Modelling of pesticide emissions for Life Cycle Inventory analysis: Model development, applications and implications

The work presented in this thesis deals with quantification of pesticide emissions in the Life Cycle Inventory (LCI) analysis phase of Life Cycle Assessment (LCA). The motivation to model pesticide emissions is that reliable LCA results not only depend on accurate impact assessment models, but also good emission inventories.

Recent LCA studies of agricultural products that take toxicity impacts into account show that pesticide emissions considerably contribute to toxicity impacts. At the same time, such conclusions are derived using a simplified approach to quantify pesticide emissions. The research presented in this thesis centers around PestLCI 2.0, a model to cal-culate pesticide emissions to air, surface water and groundwater for use in LCI.

PestLCI2.0 is an updated and expanded version of the PestLCI model, released in 2006. The boundaries between ecosphere and technosphere in the model are defined by a ‘technosphere box’, which includes the arable land where the pesticide is applied, the field soil up to 1 meter of depth and the air column above the field up to 100 meter.

When a pesticide leaves this box, it is considered an emission. The model works with a primary distribution, where the pesticide is deposited on the crop, on soil or emitted due to wind drift, followed by secondary processes that determine the pesticides’ fate. In PestLCI 2.0, most fate process modelling has been updated, most notably the modelling of pesticide volatilization from leaves and pesticide runoff.

The model was expanded by the inclusion of macropore flow, which leads to pesticide emissions to groundwater. Moreover, PestLCI 2.0s databases with active ingredients, climates and soils were updated, broadening the applicability of the model to European circumstances. A case study showed that emissions vary with variations in the climates and soils present in Europe.

Emissions of pesticides to surface water and groundwater calculated by PestLCI 2.0 were compared with models used for risk assessment. Compared to the MACRO module in SWASH 3.1 model, which calculates surface water emissions by runoff and drainage, pesticide emissions to surface water calculated by PestLCI 2.0 were generally higher, which was attributed to differences in the modelling approach between the two models.

The model comparison for emissions to groundwater showed that PestLCI 2.0 calculated higher emissions than FOCUSPEARL 4.4.4 (modelling chromatographic flow of water through the soil), which was attributed to the omission of emissions via macropore flow in the latter model. The comparison was complicated by the fact that the scenarios used were not fully identical.

In order to quantify the implications of using PestLCI 2.0, human toxicity and freshwater ecotoxicity impacts obtained with two inventory approaches were compared. The first approach was PestLCI 2.0, the second is the currently prevalent approach (the Ecoinvent approach), which assumes that 100% of the applied mass is emitted to agricultural soil.

For both impact categories it was found that the PestLCI approach results in impacts that on average are three orders of magnitude lower. This conclusion was found to be valid for characterization of the impacts with both USEtox and US-ES-LCA 2.0 characterization factors. The difference observed between these approaches will have implications for the comparison of toxicity impacts between conventional and organic agriculture.

However, the difference in pesticide use and the corresponding environmental impacts is only one of the many aspects that are relevant to assess when discussing sustainability of both types of agriculture. A second implication from these findings is that the contribution of pesticide emissions to the overall toxicity im-pacts of agricultural products may be lower than what is currently found in LCA studies.

Since the PestLCI and Ecoinvent approaches differ in both their ecosphere-technosphere boundary setting and in the modelling of fate processes within the technosphere, a hybrid approach was also used to calculate toxicological impacts. This approach combined the fate modelling of the PestLCI approach with the technosphere boundaries of the Ecoinvent approach.

The toxicological impacts of this approach showed that it is the technosphere boundaries, rather than the in- or exclusion of fate processes, that determines the differences observed between the PestLCI and Ecoinvent approaches. This technosphere-ecosphere boundary is impossible to define objectively in the case of LCAs of agricultural products: it depends on the practitioners’ values what is environment and what is man-made production system.

Therefore it is advisable to discuss what LCA should aim to protect, instead of where the boundary should be located. The first of the two applications of PestLCI 2.0 presented in this thesis is the case of pesticide emissions in conventional kiwifruit cultivation in the Western Bay of Plenty district in New Zealand.

For nine scenarios, based on different combina-tions of local soils and climates, pesticide emissions were calculated with Pes-tLCI 2.0 and subsequently characterized with characterization factors obtained using USEtox. The emissions to air showed little variation between the nine assessed scenarios.

Emissions to surface water and groundwater showed larger variations. Despite this, the differences in the freshwater ecotoxicity and human toxicity for the nine scenarios were small. In an LCA context, when considering un-certainties in emission modelling and impact assessment, these differences probably are not relevant.

For all nine scenarios, it was found that emissions of cyan-amide dominated the toxicological impacts. A second application of PestLCI 2.0 was in the comparison of the environmental impacts of barley cultivation in Denmark under current (2010) and future (2050) climatic circumstances. The functional unit of this study was 1 kg of barley at the farm gate.

Using an attributional approach, impacts of co-products were handled by economic allocation. Impact assessment was done with ReCiPe (hierarchist perspective), except for toxicity impacts, which were characterized using USE-tox. The differences between four scenarios, based on combinations of wet and dry climates, and sandy and sandy loam soils, for barley cultivation under current climatic conditions were found to be small.

Differences in impacts between cul-tivation in current and future climatic conditions were concluded to be mainly driven by differences in grain yield. The use of economic allocation was found to be a key issue, since the price levels of 2050 can’t be predicted with any reasonable certainty. Although PestLCI has been updated and expanded, further improvements are still possible.

A number of improvements and suggestions to increase the model’s applicability are discussed. These suggestions focus on both the fate modelling (for example wind drift, degradation and volatilization from leaves) and the boundary setting of the model.