Innovations in Dryland Agriculture

246

and maize (Chauhan et al. 2013 ) were used to derive the frequency of occurrence for
contrasting water-stress patterns across Australia i.e. in terms of severity and timing
with respect to critical stages for yield formation (Fig. 4 ). This work, together with
that from Hammer et al. ( 2014 ), shows potential for better matching genotypes and
managements to growing environments and seasonal conditions. A common feature
of these studies is the use of APSIM to describe contrasting environments based on
the ratio between water supply and water demand. By clustering the ratio of simu-
lated supply and demand patterns during the cropping season, major water-stress
environment types can be identified. Fig. 4 shows for maize (Chauhan et al. 2013 ),
two environments of little stress (e1) and (e2), two environments of a mild-terminal
stress (e3) and (e4), and one environment of severe-terminal stress (e5). The fre-
quency of occurrence of each of these stress patterns was shown to vary across
Australia’s northern grains region (Fig. 5 ). This information is valuable in many
respects; on the one hand, it supports breeders identify sites with a high frequency
of particular stress types to make germplasm screening more effective and efficient;
while on the other hand, by understanding the frequency of occurrence of the differ-
ent stress environments, breeders and agronomists may be able to better match
available germplasm and management to particular sites. Clearly, improvements in
the skills and lead time of seasonal climate forecasts would add considerable value
to this application, as the environments could be predicted and the right combina-
tion of genotype and management identified prior to planting.

All yrs

5th−95th percentile;

Birchip

KW(all phases)P=<0.001 KW (all phases)P = <0.001

KS

AMD

IQR

KS

AMD

IQR

CN CP RF RR NZ CN CP RF RR NZ

<0.001 <0.001

134.4 310.3 763.4 530.1 442.2

2.14 0.61 1.31 0.64 1.23

0.004 <0.001<0.001 0.013 <0.001 <0.001<0.001<0.001

583.6 126.4 312.8 276.9 104.3

2.60

(a) (b)Dooen

4000 4000
3200
2400
1600
800

SOI phases

0

3200
2400
1600
800
0

0.72 1.64 0.52 1.04

a bc abc c ab ab ac a c ab

25th−75th percentile; Median Average

CN CP RF RR NZ All yrs CN CP RF RR NZ

Fig. 3 Simulated distribution of wheat yields at Birchip and Dooen, Australia, for all years and
each SOI phase in June/July. Local soils and management conditions were used in the analysis.
Means of ENSO phases with common letters below their respective box-and-whisker plot are not
significantly different (p≤0.05). The significance of the shift in the median in each phase away
from the all-years value was determined using a Kruskal-Wallis test (KW) while the similarity of
distributions was determined using the Kolmogorov-Smirnov P-value test (KS). The shift is mea-
sured using the absolute median difference (AMD) while changes in the distribution dispersion are
measured using the inter-quartile ratio (IQR) (Source: Anwar et al. 2008 )

D. Rodriguez et al.