade for clarity. (b) Scales are given in the shaded box at the top of the figure.

Download figure to PowerPoint

While inter-annual variability appears to be a characteristic of the Australian environment, several periods of successive drought or ‘wet’ years have occurred over the last century (Fig. 8), for example, the drought period around 1940 or the wet period of the 1960s. Despite recent high rainfall in Eastern Australia (2010–2011), the current period appears to be one of the longest and most severe periods of drought for the period studied at the national level. With an average of 52% for the occurrence of severe ET3-4, the last 12 yr have only been matched in severity by the period around 1940, with a 54% occurrence of ET3-4 (Fig. 8a1, a2). Note that the simulations performed in this study do not reproduce what happened historically, but simulate the occurrence of the ETs when considering the years as independent; that is, the sowing conditions were imposed to be representative for the considered location and not specific to a particular year. Hence, the simulations approximate the potential cropping situations to which released cultivars could be exposed, and depict the effects of within-season climates (as the initial conditions were the same for each year). An alternative set of simulations has been performed where wheat crops were sown one after another, at the first sowing opportunity after a short summer fallow (Fig. 8a3). Although this does not represent current management practices, these simulations accounted for initial soil conditions resulting from the climate of the previous year. Overall, similar general patterns of drought occurrence were observed over time. In this set of simulations as well, the current drought period of the last 12 yr was found to be the most severe from the last century, apart from the 1940 drought.

Early sowing date and greater initial soil water reduced the frequency of severe drought in most regions

Management practice influenced the frequency of occurrence of each ET. In each region, later sowing reduced the occurrence of light to moderate, ET1-2 stresses while increasing the occurrence of severe, ET3-4 stresses (Fig. 9). This reflects the fact that early-sown crops usually grow in nonlimiting water conditions during the vegetative period while later-sown crops typically experienced higher temperature and more water-demanding conditions before flowering onwards. For late sowing dates (e.g. mid-June; Table 1), the occurrence of ET1 was almost zero in most regions. Although the general response of ET frequency to sowing date was similar in all regions, its magnitude varied. In regions with low ET3-4 frequency (such as Wimmera, WA zone 1 and 2, and South-eastern NSW), early sowing almost eliminated the occurrence of severe, ET3-4 stresses for the check variety studied. In contrast, early sowing had only a slight effect on the occurrence of ET4 in most regions of the East, where crops can access substantial amounts of soil water from sowing onwards (given the local management practices; Table 1).


Figure 9. Frequency of occurrence of the four ETs across the Australian wheatbelt, for each combination of sowing dates and soil water at sowing used in the simulations. Sowing dates are presented from the earliest (1) to the latest (5) each sowing date representing 20% of the simulated sowing opportunities (see Table 1 and text). Initial soil water increased from the lowest (1; most severe conditions) to the highest (5; less severe) values, each representing 20% of the simulated initial soil water (Table 1). Data were simulated for the check variety ‘Hartog’ over 123 yr of historical data, for the 22 regions of the wheatbelt (named in Fig. 1). Scales are given in the shaded box at the top of the figure.

Download figure to PowerPoint

As might be expected, increased initial soil water at any sowing date led to an increase in the frequency of ET1 and ET2 and a decrease in the frequency of ET4. The response of ET3 was dependent on the region and on the sowing date: for example, greater initial soil water was associated with a decrease in ET3 in Northern NSW, an increase in South Mallee and in WA zone 4, an increase for early sowing dates and a decrease for late sowing dates. In some regions, initial water had a greater impact on ET frequencies for earlier sowing dates, which decreased with the later sowing dates, especially for ET1 (e.g. Northern NSW). Late sowing thus diminished the advantage of increased initial water in soil in those regions (Fig. 9).

Genotype maturity × environment interactions affected drought pattern, but overall, earlier maturing crops were less impacted by drought in the wheatbelt

Early-maturing crops (simulated by the variety ‘Westonia’) were subjected to a greater proportion of light to mild, ET1-2 stresses in all except a few regions (e.g. South-western Queensland) (Fig. 10). In this study, all genotypes were simulated with the same sowing dates, meaning that the quick-maturing variety (‘Westonia’) flowered and matured earlier than the medium-maturing (‘Hartog’) and slow-maturing (‘Bolac’) varieties (Fig. 10a), and thus tended to escape stress occurring later during the season (Fig. 10b). In practice, late varieties are typically sown early to take advantage of increased time for resource acquisition and to avoid damaging frost risks that occur around flowering (e.g. Zheng et al., 2012), while quick-maturing varieties are chosen for later planting dates to avoid heat and drought during the grain-filling period.


