Previous PageTable Of ContentsNext Page

Talking about the weather: APSIM, climate change and grain farmers on the Upper Eyre Peninsula, SA.

Bronya Alexander1, Peter Hayman1, Samantha Doudle2 and Nigel Wilhelm1

1 South Australian Research and Development Institute. GPO Box 397 Adelaide SA 5001. Email
South Australian Research and Development Institute. Box 31 Minnipa SA 5654. Email


Many grain farmers in low rainfall farming regions such as the Upper Eyre Peninsula (annual rainfall 350mm, growing season rainfall 250mm) are concerned about the long term future of the region when they hear climate change projections for warming and drying. A number of locations on the Eyre Peninsula have been analysed to understand current climate variability and risks. Although they have their limitations, spatial analogues can be a powerful way to consider what a warmer and drier environment might be like, and may also identify adaptation options.

We used the crop model APSIM (Agricultural Production Systems sIMulator) to simulate current wheat yields and to transfer projections in rainfall and temperature into grain yield.

Key Words

Climate change, Eyre Peninsula, APSIM


This paper presents a regional analysis of climate on the Upper Eyre Peninsula in order to understand current climate risk as a background for climate change scenarios. The study locations were Ceduna, Minnipa, Cowell, Kimba, Lock, Pt Neill and Cummins (Figure 1).

Figure 1: Climate analysis locations on the Eyre Peninsula, SA (Adelaide also shown). The background map shows a vegetation index for September 2005, with increasing vegetation from yellow to green. Corresponding soils from Whitbread and Hancock 2008 are given along with the Plant Available Water Content (PAWC).





Minnipa red sandy clay loam



Mitcheville Flat - sandy clay loam



Buckleboo sand over clay



Lock grey calcareous loamy sand



Charra grey calcareous loamy sand


A study from CSIRO (Suppiah et al 2006) carried out climate change projections on a regional basis using 13 Global Climate Models considered most suitable for South Australia from the IPCC fourth assessment report. Table 1 shows the wide range in temperature projections for the Eyre Peninsula region using future emission scenarios based on the Special Report on Emission Scenarios (SRES). There is even more uncertainty in rainfall projections, represented by the wide range in Table 2. It is clear that all models are suggesting a warming and drying trend (Figure 2), but the extent of the warming or drying varies considerably. Even if CO2 is stabilised at 450ppm instead of 550ppm, this does not change the lower end of the projections but it does greatly reduce the upper end. The projections from Suppiah et al (2006) are not substantially different in terms of the mean and range from the latest projections released in 2007 by CSIRO and the Bureau of Meteorology (

Table 1. Range of warming (C) projected for the Eyre Peninsula using SRES scenarios.








0.4 to 1.2

0.4 to 1.3

0.4 to 1.1

0.4 to 1.2

0.4 to 1.3


0.9 to 3.5

0.8 to 4.0

0.8 to 3.5

0.8 to 3.6

0.9 to 3.8

Table 2. Range of rainfall changes in percentage projected for the Eyre Peninsula using SRES scenarios.








-10 to -1

-9 to +4

-10 to +3

-12 to -2

-20 to -2


-30 to -2

-25 to +13

-30 to +8

-35 to -4

-60 to -4

Figure 2. Projected range of changes in annual temperature (oC) and rainfall (%) in 2030 and 2070 for Eyre Peninsula if carbon dioxide concentration by 2100 was stabilised at 450 ppm, 550 ppm and the SRES range of 540 ppm to 970 ppm (drawn from tables 6 to 10 of CSIRO report, Suppiah et al 2006).


A summary of simple climatic indices including average growing season rainfall and temperature are provided for the study sites on the Eyre Peninsula. All meteorological data has come from the SILO patched point dataset ( from 1900 to 2006.

The daily crop model APSIM (Keating et al 2002) has been used as a sophisticated tool for analysing climate information. To isolate the impact of climate, APSIM was run with the same wheat variety, soil, sowing rule and starting conditions but with different climate files (daily rainfall, temperature, radiation and vapour pressure deficit) for the study locations from 1900 to 2006. The soil was selected from the Minnipa region as a sandy clay loam with water holding capacity of 89 mm, as characterised for APSIM by Whitbread and Hancock (2008). To remove N supply as a limiting factor, a high amount (250kg) of nitrogen fertiliser (N) was added at sowing. To look at the impact of soil plus climate, these simulations were then repeated using representative soils for five of the Upper Eyre Peninsula locations (see Figure 1 for soil information).

APSIM has been used for many climate change studies in Australia, including South Australia (Luo et al 2003). As a simple test of the sensitivity of simulated wheat production for the Upper Eyre Peninsula we tested temperature, CO2 and rainfall changes for Minnipa with 250kg N applied at sowing.

Because there is such a wide range in climate change projections, we used a mild, moderate, severe and extreme scenario for Minnipa, this time with a more realistic amount of nitrogen added at sowing (as per Whitbread and Hancock 2008). A key question from farmers was on the outcomes for 2030, thus up to a 1.5C temperature increase and 480ppm CO2 concentration was used as these are at the highest end of the projected ranges for 2030. The scenarios used are as follows where T = temperature increase (C) and R = reduction in rainfall (mm) with CO2 concentration at 480ppm (compared to 380ppm with no change): Mild T1 R5; Moderate T1.5 R10; Severe T1.5 R20; Extreme T1.5 R40.

Results and discussion

The study locations cover moderate (350mm) to low (190mm) growing season rainfall (GSR) and correspondingly moderate to low simulated yields (Table 4). The growing season temperature (GST) ranges by about a degree and the warmer sites generally have a higher chance of days over 35C during Sept/Oct.

