Reliability of canola production in different rainfall ... - CSIRO Publishing

6 downloads 0 Views 231KB Size Report
The area of canola in the wheat-based farming systems of the wheatbelt of Western Australia (WA) ... Due to the short history of canola production in WA, there is.
CSIRO PUBLISHING

www.publish.csiro.au/journals/ajar

Australian Journal of Agricultural Research, 2007, 58, 326–334

Reliability of canola production in different rainfall zones of Western Australia Imma Farre´ A,D , Michael RobertsonB , and Senthold AssengC A

Department of Agriculture and Food, Western Australia, 3 Baron-Hay Court, South Perth, WA 6151, Australia. CSIRO Sustainable Ecosystems, Private Bag No. 5, Wembley, WA 6913, Australia. C CSIRO Plant Industry, Private Bag No. 5, Wembley, WA 6913, Australia. D Corresponding author. Email: [email protected] B

Abstract. The area of canola in the wheat-based farming systems of the wheatbelt of Western Australia (WA) expanded rapidly during the 1990s and has subsequently decreased. Due to the short history of canola production in WA, there is little information on yield and oil content expectations in relation to rainfall, location, and soil type. In this paper we: (1) present the recent history of canola production in the context of the long-term climate record; (2) assess the effect of location, rainfall, soil type, and soil water at sowing on yield and oil content; and (3) determine cut-off sowing dates for profitable canola production. Simulations were run using the APSIM-Canola model with long-term climate records for 3 selected locations from the low-, medium-, and high-rainfall zones and different soil types. Analysis of recent trends in canola area showed that poor seasons and price volatility in the last few years have contributed to farmers’ perception of risk and hence the decline in area sown. Long-term simulations showed the importance of location, sowing date, soil type, and stored soil water at sowing on grain yield. Yield was negatively related to sowing date. Light-textured soils had lower yields and larger yield penalties with delayed sowing than heavy-textured soils. Soil water at sowing gave a yield advantage in most years in all locations studied, but especially in low- and medium-rainfall locations. Variation in oil content was most strongly affected by sowing date and location, while soil type and soil water at sowing had a minor effect. Longterm simulation analysis can be used as a tool to establish the latest possible sowing date to achieve profitable canola for different locations and soil types, given different canola prices and growing costs. Given the vulnerability of profitability to seasonal conditions, in the low- and medium-rainfall zone, the decision to grow canola should be tactical depending on stored soil water, sowing opportunities, seasonal climate outlook, prices, and costs. In contrast, in the high-rainfall zone, canola production is relatively low risk, and could become a reliable component of rotations. Additional keywords: Brassica napus L., APSIM-Canola, yield, sowing date, profitability, risk.

Introduction Growers in Western Australia (WA) currently perceive canola (Brassica napus L.) as a risky crop. Between 1990 and 1999, growers responded to the introduction of diseaseresistant cultivars, improved agronomy, favourable seasons, and favourable prices by increasing the area sown from close to nil to 900 000 ha (Oilseeds 2005). Since 2000, the area of canola sown has gradually declined to 400 000 ha in 2005. In the period 2000 to 2004 the frequent occurrence of poor seasons with late sowing opportunities, due to late starts to the rainy season after a dry summer, and volatility of canola prices created the reality of risk of growing canola among growers (Addison and Carlton 2002). Other factors have contributed to the perception of risk compared with growing cereals: (1) high costs of growing canola; (2) high variability of yield; (3) low drought resistance; (4) the need for early sowing to achieve high yields; and (5) the association between low yields and low oil content. The consequences of canola disappearing from cereal-dominated cropping systems in WA include the loss of income diversity and weed and disease breaks. © CSIRO 2007

Sowing date is an important determinant of yield in canola. Canola yields decrease markedly with delay in sowing, but sowing date varies from year to year as it is determined by the date of the opening rains. Several studies have shown yield decline in canola with delay in sowing (Robertson et al. 1999a, 2004; Hocking and Stapper 2001; Farr´e et al. 2002). However, the rate of yield decline with delayed sowing is variable. Delay in sowing also causes reduction in oil content through the effect of higher temperature and enhanced water stress during seed filling (Hocking et al. 1997; Walton 1999; Robertson and Holland 2004). Both yield and oil content determine the profitability of the crop. As a consequence, knowledge of the response of yield and oil content to delay in sowing can be used to define optimum sowing dates for different cultivars, climates, and farming systems, to achieve profitability (Hodgson 1978). In a crop validation study, done for 6 locations and 3 years, Farr´e et al. (2002) showed yield reductions ranging from 3.2 to 8.6% per week delay in sowing. Therefore, to obtain a more generalised response of canola to delay in sowing, the findings 10.1071/AR06176

0004-9409/07/040326

Reliability of canola production

Australian Journal of Agricultural Research

327

Materials and methods Analysis of trends in canola production in WA In order to put the last 15 years of canola production in the context of long-term climate, seasonal rainfall for the period 1990– 2004 was compared with the previous 90-year period (1900–89) (Table 1). Simulated sowing dates and simulated yields for the last 15 years (1990–2004) were also compared with values for the previous 90 years (1900–89) (Table 1). Comparison between the 1990s period (1990–99) and the last 5 years (2000–04) was also performed for simulated yields, sowing dates, and seasonal rainfall. T-tests were performed to compare the periods 1990–2004 with 1900–89 and to compare the periods 2000–04 with 1990–99 (Table 1). Crop statistics and prices at farm gate (corrected for inflation using 2004 as a base-line) were used to analyse the trends in the area and prices of canola sown in WA (Oilseeds 2005).