Figure 10. Distribution of flowering time (a), and frequency of occurrence of the four environment types (ETs) and the best yielding genotype (b), for an early-maturing (‘Westonia’), a medium-maturing (‘Hartog’) and a late-maturing (‘Bolac’) wheat variety. Data were simulated over 123 yr of historical data, using the same conditions for all genotypes (including the same sowing date). Results are presented for the entire Australian wheatbelt (a) and for the 22 regions (named in Fig. 1) of the Australian wheatbelt (b). For each region, a composite figure is provided (b), with (1) the frequency of occurrence of the ETs for each genotype (stacked barchart); (2) the percentage of situations (site × initial condition × year) where the considered genotype yielded more than the two others (black bars); and (3) the percentage of seasons (site × year) where the considered genotype yielded more that the two others (grey bars), i.e. for the best sowing date by maturity combination. (b) Scales are given in the shaded box at the top of the figure.

Download figure to PowerPoint

In the simulations performed here, the medium-maturing ‘Hartog’ experienced a greater (in most regions) or similar (e.g. Upper Eyre Peninsula) occurrence of severe, ET3-4 stresses compared with early-maturing crops (‘Westonia’), and was subjected to fewer or similar ET3-4 stresses than late-maturing crops (‘Bolac’). ‘Bolac’ experienced only slightly fewer ET3-4 stresses than the two other genotypes in South-western Queensland. No substantial impact of earliness on the frequency of ET1-2 vs ET3-4 stresses was seen for regions where minimal ET3-4 occurred given the high within-season rainfall (e.g. WA zone 1, averaging 278 mm), and for regions like Central Queensland, where crops mainly rely on soil-stored water for their development (only 115 mm of within-season rainfall).

In most regions, the best-yielding genotype was the one with the highest frequency of light to mild, ET1-2 stresses (black bars in Fig. 10b). However, in situations where all genotypes experienced an ET1, slow crops (‘Bolac’), with their longer developmental period, could acquire resources (e.g. light, nitrogen and CO2) and typically yielded more than medium (‘Hartog’) and quick crops (‘Westonia’). No such substantial trend was observed between genotypes all experiencing ET2 or ET3, but in ET4 environments ‘Westonia’ tended to yield more than ‘Hartog’ and ‘Bolac’ (data not shown). Note that effects of heat-shock, frost, pest, disease and lodging were not accounted for in the simulations.

When focusing only on the best practice (i.e. best sowing date × maturity combination for the year considered), ‘Hartog’ was better yielding than the other two genotypes in most regions (grey bars in Fig. 10b). However, in regions like the Wimmera, the relatively low frequency of severe, ET3-4 stresses advantaged long-season crops (‘Bolac’), while in regions like Central Queensland and South-western Queensland, in contrast, the high frequency of ET3-4 stresses was advantageous for short-season crops (‘Westonia’).

Genotype × environment interactions occurred across individual ETs (Fig. 10b). In most regions, all three genotypes were subjected to substantial frequencies of all ET patterns. The ‘Westonia’ maturity type generally experienced the most ET1 and the fewest ET4 stresses, especially in the South, South-east and parts of the East. Depending on the region, ET2 was most likely for ‘Westonia’ or ‘Hartog’, ET3 was most likely for ‘Bolac’, and ET4 occurred mostly for ‘Hartog’ and ‘Bolac’. The interactions between the timing of the rainfall, the dynamics of the crop development, the management and the soil property led to nontrivial results in terms of genotypic variations in stress duration and intensity, and ultimately yield.


