INTRODUCTION
The temperate dairy regions of Southeastern Australia predominantly utilizes pastures comprising of perennial ryegrass (Lolium perenne L.), often sown with the perennial legume white clover (Trifolium repens L.) (Read et al. Reference Read, McGowan, Reed, Ward, Lowe, Kaehne, Bartsch and Mason1991; Fulkerson & Doyle Reference Fulkerson and Doyle2001). However, many farms strategically use a range of annual and perennial forage crops to supplement the forage supply in periods of low pasture growth or nutritive value (Rawnsley Reference Rawnsley and Campbell2007), to better match animal feed demand to forage supply, to improve the productivity on a per land area basis (Garcia et al. Reference Garcia, Fulkerson and Brookes2008), or improve resource use efficiency of inputs including water and nitrogen (Garcia et al. Reference Garcia, Fulkerson and Brookes2008; Neal et al. Reference Neal, Fulkerson and Hacker2011). More recently there has been interest in fully integrating forage crops into pasture-based dairy systems with the aim of improving productivity and resilience of dairy farms in the face of a changing and variable climate (Chapman et al. Reference Chapman, Kenny, Beca and Johnson2008b , Reference Chapman, Kenny and Lane2011). The most recent survey of feed sources used on Australian dairy farms identified that, on average in the south east Australian dairy regions, forage crops contribute 19% of the forage component of the milking cow's diet (Barlow Reference Barlow2008). Despite the integral role that forage crops have on dairy farms there is a paucity of information on the likely responses of these crops to future changes in climate in southeastern Australia. Such information will be required by producers, policy makers and the industry as a whole as they attempt to adapt farming systems and practices to become resilient to current and future climatic variability, and more extreme weather events expected in the future.
Future climatic projections for the southeastern dairy regions of Australia generally indicate that the region will become warmer by between 1 and 4 °C with either a decrease or no change in annual rainfall (CSIRO & BOM 2007; Holz et al. Reference Holz, Grose, Bennett, Corney, White, Phelan, Potter, Rawnsley, Parsons, Lisson, Gaynor and Bindoff2010). These projections, however, are often uncertain due to the number of different climate models available, a range of possible future greenhouse gas emission scenarios and the large spatial and temporal resolution of the models used for these projections. While methodologies exist to downscale climatic projections to spatial and temporal scales more relevant to agricultural production (Corney et al. Reference Corney, Katzfey, McGregor, Grose, Bennett, White, Holz, Gaynor and Bindoff2010) and to help identify the most suitable models to use (Smith & Chandler Reference Smith and Chandler2010), these methods still produce projections with a large range of variability and uncertainty.
An alternative method of assessing crop production under future climates is to use the range of possible climatic variables within a biophysical modelling framework to assess the resilience or sensitivity of agricultural production to scaled changes in climatic variables (Cullen et al. Reference Cullen, Eckard and Rawnsley2012). This approach overcomes the uncertainties associated with the use of model-generated future climatic projections while still answering the questions the end user wishes to address (i.e. how resilient is a given production system to future climate change?). Using this method, Cullen et al. (Reference Cullen, Eckard and Rawnsley2012) identified that temperate pastures across southeastern Australia are generally resilient to 1–2 °C increases in temperature, while the response to more extreme changes in climate was dependent on the species composition of the pasture and the location.
The agricultural production system simulator (APSIM) is a crop simulation platform used around the world to assess complex interactions between climate, soils, crops and management (Keating et al. Reference Keating, Carberry, Hammer, Probert, Robertson, Holzworth, Huth, Hargreaves, Meinke, Hochman, McLean, Verburg, Snow, Dimes, Silburn, Wang, Brown, Bristow, Asseng, Chapman, McCown, Freebairn and Smith2003). The APSIM framework integrates sub-models describing soil, crop and farm management processes with weather data in a mechanistic manner to simulate crop growth and development as well as soil water and nitrogen dynamics (Keating et al. Reference Keating, Carberry, Hammer, Probert, Robertson, Holzworth, Huth, Hargreaves, Meinke, Hochman, McLean, Verburg, Snow, Dimes, Silburn, Wang, Brown, Bristow, Asseng, Chapman, McCown, Freebairn and Smith2003). Through integration with the livestock enterprise modules from the Grazplan and AusFarm models (McCown et al. Reference McCown, Moore, Holzworth, McDonald and Bellotti1993; Freer et al. Reference Freer, Moore and Donnelly1997), it is also capable of simulating livestock production within mixed farming systems. The major use of the APSIM framework is to explore long-term farming systems questions for broad acre cropping systems (e.g. the grazing v. harvesting of cereal crops (Bell et al. Reference Bell, Hargreaves, Lawes and Robertson2009) and the use of summer crops to prevent recharge into aquifers (Wang et al. Reference Wang, Cresswell, Yu and Verburg2008)). Recently the model has been shown to appropriately represent the factors affecting forage crop growth and development in southeastern Australian regions (Pembleton et al. Reference Pembleton, Rawnsley, Jacobs, Mickan, O'Brien, Cullen and Ramilan2013a ), and is now being used to explore risks and optimize crop management in intensive forage cropping systems (Pembleton & Rawnsley Reference Pembleton and Rawnsley2012; Pembleton et al. Reference Pembleton, Harrison, Rawnsley, Michalk, Millar, Badgery and Broadfoot2013b ). The model also has a framework to represent the nutritive value of forage (Bell et al. Reference Bell, Hargreaves, Lawes and Robertson2009). Despite popularity throughout the world only minimal effort has been expended to fully parameterize the model to reflect the effect of elevated atmospheric carbon dioxide (CO2) on crop growth, though studies on maize using APSIM to investigate agricultural production under future climates have shown promising results (Harrison et al. Reference Harrison, Tardieu, Dong, Messina and Hammer2014). The work of Reyenga et al. (Reference Reyenga, Howden, Meinke and McKeon1999) using the APSIM wheat module has also shown that such parameterization is possible and the framework exists in many of the crop modules (Wang et al. Reference Wang, Robertson, Hammer, Carberry, Holzworth, Meinke, Chapman, Hargreaves, Huth and McLean2002).
