An improved way to determine nitrogen fertiliser requirements of. sugarcane crops to meet global environmental challenges

Transcription

1 An improved way to determine nitrogen fertiliser requirements of sugarcane crops to meet global environmental challenges P. J. Thorburn 1,*, J.S. Biggs 1, A.J. Webster 2 and I.M. Biggs 1 1 CSIRO Sustainable Ecosystems and Sustainable Agriculture Flagship, 306 Carmody Rd, St Lucia, Qld, Australia. 2 CSIRO Sustainable Ecosystems and Sustainable Agriculture Flagship, PO Box Earlville BC, Cairns, Qld 4870, Australia. 1 Accepted for publication in Plant and Soil (12 April ) * Corresponding author: peter.thorburn@csiro.au Phone Fax

2 Keywords: Immobilisation, Mineralisation, Biofuel, Organic matter, N Replacement, Nitrogen use efficiency, Tropical agriculture, Water quality Abstract Nitrogen (N) fertiliser management is increasingly important in sugarcane production as imperatives to reduce environmental impacts of N escalate. In this paper we report testing of a new concept for N management in sugarcane, the N Replacement system. This system relies on soil N cycling to buffer differences in crop N needs and N fertiliser supply to individual crops, and aligns N applications with actual cane production over the longer-term rather than potential production. In 11 experiments, conducted in a wide range of environments over two to five crops, cane and sugar yields in the N Replacement treatment were similar to those achieved with the farmers conventional N management, with a trend over successive crops for yields to increase relative to conventional management. At sites where experiments ran for 1 at least 4 years, this trend resulted in cumulative sugar yields being higher in the N Replacement treatment. Average N applications were 3% lower in the N Replacement treatment, and N lost to the environment was estimated to be ~0% lower. Soil N buffering was adequate to maintain sufficient N supply to crops even when yields were up to 30% greater than expected. Thus, it is not necessary to align fertiliser applications to potential sugarcane yields, which are rarely achieved in practice. Our results show that the ecologically-based N Replacement system has promise to deliver superior environmental outcomes without significantly reducing production of sugarcane, and potentially other semiperennial crops, in the tropics and subtropics. Further evaluation of the system will be beneficial, and there is scope for determining more site-specific values of parameters in the 2

3 system. However, care must be taken to evaluate the system over sufficient time frames (e.g. > 2 crops) so that productivity improvement trends in the N Replacement system can be expressed. Introduction Nitrogen (N) fertiliser additions are an important contributor to productivity in farming systems, ensuring N supply is sufficient for crops to achieve optimum yields. However, applying N fertiliser also increases losses of N to the environment, and so intensive agricultural industries face the challenge of maintaining productivity while minimising environmental impacts of N fertiliser use. The harmful impacts of increasing N applications on water quality have long been recognised (Eickhout et al. 06), and there are now additional concerns with N applications enhancing emissions of the potent greenhouse gas nitrous oxide (Bouwman et al. 02). 1 Sugarcane (Saccharum spp.), an important tropical crop (FAO 08), is currently experiencing tensions between productivity and environmental impacts from N use (Cheesman 04). This perennial crop produces large quantities of biomass (commonly yielding ~70-90 t cane ha -1 yr -1, fresh weight), which is often associated with substantial applications of N fertiliser (Roy et al. 06). In India and China for example, recommended applications of N can be as great as 300 kg ha -1 (Roy et al. 06; Kostka et al. 09), with recommendations in the range of 10-0 kg ha -1 in a number of other countries (Guatemala, Perez et al. ; Mexico, Salgado García et al. ; Australia, Schroeder et al. ; South Africa, Meyer and Wood 1994; parts of USA, Anon 01, Wiedenfeld and Enciso 08). In 3

4 Australia, the history of high N applications to sugarcane is associated with N contamination of surface (Mitchell et al. 01; Bramley and Roth 02) and ground waters (Thorburn et al. 03a), with this contamination likely to be harming ecosystems of the Great Barrier Reef World Heritage Area (Brodie et al. 08). Sugarcane is also gaining global interest as a feedstock for biofuel production because of its potentially positive energy balance (Renouf et al. 08; Smeets et al. 09). However, sugarcane production may produce more nitrous oxide than many other broad acre crops (Thorburn et al. ), so there is a risk that nitrous oxide emissions from high N applications will reduce the greenhouse gas advantages sought from biofuel production (Crutzen et al. 08; Smeets et al. 09). As the goals of crop production evolve from maximising crop yields, to achieving economic optima, then to balancing environmental and production goals, N fertiliser recommendations also evolve leading to lower N applications. Initial fixed rates are replaced with systems based on potential N supply to crops from sources other than fertilisers (e.g. mineralisation of 1 organic N) and/or target yields which define the amount of N the crop may need (Neeteson 199). Various N recommendation systems applied in sugarcane production fall on different places on this development continuum. In common with many crops, sugarcane yields are hard to predict because of seasonal climate variability (Everingham et al. 07). Sugarcane has the added complexity of substantial variability in harvested crop age (e.g. 8 to 24 months) and hence crop size. In the face of these uncertainties farmers tend to be optimistic about the magnitude of target yields, and so over apply N relative to the size of the crop they actually grow (Meyer et al. 07). N fertiliser recommendation systems that are driven by target yields tend not to meet environmental goals (Johnson et al. 02; Beaudoin et al. 0; Derby et al. 4