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Characterization of drought as experienced by wheat crops at a continental scale over long periods of time

High spatio-temporal variations in climate and soil characteristics across the Australian wheatbelt (e.g. Fig. 2; Potgieter et al., 2002) result in wheat crops being subjected to stress events with a frequency and duration that greatly vary among seasons and across locations (Figs 3, 4; Sadras & Rodriguez, 2007). However, a set of key drought patterns can represent the variability within and across regional environments (Figs 4, 5 and 6). As might be expected, the frequency of occurrence of these ETs differed spatially and across seasons (Figs 6, 8). Only one region was consistently dominated by one ET (WA zone 1) while the others were all characterized by a substantial occurrence of all four ETs. Seasonal drought frequency highly varied spatially, with some regions experiencing a low-stress season while other parts of the wheatbelt were highly stressed. Hence, the East was regularly subjected to years dominated by severe stress, while the rest of the wheatbelt was experiencing an average or drought-free season. Also, western regions repeatedly experienced less severe drought seasons compared with the other cropping areas, which is partly attributable to El Niño-Southern Oscillation (ENSO) having a greater influence on rainfall in the eastern vs the western part of Australia (Stone et al., 1996; Potieger et al., 2002). Furthermore, low-stress years (or periods) were regularly followed by high-stress ones, at both regional and national levels. High variations in the 5-yr average of light to mild, ET1-2 stress frequency (solid line in Fig. 8) highlight a potential risk for Australian breeders, as lines are typically evaluated for only 3–5 yr between the first and final stages of testing. Hence, released varieties may be selected in a set of years that greatly differ, in terms of drought type frequency, from the 10–20 yr of their commercial life.

It is interesting to note that, despite the complexity of the environment encountered in the Australian wheatbelt, earlier flowering mostly increased the frequency of favourable ET1 conditions (Fig. 10; Ludwig & Asseng, 2010), and could be recommended to escape drought in the absence of other limiting factors such as frost around flowering (most common in the East and South-east; Zheng et al., 2012). The long-term advantage of earlier flowering has also been revealed by selection programmes, as breeders have, since the mid-1800s, tended to favour early-flowering lines, when selecting for yield (Richards, 1991). Extreme earliness is nevertheless not recommended, as lines with an excessive earliness have a limited yield potential, and poorly perform in wetter years when growers may make their most substantial returns.

Limits and advantages of such an environment characterization

In order to identify drought patterns as representative as possible of the TPE, the approach was based on numerous simulations (184 500) weighted according to local wheat cropping area and performed with a standard genotype, for representative sites, management practices and initial conditions. The quality of the characterization depends on these choices, and changing some management decisions or other factors would affect the nature and the frequency of occurrence of the identified ETs. The analysis performed for different sowing dates, soil water contents at sowing and maturity-contrasting genotypes illustrates how management strategies and genotypic characteristics can affect the drought patterns experienced by a crop (Figs 9, 10).

Summarizing the entire TPE into a few classes (‘ETs’) resulted in a high residual variability remaining within each class, as (1) a broad range of drought stresses have been included in the same smoothed and averaged ET; and (2) no environmental factors other than drought were considered. Increasing the number of environment classes would allow a better characterization of the TPE. However, four ETs seem to be an accurate trade-off in this study, as adding a fifth ET only explained < 3% more of the variability (Fig. 5, inset). Furthermore, representing the TPE with more environment classes comes at a cost for application in phenotypic evaluations. For instance, associated experimental work would need to consider more ETs (i.e. more treatments); or within a MET, an increased number of trials could be required to ensure proper representation of each ET for genotype × environment analysis (e.g. Leflon et al., 2005; de la Vega & Chapman, 2006; Chenu et al., 2011).

The ETs identified in this study only accounted for drought factors; that is, variations in crop development attributable to, for example, different seasonal temperatures, radiation or photoperiod were only considered indirectly. In order to integrate all the environmental factors, environment characterizations based on yield have been proposed (Chauhan et al., 2008; Hernandez-Segundo et al., 2009). However, such a level of integration does not necessarily help in the improvement of specific factors, such as drought tolerance. For a TPE like the Australian wheatbelt, where drought is a major limiting factor to yield, the identification of drought ETs can assist in explaining yield variation (Fig. 7; Chenu et al., 2011) and in the identification of key genotypes, traits and genes adapted to the TPE. Other major environmental factors such as temperature (e.g. Sadras et al., 2012; Zheng et al., 2012) could be added to the characterization.