of the validation work need to be extended to the long-term climate record and coupled to economics to examine profitability implications. In Mediterranean-type environments the effect of stored soil water on yield can be significant in some circumstances (Aboudrare et al. 2006). In some years, summer rainfall and/or soil water conservation practices (e.g. summer fallow weed control) may contribute to stored soil water at sowing, and alleviate the effect of below-average seasonal rainfall. In this paper we explore the effect of stored soil water at sowing on yield expectation and hence the extent to which stored soil water could be used as an indicator of the decision to sow a crop. With limited experience in growing canola in WA the current perceptions of risk of production are likely to be biased towards recent seasons. Simulation modelling offers a powerful tool to relate recent experience to the long-term context and to define risk management strategies. A validated crop model together with long-term weather data can be used to study the interaction between sowing date, location and soil type to assess risk of crop production (e.g. Asseng et al. 2001). This information can be used to obtain a realistic assessment of the prospects for canola in the wheatbelt of WA. Moreover, this assessment is needed given that during the last 30 years the region has seen a significant reduction in winter rainfall (Smith et al. 2000), and climate change may continue this trend into the future. However, the emphasis of this paper is on the effects of climate variability on canola production and not on the effects of climate change. We used the APSIM-Canola model coupled to long-term climate data from contrasting rainfall locations in WA to test the hypotheses: (1) in low-rainfall locations or marginal soil types profitable canola crops can be produced reliably in those years when favourable conditions occur (i.e. early sowing, stored soil water at sowing, and high canola prices); and (2) in high-rainfall locations or better soils profitable canola crops can be produced reliably over the long term. In order to test these hypotheses we proposed the following questions.

Model overview APSIM is a modelling framework that allows submodels (or modules) to be linked to simulate agricultural systems (Keating et al. 2003). Four modules, a specific crop module (CANOLA), a soil water module (SOILWAT2), a soil nitrogen module (SOILN2), and a residue module (RESIDUE2), are linked within APSIM v. 3.6 to simulate the cases described in this paper. The canola module has been developed using a framework described by Robertson et al. (2002), and the soil modules are described by Probert et al. (1995, 1998). APSIM has been extensively tested in the wheatbelt of Western Australia for wheat (Asseng et al. 1998) and canola (Farr´e et al. 2002) and elsewhere in Australia (Robertson et al. 1999b; Robertson and Holland 2004). The APSIM-Canola model simulates crop development (phenology), growth, yield, water uptake, and nitrogen accumulation in response to temperature, radiation, day length, soil water, and nitrogen supply. The model uses a daily timestep and is driven by daily weather inputs. It calculates the water-limited potential yield, i.e. the yield not limited by weeds, pests, and diseases, but limited only by temperature, solar radiation, water, and nitrogen supply. In our analysis the effect of waterlogging was not accounted for.

(1) Are the last 15 years of canola production in WA representative of the long-term climate expectations? (2) What is the effect of seasonal variability, location, soil type, sowing date, and plant-available water at sowing on yield and oil content? (3) What are the latest possible sowing dates to achieve profitable canola production in different rainfall locations and soil types in WA for different canola prices and growing costs?

Scenario analysis Scenario analyses were run to examine the effects of seasonal variability, location, soil type, sowing date, and plant-available water at sowing, on grain yield and oil content. Oil content was simulated using the algorithm described by Robertson and

Table 1.

Mean values of simulated yield (kg/ha), simulated sowing date (first sowing opportunity), and April–October rainfall (mm) for Kojonup, Lake Grace, and Merredin for different periods Simulated yields on a duplex soil using the sowing rule described in the text

No. of years

Period Kojonup

Yield Lake Grace

Merredin

Kojonup

Sowing date Lake Grace

Merredin

Kojonup

Apr.–Oct. rain Lake Grace Merredin

90 15

1900–89 1990–04

2568 2470

1327 1313

872 1034

139 145

147 155

150 156

453 401

268 241

233 237

10 5

1990–99 2000–04

2602 2206

1556 826

1146 810

145 144

153 160

156 157

431 341

271 182

252 207

Australian Journal of Agricultural Research

Historical trends of canola production in WA Western Australian grain growers responded enthusiastically to the release of blackleg-resistant canola cultivars in the 1990s such that the sown area increased from near zero in 1990 to a peak of 900 000 ha in 1999. Then the prices fell sharply and in the following years so did the area sown, to stabilise around 400 000 ha in the last 5 years (Fig. 1). Canola yields have fluctuated from year to year since the early 1990s, but with a general trend upwards, due in part to the concentration of canola in high-yielding environments (Fig. 1). In the 1990s there was 30, 50, and 15% of the canola area in the high-, medium-, and low-rainfall areas, respectively, whereas in the latter years this proportion has changed to 50, 40, and 5%, respectively (data not shown). In the 1990s there was a sequence of favourable seasons (early breaks and good seasonal rainfall) (Fig. 2) that encouraged the rapid increase in the area of canola grown. The 2000

1000 800 600

1000 400

Yield (kg/ha)