5 09). Therefore it is desirable to have a different approach for determining N fertiliser rates in sugarcane production to help meet environmental challenges. An alternative to traditional N recommendation systems is a more simple input-output balance approach, where N fertiliser is applied to replace the nitrogen that is removed from the system in the harvested product and (unavoidable) environmental losses (Janssen and de Willinen 06). Such a system relies on soil N reserves (organic and inorganic) to buffer differences between short-term N needs of a crop and the longer-term N balance of the soil, and represents a more complete view of soil N cycling than used in some other recommendation systems, which focus only on N mineralisation (Meyer et al. 07; Schroeder et al. ). A balance-oriented system may be applicable in agricultural systems where the soil biological and environmental conditions are conducive to dynamic N cycling (Drinkwater and Snapp 07), so the processes of mineralisation and immobilisation are rapid, and crops have well developed root system that can exploit nutrients at depth. 1 Sugarcane production systems can exhibit these characteristics. Sugarcane is potentially deep- rooted (3 m) and can take up N leached to subsoils (Smith et al. 0). Biological immobilisation of N fertiliser is also high in many sugarcane growing regions because of the warm, moist soil conditions in tropical and subtropical areas (Ng Kee Kwong et al. 1986; Basanta et al. 03; Meier et al. 06). Immobilised N is protected from environmental losses and the soil conditions facilitate the subsequent mineralisation of this N so that it is available to crops. Thorburn et al. (04) proposed such an N management system for sugarcane, the N Replacement system, and predicted that it would maintain productivity and reduce environmental losses of N compared to more traditional approaches.

6 The aim of this study was to evaluate the N Replacement (NR) system for sugarcane growing over a wide range of environments and soils types. In addition to investigating crop productivity in the NR system, N balances were determined to test some of the assumptions made in developing the system and assess the possible environmental benefits of the system. We show that this simple system maintains productivity at N rates considerably lower than those used conventionally, across a range of locations. Methods Site description Experiments were established on commercial farms in seven different regions (Figure 1) covering much of the climatic range of sugarcane production in Australia. The most southerly area, Condong, was close to the southern limit where annual harvesting of crops dominates sugarcane production. South of this region low temperature limits growth rates so that cane is 1 commonly harvested every two years. Conversely, the Burdekin region has higher cane yields (~11 t ha -1 fresh weight 1 ) than other regions (~80 t ha -1 ) because it receives high solar radiation and has plentiful supplies of irrigation water. The Innisfail, Mulgrave and Mossman regions in the wet tropics have rainfall averaging from 2,00 to 3,00 mm yr -1, but prolonged cloud cover and low solar radiation limits growth in these regions. In the Maryborough, Bundaberg and Burdekin regions irrigation is common because of low rainfall (~1,000 mm yr -1 ), and all sites in these regions were irrigated (Table 1). Sufficient irrigation water was available at the Bundaberg and Burdekin sites to fully irrigate the crops. 1 As is standard practice in sugarcane agronomy, cane yields will be given as fresh weight throughout this paper. 6

7 At Maryborough, limited amounts of irrigation water were available and only supplementary irrigation was applied to the crops to lower water stress during the growing season. The experimental sites also represented a wide range of soil conditions and sugarcane varieties (Table 1). Soil textures ranged from sandy loam to light clay, soil organic carbon (C) from < 0.8 to > 2.1% and C/N ratio from approximately 13 to 18 (methods for analyses described below). Experimental design and management A participative approach to experimental design and site management was adopted to ensure the management system was evaluated in the context of real-world farming i.e. to increase the relevance of the experiments to farmers while maintaining adequate research rigour (Carberry 01). The experimental layout at each farm was determined with input from the collaborating farmers and local farmer groups established to participate in the work. Four 1 experiments were established as complete randomised block designs with treatments replicated three times, four others were un-replicated, with the rest having duplicate plots of each treatment (Table 1). Plots were large (> 1,00 m 2 ) at all sites, except MB-2 where the area of the plots was 60 m 2. The large plot design was adopted to allow the experimental plots to be managed by the farmers in a commercial manner (as described below). The BU-1 experiment was imposed on a pre-existing N-rate experiment (Thorburn et al. 03b). Unlike the other experiments where the NR treatment plots had a pre-history of conventional N fertiliser management, the NR treatment plots in this experiment had no N applied to the previous crop, an 18 month plant crop. 7

8 All experiments were managed by the farmers using their normal crop management practices. All crops, except those on the small plots at site MB-2, were mechanically harvested with commercial harvesters. Crops at four sites (BK-1, BK-2, MB-1 and CD-1) were burnt prior to harvest. Crops at all other sites were not burnt and crop residues were left on the soil surface after harvest. Treatments Overview At all sites the NR system was compared with the farmers conventional N fertiliser management practice (termed NF). At most sites, rates in the NF treatment (Table 2) were close to those generally recommended at the time of the study (160-0 kg ha -1 on ratoon crops; Calcino et al. 00). An exception was sites BK-1 and BK-2 that had substantial amounts of N applied in the irrigation water which was not accounted for in the farmers N 1 management (described below). At five sites a lower N rate treatment was also included. N Replacement Following Thorburn et al. (04), the amount of fertiliser N applied to a crop in the NR treatment (F NR, kg ha -1 ) was calculated from: F NR = Y [ (N cane + N trash_rem ) L env ] F other, (1) where, Y is cane yield (t ha -1, expressed on a fresh weight basis) of the previously harvested crop, N cane is the N contained in the cane (kg (t cane) -1 ), N trash_rem is the N contained in any 8