Finally, as the identified drought ETs relate to a stress index that integrates the plant–environment interactions (e.g. feedback of plant growth on soil water depletion), different pedo-climatic conditions can lead to the same drought pattern, as can different genotypes. Accordingly, different traits may be needed to improve the adaptation within a single ET. For instance, deep roots might reduce the frequency of severe, ET3-4 stresses in the deep clay soils of the East which store substantial soil moisture at depth (Manschadi et al., 2006), but they may not be so valuable in sandy soils of the West, for instance, where crops rely more on in-season rainfalls. The value of such traits can be explored using functional modelling (e.g. Manschadi et al., 2006; Lilley & Kirkegaard, 2011; M. Veyradier, unpublished data).

Applications for research and the industry

While applications of environment characterization have been proposed for management decision (e.g. Lawes et al., 2009), this section will focus on applications related to crop improvement, which includes plant breeding, as well as research associated with trait and gene function. Concerning applications in plant breeding programmes, (1) knowledge about the main ETs of the TPE and their frequency is of prime interest to select trial locations and design the breeding programme (e.g. Trethowan et al., 2003; Rebetzke et al., 2013). Environment characterization is also used to design managed drought environments within the breeding programme in order to increase yield progress in highly variable drought-prone environments (e.g. Cooper et al., 1995; Campos et al., 2004; Bänziger et al., 2006). In addition, (2) environment characterization can help unravel genotype × environment interactions. Classification of METs depending on their environment has resulted in improved understanding of genotype × environment interactions for different crops and for classification based on different environmental factors (e.g. Löffler et al., 2005; Chapman, 2008; Chenu et al., 2011). (3) Characterization of METs also offers a way to weight genotype performance by the representativeness of their growing environment with respect to the TPE. Breeding-system simulations have demonstrated advantages in weighted selection strategy for variable environments, especially when genotype × environment interactions are large (Podlich et al., 1999), which is the case for rainfed Australian wheat (e.g. Rebetzke et al., 2002; Dreccer et al., 2007). Finally, (4) environment characterization can be used to identify similar growing environments around the world and exchange germplasm accordingly or breed for germplasm adapted to some world-wide ET (e.g. Braun et al., 1996; Chauhan et al., 2008; Hernandez-Segundo et al., 2009; Mathews et al., 2011).

Environment characterization also has direct application in more fundamental research. Depending on the stress intensity and on the plant stage when the water deficit occurs, different physiological processes are involved in the plant response. Adaptation to a drought type can have an adverse effect (for instance on yield) in other drought types (e.g. Trethowan et al., 2001; Collins et al., 2008; Tardieu, 2012). Some trade-off has to be made as yield is related to biomass production, which depends on photosynthesis and gas exchange, and thus on transpiration (‘water requirement’) and soil water depletion (‘water exhaustion’). For instance, increasing whole-plant transpiration by an improved drought tolerance for leaf growth can lead to exhaustion of the available soil water and result in lower yield (e.g. Chenu et al., 2009). Traits and genomic regions can have positive, null or negative additive effects depending on the drought scenario (Chapman et al., 2003; Vargas et al., 2006; Chenu et al., 2009). This reality has considerably slowed the utilization of physiological and genetics research in breeding programmes. Using representative types of drought to focus on the environments of interest can be a way to progress with more efficiency towards the discovery of genes, traits and germplasm that are adapted to the TPE. Furthermore, combining environment characterization with gene-to-phenotype simulations can help to quantify the crop-level value of such genes and traits for the main ETs targeted (e.g. Chapman et al., 2003; Hammer et al., 2006; Chenu et al., 2009; Messina et al., 2011).


This study successfully addressed the issues associated with the characterization of a complex stress (which results for the interactions between the crop and its environment) in large-scale, long-term and highly contrasting environments. Applied here to water stress in wheat for the highly variable drought-prone Australian wheatbelt, the modelling approach could be translated to other crops and other regions of the world where drought is a major limiting factor, as it could be adapted to other complex abiotic stresses (e.g. nitrogen depletion).