1500

500

200

0

04 20

02 20

00 20

98 19

96 19

94 19

92

0

19

Experiment 2. One initial soil water condition for 5 fixed sowing dates This was designed to assess the penalties associated with delays in sowing within a given season. The soil water profile was initialised (reset) each year on 1 January to the lower limit of plant-available water. Every year the canola crop was sown on 5 fixed sowing dates 20 days apart: 5 April (DOY 95), 25 April (DOY 115), 15 May (DOY 135), 4 June (DOY 155), and 24 June (DOY 175). Germination occurred when soil water was adequate (>50% plant-available water (PAW) in 0–0.05 m

Results

90

Experiment 1. Four initial soil water conditions with one sowing rule The soil water profile was initialised (reset) each year in different ways: (1) reset on 1 January to the lower limit of plant-available water, to represent maximum water use from previous crop; (2) reset on the date of sowing with 0, 20, or 40 mm of plantavailable soil water in the surface layers of the profile. Different levels of soil water at sowing were designed to represent varying degrees of plant-available water at sowing due to summer rainfall and/or incomplete water use from previous crop. As sowing dates vary due to the timing of autumn and early winter rains, sowing time was allowed to vary in the simulations each year according to a sowing rule. A sowing window was set between 1 May (DOY 121) and 15 June (DOY 166). Sowing occurred when at least 20 mm of rainfall had accumulated within 5 days. If the sowing rule was not met within the sowing window, then sowing was forced on 16 June, so that a crop was sown in every year of the climate record.

depth). Inspection of the results showed that at all locations in 50% of the seasons, seedling emergence occurred between 9 and 14 days after sowing. Model outputs for grain yield and oil content were analysed in various ways. Distributions of yield and oil content were summarised as cumulative probability distributions. The effects of sowing date and starting soil water at sowing on yield and oil content were examined. The response of grain yield to sowing date in terms of kg/ha decline per week’s delay in sowing date was calculated as the slope coefficients for the linear regressions fitted between sowing date and grain yield for every year and then the slopes were averaged for 105 years. These yield declines per week delay in sowing date were also expressed as a percent of the maximum yield. The effect of starting soil water on yield was studied by plotting cumulative probability distributions of the yield difference between sowing on a dry soil profile and sowing on 20 or 40 mm of plant-available water in the soil. The effect of variation in canola price and production costs on profitability was also examined by calculating the break-even yields for combinations of different canola prices and growing costs (production costs). The break-even yield was defined as that which is required to cover growing costs for a given grain price.

19

Holland (2004), in which oil content is calculated at the end of season as a function of average temperature and water stress during the grain-filling period, with an upper bound of 47%. Long-term simulations were run for 3 locations using the climate record for 1900–2004. The locations were selected as representative of the high- (Kojonup; 33.84◦ S, 117.15◦ E), medium- (Lake Grace; 33.10◦ S, 118.46◦ E), and low- (Merredin; 31.50◦ S, 118.22◦ E) rainfall zones of the wheatbelt of Western Australia. Long-term (1900–2004) average April–October rainfall was 446 mm in Kojonup, 264 mm in Lake Grace, and 234 mm in Merredin. Three typical soil types (Asseng et al. 2001) of the area were used in the simulations: a sandy soil with 59 mm plant-available water in the root-zone of 1.70 m, a duplex soil with 86 mm plant-available water in the root-zone of 0.80 m, and a clay soil with 116 mm plant-available water in the root-zone of 1.50 m. The choice of locations and soil types was designed to represent the extremes of environments likely to be experienced by canola crops in the wheatbelt of Western Australia. To run the simulations, previous crop residues were set each year on 1 January to 2000 kg/ha; residue type was assumed as wheat, with a C : N ratio of 70. Soil N was reset at day of sowing each year, with 100 kg mineral N/ha in the soil profile in each simulation, followed by 2 in-crop fertiliser applications of 100 kg N/ha, designed to ensure that N did not limit yields in wet seasons. Cultivar Karoo, an early–medium maturity triazinetolerant type, typical of that grown by farmers in the area, was sown to establish 80 plants/m2 . Two simulation experiments were run, differing in sowing time and plant-available soil water present in the profile.

I. Farr´e et al.

Area ('000 ha)

328

Year



Fig. 1. Trends for the period 1990–2004 in yield ( ) and area (N ) of canola grown in Western Australia.

Reliability of canola production

Australian Journal of Agricultural Research

600

400

Sand Duplex Clay

Price ($/t)

300

200 200 100

1.0 0.8 0.6 0.4

04

0.2

20

02 20

00 20

98 19

96 19

94 19

92

0 19

19

90

0

Apr.–Oct. rain (mm)

(a)