9 sugarcane crop residue (i.e. trash), expressed relative to cane yield, that is removed (kg (t cane) -1 ), L env is a factor to account for environmental losses of N and F other is the amount of N input to the field from other sources (e.g. irrigation water, see below). Expressing N fertiliser additions relative to yield (f NR, kg (t cane) -1 ) gives; f NR = [ (N cane + N trash_rem ) L env ] f other, (2) where, f other = F other / Y. Assumptions can be made to derive values for the parameters N cane, N trash_rem and L env. The value of N cane is the product of the cane N concentration (n cane, kg (t DM) -1 ) and the dry matter fraction (DMF, t DM (t cane) -1 ) of the cane: N cane, = n cane DMF. (3) 1 We assumed that n cane in sugarcane was 0.3% (or 3 kg (t DM -1 )), if not under N stress (Stevenson et al. 1992; Wiedenfeld 199; Muchow et al. 1996; Wood et al. 1996; Basanta et al. 03; Gava et al. 0) and the value of DMF was 0.3 t DM (t cane) -1 (Muchow et al. 1996), so N cane from Equation 2 was 0.9 kg (t cane) -1. The value of N trash_rem was calculated from the N concentration of the trash removed (n trash_rem, kg (t DM) -1 ), the mass of trash dry matter relative to cane yield (T, t DM (t cane) -1 ) and the proportion (P) of the trash removed: 9

10 N trash_rem = n trash_rem T P, (4) We assumed that n trash_rem was kg (t DM) -1 (Stevenson et al. 1992; Wiedenfeld 199; Gava et al. 0; Robertson and Thorburn 07) and T is 0.1 t DM (t cane) -1 (Robertson and Thorburn 07). In our experiments, trash was only removed from the experiments through the process of burning the crop prior to harvest. For this practice, the value P was assumed to be 0.7 (Ball-Coelho et al. 1993; Mitchell et al. 00; Basanta et al 03). Thus the value of N trash_rem from Equation 3 was 0.3 kg (t cane) -1. However, the value of P may be different when trash is removed by other means or for another purpose, e.g. whole crop harvesting for biofuel production (Archer et al. 08). For the value of L env, we assumed that environmental losses of N could average as low as % of fertiliser N additions at optimum N rates (Thorburn et al. 03b; Stewart et al. 06). So, the value of L env was taken as The result was that, under these assumptions, f NR from Equation 2 was 1 kg N ha -1 for systems where the crop was not burnt and 1.3 kg N ha -1 for systems where it was (Table 3), excluding inputs of N from other sources. The two sites in the Burdekin region used groundwater for irrigation. Because of nitrate contamination of groundwater in parts of the region (Thorburn et al. 03a), the irrigation water at these sites had high N concentrations, approximately 11.3 mg L -1. Irrigation applications are high (1,000 to >,000 mm yr -1 ; Charlesworth et al. 02) in the region, so considerable N could be applied through irrigation. Thus, a value of F other in Equation 1 is

11 needed to be derived to determine the amount of N fertiliser applied to NR treatment at these sites. As is normal in the region, groundwater pumps were not metered at the Burdekin sites so the mass of N applied in irrigation water could not be calculated. Even if irrigation application was known, irrigation efficiencies are often low in the region, with substantial runoff and deep drainage happening during, and soon after irrigation (Charlesworth et al. 02; Stewart et al. 06). Thus the mass of N applied in irrigation will exceed that remaining in the root zone and subsequently available to the crops. N available to the crops will be more related to the amount of irrigation water infiltrating into, and remaining in the root zone. We assumed that, over a whole crop, the amount of irrigation water remaining in the root zone approximately equalled the evapotranspiration (ET) from the crop. There is a good relationship between ET (mm) and cane yield (t ha -1 ) (Inman-Bamber and Smith 0; Stewart et al. 06): ET Y. The yield of the past crops in the experimental blocks was known, so ET could be estimated. The N available to the crops from irrigation water, approximated from the product of ET and the groundwater N concentration, was thus 148 and kg ha -1 at sites BK-1 and BK-2, respectively. This N was similar to the value of F NR for the crops at site BK-1, so no additional N was applied in fertiliser to any crop in the NR treatment at that site. At both sites the farmer s standard practice was to ignore N in irrigation water, so the NF treatments received N from both fertiliser and irrigation water. N low The lower N rate treatment (termed NL) aimed to have fertiliser N additions lower than crop N removal (as estimated from Equation 1), and so would deplete soil N stores and induce crop N stress. In the NR system, linking fertiliser N applications to yields of the previous crop will result in low amounts of N being applied after a small crop. Where a large crop follows a 11

12 small one the risk of N limiting crop yield is greatest, a situation of obvious concern to farmers. The NL treatment was included where collaborating farmers and local groups wanted to explore the time it took for yields to be reduced because of low N fertiliser application. This information would help define the risk of N stress from low amounts of N applied after a small crop. At Site BU-1, unlike the other sites, the NL treatment was imposed on plots of a pre-existing experiment (Thorburn et al. 03b) that had received lower than conventional N applications since Measurements Plant Immediately prior to the mechanical harvest, sugarcane stalks were hand-harvested from small areas within each plot (following the methods of Muchow et al. 1996). This sampling 1 involved hand-cutting stalks at ground level over a m length, equivalent to an area of 1 m 2. The total mass of these stalks was recorded. From these stalks, a subsample of 1 stalks was randomly chosen to be partitioned into: (1) millable stalk (cane), (2) non-millable stalk and green leaf sheaths (referred to as cabbage), (3) green leaf blades and (4) dead leaves and leaf sheaths (known as trash). The masses and N concentration (measured by combustion in a LECO 00CNS analyser) and total mass of N were determined for each of the four components. 12