Modelling tools offer an opportunity to consider possible changes in the frequency of different stress patterns in various situations, and in particular under climate change. Integrated modelling approaches could help breeders, geneticists and physiologists prepare for these changes and target genotypes, traits and genetic regions of specific interest for predicted climates.


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Thanks to Al Doherty for assistance in software programming, to Andries Potgieter and the Australian Bureau of Statistics for supplying maps and data concerning wheat production and planted area, and to Richard Routley, Ben Biddulph, Bill Long, Peter Hayman, Neil Fettel, Carina Moeller, Doug Abrecht, Bob French, Tim McClelland, James Hunt, Dennis van Gool, Geoff Fosbery, Jeremy Lemon, Imma Farre, Yvette Oliver, Penny Riffkin, Rebecca Byrne and Mark Silburn for providing expertise on local soil and management practices. A special thanks for Howard Cox for his time and valued agronomic expertise.


  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information
  • Bänziger M, Setimela PS, Hodson D, Vivek B. 2006. Breeding for improved abiotic stress tolerance in maize adapted to southern Africa. Agricultural Water Management 80: 212224.
  • Basford KE, Cooper M. 1998. Genotype × environment interactions and some considerations of their implications for wheat breeding in Australia. Australian Journal of Agricultural Research 49: 153174.
  • Battisti DS, Naylor RL. 2009. Historical warnings of future food insecurity with unprecedented seasonal heat. Science 323: 240244.
  • Berger B, Parent B, Tester M. 2010. High-throughput shoot imaging to study drought responses. Journal of Experimental Botany 61: 35193528.
  • Borlaug NE, Dowswell CR 2005. Feeding a world of ten billion people: a 21st century challenge. In: Tuberosa R, Phillips RL, Gale M, eds. In the wake of the double helix: from the green revolution to the gene revolution. Bologna, Italy: Avenue Media, 324.
  • Brancourt-Hulmel M. 1999. Crop diagnosis and probe genotypes for interpreting genotype environment interaction in winter wheat trials. Theoretical and Applied Genetics 99: 10181030.
  • Braun HJ, Rajaram S, vanGinkel M. 1996. CIMMYT's approach to breeding for wide adaptation. Euphytica 92: 175183.
  • Brisson N, Gate P, Gouache D, Charmet G, Oury F-X, Huard F. 2010. Why are wheat yields stagnating in Europe? A comprehensive data analysis for France. Field Crops Research 119: 201212.
  • Campos H, Cooper M, Habben JE, Edmeades GO, Schussler JR. 2004. Improving drought tolerance in maize: a view from industry. Field Crops Research 90: 1934.
  • Chapman S, Cooper M, Podlich D, Hammer G. 2003. Evaluating plant breeding strategies by simulating gene action and dryland environment effects. Agronomy Journal 95: 99113.
  • Chapman SC. 2008. Use of crop models to understand genotype by environment interactions for drought in real-world and simulated plant breeding trials. Euphytica 161: 195208.
  • Chapman SC, Cooper M, Hammer GL, Butler DG. 2000a. Genotype by environment interactions affecting grain sorghum. II. Frequencies of different seasonal patterns of drought stress are related to location effects on hybrid yields. Australian Journal of Agricultural Research 51: 209221.
  • Chapman SC, Hammer GL, Butler DG, Cooper M. 2000b. Genotype by environment interactions affecting grain sorghum. III. Temporal sequences and spatial patterns in the target population of environments. Australian Journal of Agricultural Research 51: 223233.
  • Chapman SC, Hammer GL, Meinke H. 1993. A sunflower simulation model: I. Model development. Agronomy Journal 85: 725735.
  • Chapman SC, Hammer GL, Podlich DW, Cooper M. 2002 Linking biophysical and genetic models to integrate physiology, molecular biology and plant breeding. In: Kang MS, ed. Quantitative genetics, genomics and plant breeding. Baton Rouge, LA, USA: CAB International, 167187.