Table 4: Average growing season rainfall (Av GSR), average growing season temperature (Av GST) and average simulated wheat yield using APSIM for locations on the Eyre Peninsula from 1900 to 2006. Also shown are the probabilities of getting over 35 oC during September to October for at least 1, 2 and 3 days.


Probability of getting over 35C in Sept/Oct for at least:


Av GSR (mm)

Av GST (o C)

Av simulated yield (t/ha)

1 day

2 days

3 days















Pt Neill



































Figure 3. Average April to October rainfall (mm) and simulated wheat yields using APSIM with non-limiting amounts of soil nitrogen, for case study locations on the Eyre Peninsula. Other locations include Clare (GSR 487mm), Spalding (GSR 323mm), Jamestown (GSR 342mm), Peterborough (GSR 219mm), Orroroo (GSR 225mm), Carrieton (GSR 197mm) and Hawker (GSR 202mm).

The tight relationship between average growing season rainfall and average simulated wheat yield (Figure 3), using a constant Minnipa soil and high N, suggests that for the range of locations studied, average simulated yields are closely related to average growing season rainfall (R2 = 0.986). No doubt there is an impact from other climatic parameters such as growing season temperature, evaporation rates and timing of rainfall, but the impact on average yield seems to be minor compared to the total growing season rainfall. This relationship was also tested with a range of sites across the Upper and Mid North regions of South Australia and, provided the simulation had high N and constant soil, there was a similar tight relationship between average growing season rainfall and simulated yield (R2 = 0.9 for all locations in Figure 3). The slope of this line leads to a water use efficiency (WUE) of 15-20 kg/ha/mm of rainfall. This result would not be surprising if APSIM was programmed with a simple WUE relationship. However, APSIM is a daily time step model and the WUE value is calculated after yield is simulated from a complex interaction of factors. The WUE in any one year is highly variable at a particular location, for example at Minnipa the annual WUE fluctuated between 1-20kg/ha/mm (standard deviation of 3.5kg/ha/mm). When local soils were used instead of the constant Minnipa soil, there was very little difference in yields (7% or less). When selecting these sites, farmers and agronomists expected that sites on the eastern Eyre Peninsula such as Cowell would have a higher yield for a given growing season rainfall than western sites such as Ceduna. This was thought to be due to harsher spring conditions at Ceduna. Figure 3 suggests there is little difference in simulated WUE. Furthermore, even when more appropriate soils are used, this relationship does not change very much.

The sensitivity of temperature, CO2 and rainfall changes on simulated wheat production for Minnipa is shown in Figure 4. As temperature is increased up to 4C, the lower extent of yield, including the median, stays relatively constant but we see a decrease in the extent of higher yields. The simulated wheat yields are sensitive to CO2 levels. APSIM is also sensitive to rainfall as shown by heavily reduced median and spread in yields for a 40% reduction in rainfall.

Figure 4. APSIM simulated wheat yields for Minnipa showing no climate change, an increase in temperature of 1, 2, 3 and 4C (T1, T2, T3, T4), an increase in CO2 from 380ppm to 570ppm and 760ppm (C570, C760), and reduction in rainfall by 10%, 20%, 30% and 40% (R10, R20, R30 and R40). Non-limiting amounts of nitrogen were added at sowing. Small squares show the median, large boxes show the range from the 20th to the 80th percentile and minimum and maximum amounts are indicated by the whiskers.

When a combination of climate change scenarios are run for Minnipa (Figure 5), with N supply as per Whitbread and Hancock 2008 (sowing: 7kg/ha, soil: 69kg/ha NO3, 1.5kg/ha NH4), the results show much lower yields compared to Figure 4 in wetter years due to N limitations (note change in scale of axis between Figures 4 and 5). The results also show the high sensitivity to reduction in rainfall.

Figure 5. APSIM simulations for Minnipa showing no climate change along with combinations of temperature increase (C) and rainfall decrease (percent reduction in rainfall). Small squares show the median, large boxes show the range from the 20th to the 80th percentile and minimum and maximum amounts are indicated by the whiskers. CO2 concentration under “no change” was taken at 380ppm, compared to 480ppm in all other scenarios. N application as per Whitbread and Hancock 2008.

Concluding remarks

This simulation analysis suggests a strongly precipitation-dependant system. Some of this may be an artefact of APSIM not capturing the full impact of heat stress, however APSIM does capture the effects of higher temperature on phenology and some of the impact on water use. The high sensitivity to precipitation makes it difficult to predict the future as precipitation is the most uncertain aspect of climate change projections. However a highly precipitation-dependant system gives us more confidence in using spatial and temporal analogues to understand scenarios over coming decades and discuss adaptation options.


Luo Q, Williams MAJ, Bellotti W and Bryan B (2003). Quantitative and visual assessments of climate change impacts on South Australian wheat production. Agricultural Systems, 77, 173-186.

Suppiah R, Preston B, Whetton PH, McInnes KL, Jones RN, Macadam I, Bathols J and Kirono D (2006). Climate change under enhanced greenhouse conditions in South Australia. CSIRO Marine and Atmospheric Research, Aspendale, VIC, Australia.

Whitbread AM and Hancock J (2008). APSIM modelling for the Eyre Peninsula. CSIRO Sustainable Ecosystems, Glen Osmond, SA, Australia.

Keating BA, Carberry PS, Hammer GL, Probert ME, Robertson MJ, Holzworth D, Huth NI, Hargreaves JNG, Meinke H, Hochman Z, McLean G, Verburg K, Snow VO, Dimes JP, Silburn M, Wang E, Brown S, Bristow KL, Asseng S, Chapman SC, McCown RL, Freebairn DM, and Smith CJ (2002). An overview of APSIM, a model designed for farming systems simulation. European Journal of Agronomy, 18, 267-288.

Previous PageTop Of PageNext Page