13 The mass of cane mechanically harvested from each plot was calculated from records of harvester bin weights to give crop yields at all sites, except MB-2. There, the mass of cane in the hand-harvested 1 m 2 area in the plots was determined to give crop yields. Sugar concentration of the juice samples of the harvested cane was expressed as Commercial Cane Sugar (CCS), a measure of the commercially extractable sugar that, together with cane yield, is historically the basis of payment in the Australian sugar industry. CCS was determined by standard methods in sugar mills when cane was mechanically harvested and processed at sugar mills. At site MB-2, CCS was determined on the hand-harvested cane by the methods described by Muchow et al. (1996). Leaf N concentration was measured in all plots in an effort to detect N stress. A 0 mm long sample (excluding the midrib) was obtained from the third fully expanded leaf above the top visible dewlap. Twenty samples were obtained per plot when the crop was approximately 3 1 months old. N concentration was measured by combustion in a LECO 00CNS analyser. Soil Soil mineral nitrogen (SMN), total N and total organic C were determined on all samples obtained from all plots at the commencement of the experiment, prior to application of fertiliser N. SMN was then determined after each crop was harvested but prior to fertilising. In 07, labile C (Lefroy et al. 1993) was determined in samples of surface (0-300 mm depth) soil from all treatments. All soil samples were taken to 2 m using 0 mm diameter steel tubes from four randomly chosen locations in each plot. Each core was separated into at least seven layers (0-10 mm, mm, mm, mm, mm, 1-10 mm and 13

14 10-0 mm). At each location, a core was taken from both the sugarcane row and inter-row. Samples were bulked together and then stored at 4 o C in plastic bags until analysis. Analyses for SMN were performed within seven days of sampling. SMN was determined by extraction in 2M KCl with the NO 3 -N and NH 4 -N measured colorimetrically (Henzell et al 1968; Rayment and Higginson 1992). Total N and total C in the soil were determined by combustion using a LECO 00CNS analyser. Most soils had ph < 7, (except MB-1 which has 7.6), so all the C in the soil was assumed to be organic. The volume of soil from each sample and its oven dry weight were also determined, and the data used to calculate soil bulk density. The mean (across locations and sampling times) bulk density for each plot was used to express SMN in kg ha -1 to a given depth. Soil texture was determined from samples obtained at the beginning of the experiments. Labile C was determined as the amount of soil C oxidised by 33.3 mm potassium permanganate (following the method of Lefroy et al. 1993). 1 Data analyses N balance The average N balance over the duration of the experiments was calculated for each treatment at each experiment. The N balance was defined as: N u = N a [ Y(N cane - N trash_rem ) ] N s, () where N u (kg ha -1 ) is N unaccounted for in the balance, N a (kg ha -1 ) is average N applied during the experiment through fertiliser and in irrigation water, and N s (kg ha -1 ) is average increase in N stored in the soil to 2 m depth over the experiment. Values of N cane and N trash_rem 14

15 were determined from Equations 3 and 4 using measured values of n cane and n trash_rem. N s was determined from the slope of the regression line fitted to the total soil mineral N (to 2 m soil depth) measured in the plots during the experiment. Linear regression were used to represent the change in SMN as within-year variability in the measurement of SMN was large, so comparing only the initial and final measurements would have been greatly influenced by this variability. Regressions ensured all data were used in estimating the average change in SMN. The selection of this method does not suggest that SMN varied linearly from crop to crop. This method of determining N s assumes that there is no change in soil organic N (or that the change is not measurable over the time frames of our experiments). N u is thus an estimate of the N potentially lost to the environment. Analysis of variance At sites where treatments were replicated, results were subject to analysis of variance with Statistix v 9. At each site, results from across all crops were analysed using a strip-plot 1 design (treatments x replicates, with crops as the strip). Results Given that the amount of N fertiliser applied to crops in the NR system is based on the yield of the previous crop, the N fertiliser applied each year in the NR treatments generally varied between crops (Table 2). Over all crops and sites, there was an average of 124 kg ha -1 of N fertiliser applied in the NR treatment compared with 190 kg ha -1 in the NF treatment, so on average N applications were reduced by 3% in the NR treatment. 1

16 At the sites where treatments were replicated, there was no significant (P < 0.0) difference in cane yields between the NR and NF treatments (Figure 2). At the other sites, differences were small. There were only small differences in CCS between treatments (data not shown), so the patterns that occurred in cane yield results also occurred for sugar yields (Figure 3). The similarity in cane and sugar yields between the NR and NF treatments was reflected in the leaf N concentrations (data not shown). There was a trend for mean cane yields in the NR treatments to be lower than those in the NF treatment in the first and sometimes second crops after imposition of the treatments. Sites BU-1 and MB-1 provide examples of this trend (Figure 2). However, in subsequent crops mean yields in the NR treatment were similar to or greater than those in the NF treatment, as clearly displayed by pooled data from the five sites run for the longest time (Figure 4). For these pooled data, the cumulative difference in cane yield was -2.2 t ha -1 (i.e., the NF treatment slightly out-yielded the NR treatment). The trend for yields to become higher in the NR compared with the NF treatment was also apparent in sugar yield (Figure 4), but in this case the cumulative difference in sugar yield was higher in 1 the NR treatment (by 0.16 t ha -1 ). It is notable that the improvement in yields over time occurred at the BU-1 site, at which soil N was depleted in the NR treatment plots prior to establishment of that treatment. The higher production relative to N fertiliser applied indicates greater nitrogen use efficiency in the NR treatment. Over the four crops at the five sites pooled for Figure 4, average nitrogen use efficiency was 0.82 and 0. 6 t cane (kg N) -1 in the NR and NF treatments, respectively. For sugar yield, the results were 0.12 and 0.08 t sugar (kg N) -1 in the NR and NF treatments, respectively. Cane (Figure 2) and sugar (Figure 3) yields were generally lowest in the NL treatment. At the two Maryborough sites (MB-1 and MB-2), cane yields in the NL treatment became 16

