Towards use of life cycle–based indicators to support continuous improvement in the environmental performance of avocado orchards in New Zealand

A stratified sampling strategy was used to select orchards for the study. Firstly, random sampling was carried out for orchards in each of New Zealand’s three main avocado-producing regions: the Bay of Plenty, Mid North and Far North. Only orchards with production data for a minimum of 2 years (2018–2019 and 2019–2020) were considered, and most orchards had at least 4-year data.

The selected orchards were further categorised by production practices (best practice, good practice and standard practice) and by orchard area (big (> 5 ha), medium (2.0–5 ha) and small (< 2 ha)). The orchards identified for the baseline model spanned a cumulative area of 281 ha. The production practice categories were determined by the yield of each orchard (kg/hectare, averaged for the years 2016–2020) and its irregular bearing index (IBI). Irregular or alternate bearing is the tendency of a perennial fruit tree to have an ‘On’ year of heavy fruiting, followed by an ‘Off’ year of a very light crop (Lovatt 2010; Sharma et al. 2019). The IBI of fruit trees was calculated using the following formula (Rosenstock et al. 2010):where I is equal to the sum of the absolute value of the difference in yields (y) between two consecutive years (t and t − 1), divided by the sum of the yields in the same 2 years, and then divided by the number of years minus 1 (n). For the NZ avocado sector, this value was calculated for each year between 2016 and 2020.

Questionnaires were sent to 108 sampled orchards (to collect orchard data for the period 1 November 2018 to 31 October 2019), with follow-up visits to clarify questions regarding the data and ensure consistency and accuracy. Of the 53 responses received, four ‘supra massive’² orchards were classified as young orchards with very low yields and were excluded from the baseline sample. Thus, the final sample for the baseline model comprised 49 mature orchards across the three regions (11 in the Far North, 14 in the Mid North and 24 in the Bay of Plenty).

For the baseline model, the orchard production–based weighted average impact scores for each orchard were calculated for the three regions. The regional impact scores were weighted again, based on regional production data, to provide a national production–based weighted score for each impact category.

Perennial fruit trees have much longer lifespans than field crops (10– > 50 years), depending on the fruit species. As noted in Sect. 1.2, most published fruit-related LCA studies consider only 1 year of mature orchard production and thus exclude the impacts of the unproductive years (orchard creation and destruction), and low production stages of the orchard establishment and senescence (the initial years between orchard establishment and commercial production as well as the later (low-yield) years of ageing trees). Excluding these specific stages in the orchard life cycle may lead to underestimated LCA results (Bessou et al. 2016; Brito de Figueirêdo et al. 2016; Goossens et al. 2017; Vatsanidou et al. 2020). On the other hand, including these specific stages in the orchard life misrepresents the productive stage as the LCA results will be higher due to the average annual yield being lower when calculated based on the full life cycle of the orchard. To address this aspect, Cerutti et al. (2014) recommended modelling an orchard in six life cycle stages: (1) nursery stage with sapling production,³ (2) planting and field preparation, (3) early low-production phase as the orchard matures, (4) full production, (5) declining production due to ageing trees and (6) orchard destruction and removal/disposal of trees.

For this LCA study, the full lifespan of the avocado orchard was modelled as a separate exercise, following the modular modelling approach proposed by Bessou et al. (2013). The different orchard lifespan stages and yields are shown in Table 2; inputs for the orchard establishment stage were collected from each of the four supra-massive orchards (spread over 489 ha), and their impacts were assessed separately and then averaged. Additionally, orchard infrastructure data was collected from one of the supra-massive orchards to model the orchard creation stage; this orchard was chosen as it had relatively more infrastructure than all the other orchards in the study and so was likely to represent the most extreme scenario with respect to infrastructure. For the senescent years of the orchard life, it was assumed that the inputs and yields towards the end of orchard life would be the same as the orchard establishment. The nursery stage and orchard destruction stages were not modelled due to the lack of data.

