The inventories, however, must then be structured according to the formats usable by openLCA. This modeling process also requires an abrupt switch from a very high level diagram — the flow diagram — to a detailed inventory requiring low-level information about her farm. There is no process connectivity between the two steps and information does not flow between the models. Two: Time and effort overhead. Models must be manually inspected to ensure completeness and correctness. This may result in a suite of errors, such as models that misrepresent the system, models that are missing portions of the real world system, or models that have incorrect connections between processes. Further, the modeling effort expended to create the flow diagram is not utilized to reduce effort in the LCI creation process. The majority of LCI databases are focused on industrial production systems. Those that are concerned with food are oriented toward the production of processed foods as a resultof partnerships with food processing stakeholders. Data required to conduct an LCA of any alternative agricultural systems is unavailable and puts the onus on the system owner to collect primary data. Three: Lack of flexibility. Only one type of granularity, the unit process, can be modeled. There is no support for creating logical groupings of unit processes in the form of components and no capacity to create hierarchies of such components. Alice is locked into two disconnected levels of abstraction: the high-level flow diagram, vertical grow and the low-level inventory. Even once an LCA model is created, current mechanisms only support the sharing of unit process data is shared.
No other reusable portions of the model are explicitly supported. This also means that as things on the farm change, an entirely new LCA model would need to be created.Life cycle assessment has enabled people to investigate, quantify, and understand the environmental impacts of agricultural and industrial systems. On the whole, LCA is useful, but it is not without its limitations. LCA requires meticulous and tedious data collection for every single process within a system. A good LCA is comprehensive and results in detailed models, however it can be a time consuming and cost-prohibitive process, depending on the size, complexity, and novelty of the system under analysis. The development of software also involves modeling, and there has been substantial research into improving the software modeling workflow. In the following analysis, I call on software engineering research and practice to tease apart some of the modeling challenges faced in modeling the environmental impacts of agricultural systems, as well as to propose opportunities for future work. Fred Brooks discriminates between essential difficulties with software — those relating to intrinsic characteristics — and the accidental difficulties — those relating to temporary or circumstantial characteristics. He goes on to state that to address accidental difficulties, one must promote incremental improvements, but to solve essential difficulties, one must promote revolutionary improvements. The challenges faced in the modeling of the environmental impacts of agricultural systems involve addressing both the essential and accidental difficulties. LCA is one of many environmental impact assessment methods that aim to address these difficulties. LCA methods, tools, and data have undergone immense incremental improvements . However, several essential difficulties remain.
Essential difficulties faced when modeling the environmental impacts of agricultural systems involve the representation, or capture, of complexity, change, and context of such systems. Interestingly, some of these mirror the essential difficulties or characteristics of software as defined by Brooks: complexity, conformity, and changeability.Only some of these attributes are represented in current LCA models: the flow diagram informally represents the scope and boundaries of the system and data documentation formats provide structure when representing unit processes. Dependencies, subsystems, and other potentially relevant attributes are linguistically represented, i.e., they are often described in reports that accompany LCA models. Through the structuring of unit processes as prescribed by DDFs and the list-of-unit processes LCI inventory, LCA results in a powerful declarative model: one where within the system boundary, unit processes represent all major flows of material and energy withinthe system, and are connected through the functional unit.Current LCA modeling is similar, in that it is effective at retroactively assessing static systems. Agricultural systems have a symbiotic relationship with the environment, and are closely tied to the health of the environment due to the interconnectedness with many natural systems. These assessment of agricultural systems is unique as not only is one assessing a system that is directly dependent on the natural environment, but that also has a large component of it that are not of human construction, i.e., they are a part of nature: plants, animals, and the land. The quality of the food produced on a farm is related to the quality of the land, air, and water used. These are seasonal systems that are constantly producing, consuming, and evolving over time.