400

329

Year

0.0



good season of 1999 was followed by 3 consecutive years with late breaks and low seasonal rainfall (Fig. 2). The low price of 1999, followed by 3 poor seasons, was associated, among other reasons, with a gradual decline in the area sown from 900 000 to 300 000 ha. The price of canola had an upward trend from the early 1990s to the mid 1990s, followed by a plateau to the late 1990s. The price decreased markedly in 1999 and has been volatile since then, with marked fluctuations (Fig. 2). Simulations for different periods for the 3 locations studied showed that the last 15 years were similar to the previous 90-year period in terms of water-limited potential yields, sowing dates, and seasonal rainfall, except for Kojonup where the seasonal rainfall in the last 15 years was lower than previously (Table 1). The t-test analysis showed that the 1990s period (1990–99) was not significantly different from the previous 90-year period (1900–89). The period 2000–04 was not significantly different from the previous 10 years (1990–99) in sowing dates. However, expected yields in Lake Grace and seasonal rainfall in Kojonup and Lake Grace were significantly lower in the last 5 years than in the previous 10 years (Table 1). The last 15 years are, therefore, representative of the long-term climate expectations. The last 5 years were drier than previously. Production risk Simulated grain yields for 3 locations and 3 soil types sown according to a sowing rule are summarised in Fig. 3 as cumulative distribution functions (CDFs). Overall the results demonstrate the large variability in yields generated by climate variability in the region. Among locations, average yields were highest at the high-rainfall location Kojonup and lowest at the low-rainfall location Merredin. Among soil types, average yields were highest on the clay soil and lowest on the sandy soil. Maximum yields varied from 1900 kg/ha on the sandy soil at Merredin to 3500 kg/ha on the clay soil at Kojonup. Minimum yields varied from zero at Lake Grace and Merredin on a clay soil to 200 kg/ha at Kojonup on a clay soil. At Kojonup the median yield ranged from 2100 kg/ha on the sandy soil to 2700 kg/ha on the duplex soil. At Lake Grace the median yield varied from 1000 kg/ha on the sandy soil to 1300 kg/ha on the duplex soil. At Merredin the median yields ranged from 550 kg/ha on the sandy and clay soil to 800 kg/ha on the duplex soil. The results indicate

(b) 1.0

Cumulative probability

Fig. 2. Trends for the period 1990–2004 in canola price ( ) in Western Australia, adjusted to a base-line of 2004 using the consumer price index and seasonal rainfall (April–October) (bars) for Lake Grace Shire.

0.8 0.6 0.4 0.2 0.0 (c) 1.0 0.8 0.6 0.4 0.2 0.0 0

1000

2000

3000

4000

Yield (kg/ha) Fig. 3. Simulated cumulative distribution functions of simulated grain yield for (a) Kojonup, (b) Lake Grace, and (c) Merredin for 3 soil types. Each curve is composed of 105 seasons.

the importance of rainfall, location, and soil type in determining the overall level of production. Importance of sowing date The simulations with fixed sowing dates showed that different locations had different responses of grain yield to delayed sowing. In general, all locations showed a decline in yield with delayed sowing. For Kojonup there was no yield decline between the first 2 sowing dates at the beginning and end of April and there was a linear decline for sowing dates after the end of April. The linear decline in grain yield ranged from 141 kg/ha.week in Kojonup to 71 kg/ha.week in Merredin (Table 2). Expressing the yield decline as a percentage of the maximum yield, it varied between 5.8%/week on a sandy soil in Lake Grace and 4.2%/week on a clay soil in Kojonup (Table 2).

330

Australian Journal of Agricultural Research

I. Farr´e et al.

Table 2. Simulated response of grain yield to sowing date in terms of kg/ha decline per week’s delay in sowing date and yield reduction (%) expressed as a per cent of the maximum yield Yield declines are expressed as the average of yield declines for the 105 seasons Location

Soil

Kojonup

Lake Grace

Merredin

kg/ha.week

% Yield/week

Sand Duplex Clay

141 140 131

5.6 4.3 4.2

Sand Duplex Clay

108 116 100

5.8 5.0 5.0

Sand Duplex Clay

80 94 71

5.6 5.5 4.5

As a consequence, yields for different seasons tended to converge with delayed sowing. Among soil types, yield reductions with delayed sowing were highest on sandy soil and lowest on clay soil (Table 2). On sandy soils, yield reductions as a percentage were similar in all 3 locations, whereas on duplex and clay soils, percentage yield reductions were smaller at Kojonup. Simulations using a sowing rule showed that within a location and soil type, grain yield was moderately correlated with sowing date (Table 3), with R2 values less than 0.5 in all cases. The slope and the intercept of the relationship between yield and sowing date were greater at Kojonup than at Lake Grace or Merredin, indicating that Kojonup had higher yields for early sowings, but yields tended to converge at the 3 locations for late sowings. Oil content Simulated oil content was affected by rainfall and location. There was little effect of starting soil water or soil type on oil content (data not shown). Simulated oil content ranged from 37% to the upper limit in the model of 47% across locations and soil types. Median oil content was 47, 44, and 42% for Kojonop, Lake Grace, and Merredin, respectively. The cut-off for a bonus/penalty in grain price for canola is currently 42% oil content. The probability of exceeding this cut-off was 100, 84, and 45% of seasons for Kojonop, Lake Grace, and Merredin, respectively. Oil content declined with a delay in sowing date in all locations. The slope of the regression between oil content and sowing date was lower in Kojonup (–0.07) than in Lake Grace (–0.10) or Merredin (–0.11). The correlation between oil Table 3.