Data on agrichemical use was obtained from the spray diaries completed by orchards as part of NZ Avocado’s AvoGreen programme. The calculation and analysis of agrichemical emissions was based on the AI of the pesticide products (SI Table 2), whereas production impacts were based on the inputs for the formulated products. For insecticides, there are very limited inventory datasets available, and none was found for the AIs of insecticides used for avocado production in New Zealand. Therefore, the closest groups of chemical families related to the AIs used in New Zealand avocado production were selected and used to create an ‘average insecticide production’ dataset. For fungicides, five copper-based fungicides are used on avocado orchards in New Zealand (formulations of copper oxide, copper hydroxide, copper oxychloride, cuprous oxide and tribasic copper sulfate) (SI Table 3); their impacts were calculated based on the copper content as the AI. As many agrichemical products for NZ avocado growing are imported from Europe, the model assumed truck transport from a manufacturing plant to the Port of Hamburg (100 km), ocean shipping to the Port of Auckland (27,663 km), truck transport to a regional storehouse in Tauranga/Whangarei (~ 200 km) and, finally, truck transport to the orchard (100 km). As mentioned in Sect. 3.2, all pesticide AIs were modelled for 100% emission to agricultural soil.

Fertiliser and soil conditioner use data was obtained from the spray diaries completed by NZ avocado growers as part of the AvoGreen programme and was supplemented with additional data from the questionnaires filled in by the growers. Most of the products used on New Zealand avocado orchards are straight/simple fertilisers (SI Table 4). For blended and complex fertilisers, the component quantities were calculated separately and used as inputs to the orchard processes, using the straight fertiliser datasets (SI Table 5, Table 6). Like agrichemicals, all fertilisers were modelled to include product formulation, except for those which were entirely made of the constituent chemical.

For production-related climate change impacts, the European ecoinvent datasets were used where they were available, and otherwise, global datasets were used. However, Brentrup et al. (2018) recommended updated carbon footprint values for the production of certain commonly used fertiliser products in different regions of the world. Therefore, the ecoinvent datasets for the production of these relevant fertilisers were updated to reflect the global warming potential (GWP) values in Brentrup et al. (2018).

Transport of fertiliser products was modelled in the same way as for agrichemicals (produced in Germany and transported to NZ) (Sect. 3.1.1) with a few exceptions (SI Table 7):

Nitrogen and phosphorous emissions from the use of fertilisers were modelled using the guidelines in EPD International (2019). For CO₂ emissions, following the national guidelines for New Zealand (Ministry for the Environment 2020), the following emission factors were used:

Compost was used on two orchards. The emission factors for compost production were 0.004 kg CH₄/kg of compost and 0.00024 kg N₂O/kg of compost (Hergoualc’h et al. 2019). Assuming compost producers in New Zealand use best practice methods of windrow composting, only a net 2% of the CH₄ emission factor for composting process was modelled as emitted to air (Jones et al. 2020). In addition, compost also releases nitrous oxide emissions after application; an emission value of 0.047 kg N₂O/kg of compost was used (Mithraratne et al. 2010). With regard to heavy metal emissions from fertiliser application, the commonly used SALCA-HM method (Nemecek et al. 2020) requires detailed site-specific data. Some horticultural LCA studies exclude heavy metal emissions from fertilisers either due to lack of appropriate models for site-specific analysis (Gentil et al. 2020; Martin-Gorriz et al. 2020); others note that heavy metal–related ecotoxicity impacts are usually less significant than other sources in conventional fruit production (Milà I Canals, 2003; Milà I Canals et al. 2006; Meier 2019; Vatsanidou et al. 2020). Moreover, EPD International (2019), which was used as a guideline for this study, did not require heavy metal emissions from fertiliser application to be included. Therefore, these were excluded from the current study.

The main activities carried out on orchards are listed in Table 3. Direct energy use for these activities includes fossil fuels (petrol, diesel and lubricants like grease and oil) and electricity. These fuels are used by both the growers themselves and contractors for specific operations.

All growers provided overall diesel and petrol use on their orchards and also specified the activities undertaken by contractors. The fuel use rates for specific activities were provided by some of the growers and were used to determine average fuel use rates for these activities (Table 3); these values were used as proxy plug-in data for fuel use on orchards for contractor activities. In some cases, the overall fuel use data provided by the growers did not match their disaggregated data by activity; in these cases, the overall fuel use was recalculated based on their activities and the disaggregated fuel use.

