Part 3: Approaches and methods to integrate diverse socio-environmental data

[Music] Welcome to our series of videos focusing on interdisciplinary socio-environmental team synthesis. I’m Margaret Palmer, Professor at the University of Maryland and Director of SESYNC, the National Socio-environmental Synthesis Center. These videos were designed as a tool to help understand and, hopefully, solve socio-environmental problems. After all, environmental problems are by definition – social problems. The first video in the three-part series presents examples of how researchers use a socio-environmental lens to study environmental problems. In it, we introduce the concept of socio-environmental systems as dynamical systems characterized by feedback loops both within and between social and ecological subsystems. We also describe several examples of socio-environmental synthesis research. The second video in the series provides an overview of the synthesis research process, as well as, the concepts and approaches used by synthesis teams. We explore the importance of integrating diverse forms of social and environmental knowledge when addressing socio-environmental problems. And we also describe some of the findings from on-going work on what fosters synthetic team research. In this third and final video, we focus on the steps involved in the synthesis process. Development of a shared conceptual model of the problem, including the components and drivers of change in the socio-environmental system being studied, is the critical first step. Data collection, integration and analysis follow. In this video, we provide an overview of the diverse types of data and methods used in socio-environmental synthesis. Heterogeneous data from across the natural and social sciences forms the basis of socio-environmental synthesis, and we discuss in particular the increased interest in integrating qualitative and quantitative data. Common methods include statistical and spatially explicit modeling, dynamical systems modeling, and agent-based modeling. These and many other specific methods reflect the general analytical approach of synthesis research, which requires teams to articulate a shared conceptual understanding of a socio-environmental system, identify diverse data sources to measure social and environmental dimensions of the system, and integrate those data into complex analyses. We open this video with an example of a synthesis project on human-wildlife conflicts and we focus on how the team developed a conceptual model to guide their research. Few wildlife species face more potential conflicts with humans, than tigers, which require large areas for hunting and raising their young, and inhabit some of the most densely populated regions of the world. Humans and tigers compete for space – the result? Most of their historic range has been lost. Hunting and habitat loss have also resulted in a decline in their ungulate prey populations. Once numbering more than 100,000 worldwide, today, as few as 3,000 tigers remain in the wild. India is home to most of these and despite major efforts to combat tiger poaching, the numbers are still high. Skins, bones, claws, and other parts, mostly going to China, demand a high price. Shown here mounting a motion sensitive camera on an animal trail in the Chitwan National Park, Dr. Neil Carter of Boise State University, has been studying the interactions between humans and large carnivores. He and his colleagues formed a SESYNC synthesis team to develop a conceptual socio-environmental system framework. This framework now serves as a common platform for guiding research on wildlife poaching feedbacks, to better inform policy making and enforcement. Their conceptual framework integrates factors related to both human motivations for poaching and animal vulnerability. Poaching is viewed as occurring within a nested, multiple level socio-environmental system. The inner most levels are closest to the physical act of illegal killing and reflect social characteristics that immediately effect the opportunity to poach at a given time and place, for instance, lack of enforcement patrols. Factors in the intermediate levels reflect the individual characteristics that may directly motivate a person to poach, or increase an animal’s vulnerability. The outer-most levels, include broader social and ecological contexts in which human societies and wildlife interact. The Carter synthesis team applied their framework to two cases for which there were empirical data: tigers in Laos and wolverines that prey on livestock in Sweden. For the tigers, global factors like market demand, motivates humans to poach. When wildlife enforcement patrols increased in the past, local factors such as hunting methods changed, so that tigers are now largely caught in difficult to detect wire snares. Enforcement patrols did lead to reduced hunting of the tiger’s ungulate prey which should be beneficial for tiger populations; however, despite this, tigers continued to decrease and their rarity will likely increase their market value. The development of this conceptual model for human-wildlife socio-environmental systems, is an example of what is typically the first step in the synthesis process. As we discussed in video 1 of this series, as with all complex systems research, synthesis teams must consider many components and their interactions, including non-linear feedbacks that may span scales. Developing such a framework lays the foundation for additional work, including how to structure research focused on a class of problems, what type of data need to be synthesized, which feedbacks may be driving overall system behavior and what is needed to take a systems dynamics research approach. Conceptual