In the present study reported, parameters to enable the APSIM crop models to reflect the influence of elevated atmospheric CO2 concentrations on forage crop production were developed. The APSIM model was then used to examine the resilience of a range of forage crops to projected changes in climate at three locations in the southeastern Australian dairy regions.
MATERIALS AND METHODS
Sites and cropping systems
The locations in southeastern Australia used for the current study were Dookie in northern Victoria, Terang in southwestern Victoria and Elliott in northwestern Tasmania. These locations were chosen as being representative of the broad climatic conditions that the southeastern Australian dairy regions encompass. The prevailing climatic and edaphic conditions at each location are provided in Table 1.
Table 1. Soil type, drained upper limit (DUL) and lower limit (LL; soil water content at −1500 kpa) in the surface 1200 mm of soil at each location and the average daily maximum and minimum temperature, monthly rainfall and evaporation, during the period of the simulation study (1971–2010)
* Isbell (Reference Isbell2002).
† Pembleton et al. (Reference Pembleton, Rawnsley and Donaghy2011); APsoil (2010).
‡ Calculated from SILO patched-point data (www.longpaddock.qld.gov.au/silo).
At each location the growth of forage crops (wheat (Triticum aestivum L.), oats (Avena sativa L.), annual ryegrass (Lolium multiflorum Lam.), forage rape (Brassica napus L.), sorghum (Sorghum bicolor (L.) Moench.), maize (Zea mays L.) and lucerne (Medicago sativa L.)) was simulated using APSIM (version 7.3) over 40 years using climate data from the years 1971 to 2010 and management rules that were developed in consultation with local agronomists and dairy forage researchers working at the locations (Table 2). Lucerne was simulated as a perennial crop (i.e. sown once at the start of each simulation). For the lucerne simulations, the crop was defoliated when it reached the flowering growth stage irrespective of the time of year. When lucerne was simulated under irrigated conditions, a winter-active type was used, while under dryland conditions a winter-dormant type was used. This is reflective of the specific adaptation of winter-dormant types of lucerne to dryland conditions and the specific adaptation of winter active types of lucerne to irrigated conditions (Pembleton et al. Reference Pembleton, Donaghy, Volenec, Smith and Rawnsley2010a , Reference Pembleton, Smith, Rawnsley, Donaghy and Humphries b ). Irrigation of the irrigated lucerne crops were scheduled using a soil water deficit of 30 mm as the irrigation trigger.
Table 2. Crop agronomic management for the annual forage crops simulated for Dookie and Terang, Victoria and Elliott, Tasmania as part of the study
DAS, days after sowing; NA, not applicable; SWD, soil water deficit, GS, growth stage according to Zadoks et al. (Reference Zadoks, Chang and Konzak1974).
* See Pembleton et al. (Reference Pembleton, Rawnsley, Jacobs, Mickan, O'Brien, Cullen and Ramilan2013a ) for details.
Wheat and oat crops at all locations were simulated with their corresponding crop modules (Wang et al. Reference Wang, van Oosterom, Meinke, Asseng, Robertson, Huth, Keating, Probert, Unkovich and O'Leary2003; Peake et al. Reference Peake, Whitbread, Davoren, Braun, Limpus and Unkovich2008) under dryland conditions, while maize crops were simulated under irrigation conditions with the maize module (Carberry et al. Reference Carberry, Muchow and McCown1989). Lucerne growth was simulated at Elliott and Dookie with the lucerne module (Robertson et al. Reference Robertson, Carberry, Huth, Turpin, Probert, Poulton, Bell, Wright, Yeates and Brinsmead2002) under both irrigated and dryland conditions at Dookie and Elliott, but only under dryland conditions at Terang. Forage sorghum crops were simulated under dryland conditions at Dookie and Terang and under both irrigated and dryland conditions at Elliott. The canola module (Robertson et al. Reference Robertson, Holland, Kirkegaard and Smith1999) with the forage rape cultivar described in Pembleton et al. (Reference Pembleton, Rawnsley, Jacobs, Mickan, O'Brien, Cullen and Ramilan2013a ) was used to simulate the growth of forage rape crops, while annual ryegrass crops were simulated with the annual ryegrass ecotype in the weed module (Deen et al. Reference Deen, Cousens, Warringa, Bastiaans, Carberry, Rebel, Riha, Murphy, Benjamin, Cloughley, Cussans, Forcella, Hunt, Jamieson, Lindquist and Wang2003). Forage rape growth was simulated under both irrigated and dryland conditions at Dookie and Elliott but only under dryland conditions at Terang. Annual ryegrass was simulated under dryland conditions at all locations. To initialize soil carbon, nitrogen and water conditions, each simulation had a 10-year lead-in period in which the growth of a dryland pasture was simulated with the AgPasture module (Li et al. Reference Li, Snow and Holzworth2011) using baseline climate data from the period of 1961 to 1970. The simulation results from this period were not used in any subsequent analysis.
Model modifications
Prior to the commencement of the simulation study, several modifications to the APSIM model were undertaken. These included the addition of parameters to the oat and canola modules to allow them to be grazed by the livestock module and the development of a forage-specific cultivar within the canola module. These additions are described in Pembleton et al. (Reference Pembleton, Rawnsley, Jacobs, Mickan, O'Brien, Cullen and Ramilan2013a ). When lucerne growth was simulated, specific manager rules to control the expression of winter dormancy were included as described in Pembleton et al. (Reference Pembleton, Rawnsley and Donaghy2011).