Most growers also provided the electricity use for the orchard (i.e. disaggregated from other uses, for example house on the property). In cases where disaggregation of electricity use was not possible, grower estimates were used. Some growers were unable to provide the total electricity use, but instead provided the type of electric motor used (e.g. 1.6 kWh) and the run time of the motor. These data were then used to calculate electricity use per hectare for orchard activities during the study period.

Water used on orchards is sourced either from the public water supply network or from groundwater reservoirs via bores (and in some rare cases, surface water bodies on the property). Activities include irrigation, spraying pesticides and foliar fertilisers, frost protection and fertigation. All growers provided data for total water use on their orchards and the source of water (town supply, groundwater, surface water, etc.). Some growers also provided disaggregated water use data for the different activities. For orchards where water was obtained from the public water supply network, the associated electricity used for treating (purifying) and distributing water was included in the model. For water pumped from underground bores or a surface water body, the electricity or diesel used for pumping water was included in the model.

The main contributing sub-stages/inputs to the climate change impact category in all three regions are fuel and fertilisers (Fig. 1). The fuel use impacts are primarily due to CO₂ emissions from fuel combustion (mainly diesel) in agricultural machinery used on orchards.

All fertiliser-related impacts are due to emissions of CO₂ and N₂O to air. These impacts are related to both ‘non-application’ activities (production/formulation, and transport) and application on the orchard. Of these, the non-application activities account for > 61% of the total fertiliser-related impacts (Fig. 2). Figure 3 shows the variability between relative contributions (%) of production and transport of individual fertilisers and soil conditioners to overall non-application climate change impacts.

McNally and Gentile (2021) report that avocado trees and associated shelterbelt vegetation can act as a carbon sink for the first 28 years of the orchard’s life. Using their data on carbon storage in avocado trees and shelterbelts, an avocado orchard could store between 0.051 and 0.077 kg CO₂ eq./kg avocados produced in an orchard with topped and untopped shelterbelts, respectively. Although this carbon storage was not included in the baseline modelling for this study, the values are equivalent to a carbon offset of 12% or 18% for an orchard with topped or untopped shelterbelts, respectively, when using the national climate change score for NZ avocados developed in this study (0.43 kg CO₂ eq./kg avocados). Any such calculation would also need to account for potential soil carbon changes associated with establishment of avocado orchards, hence accounting for impacts related to land use change, which was outside the scope of this study (Sect. 2.1).

Fertiliser use and soil conditioner use contribute the most to the eutrophication impact category in all three regions (Fig. 4), followed by fuel use. The majority of the impacts are from fertiliser application (Fig. 5) and are due to emission of nitrates, phosphates and phosphorous to freshwater. The remaining impacts are from emissions of ammonia, nitrous oxide, nitrogen oxides and other nitrogen compounds to air.

The water use impact scores for the orchard stage are highest in the Far North followed by the Mid North and Bay of Plenty (0.75 m³ world equivalent/kg avocados, 0.2 m³ world equivalent/kg avocados and 0.14 m³ world equivalent/kg avocados, respectively). Figure 6 shows the contribution (%) of direct and indirect water uses to the total water use impact score in the three regions. Indirect water use contributes the most to this impact category in all three regions.

While agrichemical use is the largest contributor to freshwater ecotoxicity in the Far North and Bay of Plenty, fertiliser and soil conditioner inputs contribute slightly more to the impact score in the Mid North (Fig. 7). The difference in relative contributions of these two types of input largely reflects the varying quantities of copper-based fungicides used on orchards in the different regions.

Terrestrial ecotoxicity impacts are mainly due to emissions of heavy metals (particularly copper) to air during the manufacture of agrichemicals and fertilisers (Fig. 8). Fuel use is also a contributor to these impact category results.