models are based on existing knowledge among team members, or others in relevant fields. Once the interdisciplinary wildlife-human interaction team was formed, they co-developed a model that was informed by research findings and theory from many fields including: situational crime prevention, human psychology and animal behavior, effectiveness of education protocols, studies on social norms and political science, as well as wildlife population and community behavior. The model formally expressed how the team believes the system works, and how it is structured, which helps them identify critical research questions and data needs. The team is now moving forward to synthesize data and use a variety of analysis methods to evaluate how different policies and practices can influence feedbacks to reduce poaching. Regardless of the specific topic, socio-environmental synthesis research relies on many different forms of information. This might include data collected by quantifying an event or outcome, running a computer simulation, collecting photographs, transcribing interviews, or capturing social media activity. In other words, highly heterogeneous data. However, if all of the data are well structured, they can be organized in a defined way like tables of numbers or textual categories, and they are easier to integrate for a synthesis project than unstructured data. The latter is often text heavy but also includes elements like images, media content, and material “scraped” from websites. The growing availability of data and new ways to extract data really fueled synthesis research as well as interdisciplinary collaborations to address research problems that really can’t be addressed by disciplinary approaches alone. Work by sociologist Dr. Heather Randell and geographer Dr. Clark Gray, provides an example of an interdisciplinary synthesis project focused on communities in rural Ethiopia. They integrated data on climatic conditions with socioeconomic, demographic, and educational attainment data to explore factors influencing educational outcomes. We used the synthesis method for this research and what that involved was combining temperature and precipitation data from Ethiopia with longitudinal household survey data on children’s ages and schooling completed… We also controlled in our analysis for a number of other factors, including gender, family composition and socio-economic status as indicated by land ownership. Heather and Clark’s conceptual model depicted potential relationships between environmental change, agricultural production, and educational attainment for children living in farming households. It helped them to fine-tune their specific research questions and the type of data they needed. We first assembled a relational data base of the climate and social data – and then used logistic regression to test the hypothesis that climate variability influences education outcomes among children from rural agricultural households in Ethiopia. Research like this is important because many people would think that climate change is affecting education outcomes. But education is really important for multiple factors including adaptation to future climate change, so it’s important to think about how we can mitigate this relationship and ensure that kids stay in school in light of increasing climate variability and increasing extreme weather. Their data ranged from economic indicators, to demographic variables, to characteristics of behavior – as well as climatic measures in gridded form. While heterogeneous, the various datasets had one variable in common – household location – so it was fairly straightforward to build a relational database. Their data were all quantitative, which is not always the case for many socio-environmental problems. The team did have to make decisions on how to disaggregate the data that were collected at scales larger than the household. As the team worked to organize and standardize such diverse data, they didn’t have to assemble all of the data sets in the same place – that is, the datasets did not all have to reside on the same server. Instead, the datasets could be interconnected by a computer network and a data model that specifies how to link them; this is known as a federated database. Federated databases are especially useful for teams that are geographically dispersed and that are bringing together many different data sets into a single analysis. The household level survey data from Heather and Clark’s study were all in categorical or numeric form, making it feasible to analyze their data and test their hypotheses using widely available statistical software. However, some teams that have qualitative information such as that from textual sources or survey responses; are able to convert it to quantitative data, as was the case with a synthesis project on marine conservation. The goal of the synthesis project that we undertook was to get a better understanding of when marine protected areas are effective or how can we make them more effective. And so, a lot of work has been done looking to see how we design marine protective areas to be effective but then there hasn’t been much work to understand how management and governance can lead to more successful marine protected areas. To do the analysis, we were, after many months we were able to compile over 10, around 10, possibly to 20 different datasets and databases and these ranged from Excel files, these ranged from remote sensing data, they or even Word documents or excel files that had to be converted and brought together to into a single database. So the way we combined the qualitative data to our quantitative datasets was to convert the qualitative scores into ordinal data. And so