17 increasingly lower relative to the other two treatments over time, possibly indicating a depletion of soil N stores, although there was no evidence of such depletion in soil mineral N at these sites (Figure ). The relative reduction in yields through time was not evident at the BU-1, CD-1 or IN-1 sites. This result is understandable for the first of these sites, as the NL and NF treatments had been in place for a number of years before the imposition of the NR treatment in the experiment, and so N demand and supply may have been approaching a steady state in the plots. However, the reason for the result is not apparent for the other two sites. While soil mineral N (SMN) varied through time at the sites (example data for three sites shown in Figure ), there was generally little difference in SMN between treatments. The most dramatic difference in SMN occurred at the IN-3 site, where the NR treatments plots commenced with higher SMN than the NF plots (indicated by the slopes of the SMN regressions in Figure ). However, at all sites slopes of the regressions were not significantly 1 different (P > 0.0) between treatments. As found in previous studies (Meier et al. 06; Thorburn et al. 03b), there were substantial amounts ( kg ha -1 ) of SMN in the profile (to 2 m soil depth) at some of the sites, for example IN-3 in the wet tropics and MB-2 in the dry sub-tropics (Figure ). However, SMN in the profile could also be low, e.g. less than 0 kg ha -1 at sites MB-1 (Figure ), CD-1 and BK-2 (data not shown for the latter two). These low SMN stores did not predispose yields at these the sites to being more sensitive to N fertiliser application rate (Figures 2 and 3). 17

18 As well as little difference in SMN, there was no evidence of total soil C or labile C being lower in the NR (or NL) treatments than the NF at individual sites (data not shown) or over all sites (Table 4). Similarly, there was no evidence in decline in labile C as a proportion of total soil C. In the N balance, there was substantial variation in N unaccounted for across all sites. There was also more N unaccounted for in the NF treatment than the NR at every site except CD-1 (Figure 6). At this site the farmer reduced his N applications across his farm, and hence in the NF treatment, so average N fertiliser applications were similar in both the NF and NR treatments; hence the N balances were similar in these two treatments. Across all treatments, the N unaccounted for averaged 128 (s.e. 19.6) kg ha -1 crop -1 in the NF treatment and 66 (s.e. 12.9) kg ha -1 crop -1 in the NR treatment. This halving of N unaccounted for compares with a 3% reduction in the N applied in the NR treatment. In the NL treatment, the N not accounted for ranged from 8 to -32 kg ha -1 crop -1 (Figure 6) with a mean of 12 (s.e. 14.9) kg ha -1 crop Concentrations of N in the cane were variable, particularly between sites (Figure 7). N concentrations have been previously found to be responsive to N fertiliser applications (Stevenson et al. 1992; Wiedenfeld 199; Muchow et al. 1996; Wood et al. 1996). However, at sites where treatments were replicated, a similar response was neither significant (P > 0.0) nor consistent (e.g. BU-1 08 crop v. the 06 crop). This lack of correlation between cane N concentrations and N fertiliser treatments likely reflects the lack of treatments effects on yields differences at many of the sites. The highest N concentrations occurred in the oldest varieties (Q96, Tellus and Q117) in the NF treatment at the two Burdekin sites (BK-1 and BK-2, Table 1). However, other varietal effects were inconsistent. For example, cane N 18

19 concentrations were greater at site BU-1 than MB-2, yet the variety Q138 was grown at both sites. Discussion The results from the 11 experiments suggest the N replacement concept has promise for meeting the productivity needs of N fertiliser management in sugarcane while reducing potential environmental losses of N, and hence to meet global environmental challenges in sugarcane production. Cane (Figure 2) and sugar (Figure 3) yields were similar to those with higher and more conventional (Calcino et al. 00; Roy et al. 06) applications of N, including recent recommendations developed in Australia ( 6ES, Table 2), and so N use efficiency was higher in the NR treatments. Hence, potential environmental losses of N were reduced (Figure 6) by ~0% compared with conventional N management. As well, from the 3 rd crop after it was imposed the N replacement system had productivity benefits (Figure 4), 1 with these benefits exceeding trends to lower productivity in the first two crops for sugar production. The overall results suggest the NR approach to managing N may help relieve current tensions between productivity and environmental impacts in sugarcane (Cheesman 04), and make the crop a more attractive feedstock for biofuel production (Renouf et al. 08; Smeets et al. 09). There are implications of our results for both further applications of the NR system and more general N recommendations for sugarcane and similar crops, which we will now consider. Implications for applying the N Replacement system 19

20 The results of our experiments allow us to examine the assumptions originally made by Thorburn et al. (04) in conceiving the NR system. Overall, N contained in cane in the experiments was lower than that originally assumed, while N lost to the environment was higher. Summarising the results in terms of Equations 1 and 3, average cane N concentrations (i.e. n cane values) in the NR treatment were 0.17% (Figure 7), which gives a value for N cane from Equation 3 of 0.1 kg (t cane) -1 (Table 3). Given that approximately half of the N applied in the NR treatment was unaccounted for and hence potentially lost to the environment, the average value of L env in the experiments was ~2 compared with the original assumption of 1.1 (Table 3). These results illustrate the conceptual nature of the NR system. It provides an explicit framework in which to consider the fate of N applied to crops, and so parameter values in Equation 1 may well need to be fine tuned for different circumstances. There are two examples of possible fine tuning of the NR system. The first is where cane N concentrations are as high as originally assumed in developing the NR system (i.e. the value 1 of N cane is 0.9 kg (t cane) -1 ) and environmental losses were as high as those found in this study (i.e. a value of L env of 2). In this hypothetical example, the value of f NR would 1.8 kg (t cane) -1 if trash was not removed and 2.4 kg (t cane) -1 is the crop was burnt pre-harvest (Table 3). The second example of fine tuning is the potential to define a more site-specific value for the parameter L env (Equation 1). The value of L env may depend on the interactions between soils, environments and management practices. In places where environmental losses of N may be high, a higher value of L env may be used. Our experiments included a number of locations in such environments, i.e. the wet tropics (Sites IN-1 and IN-3) where rainfall averaged ~3,700 mm yr -1 during the experiments and the Burdekin region where crops were fully irrigated