content and sowing date was greater in the medium- and low-rainfall (r2 = 0.66–0.68) location than in the high-rainfall (r2 = 0.41–0.42) location (Table 3). Low-yielding seasons and late sowings were associated with low oil contents and low yield, causing reduced profitability of canola crops. The correlation between yield and oil content was greater in the high-rainfall (r2 = 0.57–0.64) location than in the medium- and low-rainfall (r2 = 0.34–0.40) location (Table 3). Importance of starting soil water Starting soil water, defined as plant-available water at sowing, produced a yield advantage over sowing on a dry soil in all 3 locations and soil types (Fig. 4). The yield advantage due to starting soil water was greater on the duplex soil than on the sandy soil, due to the lower plant-available water capacity of the sandy soil. Starting soil water produced a yield advantage in every year in the low- and medium-rainfall locations, but in the high-rainfall location produced a yield penalty of 15–40 kg/ha in ∼5% of the years, due to increased losses of nitrogen by leaching. Starting soil water of 20 mm provided a yield advantage in all locations. Starting soil water of 40 mm provided additional yield advantage in Lake Grace and Merredin, but not in Kojonup. In Kojonup, starting soil water gave a median yield advantage of 22–66 kg/ha across soil types, representing a median yield increase of 1–2% compared with a dry soil profile. In Lake Grace, the median yield advantage from initial soil water was 217–475 kg/ha, representing a median yield increase of 36–71%. In Merredin, the median yield advantage from starting soil water was 142–368 kg/ha, representing a 47–97% increase in yield. The results indicated the importance of starting soil water in determining overall level of production, especially in the lowand medium-rainfall locations. Table 4 shows the frequency of having initial water at sowing. In Kojonup, more than 80% of the years had more than 20 mm of plant-available water at sowing and more than 40% of the years had more than 40 mm. In Lake Grace and Merredin, more than 50% of the years had more than 20 mm and more than 20% of the years had more than 40 mm. Among soil types, the clay soil had a lower proportion of years with high plant-available water than the sandy or duplex soils due to higher evaporation losses on the clay soil. Break-even yields for different canola prices and growing costs Growing costs (i.e. variable costs) of canola are higher than of wheat. Growing costs for canola production in the 3 shires

Associations between simulated grain yield (kg/ha) and sowing date, oil content (%) and sowing date, and oil content and grain yield for Kojonup, Lake Grace, and Merredin, on sandy and duplex soils

Location

Soil R2

Yield v. sowing day Slope Intercept

R2

Oil v. sowing day Slope Intercept

R2

Oil v. yield Slope Intercept

Kojonup

Sand Duplex

0.44 0.44

–30.5 –28.5

6252 6544

0.42 0.41

–0.07 –0.07

56 56

0.57 0.64

0.002 0.002

42.53 41.09

Lake Grace

Sand Duplex

0.37 0.29

–19.5 –20.8

3906 4409

0.66 0.66

–0.10 –0.10

58 58

0.40 0.34

0.002 0.002

41.38 41.40

Merredin

Sand Duplex

0.34 0.30

–15.6 –20.3

3021 3953

0.67 0.68

–0.11 –0.11

59 59

0.40 0.35

0.003 0.002

40.03 40.25

Reliability of canola production

Australian Journal of Agricultural Research

(a)

(b)

20 mm

1.0

331

40 mm 0.8 0.6 0.4 0.2

Cumulative probability

0.0 (c)

(d )

(e)

(f )

1.0 0.8 0.6 0.4 0.2 0.0 1.0 0.8 0.6 0.4 0.2 0.0 0

200

400

600

800

0

200

400

600

800

1000

Yield advantage from soil water at sowing (kg/ha) Fig. 4. Cumulative distribution functions of simulated yield advantage of 20 and 40 mm of plantavailable water at sowing compared with lower limit at sowing for (a, b) Kojonup, (c, d ) Lake Grace, and (e, f ) Merredin on a sandy soil (a, c, e) and on a duplex soil (b, d, f ). Each curve is composed of 105 seasons. Table 4. Simulated % of years with available soil water at sowing greater than 20 or 40 mm for the 3 locations and 3 soil types Soil

% years with starting soil water Kojonup Lake Grace Merredin

>20 mm

Sand Duplex Clay

94 96 84

80 81 66

76 77 58

>40 mm

Sand Duplex Clay

57 58 44

41 41 26

41 42 27

studied ranged from 230 $/ha in Merredin to 290 $/ha in Kojonup in 2005 according to the ‘Gross margins guide 2005 WA’ prepared by regional economists of the WA Department of Agriculture. Variation among regions is mainly a function of fertiliser inputs. In comparison, growing costs for wheat production in the same shires ranged from 180 to 220 $/ha.

Canola price has been variable in the last few years (Fig. 1). The latest (2006) estimate of net price on-farm for canola is around 400 $/t. Growing costs and the price of canola determine the minimum yield required to cover costs (the ‘break-even’). Table 5 shows minimum yields required to break even for combinations of different canola prices and growing costs. In this calculation, canola price did not account for oil content. If we consider an average canola price for the last few years of ∼300 $/t and growing costs of 250–300 $/ha, depending on yield expectations, the break-even yield is 800–1000 kg/ha. The probability of achieving this break-even yield depends on location, soil type, and sowing date (Fig. 5). For example, in Kojonup, 1000 kg/ha can be achieved in more than 80% of years on a sandy soil if sown before the end of May and in more than 95% of years on a duplex soil for any sowing date within the sowing window. In Lake Grace, 1000 kg/ha can be achieved in 80% of years on a sandy soil if sown before the end of April and on a duplex soil if sown before the middle of May. In contrast, in Merredin on a sandy soil, 1000 kg/ha can be achieved in 50%

332

Australian Journal of Agricultural Research

Table 5.