Crop and pasture responses to growth under elevated atmospheric carbon dioxide concentrations
The parameterization of crop responses (photosynthesis, transpiration efficiency and plant nitrogen content) to growth under elevated atmospheric CO2 concentration in APSIM has been fully undertaken for wheat (Reyenga et al. Reference Reyenga, Howden, Meinke and McKeon1999). However, for the other species used in the current study, only the influence of elevated atmospheric CO2 concentration on photosynthesis has been parameterized (Wang et al. Reference Wang, Robertson, Hammer, Carberry, Holzworth, Meinke, Chapman, Hargreaves, Huth and McLean2002). To develop the crop-specific parameters to describe changes to transpiration efficiency and nitrogen concentration for the crops other than wheat, a review of previously published studies investigating the transpiration efficiency and crop nitrogen concentration response of each crop species to growth under elevated CO2 was undertaken. In this review, preference was given to studies investigating the response of swards rather than individual or spaced plants. From the range of responses observed, functions were developed to account for the relative increase/decrease in transpiration efficiency and plant nitrogen concentration against the increase in atmospheric CO2 concentration. The shape of these functions was chosen based on the functions described by Reyenga et al. (Reference Reyenga, Howden, Meinke and McKeon1999) for the APSIM wheat module. Tables 3 and 4 present a summary of the literature review and the modifier functions developed for each crop species. The modifier functions developed were then incorporated into each crop module as for those incorporated for wheat described by Reyenga et al. (Reference Reyenga, Howden, Meinke and McKeon1999).
Table 3. Published values of the relative increase in crop transpiration efficiency with increases in atmospheric CO2 concentration and the regressions developed to modify crop transpiration efficiency in APSIM. Functions were developed by fitting regressions to the reported increase in transpiration efficiency to the increase in atmospheric CO2
* Values for oilseed rape (canola).
† Values for Brassica juncea.
‡ Values for Brassica nigra.
§ Values for Brassica carinata.
¶ Values for grain sorghum.
‖ Values for perennial ryegrass.
Table 4. Published values of the relative decrease in plant nitrogen concentration with increases in atmospheric CO2 concentration and the modifier functions developed from them to modify plant nitrogen concentration in APSIM. Functions were developed by fitting regressions to the reported decrease in plant nitrogen concentration to the increase in atmospheric CO2
* Values from Brassica juncea.
† Values from Brassica oleracea.
‡ Values from grain sorghum.
§ Values from perennial ryegrass.
¶ Modifier function derived from the wheat module in APSIM.
To evaluate the validity of the modifier functions that were developed, simulations were undertaken to compare the impact of elevated atmospheric CO2 on forage crop production without additional changes in temperatures and rainfall to those observed in free air CO2 enrichment (FACE) experiments containing these crop species (online supplementary Table S1, available from http://journals.cambridge.org/AGS). The proportional change in dry matter (DM) yield, leaf area index (LAI), tissue nitrogen (N) concentration, transpiration and DM digestibility (DMD) under elevated CO2 compared with simulations undertaken with ambient CO2 concentrations was then compared to the proportional change observed in previously published FACE studies of each crop species or closely related crop species (online supplementary Table S1). Where no species-specific responses have been reported in the literature, the modelled response was compared with the generic responses based on species functional group reported in Ainsworth & Long (Reference Ainsworth and Long2005).
Climatic scaling
Climatic datasets used in the current study were obtained as a patched-point dataset in the APSIM file format from the SILO database (www.longpaddock.qld.gov.au/silo), which has been developed based on the methods described by Jeffrey et al. (Reference Jeffrey, Carter, Moodie and Beswick2001). This format includes data for daily minimum and maximum temperatures, daily rainfall and daily solar radiation. For each simulation this data were scaled with a 0, 1, 2, 3 or 4 °C increase in daily minimum and maximum temperatures (based on the findings reported in CSIRO & BOM (2007)) with corresponding atmospheric CO2 concentrations of 380, 435, 535, 640 and 750 ppm, respectively (creating the scenarios from here on referred to as baseline, T1R0, T2R0, T3R0 and T4R0). This was based on the predictions by the Intergovernmental Panel on Climate Change (IPCC 2000). These scenarios were run either with no change in daily rainfall or with a respective 10% decrease in daily rainfall for every 1 °C increase in temperature up to a 30% decrease in daily rainfall (creating the scenarios from here on referred to as T1R10, T2R20, T3R30 and T4R30). These changes in temperature and rainfall were selected to be consistent with projections for climate change in southern Australia (CSIRO & BOM 2007). The scaling was applied evenly to all days in the year. Daily solar radiation was not scaled in any of the future climate scenarios.
Simulation outputs and calculations
For each simulation the outputs included grazing yield, silage yield, total biomass yield and, if applicable, total irrigation inputs. Dry matter digestibility was also an output for the crop modules that had the capacity to simulate that process (wheat and forage sorghum). Grazing yield was calculated as the sum of the growth rate from the date that the crop becomes available for grazing until the crop is terminated or locked up for silage and the forage available for grazing at the start of this period. Silage yields (if applicable) were the crop biomass on the date of the silage harvest. Irrigation input was the sum of the irrigation water applied to the crop from sowing to conclusion.
RESULTS
Response of crops simulated with APSIM to elevated atmospheric carbon dioxide concentration
The effects of elevated CO2 on crop biomass, LAI, crop N concentration, transpiration and DMD for the species are presented in Figs 1–5. For wheat the simulated biomass, transpiration and LAI responses were similar to those observed in FACE experiments, while the crop N concentration and crop DMD modelled responses were within the range reported from FACE experiments (Fig. 1).