we were able to combine these qualitative values from multiple sources into a single range of Likert scores and with these scores we were then able to use these within a model to test them and their interactions with ecological outcomes. Qualitative data can also be kept in its original form and used to reveal connections, interpret quantitative data, and understand contextual factors. It is particularly useful in asking “how” or “why” questions, especially those focused on governance and long-term or unexpected processes, as SESYNC’s Dr. Steve Alexander explains. But then there’s other variables that don’t lend themselves to being quantified. A lot of those variables have to do with governance, so when you think about things like policies, or rules and regulations, or kind of the, the different types and nature of actors who are involved, those things are much harder to quantify. And so the result is, in that project, we’ve had to take a very different approach. For the qualitative data, what we’ve focused on is using an approach known as process tracing, which we’re borrowing from political science. And this is a historical qualitative analytical approach, where you can draw on a diverse array of qualitative data to start to get at understanding causality, and the interaction between multiple variables and significant events. Qualitative sources may provide information on contextual factors, such as the timing of major socio-cultural transitions, which can be valuable. Recent work by Dr. Alan Taylor and his colleagues focused on changes in the frequency and extent of large wildfires in California over the last 400 years. They integrated climate data, tree-ring data on fire history and records of the area burned – and used written documents, some of which were ethnographic, to show that changes in wildfire activity coincided with major socio-cultural changes. After the Spanish missionaries arrived and there was depopulation of Native Americans, fires increased dramatically and did not decline until the Gold Rush and Euro-American settlement. By the early 1900’s fire patterns again change when a policy of fire suppression was put in place. Qualitative information such as that in texts and ethnographies can be analyzed to identify themes. Researchers can use theory or a priori knowledge to explore and code text themselves, or they can use computational methods like topic modeling for identifying patterns and discovering themes in a large number of documents. In addition to texts, themes and meaning can be extracted from other forms of unstructured data. The biggest growth today is in the use of social media, such as content from tweets, Facebook postings, and Flickr images. Not only are tools available for scraping this type of data from the web, but there are also analytics tools for detecting patterns, identifying connections, and building networks. Sentiment analysis can be used to identify opinions, attitudes, and preferences expressed in social media data, and because much of the information is geocoded, material can be linked to other spatial data. One of the most exciting things about this project are the different kinds of data we’re bringing together, so we have a huge dataset of lake attributes. So, everything from nitrogen, phosphorus, lake clarity, lake size, and characteristics about the watersheds of different lakes. So we have a dataset of lakes that crosses 17 states and 50,000 lakes. And we’re integrating that limnological data with data on social media that reflects where people visit those lakes. So we’ve used twitter and Flickr to understand where people visit lakes across that huge gradient to understand if people are visiting clear lakes more often than less clear lakes and from that we’ll be able to understand the patterns of visitation and how peoples’ preferences for water quality vary. Once teams have assembled and harmonized their data, there is an endless array of methods for analysis. And the methods toolbox is growing every day – in part, this is being fueled by the rapid growth in available data, the desire to identify underlying patterns and the interest in combining data with other information in novel ways. Andrew Jorgenson, from Boston College, talks about various methodological approaches commonly used by environmental sociologists. That in terms of statistical techniques or quantitative techniques, we use a lot of the same statistical modeling techniques that are employed by our colleagues in economics and political science but also those that are used by natural scientists. And so some of the more cutting edge methods, statistical methods that we use would be dynamic longitudinal methods using different kinds of fixed effects modeling techniques, multi-level modeling techniques which are also known as hierarchal linear modeling. Well, we also use structural equation model techniques and increasingly we are using spatial analysis techniques as well. We also use a variety of qualitative methods such as ethnographic methods, that sort of overlap of anthropology, historical methods, comparative case study methods, and, you know, interviews and text analysis so we really have a very large sort of suite of both quantitative and qualitative methods as well as hybrid methods. Actually, qualitative comparative analysis I think is a nice example of what I and some others would be, could call, like a hybrid methodology. It’s a very flexible method at dealing with complex interactions much more so than statistical models, regression models. Recent work on global water problems provides an example of the use of Qualitative Comparative Analysis by a team of hydrologists. Using QCA, the team showed that while there is no single metric associated