There was considerable variability between orchards in terms of inputs (Sect. 4.1 and Tables 5, 6 and 7) and impacts (Sect. 4.3 and Table 9). To further illustrate this point, Fig. 13 shows the fuel use across the 49 orchards assessed in the study. It can be seen that there is high variability between individual orchards as well as between regions. Orchards in the Far North, for example, use more fuel than the other two regions. This is because the management practices are different in the Far North: the orchards in this region are larger, more intensive and commercially oriented than the traditionally family-owned and hobby-farmed smaller orchards in the Bay of Plenty.

This variability can be attributed to land, soil and climatic conditions and may also be related to economies of scale on larger orchards, as well as different management practices. For example, the largest input-related variability was noted in soil conditioners—this is potentially because unlike other inputs, soil conditioners are typically used only once every 2 years or 3 years on any orchard block. Thus, within the data collection period for this study, some growers had used them in varying quantities, while others had not. Variability in both inputs and impacts was more pronounced in the Bay of Plenty, where most of the growers are smallholders and orchard management practices can be very different between growers. This contrasts with the Far North and Mid North, where avocado orchards are generally larger, have more commercial operations and therefore may have more homogenous practices.

Such variability in agricultural systems has been noted elsewhere in the literature (Astier et al. 2014; Bojacá et al. 2014; Yan et al. 2016; Notarnicola et al. 2017; Poore and Nemecek 2018; Cucurachi et al. 2019; Green et al. 2021). Therefore, careful consideration should be given to the variability amongst NZ avocado orchards when choosing indicators to drive continuous improvement at individual orchard or regional level. The most feasible improvement options may vary between regions and management regimes (e.g. the large commercial operations in Northland and the smaller family run businesses in the Bay of Plenty).

The co-relation analysis between orchard productivity and environmental impact scores (Sect. 4.3) showed that climate change and eutrophication scores (per kg avocados) decreased with increasing yields. However, the other three impact category scores increased with increasing productivity up to 13,000–15,000 kg/ha, after which they declined. This suggests that increased orchard productivity may be associated with improved environmental performance, particularly at yields higher than 13,000–15,000 kg/ha. Further research is required to better understand the relationship between these variables. However, one inference is that the environmental performance of NZ avocado orchards can be improved by increasing productivity sustainably. This concept of ‘sustainable intensification’ has been discussed in horticultural and agricultural literature (Dicks et al. 2019; Hasler et al. 2015; Iglesias & Echeverria 2022; Li et al. 2022) and has even been adopted as a targeted policy goal (Garnett et al. 2013). Sustainable intensification can be achieved by improving orchard management practices (Yan et al. 2016), possibly with little to no increase in inputs (Svanes & Johnsen 2019). Therefore, orchard productivity is a possible indicator for supporting continuous improvement, alongside environmental impact scores. Extending this idea, eco-efficiency is an approach which considers the economic value of a product in relation to its environmental impacts, most commonly by dividing the economic value of the product by its environmental ‘score’. For example, Müller et al. (2015) divided the net profit of kiwifruit (in NZD per hectare) by the GWP value (per hectare) to investigate the eco-efficiency of NZ-produced kiwifruit. This indicator could provide combined environmental and economic information to growers.

It is important to note that alternate bearing is a challenge to using either of these indicators on an annual basis in the NZ avocado sector. However, when calculating the sector’s IBI (see Sect. 2.3.1), if the orchard yield increases 3 years in a row, then it is considered to have an IBI of 0 as the increasing yield is seen as an improvement in production rather than irregular bearing (Pers. Comm., NZ Avocado, September 2021). Thus, tracking the results through both ‘on’ and ‘off’ years and benchmarking against the sector’s rolling average score over (at least), a 2-year period can help the growers understand their eco-efficiency through the irregular bearing cycle. This will also ensure appropriate representation in the indicator results of infrequent activities such as application of soil conditioners.