I. Farr´e et al.

Break-even yield (kg/ha) for combinations of different canola prices and growing costs

Canola price ($/1000 kg)

200

225

Growing costs ($/ha) 250 275

250 300 350 400 450 500

800 670 570 500 440 400

900 750 640 560 500 450

1000 830 710 630 560 500

1100 920 790 690 610 550

300

325

1200 1000 860 750 670 600

1300 1080 930 810 720 650

end of May on the sandy soil and end of June on the duplex soil (Fig. 3). With the sowing rule used in the simulations, the earliest possible sowing opportunity in 50% of the years (median) is mid to end of May for Kojonup, and end of May to mid June for Lake Grace and Merredin. With these sowing dates, a breakeven yield of 1000 kg/ha would be achieved in most of the years in Kojonup but in Merredin it would be achieved in less than 20% of the years (Fig. 5). Discussion Production risk in different rainfall zones High rainfall

(b)

(a) 100 80 60 40

% Years

20 0 (c)

(d )

100 80 60 40 20 0 21/3 10/4 30/4 20/5

9/6

21/3 10/4 30/4 20/5

9/6

29/6

Sowing date Fig. 5. Proportion of years (%) with simulated grain yield greater than 500 ( ), 1000 ( ), 1500 (䉲), 2000 (䉮), 2500 (䊏), 3000 (䊐), 3500 ( ) and 4000 () kg/ha v. sowing date for (a, b) Kojonup and (c, d ) Merredin on a sandy (a, c) and duplex (b, d ) soil.







of years if sown before the end of April and in less than 10% of years if sown before the end of May. In Merredin on a duplex soil, 1000 kg/ha can be achieved in 70% of years if sown before the end of April and in 30% of years if sown before the end of May. The latest possible sowing date to achieve profitability in canola production will depend on the canola price and growing costs for the season, which will determine the break-even yield for that year. Figure 5 shows that if we assume, for example, a break-even yield of 1000 kg/ha, with the soil types and the cultivar used in the simulations, the latest possible sowing date to achieve a yield of 1000 kg/ha in 50% of the years in Kojonup is mid June on the sandy soil and end of June on the duplex soil. In Merredin the latest possible sowing date to achieve a yield of 1000 kg/ha in 50% of the years is end of April on the sandy soil and mid June on the duplex soil. If we assume a year with high canola price and/or low growing costs so that the break-even yield is 500 kg/ha, the latest possible sowing dates to achieve a yield of 500 kg/ha in 50% of the years in Merredin would be

The results presented here show that the reliability of canola production varies widely among the different rainfall zones in the WA wheatbelt. At Kojonup, typical of the high-rainfall zone, the simulated median yield varies between 2000 and 3000 kg/ha (Fig. 3), depending upon soil type. With break-even yields of 500, 1000, and 1300 kg/ha (Table 5), which represent the expected range with historical prices of $250–500/t (Fig. 1), there is a better than 80% chance that canola will be profitable in this zone. Kojonup typically has earlier opportunities to sow than the low- or medium-rainfall zones (Table 1), and lesser penalties from delayed sowing after April. Despite stored soil water at sowing being at least 20 mm in 80–90% of seasons (Table 4) there is no advantage for yield in 70% of seasons (Fig. 4). In such a high-rainfall environment, in-crop rainfall is enough to meet the environmental potential in most seasons. These simulations do not account for waterlogging effects, and it is known that in above-average rainfall seasons on duplex soils, transient waterlogging can affect canola production in the highrainfall zone (Zhang et al. 2004). Medium rainfall At Lake Grace, typical of the medium-rainfall zone, the median yield varies between 1000 and 1500 kg/ha, depending upon soil type (Fig. 3). With break-even yields of 500, 1000, and 1300 kg/ha (Table 5), reliability of canola is considerably less than at Kojonup, with a better than 70, 50, and 40% chance that canola will be profitable at these break-even yields, respectively. Lake Grace has similar timing of sowing opportunities to the low-rainfall zone. The reliability of canola in this medium-rainfall zone will be under threat if canola prices fall, costs continue to rise, and the current downward trend in winter rainfall is maintained. In such circumstances, risk minimisation strategies are needed to improve the reliability of yield (see below). Low rainfall At Merredin, typical of the low-rainfall zone, the median yield varies between 500 and 800 kg/ha, depending upon soil type (Fig. 3). With break-even yields of 500, 1000, and 1300 kg/ha (Table 5), reliability of canola is considerably less than at Lake Grace, with only a better than 50, 30, and 10% chance, respectively, that canola will be profitable. This highlights that the profitability of canola in the low-rainfall zone is highly sensitive to changes in costs and prices and is currently