with all types of water crises, distinct resource utilization patterns were associated with distinct causal factors. The work suggests that predicting water crises and finding solutions will require water researchers to look beyond demand for water, to the drivers of demand – as well as the institutions that govern water use. A case study approach like this team used, can be a valuable addition to synthesis projects, many of which use modeling approaches outlined in the next example. Synthesis teams make extensive use of a diverse array of modeling approaches. We have already mentioned a few statistical modeling approaches, which are commonly used to understand the dynamics of components of a socio-environmental system. Other types of models are often used to produce outputs that will be integrated with other types of data to test hypotheses. Examples include outputs from climate change and hydrologic models. Outputs from global climate models were used in a recent synthesis project to estimate carbon dioxide-driven ocean acidification and its effects on shellfish, and local economies that depend on shellfish for livelihoods. While elevated atmospheric CO2 from climate change is the primary cause, acidification may be amplified in areas experiencing higher than average levels of carbon dioxide due to processes internal to the ocean. For example, eutrophic or upwelling regions of the ocean have elevated carbon dioxide due to high levels of organismal respiration. Focusing on the coastal U.S., Julie Ekstrom and her team, combined spatially-explicit model outputs on ocean acidification with data from local regions on the amplifying factors. For each region, the team also gathered data on shellfish harvesting, associated economics, and adaptive capacity. The ability to adapt was represented by three classes of indicators: status of state government climate and acidification policies, local employment alternatives, and availability of science. The study found different coastal regions, face varying combinations of risk factors, making them unique “hot zones” of vulnerability to ocean acidification. Many of the regions most economically dependent on shellfish, are currently the least prepared to respond, and the authors discuss the types of action that can help at-risk communities while protecting our environment. In a second example of ongoing work using modeling approaches, a synthesis team is combining modeling, data integration and multiple forms of analysis to distinguish the human and ecological dimensions of drought vulnerability. This will allow them to identify opportunities for mitigating and/or adapting to drought – that is to say, informing future decisions. Their project relies on synthesizing heterogeneous data types – time series, gridded historical data, images from remote sensing, as well as qualitative information from focus groups, interviews, and drought plans. But it also involves modeling past ecological impacts using machine learning techniques and correlative niche modeling to parse out the strongest drivers of drought vulnerability. Studies such as the drought example often use a systems dynamics approach, which is helpful to identify and articulate the environmental and social pressures, outcomes, and feedback on systems, subsystems and components. In the formal sense this is a simulation model of coupled, nonlinear, differential equations. These models can be used to test hypotheses and explore outcomes under different assumptions – something that is hard or impossible to do empirically for socio-environmental systems. Ecologist SESYNC scholar Ginger Allington uses systems modeling in her synthesis research. Systems dynamics modeling is a way to try to understand behavior of complex systems. It was originally developed for trying to understand industrial systems, and there’s the movement of materials through manufacturing and industrial processes. But one of the things that’s really useful for is incorporating feedback loops and non-linear dynamics, and we can put together an entire system to the best that we can with the data that we have to really look at the behavior of complex systems over time. System dynamics models are most useful in my opinion particularly within the context of socio-environmental systems as a way to explore system behavior and to compare potential future trajectories under alternate futures. However, if we are trying to model the behavior of the fully coupled socio-environmental system – that is to say, the behavior that emerges from all the components and their interactions – things get very challenging. The many nonlinear feedbacks, processes operating at different temporal or spatial scales, and major state changes are difficult to model using traditional systems dynamics approaches. This is particularly the case when systems undergo major changes. Despite the challenges, there has been a great deal of interest in developing methods to predict when systems will change states. The California fire history example shown earlier, is an example of a socio-environmental state change, as was the Baltic sea cod fishery example shown in the first video of this series. Models are built based on what we know about a system from its past behavior. System dynamics models are built based on the system’s structure and use differential equations to describe processes within. But changing the values of parameters in the equations, or changing the strength of feedbacks, will not be adequate if a system is undergoing a major change. Gary Polhill and his colleagues, provide a clear overview of the problem and paths forward. They emphasize that when using a systems dynamics modeling approach to study systemic changes, the model’s