The global warming potential of a farm in Northern California over the Spring of 2014, when they were growing peas and corn, will not be the same five years from now. For this reason, LCA models are time, location, and system sensitive. Dynamic models of agricultural systems that represent changing resource flows, however, are not available in the LCA modeling paradigm to date. Existing LCA models and tools do not capture change effectively. They provide functionality to increment a model version number, but updating and maintaining models to keep up with changes in the real world is difficult. Excel has the “track changes” functionality,which is a helpful collaboration aid that provides different authors with awareness of how a document is changing, but is not explicitly geared toward capturing changes in LCA models. GaBi does something similar. The tool logs all changes to each object at every save . Once again, the goal is to keep track of which user changed the model to support collaboration and not necessarily to augment the capacity of the model to reflect changes in the world. openLCA simply provides a version number for each object in the file that can be incremented as changes are made.State of the art LCA modeling tools do not provide the ability to revert to previous models, compare models over time, or maintain a model history that reflects the changing world. If one is to update and maintain a model over time, it would involve a manual process of backing up and continuing from where one left off. While this is a valid approach, grow vertical given the scale to which some LCA models can grow and the increasing complexity of agricultural systems, more advanced tool support is required.We are constantly producing and trying to connect different types of information in the real world, as well as in our models. A model has two aspects: the representation of a system, and perspective on the system. When these models are treated in isolation, they provide representation without context. Such models cannot be connected with each other to produce meta-models, nor can existing models be reused. When a model is created of an agricultural system, an artificially created system boundary sharply closes off the rest of the world from the system under analysis. The flow diagram tries to capture some contextual information by showing materials and energy flowing into and out of the system as a whole, but, as discussed earlier, the flow diagram is quite disconnected from the eventual LCA model. Trying to capture the context within which a system exists is a difficult problem. After all, one must scope a modeling problem appropriately lest they end up trying to simulate the world. The question remains, however, how does one capture the context of a system? Standalone LCAs are still common as they are conducted when stakeholders of a particular system are interested in self-evaluation and improvement. Comparative LCAs are conducted when stakeholders in an industry are interested in pitting one set of production techniques against another, to analyze whether or not a new or alternative concept is actually better for the environment than an older or standard approach, or any of the other reason discussed previously in this paper. Conducting an LCA that is wide in scope is difficult, and so, when people are interested in understanding wider or more far reaching impacts, existing LCA studies are used to conduct retrospective meta-analyses. Environmental consulting agencies also tend to conduct these analyses, as part of their business model is LCA as a service , where they charge for their expertise, time, and effort. Depending on the intellectual property issues at play, they may own the model details and are not beholden to reveal the raw data to anyone else.
LCA models provide perspective on a specific system without any context. They are difficult to connect with each other to provide more holistic environmental impact assessments, and it is difficult to reuse partial models. People all over the world are producing extremely detailed models, with a lot of time, money, and effort going into inventorying complex systems. A vast amount of data and a large number of models are produced through LCA. We are unfortunately, producing, but not connecting, environmental impact models. No explicit interfaces or connectors are available to connect entire LCA models.LCA models are intended to provide efficient and convenient access to information about the environmental performance of production systems, such as agricultural systems. However, due to the mismatch between the current LCA modeling language, workflow, tool, and the systems represented, I believe that an important opportunity has been missed to capture the complexity, change, and context of agricultural systems. In this section, I describe a potential avenue through which to address these modeling difficulties. In software engineering, the Object-Oriented paradigm came about as an attempt at capturing the essential complexity of software. Booch describes two kinds of decomposition: Algorithmic and Object-Oriented decomposition. The algorithmic results in a top down structured model, declarative and directed. On the other hand, in the OO approach, one decomposes the systems using the key abstractions. Booch notes: “[In object oriented modeling,] we view the world as a set of autonomous agents that collaborate to perform some higher level behavior”. Current LCA modeling utilizes a hybrid of the two, where the flow diagram is created using an OO approach, but is then quickly abandoned for a mostly algorithmic decomposition in subsequent models. Given the modeling challenges described in this chapter, I propose that a potential means to capturing the inherent complexity of agricultural systems is through a new modeling language that allows for a more object-oriented approach.A new modeling language would not be a silver bullet. It would not necessarily capture all of the complexities, changes, or context of agricultural systems, nor would it remove all accidental misrepresentation of systems. What it would hopefully do, is make the process of representing the environmental performance of such systems easier and less tedious, providing clear and distinct ways to represent and connect modular system models. They are prone to change over time due to rising environmental issues , advances in agricultural practices and technology, changing needs of an ever-growing human society, and a dynamic economic context. Our farms are part of the highly interconnected web of industrial civilization, with dependencies on many other systems. Environmental impact assessments are conducted, regulations written to govern production, tools built to guide the flows of materials and products through supply chains, and food labels created to assist in consumer decision making. At the core of these issues is a mismatch between existing environmental assessment tools and the needs of small- to medium-scale farmers , who do not engage in large-scale industrialized agricultural practices. In Chapter 3, I described the essential difficulties faced when modeling agricultural systems.The Grounded Theory Method is methodology used to develop theory about phenomena through iterative interrogation of data. GTM offers a means to explore new territory, particularly when there is a lack of dominant theory. While the goal of this study was domain understanding, I used GTM as a means of structuring the design of this study and utilize GTM techniques for subsequent data analysis. Through use of GTM, my broader goal was to develop theories of design for sustainable agriculture. Muller writes that one must “remain faithful to the data, and to draw conclusions that are firmly grounded in the data” . I do this by iterating between recruitment, data collection, and analysis, exploring the different concepts at play, while gradually developing a theory of how to design a consistent mechanism for modeling sustainable farms and their interactions with the environment.