Reliability of canola production

a risky proposition in most seasons, particularly with late sowing opportunities. Latest sowing date The latest profitable sowing date for a given break-even, depends on location and soil type. For example, the latest possible sowing dates to achieve a yield of 1000 kg/ha in more than 50% of the years are mid to end of June for Kojonup, mid to end of May for Lake Grace, and end of April for Merredin. Historical trends of canola production in WA Given the risk profiles described above for the different rainfall zones it is not surprising that the last 15 years of canola production in WA have seen wide fluctuations in the area under production and a heightened awareness of the risks of growing the crop. The effect of the recent sequence of belowaverage seasons on canola area can be seen by examining the risk of production in the different rainfall locations. Yield expectations from the 2000–04 seasons for the medium- (Lake Grace) and low-rainfall (Merredin) zones were around 50 and 70%, respectively, of the yields expected in the 1990–99 period (Table 1) when the canola industry was expanding rapidly in WA. The lower yields in 2000–04 were associated with a 4–10 day later start to the season and considerably lower April–October rainfall than in the 1990–99 period. Comparing 2000–04 with 1990–99, the lower April–October rainfall in Kojonup in the last 5 years had little effect on expected yields. In Lake Grace, the lower April–October rainfall resulted in lower expected yields. In Merredin there were no differences in rainfall or yields for the 2 periods. Clearly, recent climate variability in WA has most severely affected the low- and medium-rainfall zones. This effect is likely to worsen given that since 1970, winter rainfall has declined by 10–20% in the south-western part of Australia and future climate change scenarios predict more rainfall reductions for this region (Smith et al. 2000). Risks associated with seasonal variability have been reinforced by fluctuations in the canola price. Large changes in prices of $100/1000 kg between one season and the next (Fig. 1) will shift the break-even yield for a set cost of production by 200–300 kg/ha. A shift of this magnitude at the 2 riskiest locations of Merredin and Lake Grace will translate into about a 30% change in the risk of breaking even. At Kojonup, the risk of not breaking even is so small that a shift of $100/1000 kg in the canola price will not materially change the risk profile in this rainfall zone. Despite the recent sequence of poor seasons, the last 15 years of canola production in WA are representative of the long-term climate. Risk management strategies Given the vulnerability of profitability in canola to changes in climate and price in the low- and medium-rainfall zones it is useful to explore the options for risk minimisation. Availability of stored soil water at sowing may provide a trigger for those seasons when risks will be less than average. At Lake Grace, stored soil water at sowing is estimated to be at least 20 mm in 65–80% of seasons, assuming good summer stubble cover and a weed-free fallow. In 50% of seasons, 20 mm of stored soil water will increase yield by at least 200 kg/ha. In only half of

Australian Journal of Agricultural Research

333

the seasons will an additional 20 mm (40 mm total) make any further difference. At Merredin, stored soil water at sowing is estimated to be at least 20 mm in 60–75% of seasons. Similarly to Lake Grace, in 50% of seasons, 20 mm of stored soil water will increase yield by at least 200 kg/ha. In only half of the seasons will an additional 20 mm (40 mm total) make any further difference. At both locations an extra 200 kg/ha of yield will translate into about a 30% change in the risk of breaking even (Fig. 2). This indicates that in the low- and medium-rainfall zones, maintenance of weed-free fallows with good stubble cover will return benefits in reducing the risk of low canola yields. Within a location, yield advantage of starting soil water was greater on the duplex soil. The simulations reported here also emphasised the value of timely sowing in risky environments. In dryland farming systems, timely sowing at the opening rains is not always possible because of the need for tillage, weed control, and sowing of higher priority crops. The relationship between yield and delayed sowing across all rainfall zones gave a rate of loss of 4–6% per week. These rates are similar to those reported in other Australian studies (Robertson et al. 1999a, 2004; Farr´e et al. 2002). At the riskier 2 locations of Lake Grace and Merredin, a delay of 2 weeks would translate into 150–200 kg/ha lower yield potential (Table 2). As seen above, this is equivalent to having ∼20 mm of stored soil water at sowing, or a change of $100/1000 kg in the canola price. At current costs and prices, the latest possible sowing date to achieve a 50% chance of producing a breakeven yield of 800–1000 kg/ha is the end of April at Merredin on a sandy soil and by mid May on a duplex soil (Fig. 5). In contrast, at Kojonup, with current growing costs and prices, break-even yield would be achieved in more than 80% of years on both soil types. It therefore appears that early arrival of the opening rains in the low- and medium-rainfall locations provides a signal for lower-than-average risk in those seasons and could be used as a trigger to indicate opportunistic use of canola in the cropping system. The decision of whether to sow canola or not of course should be evaluated against the profitability of other viable alternatives. The effect of delayed sowing on profitability will be greater than that reported here due to the lower oil content that is associated with later sowings. For a delay of 2 weeks in sowing at Lake Grace and Merredin, this would translate to a 1.4% fall in oil content (Table 3). Under the Australian payment system for canola, grain with oil content of greater than 42% attracts a bonus of 1.5%, while that below 42% attracts a penalty of 1.5%. Conclusions This analysis has highlighted the vulnerability of profitable canola production to climate variability, costs of production, prices for canola grain, and management in the low- and medium-rainfall zones of WA. The recent declines in the area of canola production have been associated with poorer seasons and fluctuating prices. However, the last 15 years have been found representative of the long-term climate expectations for WA. This study has quantified the risks to production in the high-, medium-, and low-rainfall zones of WA so that objective measures can be used to judge the riskiness of the crop in the farming system.

334

Australian Journal of Agricultural Research

The reliability of canola production varies widely among the different rainfall zones in the WA wheatbelt. In the highrainfall zone, there is a better than 80% chance that canola will be profitable for a wide range of canola prices and growing costs. Profitability of canola in the medium- and low-rainfall zones is highly sensitive to changes in costs and prices and is currently a risky proposition in most seasons, particularly with late sowing opportunities. In the medium- and low-rainfall zones of the WA wheatbelt, optimal management (i.e. early sowing, high prices, low growing costs) can go some way towards mitigating the risks of profitable canola production. Timing of the opening rains or sowing (before end of May in medium-rainfall and before end of April in low-rainfall locations) and level of stored soil water at sowing (more than 20 mm) are indicators that could be used to trigger the sowing of canola on an opportunistic basis in the lowand medium-rainfall zones. Such strategies will be necessary to avoid the possibility of canola disappearing as a permanent feature of cereal-dominated cropping systems in WA with the attendant negative consequences such as increase of weeds and diseases. Acknowledgments We thank Graham Walton for providing access to data on canola production and area.

References Aboudrare A, Debaeke P, Bouaziz A, Chekli H (2006) Effects of soil tillage and fallow management on soil water storage and sunflower production in a semi-arid Mediterranean climate. Agricultural Water Management 83, 183–196. doi: 10.1016/j.agwat.2005.12.001 Addison B, Carlton P (2002) Getting the best out of canola in the low rainfall central wheatbelt. Western Australia Crop Updates, Department of Agriculture WA. Asseng S, Fillery IRP, Dunin FX, Keating BA, Meinke H (2001) Potential deep drainage under wheat crops in a Mediterranean climate. I. Temporal and spatial variability. Australian Journal of Agricultural Research 52, 45–56. doi: 10.1071/AR99186 Asseng S, Keating BA, Fillery IRP, Gregory PJ, Bowden JW, Turner NC, Palta JA, Abrecht DG (1998) Performance of the APSIM-wheat model in Western Australia. Field Crops Research 57, 163–179. doi: 10.1016/S0378-4290(97)00117-2 Farr´e I, Robertson MJ, Walton GH, Asseng S (2002) Simulating phenology and yield response of canola to sowing date in Western Australia using the APSIM model. Australian Journal of Agricultural Research 53, 1155–1164. doi: 10.1071/AR02031 Hocking PJ, Kirkegaard JA, Angus JF, Gibson AH, Koetz EA (1997) Comparison of canola, Indian mustard and Linola in two contrasting environments. I. Effects of nitrogen fertilizer on dry-matter production, seed yield and seed quality. Field Crops Research 49, 107–125. doi: 10.1016/S0378-4290(96)01063-5 Hocking PJ, Stapper M (2001) Effects of sowing time and nitrogen fertiliser on canola and wheat, and nitrogen fertiliser on Indian mustard. I. Dry matter production, grain yield, and yield components. Australian Journal of Agricultural Research 52, 623–634. doi: 10.1071/AR00113

I. Farr´e et al.

Hodgson AS (1978) Rapeseed adaptation in northern New South Wales. II. Predicting plant development of Brassica campestris L. and Brassica napus L. and its implications for planting time, designed to avoid water deficit and frost. Australian Journal of Agricultural Research 29, 711–726. doi: 10.1071/AR9780711 Keating BA, Carberry PS, Hammer GL, Probert ME, Robertson MJ et al. (2003) An overview of APSIM, a model designed for farming systems simulation. European Journal of Agronomy 18, 267–288. doi: 10.1016/S1161-0301(02)00108-9 Oilseeds WA (2005) Canola production benchmarks WA. http://www. australianoilseeds.com (accessed 26 September 2005). Probert ME, Dimes JP, Keating BA, Dalal RC, Strong WM (1998) APSIM’s water and nitrogen modules and simulation of the dynamics of water and nitrogen in fallow systems. Agricultural Systems 56, 1–28. doi: 10.1016/S0308-521X(97)00028-0 Probert ME, Keating BA, Thompson JP, Parton WJ (1995) Modelling water, nitrogen, and crop yield for a long-term fallow management experiment. Australian Journal of Experimental Agriculture 35, 941–950. doi: 10.1071/EA9950941 Robertson MJ, Carberry PS, Huth NI, Turpin JE, Probert ME, Poulton PL, Bell M, Wright GC, Yeates SJ, Brinsmead RB (2002) Simulation of growth and development of diverse legume species in APSIM. Australian Journal of Agricultural Research 53, 429–446. doi: 10.1071/AR01106 Robertson MJ, Holland JF (2004) Production risk of canola in the semi-arid subtropics of Australia. Australian Journal of Agricultural Research 55, 525–538. doi: 10.1071/AR03219 Robertson MJ, Holland JF, Bambach R (2004) Response of canola and Indian mustard to sowing date in the grain belt of north-eastern Australia. Australian Journal of Experimental Agriculture 44, 43–52. doi: 10.1071/EA02214 Robertson MJ, Holland JF, Bambach R, Cawthray S (1999a) Response of canola and Indian mustard to sowing date in risky Australian environments. In ‘Proceedings of the 10th International Rapeseed Congress’. Canberra, ACT. (CD-ROM) Robertson MJ, Holland JF, Kirkegaard JA, Smith CJ (1999b) Simulating growth and development of canola in Australia. In ‘Proceedings of the 10th International Rapeseed Congress’. Canberra, ACT. (CD-ROM) Smith IN, McIntosh P, Ansell TJ, Reason CJC, Mcinnes K (2000) Southwest Western Australian winter rainfall and its association with Indian Ocean climate variability. International Journal of Climatology 20, 1913–1930. doi: 10.1002/1097-0088(200012)20:153.0.CO;2-J Walton GH (1999) Cultural practices and their effects on canola yield and oil in Western Australia. In ‘Oilseeds crop updates’. (www.agric.wa.gov.au/cropupdates/1999/oilseeds/Walton.htm) Zhang H, Turner NC, Poole ML (2004) Yield of wheat and canola in the high-rainfall zone of south-western Australia in years with and without a transient perched watertable. Australian Journal of Agricultural Research 55, 461–470. doi: 10.1071/AR03122

Manuscript received 28 May 2006, accepted 22 January 2007

http://www.publish.csiro.au/journals/ajar