ontology itself may need to change. New state variables, feedbacks, or system boundaries may be needed. Thus, some researchers shift to a different modeling approach. A somewhat more flexible modeling approach is Agent Based Modeling, or ABM. Rather than assuming a systems’ structure up front like system dynamics models, ABM begins by focusing on the behaviors of the individual components, or “agents”, in the system. These agents shape and change the state and structure of the system over time. Importantly, it allows for diversity in behaviors for a given agent type. The rules that govern an agent’s behavior are generally drawn from a distribution function. ABMs also allow for spatial structure and processes so it is useful in studying dynamics across landscapes or networks. A useful example can be found in the work of Nick Magliocca and his colleagues, who used an economic agent-based model of land use to explore the importance of economics in the spatial patterns of development over time. Understanding dispersed patterns of urban development is important for designing policies that take into account the environmental Impacts of sprawl. Our model had three types of agents – consumers who purchase houses of different types in different locations, a developer who converts undeveloped land into residential housing, and farmers who choose between farming and selling their land. The model simulated a series of market events including buying land, building houses, and setting asking and bid prices. Agents decisions were governed by a set of mathematical algorithms – these are learning rules based on cognitive models so that consumers, developers, and land owners could adapt over time as they would in the real world. We found that early development patterns are determined by agricultural land values, but eventually over time become dominated by farmers’ land price expectations. This work pointed to the importance of people’s expectations and behaviors – it highlighted the needs to take into account more than just economics. As the brief sections on modeling socio-environmental system behavior indicate, the study of these systems is not always easy. Socio-environmental research is a growing field and it is challenging researchers to think outside the box, to learn new vocabularies and methods, and to develop novel approaches and tools for analyzing these complex systems. But the payback will be immense. The difficult social and environmental problems we face today are inextricably linked to one another. And solving them will require new insights and unparalleled collaborations – not only with other scholars, but also with stakeholders from diverse sectors. The good news is that this presents a major opportunity. There’s lots of really sweet spots for synthesis right now, and they span not only academic disciplines, but also engage nonprofits or industry sector. Because data is everywhere, and we’re in this era where we’re collecting information from all different places. There’s a real opportunity to integrate information, and ideas and data across multiple sectors, to essentially address major challenges in understanding global change, and how societies are going to adapt. Synthesis research may involve many types of analytical approaches including: those that involve statistical analysis; those that involve the integration of quantitative and qualitative data; OR those that involve the development of new modeling approaches for dealing with feedbacks that operate on different spatial or temporal scales. These methods are complemented by the team science process, through which diverse groups of individuals and their knowledge are brought together to develop a shared conceptual framework that can then be used to guide the analysis of heterogeneous data and information. While this video completes SESYNC’s 3-part introductory series on socio-environmental synthesis, it does not end our ongoing search for new and better ways to discover, integrate, and analyze diverse forms of social and environmental knowledge. Throughout this series, we have presented a range of synthesis projects that are meant to highlight the linkages between humans and their environment as well as highlight some of the most common methods and approaches used in synthesis research. We can’t emphasize enough the importance of team diversity to socio-environmental research. This includes scholars from many disciplines and often includes individuals from stakeholder communities, NGOs, public agencies, or the business sector. As this series has emphasized, society’s most challenging and complex environmental problems are rooted in the deeply interconnected relationship between humans and the ecosystems in which they live. By taking the time to understand the different perspectives, methods, and cultures of all actors involved in socio-environmental systems, we can better understand the complexity of an environmental issue and then work to resolve it and to improve the lives of people that are affected. As we said at the beginning of the series all environmental problems are social problems. The new and growing field of socio-environmental research works to contribute to the resolution of these problems through research and analysis that recognizes and emphasizes the connections within and across socio-environmental systems. All of these videos can be found at the url that follows. And we encourage you to visit our site to explore some the many examples of socio-environmental synthesis that SESYNC has supported as well as to explore the many opportunities, tools, and learning resources we have developed. Thank you. [Music]

Leave a Reply

Your email address will not be published. Required fields are marked *