  • Chauhan Y, Wright G, Rachaputi N, McCosker K. 2008. Identifying chickpea homoclimes using the APSIM chickpea model. Australian Journal of Agricultural Research 59: 260269.
  • Chelle M. 2005. Phylloclimate or the climate perceived by individual plant organs: What is it? How to model it? What for? New Phytologist 166: 781790.
  • Chenu K, Chapman SC, Tardieu F, McLean G, Welcker C, Hammer GL. 2009. Simulating the yield impacts of organ-level quantitative trait loci associated with drought response in maize: a “Gene-to-phenotype” modeling approach. Genetics 183: 15071523.
  • Chenu K, Cooper M, Hammer GL, Mathews KL, Dreccer MF, Chapman SC. 2011. Environment characterization as an aid to wheat improvement: interpreting genotype-environment interactions by modelling water-deficit patterns in north-eastern Australia. Journal of Experimental Botany 62: 17431755.
  • Chenu K, Franck N, Lecoeur J. 2007. Simulations of virtual plants reveal a role for serrate in the response of leaf development to light in Arabidopsis thaliana. New Phytologist 175: 472481.
  • Chenu K, Rey H, Dauzat J, Lydie G, Lecoeur J. 2008. Estimation of light interception in research environments: a joint approach using directional light sensors and 3d virtual plants applied to sunflower (Helianthus annuus) and Arabidopsis thaliana in natural and artificial conditions. Functional Plant Biology 35: 850866.
  • Collins NC, Tardieu F, Tuberosa R. 2008. Quantitative trait loci and crop performance under abiotic stress: where do we stand? Plant Physiology 147: 469486.
  • Comstock RE 1977. Quantitative genetics and the design of breeding programs. In: Pollack E, Kempthorne O, Bailey TB, eds. Proceedings of the international conference on quantitative genetics. Iowa State University. Press, Ames, USA, 705718.
  • Cooper M, Fox PN. 1996. Environmental characterization based on probe and reference genotypes. In: Cooper M, Hammer GL, eds. Plant adaptation and crop improvement. Wallingford UK: CAB International, 529547.
  • Cooper M, Woodruff DR, Eisemann RL, Brennan PS, Delacy IH. 1995. A selection strategy to accommodate genotype-by-environment interaction for grain yield of wheat: managed-environments for selection among genotypes. Theoretical and Applied Genetics 90: 492502.
  • Coumou D, Rahmstorf S. 2012. A decade of weather extremes. Nature Climate Change 2: 491496.
  • Daniells I, Manning B, Pearce B. 2002 Profile descriptions: district guidelines for managing soils in north-west NSW. Tamworth, NSW, Australia: Grains Research and Development Corporation/NSW Agriculture/Natural Heritage Trust.
  • Dreccer MF, Borgognone MG, Ogbonnaya FC, Trethowan RM, Winter B. 2007. CIMMYT-selected derived synthetic bread wheats for rainfed environments: yield evaluation in Mexico and Australia. Field Crops Research 100: 218228.
  • Fischer RA. 2011. Wheat physiology: a review of recent developments. Crop & Pasture Science 62: 95114.
  • Fleury D, Jefferies S, Kuchel H, Langridge P. 2010. Genetic and genomic tools to improve drought tolerance in wheat. Journal of Experimental Botany 61: 32113222.
  • Furbank RT, Tester M. 2011. Phenomics - technologies to relieve the phenotyping bottleneck. Trends in Plant Science 16: 635644.
  • Granier C, Aguirrezabal L, Chenu K, Cookson SJ, Dauzat M, Hamard P, Thioux JJ, Rolland G, Bouchier-Combaud S, Lebaudy A et al. 2006. PHENOPSIS, an automated platform for reproducible phenotyping of plant responses to soil water deficit in Arabidopsis thaliana permitted the identification of an accession with low sensitivity to soil water deficit. New Phytologist 169: 623635.
  • Hammer G, Cooper M, Tardieu F, Welch S, Walsh B, van Eeuwijk F, Chapman S, Podlich D. 2006. Models for navigating biological complexity in breeding improved crop plants. Trends in Plant Science 11: 587593.
  • Hammer GL, Jordan DR. 2007. An integrated systems approach to crop improvement. In: Spiertz JHJ, Struikc PC, van Laar HH, eds. Scale and complexity in plant systems research: gene-plant-crop relations. Wageningen UR-Frontis Series No. 21. Dordecht, the Netherlands: Springer, 4561.