21 (with irrigation application up to,000 mm yr -1 ; Charlesworth et al. 02). There was no significant loss of productivity in the NR treatments at these sites (Figures 2 and 3), hence the partitioning of N fertiliser between the crop and environmental losses was appropriate. This result may imply that a lower value of L env may have been appropriate at other locations, where N loss processes were less aggressive. For this instance, the results from site MB-2 illustrate that a lower value of L env may have been appropriate at this site. Halving N fertiliser application in the NL treatment relative to the NR (Table 2) had little impact on yields (Figure 2) or cane N concentration (Figure 7), and hence N exported in the cane. Thus from the N balance (Equation ) the environmental losses would have been substantially lower in the NL treatment: Substituting measured value of N cane, Y and F NR into Equations 1 and 2 for this treatment gives a value of L env of 1.1, which results in a hypothetical value of f NR of 0.6 (Table 3). The yield reduction in the third crop in the NL treatment (Figures 2 and 3) suggests that the low N applications may not 1 sustain production in the long term at the site. But the results of this example suggest that a sustainable value of L env at the site may be lower than 2, the average value across the NR treatments. Another instance of where L env may benefit from fine tuning is where certain management practices reduce losses of N to the environment. For these types of practices a lower value of L env in the NR system may be appropriate. For example, the value of L env might be lowered with precision agriculture as fertiliser applications can be adjusted to low and high yield zones within a field (Bramley et al. 08). Split applications of N or slow release fertilisers are other examples. 21

22 Finally, apart from fine tuning the NR system, there are unexpected results that would benefit from further investigation. The productivity benefits seen from the 3 rd crop after the imposition of the NR treatment (Figure 4), despite the lower N applications (Table 2), are not seen in more conventional N rate experiments where a constant amount of N is applied in each treatment (Wiedenfeld 199; Muchow et al. 1996; Thorburn et al. 03b; Wiedenfeld and Enciso 08). This contrast suggests the crops were responding to the variable N inputs. Sugarcane root have been observed exploiting SMN deep in the profile when the crop is under N stress (Thorburn et al. 03b; Smith et al. 0), and such a physiological response may have been occurring in the NR treatments. The extent to which the NR system can promote greater yields (Figure 4) and the reasons for this response also need to be resolved. Until this has happened, care needs to be taken to neither predict the outcome of the NR system using data from constant N rate experiments nor evaluate the success of the NR system over short time frames, e.g., one or two crops. 1 General implications The improved outcome of the NR system over the farmers conventional management was potentially due to a number of factors which have general implications for sustainable management of N in sugarcane. Firstly, cane yields in most experiments (Figure 2) were generally lower than potential yield benchmarks (i.e t ha -1 at most sites, except in the Burdekin where they were 180 t ha -1 ; Schroeder et al. ) which drive thinking on farmers N management. Thus our results support those from other studies (Johnson et al. 02; Beaudoin et al. 0; Derby et al. 09) that such a strategy clearly wastes fertiliser when actual yields do not realise their potential, leading to greater environmental impacts. 22

23 Our experiments also allow us to explore the importance of matching N fertiliser applications to expected yields. Since in the NR system the amount of N applied is based on yield of the previous crop, N applications will only be equal to the needs of the coming crop if yields are constant over a number of crops. Where yields increase through time N applications will be lower than those needs, as would be the case when yields are higher than expected. This situation happened in the NR treatment at site MB-2. Cane yields of the 06 and 07 crops were 2 and 30% higher than the previous crop (Figure 2), resulting in actual N applications of ~0.8 kg N t -1 cane -1 relative to the achieved yield. Yet, applying extra N fertiliser (average of 47 kg ha -1 crop -1 ) in the NF treatment in these years did not significantly increase yield, particularly in the 07 crop. Additionally, the NR treatment at the BU-1 site was established following a crop that received no N yet, after two crops, yields had recovered to averages greater than those of the NF treatment that had received 4 kg ha -1 crop -1 extra N fertiliser. And the soil at this site was relatively low in organic matter (Table 1). Hence, from crop to 1 crop, actual yields can be considerably higher than expected yields without N supply being limiting showing that it is less important to fertiliser for expected yields than commonly thought. The success of the NR system also shows that, as hypothesised in conceiving the NR concept (Thorburn et al. 04), the soil N cycling processes buffered the difference between short- term crop N needs in the N replacement system in the short term, and allowed N applications to be aligned with longer-term actual production. There were considerable stores of SMN throughout the experiment sites, e.g. MB-2 and IN-3 (Figure ), as have been observed previously in the Australian sugarcane regions (Thorburn et al. 03b; Meier et al. 06), 23