The results of the impact assessment across the whole orchard life (Sect. 4.4) demonstrated that it is worthwhile to consider the entire lifespan of the orchard when considering improvement options. But, orchard establishment and the early low production stage are ‘in the past’ for most growers and its assessment is irrelevant in terms of improvement options. Also, as mentioned in Sect. 1.2, young orchards (in the orchard creation and establishment stage) have very low yields, so measuring their environmental impacts using a mass-based FU and then benchmarking these scores against the sector average would be an unfair comparison. Therefore, a solution could be to assess the young orchards (3 years old or less) separately using an area-based FU (1 ha of worked avocado orchard); this could be benchmarked against the industry average of the other young orchards in the region using the same area-based FU. Figures 14, 15, 16, 17 and 18 demonstrate this approach for the four supra-massive orchards in the Far North and also show how they compare with the other orchards (in commercial production) in the Far North on an average per hectare basis for climate change. It can be seen that the average climate change score of the four supra-massive orchards is lower than that of the orchards in commercial production when calculated on an area basis. In contrast, the average eutrophication, water use and ecotoxicity impact scores of the four orchards are higher than the corresponding average impact scores of the orchards in commercial production. At an individual orchard level, the high impact scores (except for water use) of orchard ‘a’ relative to orchards ‘b’, ‘c’ and ‘d’ are particularly noteworthy; orchards b, c and d are owned and managed by the same company, and orchard a is under a different management.

The environmental impact category indicators used in this study can be refined by addressing certain methodological limitations.

For the climate change indicator, carbon sequestration in avocado orchards could be included by calculating the carbon stored in the avocado trees, in the topped and untopped shelterbelt vegetation on the orchard, and any changes in soil carbon levels. Biogenic carbon stored in wooden posts should also be considered for carbon sequestration potential.

With respect to eutrophication impacts, current methods are not particularly relevant for New Zealand (Sect. 2.2). However, in the absence of a suitable site-specific method, the CML 2001 method is recommended as a precautionary approach (Payen and Ledgard 2017).

For water use, watershed-level characterisation factors are most appropriate when conducting water footprints because national-level characterisation factors can result in large uncertainty in the results, particularly for countries where watersheds have quite different water scarcity values (Boulay and Lenoir 2020). Moreover, an annual water use value was collected for this study period of 2018–2019, as monthly and seasonal values were unavailable. However, water availability changes with seasons through the year and therefore incorporating the temporal aspect in water footprints would contribute to a more accurate impact assessment of water use. Therefore, while the water use results of the current study are a good starting point for indicator development, it is recommended that the water input data are updated in the future to offer more temporal and spatial resolutions. These characterised water scarcity indicator results (using sub-national-level CFs) should be used by the avocado sector when communicating its environmental profile to external stakeholders. However, as water use on orchards is under the direct control of farmers, it may be more effective to use on-orchard volumetric water use as an indicator to support continuous improvement initiatives rather than (or in addition to) a characterised water scarcity indicator that includes indirect water use elsewhere in the supply chain (Sect. 5.1).

For ecotoxicity, there is a general lack of consensus about the proportion of pesticide emissions reaching different environmental compartments (air, water, soil, etc.) when modelling agricultural systems (Fantke 2019). The most widely used current approaches in agricultural LCAs base their models on the assumption that 100% of the applied pesticides (AIs) are emitted directly to agricultural soil (Nemecek and Schnetzer 2012; Christel et al. 2014; Fantke 2019; Nemecek et al. 2020), and this is the approach applied in this study. Also, agricultural soil is often not considered an environmental compartment for emissions (Birkved and Hauschild 2006; Christel et al. 2014; Rosenbaum et al. 2015) and so the fraction of pesticides that reaches, and remains in, agricultural soil is not assessed. It would be preferable in the future to update the ecotoxicity assessment to account for pesticide emission fractions reaching different environmental compartments and to include toxicity impacts in orchard soil as a separate indicator. Heavy metal emissions from fertiliser application should also be included.

Future studies could also consider other impacts relevant to perennial crop production identified in the literature, including acidification, biodiversity loss and agricultural land transformation (which is closely related to the soil carbon change impacts noted above). Novel impact categories could also be explored (e.g. inclusion of the nutritional aspect as an impact category) (McAuliffe et al. 2023; McLaren et al. 2021b). Finally, additional targeted data collection is recommended for any future avocado-related research in NZ. This includes data related to the ‘source of origin’ of fertilisers and agrichemicals used on NZ avocado orchards, as well as collecting soil conditioning data over several years which can then be used to calculate average annual inputs.