  • Hernandez-Segundo E, Capettini F, Trethowan R, van Ginkel M, Mejia A, Carballo A, Crossa J, Vargas M, Balbuena-Melgarejo A. 2009. Mega-environment identification for barley based on twenty-seven years of global grain yield data. Crop Science 49: 17051718.
  • Hodson DP, White JW. 2007. Use of spatial analyses for global characterization of wheat-based production systems. Journal of Agricultural Science 145: 115125.
  • IPCC 2007. Core Writing Team, Pachauri RH, Reisinger A, eds. Synthesis report. Contribution of working groups I, II and III to the fourth assessment report of the intergovernmental panel on climate change. Geneva, Switzerland: IPCC.
  • Jeffrey SJ, Carter JO, Moodie KB, Beswick AR. 2001. Using spatial interpolation to construct a comprehensive archive of Australian climate data. Environmental Modelling and Software 16: 309330.
  • Keating BA, Carberry PS, Hammer GL, Probert ME, Robertson MJ, Holzworth D, Huth NI, Hargreaves JNG, Meinke H, Hochman Z et al. 2003. An overview of APSIM, a model designed for farming systems simulation. European Journal of Agronomy 18: 267288.
  • Lawes RA, Oliver YM, Robertson MJ. 2009. Integrating the effects of climate and plant available soil water holding capacity on wheat yield. Field Crops Research 113: 297305.
  • Leflon M, Lecomte C, Barbottin A, Jeuffroy MH, Robert N, Brancourt-Hulmel M. 2005. Characterization of environments and genotypes for analyzing genotype × environment interaction. Journal of Crop Improvement 14: 249298.
  • Lilley JM, Kirkegaard JA. 2011. Benefits of increased soil exploration by wheat roots. Field Crops Research 122: 118130.
  • Löffler CM, Wei J, Fast T, Gogerty J, Langton S, Bergman M, Merrill B, Cooper M. 2005. Classification of maize environments using crop simulation and geographic information systems. Crop Science 45: 17081716.
  • Ludwig F, Asseng S. 2010. Potential benefits of early vigor and changes in phenology in wheat to adapt to warmer and drier climates. Agricultural Systems 103: 127136.
  • Manschadi AM, Christopher J, Devoil P, Hammer GL. 2006. The role of root architectural traits in adaptation of wheat to water-limited environments. Functional Plant Biology 33: 823837.
  • Mathews KL, Trethowan R, Milgate AW, Payne T, van Ginkel M, Crossa J, DeLacy I, Cooper M, Chapman SC. 2011. Indirect selection using reference and probe genotype performance in multi-environment trials. Crop & Pasture Science 62: 313327.
  • Messina CD, Podlich D, Dong Z, Samples M, Cooper M. 2011. Yield-trait performance landscapes: from theory to application in breeding maize for drought tolerance. Journal of Experimental Botany 62: 855868.
  • Muchow RC, Cooper M, Hammer GL 1996. Characterizing environmental challenges using models. In: Cooper M, Hammer GL, eds. Plant adaptation and crop improvement. Wallingford, UK: CAB International, 349364.
  • Nelson DE, Repetti PP, Adams TR, Creelman RA, Wu J, Warner DC, Anstrom DC, Bensen RJ, Castiglioni PP, Donnarummo MG et al. 2007. Plant nuclear factor Y (NF-Y) B subunits confer drought tolerance and lead to improved corn yields on water-limited acres. Proceedings of the National Academy of Sciences, USA 104: 1645016455.
  • Podlich DW, Cooper M, Basford KE. 1999. Computer simulation of a selection strategy to accommodate genotype-by-environment interaction in a wheat recurrent selection program. Plant Breeding 118: 1728.
  • Potgieter AB, Hammer GL, Butler D. 2002. Spatial and temporal patterns in Australian wheat yield and their relationship with ENSO. Australian Journal of Agricultural Research 53: 7789.
  • R Development Core Team. 2011. R: a language and environment for statistical computing. Vienna, Austria: R foundation for statistical computing. ISBN 3-900051-07-0, url http://www.R-project.Org
  • Rebetzke GJ, Chenu K, Biddulph B, Deery DM, Moeller C, Rattey AR, Bennett D, Barrett-Lennard EG, Mayer J. 2013. A multi-site, Managed Environment Facility (MEF) for targeted trait and germplasm phenotyping. Functional Plant Biology 40: 113.
  • Rebetzke GJ, Condon AG, Richards RA, Farquhar GD. 2002. Selection for reduced carbon isotope discrimination increases aerial biomass and grain yield of rainfed bread wheat. Crop Science 42: 739745.
  • Reyenga PJ, Howden SM, Meinke H, Mckeon GM. 1999. Modelling global change impacts on wheat cropping in south-east Queensland, Australia. Environmental Modelling and Software 14: 297306.
  • Richards RA. 1991. Crop improvement for temperate Australia - future opportunities. Field Crops Research 26: 141169.
  • Richards RA, Rebetzke GJ, Watt M, Condon AG, Spielmeyer W, Dolferus R. 2010. Breeding for improved water productivity in temperate cereals: phenotyping, quantitative trait loci, markers and the selection environment. Functional Plant Biology 37: 8597.
  • Robertson MJ, Fukai S, Ludlow MM, Hammer GL. 1993. Water extraction by grain sorghum in a sub-humid environment. I. Analysis of the water extraction pattern. Field Crops Research 33: 8197.
  • Sadras VO, Lake L, Chenu K, McMurray LS, Leonforte A. 2012. Water and thermal regimes for field pea in Australia and their implications for breeding. Crop & Pasture Science 63: 3344.
  • Sadras VO, Rodriguez D. 2007. The limit to wheat water-use efficiency in eastern Australia. II. Influence of rainfall patterns. Australian Journal of Agricultural Research 58: 657669.
  • Slafer GA. 2003. Genetic basis of yield as viewed from a crop physiologist's perspective. Annals of Applied Biology 142: 117128.
  • Snape JW 2004. Challenges of integrating conventional breeding and biotechnology: a personal view! In: Proceedings of the 4th international crop science congress. Brisbane, Australia, 26 September-1 October, 2004:
  • Stone RC, Hammer GL, Marcussen T. 1996. Prediction of global rainfall probabilities using phases of the southern oscillation index. Nature 384: 252255.
  • Tardieu F. 2012. Any trait or trait-related allele can confer drought tolerance: just design the right drought scenario. Journal of Experimental Botany 63: 2531.
  • Trethowan RM, Crossa J, van Ginkel M, Rajaram S. 2001. Relationships among bread wheat international yield testing locations in dry areas. Crop Science 41: 14611469.
  • Trethowan RM, van Ginkel M, Ammar K, Crossa J, Payne TS, Cukadar B, Rajaram S, Hernandez E. 2003. Associations among twenty years of international bread wheat yield evaluation environments. Crop Science 43: 16981711.
  • Vargas M, van Eeuwijk FA, Crossa J, Ribaut JM. 2006. Mapping QTLs and QTL × environment interaction for cimmyt maize drought stress program using factorial regression and partial least squares methods. Theoretical and Applied Genetics 112: 10091023.
  • de la Vega AJ, Chapman SC. 2006. Defining sunflower selection strategies for a highly heterogeneous target population of environments. Crop Science 46: 136144.
  • Wang E, Robertson MJ, Hammer GL, Carberry PS, Holzworth D, Meinke H, Chapman SC, Hargreaves JNG, Huth NI, McLean G. 2002. Development of a generic crop model template in the cropping system model APSIM. European Journal of Agronomy 18: 121140.
  • Williams JH, Hook RA, Hamblin A. 2002 Agro-ecological regions of Australia. Methodologies for their derivation and key issues in resource management. Canberra, ACT, Australia: CSIRO Land and Water.
  • Zheng B, Chenu K, Dreccer MF, Chapman SC. 2012. Breeding for the future: what are the potential impacts of future frost and heat events on sowing and flowering time requirements for Australian bread wheat (Triticum aestivium) varieties? Global Change Biology 18: 28992914.

Supporting Information

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office.

nph12192-sup-0001-FigS1.docWord document720KFig. S1 Stress duration depending on the number of stress events occurring during the crop cycle.