24 which would have provided buffering. However, the success of the NR treatments at sites with little SMN (e.g. MB-2, Figure ) suggests it was not just SMN stores that contributed to buffering of N supply to the crop. Buffering may be greater in soils with higher organic matter and hence greater capacity to immobilise and mineralise N. However, there was no difference in the relative productivity of the NR treatment (Figures 2 and 3) between sites with the lowest (BK-1, BK-2, BU-1) and highest (CD-1, IN-3) soil organic matter (Table 1). The similarity in results at these sites suggests that high organic matter levels are not necessary to provide sufficient buffering for the NR system to be successful and so the buffering concept may be more widely applicable than otherwise thought. Over-application of N to sugarcane may also be driven by the perception of high losses of N where crops grow in environments with high rainfall and/or irrigation. As discussed above, we had a number of experiments (BK-1, BK-2, IN-1 and IN-3) located in such environments that showed no significant loss of productivity with the lower N rates in the NR treatment 1 (Figures 2 and 3). While potential losses of N were substantial under conventional N management at these sites (Figure 6), it is unlikely that the environment was the cause. Potential losses of N from Sites IN-1 and IN-3 were not notably different from most other sites. And while potential losses of N were greatest at the Burdekin sites (BK-1 and BK-2) these sites had the greatest applications of N due to the contribution of N in the irrigation water, so irrigating itself may not have been responsible. However, overall, potential N losses were lowered by ~0% through reducing N rates. Hence high losses were more due to the high amounts of N applied in farmers conventional practices than the environment itself. This study represents the broadest attempt to estimate losses of N to the environment in sugarcane production and there needs to be a greater number of measurements of N losses 24

25 under different N rates to get a better understanding of the cause of these losses. Such understanding would allow a refinement of the value of L env in Equation 1, as discussed above. Another generally relevant result of our study is the lower than expected cane N concentrations found in our experiments (Figure 7). Similar results were found in Brazil (2 nd ratoon crop of Basanta et al. 03) and more recently in Australia (Bell et al. 06, ). While cane N concentrations may differ between varieties, there was no obvious link between varieties and N concentrations in our results. However, our results and the N Replacement framework show the importance of basing N recommendation systems on relevant information about cane N concentrations. Finally, while we have evaluated the NR strategy on sugarcane, the approach should be applicable to other crops that exhibit similar characteristics that is crops that have well 1 developed root system that can exploit nutrients at depth and grow in environments where the processes of mineralisation and immobilisation are rapid. Perennial and semi-perennial horticultural tree crops growing in the tropics and sub-tropics are likely to fulfil these criteria. Thus a more ecologically-based approach to N management, focussing more on having fertiliser applications maintaining soil N stores (Drinkwater and Snapp 07), may have broad applicability for management of N fertiliser, and reducing the long-recognised harmful impacts of increasing N applications on the environment (Eickhout et al. 06). 2

26 Acknowledgements We acknowledge the generous support of collaborating farmers and milling companies for this work and funding from the Australian Government and sugar industry through the Sugar Research and Development Corporation. We also thank Dr Jo Stringer (BSES Ltd.) for statistical advice. Steve Attard and Mike Spillman (CSIRO) provided valuable support for activities in the Burdekin region. References Archer AA, Thorburn PJ, Hobson PA, Higgins AJ (08) Evaluating alternate strategic options for agricultural value chains. J Chain Network Sci 8: Basanta MV, Dourado-Neto D, Reichardt K, Bacchi OOS, Oliveira JCM, Trivelin PCO, Timm LC, Tominaga TT, Correchel V, Cássaro FAM, Pires LF, Macedo LF (03) Management effects on nitrogen recovery in a sugarcane crop grown in Brazil. 1 Geoderma 116: Beaudoin N, Saad JK, Van Laethem C, Machet JM, Maucorps J, Mary B (0) Nitrate leaching in intensive agriculture in Northern France: Effect of farming practices, soils and crop rotations. Agric Ecosyst Environ 111: Bell MA, Garside AL, Halpin NV, Slater B, Moody PW, Park G () Interactions between rotation breaks, tillage and n management on sugarcane grown at Bundaberg and Ingham. Proc Aust Soc Sugar Cane Technol 30: in press. Bell MA, Moody PW, Halpin NV, Garside AL (06) Impact of management of cane trash and legume residues on N mineralisation and crop N uptake. Proc Aust Soc Sugar Cane Technol 28:

27 Bouwman AF, Boumans LJM, Batjes NH (02) Emissions of N 2 O and NO from fertilized fields-summary of available measurement data. Glob Biogeochem Cycles 16:8 doi gb Bramley RGV, Roth CH (02) Land-use effects on water quality in an intensively managed catchment in the Australian humid tropics. Marine Freshw Res 3, Bramley RGV, Hill PA, Thorburn PJ, Kroon FJ, Panten K (08) Precision Agriculture for improved environmental outcomes: Some Australian perspectives. Landbauforschung - vti Agric For Res 8: Brodie J, Binney J, Fabricius K, Gordon I, Hoegh-Guldberg O, Hunter H, O Reagain P, Pearson R, Quirk M, Thorburn P, Waterhouse J, Webster I, Wilkinson S (08) Synthesis of evidence to support the Scientific Consensus Statement on Water Quality in the Great Barrier Reef. The State of Queensland (Department of the Premier and Cabinet), Reef Water Quality Protection Plan Secretariat, Brisbane, p 84. Calcino DV, Kingston G, Haysom M (00) Nutrition of the plant. In: Hogarth DM and 1 Allsopp PG (eds), Manual of cane growing. Bureau of Sugar Experiment Stations, Indooroopilly p Carberry PS (01) Are science rigour and industry relevance both achievable in participatory action research. Agric Sci 14: Charlesworth PB, Chinn C, Bristow KL, Ham G (02) Healthy crop and healthy groundwater: Sugarcane in the Burdekin delta. In: Proc Nat Conf Irrig Assoc Aust May 02, Irrig Assoc Aust, Sydney, (on CD). Cheesman OD (04) The environmental impacts of sugar production. CABI Publishing, Wallingford. 27