Fig. 1. The mean simulated effect (error bars represent the range in values) of elevated carbon dioxide (CO2) (c. 600 ppm) on crop biomass, leaf area index, tissue nitrogen (N) concentration, crop transpiration and crop dry matter digestibility (DMD) of dryland wheat and oats crops compared with previously published effects observed in FACE experiments as outlined in online supplementary Table S1 (represented by the grey areas) and the effect reported for the most relevant function plant group reported in the meta-analysis and literature review undertaken by Ainsworth & Long (Reference Ainsworth and Long2005) (represented by the vertical dotted lines) where that data were available. An effect less than 1 indicates a decrease, while an effect greater than 1 indicates an increase.
Fig. 2. The mean simulated effect (error bars represent the range in values) of elevated carbon dioxide (CO2) (c. 600 ppm) on crop biomass, leaf area index, tissue nitrogen (N) concentration, and crop transpiration of irrigated maize and dryland annual ryegrass crops compared with previously published effects observed in FACE experiments as outlined in online supplementary Table S1 (represented by the grey areas and closely related species represented by the vertical broken lines) and the effect reported for the most relevant function plant group reported in the meta-analysis and literature review undertaken by Ainsworth & Long (Reference Ainsworth and Long2005) (represented by the vertical dotted lines) where that data were available. An effect less than 1 indicates a decrease, while an effect greater than 1 indicates an increase.
Fig. 3. The mean simulated effect (error bars represent the range in values) of elevated carbon dioxide (CO2) (c. 600 ppm) on crop biomass, leaf area index, tissue nitrogen (N) concentration, and crop transpiration of irrigated and dryland forage rape crops compared with previously published effects observed in FACE experiments as outlined in online supplementary Table S1 (represented by the grey areas) and the effect reported for the most relevant function plant group reported in the meta-analysis and literature review undertaken by Ainsworth & Long (Reference Ainsworth and Long2005) (represented by the vertical dotted lines) where that data were available. An effect less than 1 indicates a decrease, while an effect greater than 1 indicates an increase.
Fig. 4. The mean simulated effect (error bars represent the range in values) of elevated carbon dioxide (CO2) (c. 600 ppm) on crop biomass, leaf area index, tissue nitrogen (N) concentration, crop transpiration and crop dry matter digestibility (DMD) of irrigated and dryland forage sorghum crops compared with previously published effects observed in FACE experiments as outlined in online supplementary Table S1 (represented by the grey areas and closely related species represented by the vertical broken lines) and the effect reported for the most relevant function plant group reported in the meta-analysis and literature review undertaken by Ainsworth & Long (Reference Ainsworth and Long2005) (represented by the vertical dotted lines) where that data were available. An effect less than 1 indicates a decrease, while an effect greater than 1 indicates an increase.
Fig. 5. The mean simulated effect (error bars represent the range in values) of elevated CO2 (c. 600 ppm) on crop biomass, LAI, tissue nitrogen (N) concentration, and crop transpiration of irrigated winter active and dryland winter-dormant lucerne crops compared with previously published effects observed in FACE experiments as outlined in online supplementary Table S1 (represented by the grey areas) and the effect reported for the most relevant function plant group reported in the meta-analysis and literature review undertaken by Ainsworth & Long (Reference Ainsworth and Long2005) (represented by the vertical dotted lines) where that data were available. An effect less than 1 indicates a decrease, while an effect greater than 1 indicates an increase.
The mean simulated change in oat biomass was greater than the accepted response for this crops functional species group. However, the range in modelled responses overlapped this accepted range. The modelled change in oats LAI was within the accepted response range for this crop's functional species group. The modelled change in tissue N concentration overlapped the range reported by Ainsworth & Long (Reference Ainsworth and Long2005).
The modelled response of annual ryegrass biomass and LAI to elevated CO2 was similar to the responses observed in the field under FACE (Fig. 2). For this species there was also an overlap in the modelled values and observed values for crop transpiration and crop N content. This held true across the three locations.
The modelled biomass response for maize was greater than that observed in the field but was within the accepted response range for annual C4 grasses (Ainsworth & Long Reference Ainsworth and Long2005). The results of simulations of maize grown at Elliott reflect the observed FACE response for LAI. However, at the other two locations the modelled increase in LAI was greater than that observed under FACE conditions. The simulated change in tissue N concentration for maize crops grown at Elliott and Terang were within the range of responses observed under FACE conditions, while at Dookie this response was within the accepted range for the crop's functional species group.
The response of dryland forage rape biomass to elevated CO2 was similar to that observed under FACE (Fig. 3). The response of the irrigated forage rape was lower than the FACE observations, but was within the range of previously published responses for this species. The change in LAI of irrigated forage rape grown at Terang fell within the range observed under FACE conditions, while the mean response observed for dryland forage rape at Terang and both dryland and irrigated forage rape at Dookie was above the observed FACE range. However, the range of the modelled response overlapped with the range of observed values. Only the crop N concentration of forage rape grown at Terang and the dryland forage rape grown at Elliott had an overlap between the modelled and accepted response for the plant functional group to which forage rape belongs.
The modelled response to elevated CO2 of irrigated sorghum biomass was similar to that observed in FACE experiments (Fig. 4). The response of dryland forage sorghum was lower than the FACE observations for this crop species but was within the range of observations for annual C4 grasses. The simulated LAI response to elevated CO2 of irrigated forage sorghum at Dookie and Terang and dryland forage sorghum at Elliott and Terang were in line with observations from FACE experiments. For the other irrigation and location combinations the response was within the accepted response range of functional species group this crop belongs to. For the forage sorghum grown at Terang and at Dookie under elevated CO2, the decrease in tissue N concentration was also within the accepted range defined by Ainsworth & Long (Reference Ainsworth and Long2005) for annual C4 grasses. At Elliott, the response was greater than the accepted range. However, the decrease in tissue N concentration was less than this range for the crops grown at Elliott. Similar to the tissue N response, the change in transpiration of forage sorghum grown at Dookie and Terang was within the range observed in past FACE experiments. The modelled change in forage DMD with elevated CO2 was within the range reported for this crop from FACE experiments.
At all locations the simulated change in irrigated lucerne biomass was similar to that observed under FACE conditions (Fig. 5). In contrast, the change in the simulated dryland lucerne biomass was twice that of the response observed in the FACE experiments. However, the range of responses observed for Terang and Elliott overlapped the response observed under FACE. The modelled LAI response of dryland winter-dormant lucerne was above the range for temperate legumes defined by Ainsworth & Long (Reference Ainsworth and Long2005), while the change in LAI for winter-active lucerne fell within this range. The range in the predicted response of tissue N concentration of both winter-dormant and winter-active lucerne overlapped the accepted response range for temperate legumes.
Crop responses to the future climate scenarios
Forage crop yield
The mean simulated yield of each crop grown at each location under the baseline climate scenario is presented in Tables 5 and 6. Under dryland conditions wheat, oats and annual ryegrass had greater yields than forage rape and forage sorghum at all three locations (Table 5). Irrigation improved crop yield of forage rape and lucerne at Dookie and Elliott, and the yield of forage sorghum at Elliott (Table 5 compared with Table 6).
Table 5. Mean simulated annual yield (t DM/ha) under the baseline climate scenario and the change in the yield relative to the baseline yield of dryland forage crops grown at Elliott Tasmania, Dookie Victoria and Terang Victoria under the future climate scenario of a 1, 2, 3 and 4 °C increase in temperature and no change in rainfall and a 1, 2, 3 and 4 °C increase in temperature with a 10, 20, 30 and 30% respective decrease in rainfall. Increases in air temperatures of 1, 2, 3 and 4 °C were associated with atmospheric CO2 concentrations of 435, 535, 640 and 750 ppm, respectively, while the baseline scenario had an atmospheric CO2 concentration of 380 ppm. Values in parenthesis are the coefficients of variation (CV)
Table 6. Mean simulated annual yield (t DM/ha) under the baseline climate scenario and the change in yield relative to the baseline yield of irrigated forage crops grown at Elliott Tasmania and Dookie Victoria under the future climate scenario of a 1, 2, 3 and 4 °C increase in temperature and no change in rainfall. Increases in air temperatures of 1, 2, 3 and 4 °C were associated with atmospheric CO2 concentrations of 435, 535, 640 and 750 ppm, respectively, while the baseline scenario had an atmospheric CO2 concentration of 380 ppm. Values in parenthesis are the coefficients of variation (CV)
Lucerne and annual ryegrass were the crops that were most responsive to changes in climate, while forage sorghum and forage rape were the least responsive. At all locations, the total yield of wheat, oats, annual ryegrass and lucerne increased under all future climate scenarios (Table 5). At Elliott, for the T1R0, T2R0, T3R0 and T4R0 scenarios, forage sorghum yield increased above the baseline. Under the same climate scenarios there was no change in forage sorghum yield at Terang, and at Dookie the yield of this crop decreased compared with the baseline scenarios. For the T1R10, T2R20, T3R30 and T4R30 scenarios the yield of forage sorghum decreased at all locations. The yield of forage rape grown at Elliott and Terang increased in the T1R0, T2R0, T1R10 and T2R20 scenarios. At Dookie, forage rape yield increased in T1R0, T2R0, T3R0 and T4R0 scenarios. However, in the T1R10, T2R20, T3R30 and T4R30 scenarios forage rape yield decreased.
Inter-annual variability in dryland crop yield (as indicated by the coefficient of variation (CV)) at Terang or Dookie remained static or decreased in the T1R0, T2R0, T3R0 and T4R0 scenarios. At Elliott under these conditions, inter-annual yield variability of winter-dormant lucerne decreased, while that of forage rape increased. A similar response occurred at Elliott for the T1R10, T2R20, T3R30 and T4R30 scenarios. This response was reversed at Terang and Dookie. Furthermore at both locations the inter-annual variability in the yield of annual ryegrass increased as rainfall decreased.
As temperatures and atmospheric CO2 concentrations increased, the yield of irrigated forage sorghum and winter-active lucerne grown at Elliot and winter-active lucerne grown at Dookie increased above the baseline yield (Table 6). At Elliott, the yield of irrigated forage rape and maize initially increased above the baseline yields for T1R0 and T2R0 scenarios but then decreased in the T3R0 and T4R0 scenarios. At Dookie, the yield of irrigated forage rape was unresponsive to changes in temperature and CO2 concentration. The yield of maize crops grown at Dookie decreased below the baseline yield in the T1R0, T2R0, T3R0 and T4R0 scenarios. At Elliott, the CV of total yield decreased for irrigated forage sorghum and increased for forage rape under the future climate scenarios compared with the baseline scenarios. At Dookie, the CV of crop yield of all irrigated crops remained consistent across the future climate scenarios.
Seasonality of forage supply
For both wheat and oats, the proportion of yield that was grazed during the winter months increased above the baseline in the future climatic scenarios (Fig. 6). This trend was consistent across all three locations. The proportion of yield that was grazed was slightly lower in the T1R10, T2R20 and T3R30 scenarios compared to the T1R0, T2R0 and T3R0 scenarios.
Fig. 6. The proportion of total annual production that is grazed (black) or ensiled (grey) of forage wheat and oats for the baseline climate and scenarios with +1, +2, +3 °C, +1 °C with −10% rain, +2 °C with −20% rain and +3 °C with −30% rain at Dookie, Victoria (left panels), Terang, Victoria (middle panels) and Elliott, Tasmania (right panels). Increases in air temperatures of 1, 2 and 3 °C were associated with atmospheric CO2 concentrations of 435, 535 and 640 ppm, respectively, while the baseline scenario had an atmospheric CO2 concentration of 380 ppm.
Spring and summer were consistently the periods with the greatest lucerne growth across all locations, irrigation conditions and climate scenarios (Fig. 7). While small, the proportion of lucerne yield grown during winter increased as temperatures were increased above the baseline scenarios for lucerne grown under dryland conditions at all three locations and for the lucerne grown under irrigation at Dookie. This increase in winter growth was not observed for the lucerne grown at Elliott under irrigated conditions.
Fig. 7. The proportion of DM yield that is available in summer (black), autumn (white), winter (dark grey) and spring (light grey) of lucerne crops grown at Dookie under dryland conditions (far left panel), Dookie under irrigated conditions (inner left panel), Terang under dryland conditions (middle panel), Elliott under dryland conditions (inner right panel) and Elliott under irrigated conditions (far right panel). When lucerne growth was simulated under dryland conditions the winter-dormant genotype was used while under irrigated conditions a winter active genotype was used. Increases in air temperature of 1, 2 and 3 °C were associated with atmospheric CO2 concentrations of 435, 535 and 640 ppm, respectively, while the baseline had an atmospheric CO2 concentration of 380 ppm.
Forage digestibility under future climate scenarios
There was only a very minor influence of future climate scenarios on the DMD of forage sorghum and wheat. The greatest decrease in forage sorghum digestibility, a 6% decrease in DMD from the baseline scenario, was observed at Elliott under the T4R30 scenario. Changes in DMD were small for forage wheat (<0·5%).
Changes in irrigation requirements of crop species
Compared with the baseline scenario, the irrigation requirement at Dookie decreased under the future climate scenarios (Table 7) for all crops, except for lucerne grown under the T1R10 scenario and forage rape grown under the T3R30 scenario. A similar response was observed for irrigated lucerne grown at Elliott. For forage sorghum and maize grown at Elliott, there was an increase in the irrigation requirement for all the future climate scenarios and this increase was greatest for the scenarios that had a decrease in rainfall. For forage rape grown at Elliott, the irrigation requirement decreased for the T1R0, T2R0 and T3R0 scenarios. However, for the T1R10, T2R20 and T3R30 scenarios there was an increase in the crops irrigation requirement.
Table 7. The change in irrigation requirement (%) from that of the baseline scenario (mm) of irrigated forage crops grown at Dookie, Victoria and Elliott Tasmania under the +1, +2, +3 °C, +1 °C with −10% rain, +2 °C with −20% rain and +3 °C with −30% rain climate scenarios. The baseline, +1, +2 and +3 °C scenarios were associated with an atmospheric CO2 concentrations of 380, 435, 535 and 640 ppm, respectively
DISCUSSION
Overall the forage crop species examined in the current study with the exception of forage sorghum and forage rape showed resilience to potential future changes in climate with either an increase in yield with increasing temperature and atmospheric CO2 concentrations and either no change or a minimal decrease in DM yield with decreasing rainfall. Forage sorghum yield decreased with a 10% or greater reduction in rainfall while forage rape yield decreased with a 30% decrease in rainfall. Consequently it can be concluded that annual forage crops examined in the current study will remain viable forage options for the southeastern Australian dairy regions into the future. However, before this conclusion is accepted the underlying assumptions within the model should be considered. As part of the study, parameters were developed to describe the species specific adaptation to increasing atmospheric CO2 concentrations, namely the responses of crop transpiration efficiency and crop tissue N concentration. Prior to modification in the current work, the model already represented the response of photosynthesis in C3 and C4 plants to increase in atmospheric CO2 concentrations (Reyenga et al. Reference Reyenga, Howden, Meinke and McKeon1999). An increase in transpiration efficiency in plants exposed to elevated atmospheric CO2 concentrations occurs because the stomata do not have to open as wide for leaf internal CO2 concentration to be optimum for photosynthesis and hence reduce water loss (Nie et al. Reference Nie, He, Kirkham and Kanemasu1992). Tissue N concentration decreases with increasing atmospheric CO2 concentrations due to changes in the balance of the photosynthetic carbon reduction cycle and the photorespiratory cycle (Conroy & Hocking Reference Conroy and Hocking1993) and the dilution of N in the additional biomass grown. In APSIM, these responses are incorporated into the model through modifier functions for transpiration efficiency and plant N concentration (Reyenga et al. Reference Reyenga, Howden, Meinke and McKeon1999; Wang et al. Reference Wang, Robertson, Hammer, Carberry, Holzworth, Meinke, Chapman, Hargreaves, Huth and McLean2002). While it was possible to develop these functions from published data, the number and spread of data points for crop transpiration efficiency of annual ryegrass, forage sorghum and oats and tissue N concentration of maize, annual ryegrass and oats were limited. This is a clear gap in the literature and increasing the amount of information relating to the responses of these crops to elevated CO2 should be a focus of future research effort. Most of the data used to develop the relationships were from atmospheric CO2 concentrations between 350 and 750 ppm. Consequently, model use should be constrained to atmospheric CO2 concentrations between these values.
A visual assessment of the validity of the modifier functions for transpiration efficiency and plant N concentration modifiers was made by comparing the relative response of crop DM yield, LAI, tissue N concentration and transpiration to elevated atmospheric CO2 (without changes in other climatic parameters) to the relative response observed in FACE studies. For all the annual crops except maize and forage rape, the responses observed in the FACE studies were similar to those predicted by APSIM at each of the locations. Even when there was no data from FACE studies available for comparison, the responses predicted by APSIM were in agreement with the generally accepted responses of plants to elevated atmospheric CO2 concentrations (Kimball et al. Reference Kimball, Kobayashi and Bindi2002; Long et al. Reference Long, Ainsworth, Rogers and Ort2004; Ainsworth & Long Reference Ainsworth and Long2005). This indicates that the responses of these crops modelled under the future climate scenarios can be taken as legitimate even if the relationships describing these responses were developed using limited data. While the mean maize and forage rape responses were outside of those observed in the FACE experiments, the available FACE data for each crop were limited to one growing season and one location. However, the range in most responses of maize crossed the range defined by Ainsworth & Long (Reference Ainsworth and Long2005) for C4 grasses. For dryland forage rape, there was better alignment with the observed FACE responses and the defined range for this plant's functional group. This reflects the dryland conditions of the FACE experiment and the fact that more data from dryland conditions compared with irrigated conditions was available to Ainsworth & Long (Reference Ainsworth and Long2005) when they defined the expected response range. For lucerne, the response predicted for the irrigated crops was similar to that observed in FACE, which was undertaken in dryland conditions (Luscher et al. Reference Luscher, Hartwig, Suter and Nosberger2000), while the response was over-predicted in the dryland simulations. The FACE study in which the response of lucerne growth to elevated CO2 was determined was undertaken in a high-rainfall (1100 mm per year) summer-dominant rainfall environment, while the sites used for the simulations have winter-dominant rainfall patterns with a range of 567–1196 mm in annual rainfall. Potentially a greater response in DM yield to elevated CO2 concentration for lucerne in the FACE experiment was masked by the availability of water, making this dryland study closer to an irrigated study in terms of the observed response. The transpiration efficiency of legumes is known to be more responsive to changes in atmospheric CO2 concentration relative to other species (Ainsworth & Long Reference Ainsworth and Long2005). For forage sorghum and wheat, the response to elevated atmospheric CO2 concentrations agreed with the observations from FACE experiment. This finding supports the use of APSIM for assessing the impact of elevated atmospheric CO2 on forage nutritive value as well as forage yield, an important consideration in designing forage cropping systems under future climates where the overall objective is to convert forage into animal product.
A consistent trend across all crops and locations was that the yield response to elevated temperature and atmospheric CO2 was mediated by a decrease in rainfall. However, the extent of this mediation was dependent on crop type, with the winter-grown crops of forage wheat, oats and annual ryegrass having a minimal decrease in yield compared with the spring/summer-grown crops of forage rape and forage sorghum. Soil water holding capacity will also impact the resilience summer crops to future decreases in rainfall since there will be a greater reliance on stored soil moisture to support crop growth. There is considerable uncertainty surrounding the magnitude of changes in rainfall in the future climate projections for Australia (CSIRO & BOM 2007). However, if a large decrease in rainfall is received, a shift in dryland forage cropping from summer to winter could be expected.
With the exception of maize and forage sorghum grown at Dookie and Terang, yields of crops increased with an initial increase in temperature and atmospheric CO2 concentration. This was due to an increase in temperature to those more favourable for crop growth, the fertilization effect of increasing atmospheric CO2 concentration on photosynthesis and, for the dryland-grown crops, the increase in water use efficiency associated with increasing atmospheric CO2 concentration (Bunce Reference Bunce2004). However, under the T3R0 and T3R30 scenarios, forage rape yield at all three locations and maize yield at Elliott decreased. This response plus the decrease in forage sorghum and maize yield with any increase in temperature above the baseline at Dookie and Terang was potentially due to an increase in the rate of maturity with increasing temperatures for these species. Furthermore, the decrease in plant N concentration below that required by the crops for optimum growth with increasing atmospheric CO2 concentration would have limited growth (Long et al. Reference Long, Ainsworth, Rogers and Ort2004), a response observed in rice (Oryza sativa L.) by Makino et al. (Reference Makino, Harada, Sato, Nakano and Mae1997). While an increasing rate of soil N mineralization may be expected to occur with increasing temperature, the current authors have previously identified that in high N input forage cropping systems, soil N has minimal influence on overall yield (Pembleton et al. Reference Pembleton, Rawnsley, Jacobs, Mickan, O'Brien, Cullen and Ramilan2013a ). Increasing the application of nitrogen fertilizers to these crops under these scenarios could prevent this response (Farage et al. Reference Farage, McKee and Long1998) and ensure that the full yield benefits of a CO2 enriched atmosphere are captured.
The crops with the greatest yield improvements were annual ryegrass and lucerne. The long growing season of annual ryegrass and the year-round growth of lucerne meant these crops took the greatest advantage of the improved daily growth rate from increased autumn, winter and spring temperatures and improved water use efficiency associated with the increase in atmospheric CO2. With climate change similar to the scenarios used in the present study, Hatfield et al. (Reference Hatfield, Boote, Kimball, Ziska, Izaurralde, Ort, Thomson and Wolfe2011) suggested a 30% improvement in soybean yield due to an increase in temperature and atmospheric CO2 concentration. Lucerne was also advantaged by the increase in temperature reducing the time the plant spends in dormancy over winter.
Yield improvements of the annual crops were in line with studies undertaken for other regions with greater yield improvements for C3 crops compared to C4 crops (Hatfield et al. Reference Hatfield, Boote, Kimball, Ziska, Izaurralde, Ort, Thomson and Wolfe2011). In grain crops, increases in temperature can lead to stress during critical growth stages (e.g. flowering or grain fill), negating the benefits from CO2 fertilization and improved water use efficiency (Hatfield et al. Reference Hatfield, Boote, Kimball, Ziska, Izaurralde, Ort, Thomson and Wolfe2011). However, forage crops that are harvested while still vegetative will avoid these negative consequences of increased temperatures.
Variability, as described by the CV, in crop yield decreased or remained stable for the crops investigated with the exception of dryland lucerne and annual ryegrass. Both these crops have a longer growing season compared with the other crops and hence have a greater chance of exposure to water deficits severe enough to counteract the improvements in transpiration efficiency. This is also the reason for the large range in the modelled responses observed for dryland winter-dormant lucerne, forage rape and forage sorghum when CO2 concentration was increased without additional scaling of temperature and rainfall.
In the forage wheat and oat crops as well as the lucerne crops, the proportion of total yield that was grown and available for grazing during the winter increased as temperature and atmospheric CO2 concentrations increased for all locations. For lucerne this increase was between 1·8 and 12·3%, while for wheat and oats this increase was between 2·2 and 26·8%. The increase in production during winter was due to an extension of the growing season of the crops longer into the cooler months by the warmer temperatures. An increase in the proportion of the predicted growth towards winter of perennial pastures in southeastern Australia under future climate scenarios has also been reported (Cullen et al. Reference Cullen, Eckard, Callow, Johnson, Chapman, Rawnsley, Garcia, White and Snow2008, Reference Cullen, Eckard and Rawnsley2012). While this could help alleviate a common feed deficit on pasture-based dairy farms in Australia (Rawnsley et al. Reference Rawnsley, Donaghy, Stevens, Chapman, Clark, Macmillan and Nation2007), the results of Cullen et al. (Reference Cullen, Eckard and Rawnsley2012) suggested that this will be at the expense of summer pasture growth.
Forage sorghum and wheat DMD changed little from the baseline values under the future climate scenarios. This is in contrast to the predictions for timothy (Phleum pratense L.) made by Jing et al. (Reference Jing, Belanger, Qian and Baron2013) that suggested a decrease in digestibility with future climate change. The longer growing season of perennial pastures increases the periods when the plant is exposed to temperatures high enough to decrease forage digestibility. An analysis of the nutritive value of forage harvested from FACE experiments has also indicated little influence from elevated CO2 relative to the other experimental factors (i.e. water deficit or stage of harvest) (Akin et al. Reference Akin, Kimball, Mauney, LaMorte, Hendrey, Lewin, Nagy and Gates1994, Reference Akin, Kimball, Windham, Pinter, Wall, Garcia, LaMorte and Morrison1995; Porteaus et al. Reference Porteaus, Hill, Ball, Pinter, Kimball, Wall, Adamsen, Hunsaker, LaMorte, Leavitt, Thompson, Matthias, Brooks and Morris2009). Based on this simulation analysis and the results of past research, it is likely that there will be minimal change to the digestibility of annual forage crops grown in the southeast Australian dairy regions under future climates.
Irrigation inputs increased for the summer crops grown at Elliott under the future climate scenarios, while at Dookie the irrigation required decreased under the future climate scenarios, even with a reduction in rainfall. This was due to a reduced time to harvest maturity (e.g. between 8 and 13 days for wheat), leading to warmer growing conditions and the CO2-driven improvements in transpiration efficiency leading to an increase in crop water use efficiency. While this analysis has identified that yield of irrigated forage crops will remain relatively consistent or increase with possible future climate change, the change in the irrigation requirements and the availability of irrigation water to grow such crops (a factor not considered in the current analysis) in the future will determine if they continue to be utilized within dairy systems. Certainly the decrease in irrigation inputs of the irrigated crops grown at Dookie bodes well for the continuation of irrigated forage production in that region.
The approached used in the present study was to explore the resilience of forage crops to concurrent increases in temperature and atmospheric CO2 concentration with or without decreases in rainfall. This approach does not rely on uncertain predictions of future climates while still providing the information industry needs to develop adaptation strategies. Consequently the results of the individual climate scenarios explored should not be interpreted as a definitive forecast; rather it is the trends between the scenarios and the baseline that can be used. The method of climatic scaling used to explore the growth of forage crops under future climates used in this current study fails to take into account the increasing frequency extreme climatic events (e.g. floods and heatwaves) that are expected into the future (Alexander & Arblaster Reference Alexander and Arblaster2009). These events will also influence forage crop production, particularly the risks associated with relying on annual forage crops to supply forage within a dairy system (i.e. crop failure). The analysis employed also did not account for the increase in weed growth and increased incidence of pests and diseases that are also predicted to occur under future climates (Hatfield et al. Reference Hatfield, Boote, Kimball, Ziska, Izaurralde, Ort, Thomson and Wolfe2011). The potential consequences of extreme climatic and biotic events should not be ignored. Furthermore, the present study has only considered the resilience of forage crops on an individual basis. Forage crops are often grown within a system to achieve a set of desired outcomes and to address the systems limitations of perennial grass pastures (i.e. improvement in water and nutrient use efficiency, increase production per unit area) (Garcia & Fulkerson Reference Garcia and Fulkerson2005; Chapman et al. Reference Chapman, Kenny, Beca and Johnson2008a ). Even small changes in the growth and nutritive value of one crop within these systems can have large implications to the forage components within a farms feedbase and the dairy system overall (Rawnsley et al. Reference Rawnsley, Chapman, Jacobs, Garcia, Callow, Edwards and Pembleton2013). As such, any consideration of the future role that forage crops will play on dairy farms needs to be evaluated from a systems perspective.
The supplementary material for this paper can be found at http://dx.doi.org/10.1017/S0021859615001185
The authors wish to acknowledge the advice from Dr Joe Jacobs, Dr Kithsiri Dassanayake, Mr Frank Mickan, Mr Greg O'Brien and Mr Roby Zeissig in the design of the forage crop systems simulated in this study. The authors also acknowledge the Australian Department of Agriculture, Fisheries and Forestry for its financial support for this study.