28 Crutzen PJ, Mosier AR, Smith KA, Winiwarter W (08) N 2 O release from agro-biofuel production negates global warming reduction by replacing fossil fuels. Atmos Chem Phys 8: Derby NE, Casey XM, and Knighton RM (09) Long-term observations of vadose zone and groundwater nitrate concentrations under irrigated agriculture. Vadose Zone J 8: Drinkwater LE, Snapp SS (07) Nutrients in agroecosystems-rethinking the management paradigm. Adv Agron 92: Eickhout B, Bouwmana AF, van Zeijts H (06) The role of nitrogen in world food production and environmental sustainability Agric Ecosyst Environ 116:4-14. Everingham YL, Inman-Bamber NG, Thorburn PJ, McNeill TJ (07) A bayesian modelling approach for improving long lead sugarcane yield forecasts. Aust J Agric Res 8:87-94 FAO (06) Fertilizer use by crop. FAO Fertilizer and Plant Nutrition Bulletin 17. Food and Agriculture Organization of the United Nations, Rome. 1 Gava GJ, Trivelin PCO, Vitti AC, Oliveira MW (0). Urea and sugarcane straw nitrogen balance in a sol-sugarcane crop system. Pesq Agropec Bras 40: Henzell EF, Vallis I, Lindquist JK (1968) Automatic colorimetric methods for the determination of nitrogen in digests and extracts of soil. Trans. 9 th Internat. Cong. Soil Sci, Adelaide, p 13. Inman-Bamber NG, Smith DM (0) Water relations in sugarcane and response to water deficits. Field Crops Res 92:18-2. Janssen BH, de Willinen P (06) Ideal and saturated soil fertility as bench marks in nutrient management 1. Outline of the framework. Agric Ecosyst Environ 116:

29 Johnson PA, Shepherd MA, Hatley DJ, Smith PN (02) Nitrate leaching from a shallow limestone soil growing a five course combinable crop rotation: the effects of crop husbandry and nitrogen fertilizer rate on losses from the second complete rotation. Soil Use Manag 18: Lefroy RDB, Blair GJ, Strong WM (1993) Changes in soil organic matter with cropping as measured by organic carbon fractions and 13C natural isotope abundance. Plant Soil 16: Meier EA, Thorburn PJ, Wegener MK, Basford KE (06) The availability of nitrogen from sugarcane trash on contrasting soils in the wet tropics of North Queensland. Nutr Cycl Agroecosys 7: Meyer JH, Schumann AW, Wood RA, Nixon DJ, Van Den Berg M (07) Recent advances to improve nitrogen use efficiency of sugarcane in the South African sugar industry. In: Hogarth DM (ed) Int Soc Sugar Cane Technol Proc XXVI th Congress, Durban July 07, S Afr Soc Sugar Cane Technol Assoc and XXVI th ISSCT Organising Committee, 1 pp (on CD). McDonald RC, Isbell RF, Speight JG, Walker J, Hopkins MS (1990) Australian soil and land survey field handbook 2nd edn, Inkata Press: Melbourne. Mitchell RDJ, Thorburn PJ, Larsen P (00) Quantifying the loss of nutrients from the immediate area when sugarcane residues are burnt. Proc Aust Soc Sugar Cane Technol 22: Mitchell AW, Reghenzani JR, Furnas MJ (01) Nitrogen levels in the Tully River - a longterm view. Water Sci Technol 43:99-. Muchow RC, Robertson MJ, Wood AW, Keating BA (1996) Effect of nitrogen on the timecourse of sucrose accumulation in sugarcane. Field Crops Res. 47:

30 Neeteson JJ (199) Nitrogen management for intensively grown arable crops and field vegetables. In: Bacon PE (ed) Nitrogen fertilisation and the environment. Marcel D, New York, USA, pp Ng Kee Kwong KF, Deville J, Cavalot PC, Riviere V (1986) Biological immobilization of fertilizer nitrogen in humid tropical soils of Mauritius. Soil Sci 141: Rayment GE, Higginson FR (1992) Australian Laboratory Handbook of Soil and Water Chemical Methods. Inkata Press, Melbourne. Renouf MA, Wegener MK, Nielsen LK (08) An environmental life cycle assessment comparing Australian sugarcane with US corn and UK sugar beet as producers of sugars for fermentation. Biomass Bioener 32: Robertson FA, Thorburn PJ (07) Decomposition of sugarcane harvest residue in different climatic zones. Aust J Soil Res 4:1-11. Roy RN, Finck A, Blair GJ, Tandon HLS (06) Plant nutrition for food security-a guide for integrated nutrient management. FAO Fertilizer and Plant Nutrition Bulletin 16, FAO, 1 Rome, 06. Schroeder BL, Hurney AP, Wood AW, Moody PW () Concepts and value of the nitrogen guidelines contained in the Australian sugar industry s six easy steps nutrient management program. Proc Int Soc Sugar Cane Technol 27, 11 pp. Smeets EMW, Bouwmanw LF, Stehfest E, van Vuuren DP, Posthuma A (09) Contribution of N 2 O to the greenhouse gas balance of first-generation biofuels. Glob Change Biol 1:1-23. Smith DM, Inman-Bamber NG, Thorburn PJ (0) Growth and function of the sugarcane root system. Field Crops Res 92: