Lecture 4: Scenarios for the Future and Land Use Change Modeling
Reading: Hurtt et al. (2020) and Popp et al. (2017)
Slides as Powerpoint: Download here
Video link: On Youtube
Content
Introduction and Course Context
This lecture focuses on scenarios for the future and land use change modeling, building upon the theoretical framework established in previous sessions while moving toward more specific applications and datasets. The session emphasizes practical implementation through hands-on work with QGIS software and the Land Use Harmonization 2 (LUH2) dataset, marking a transition in the course toward more applied methodologies in earth economy modeling.
The course trajectory has evolved from introducing earth economy modeling and examining its origins in national sustainability models and general equilibrium approaches to now focusing on the specific scenarios that serve as inputs to these complex models. This progression reflects the pedagogical approach of moving from theoretical foundations to practical applications, ensuring students understand both the conceptual frameworks and their real-world implementations.
Looking ahead in the course schedule, the next session will shift focus toward understanding integrated assessment models for climate change through hands-on exploration. This will include examination of the DICE model developed by William Nordhaus, which despite its controversial use by climate skeptics to minimize climate change impacts, remains a foundational model in the field. The course will also explore the Green DICE Model update from Frances Moore, continuing the emphasis on practical application through direct engagement with these modeling tools. Subsequent lectures will cover inclusive wealth concepts and ecosystem services, maintaining the balance between theoretical understanding and practical implementation that characterizes this applied course approach.
Data Organization and File Management
Establishing a Standardized File Structure
The importance of proper file organization cannot be overstated in computational modeling and data analysis. The majority of programming errors encountered by students, particularly those new to coding, stem from incorrect file paths and misplaced data files. This fundamental issue, while seemingly trivial, can consume significant time and create frustration that detracts from learning the substantive content of the course. To address this challenge proactively, the course implements a standardized file organization system that all students must adopt.
The standardization follows best practices established in the programming community, utilizing the user directory as the foundation for file organization. This approach ensures cross-platform compatibility across Windows, Mac, and Linux operating systems, eliminating potential issues that arise from system-specific file path conventions. The user directory, typically located at C:[username] on Windows systems or /Users/[username] on Mac systems, serves as the root location for all course-related files.
Within the user directory, students create a folder called “Files” to maintain separation from system configuration files and other program-generated content that typically populates the user directory. This Files folder serves as the container for all course materials, providing a clean organizational structure that prevents confusion with system files while maintaining accessibility for course work.
Implementing the Base Data Structure
The core of the file organization centers on the BaseData folder, which resides within the Files directory. This folder mirrors the structure of the shared Google Drive repository, ensuring consistency between the cloud-based source and local implementations. The mirroring approach means that any file path referenced in course materials or sample code will work identically for all students, regardless of their individual system configurations.
When downloading files from the shared Google Drive repository, students must replicate the exact folder hierarchy found in the cloud storage. For instance, the LULC_CurrentWillamette.tiff file resides in the path BaseData > Invest Sample Data > Carbon. Students must create each of these folders in sequence, maintaining the exact naming conventions, including capitalization and spacing. This precision in folder naming prevents errors that could arise from minor variations in folder names when code attempts to access specific files.
The Google Drive Desktop Sync application offers an automated alternative to manual file downloading and organization. Students who install this application benefit from automatic synchronization that maintains the proper folder structure without manual intervention. However, recognizing that not all students may have access to sufficient storage space or prefer to install additional software, the manual download method remains the primary approach taught in the course.
The IPBES Framework and Scenario Typology
Understanding the Intergovernmental Science-Policy Platform
The Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services (IPBES) represents a crucial international body dedicated to synthesizing current understanding of biodiversity and ecosystem services. This organization serves as the biodiversity equivalent to the more widely known Intergovernmental Panel on Climate Change (IPCC), addressing the historical imbalance in attention between climate change and biodiversity conservation. IPBES combines, funds, and supports scientific research in biodiversity and ecosystem services, providing a coordinated international approach to understanding and addressing biodiversity loss.
The organization gained significant recognition through the publication by Sandra Diaz and colleagues in 2015, which presented the IPBES conceptual framework. While the framework itself contains substantial theoretical complexity that extends beyond the scope of this course, its practical contribution lies in establishing a useful typology for scenarios that has become standard in the field. This typology provides a structured approach to thinking about different types of future projections and their applications in sustainability science.
Four Types of Scenarios in Sustainability Science
The IPBES framework identifies four distinct types of scenarios, each serving different purposes in sustainability analysis and decision-making. Exploratory scenarios represent the most common understanding of scenario analysis, where researchers examine the space of possible futures from a current starting point. These scenarios explore various trajectories without prescribing specific pathways, allowing for broad investigation of potential outcomes. For instance, given observed declines in nature’s benefits to people, exploratory scenarios investigate various ways this trend might continue, stabilize, or reverse, without necessarily specifying the mechanisms for achieving particular outcomes.
Target-seeking scenarios take a fundamentally different approach by starting with a specific desired outcome and working backward to identify feasible pathways to achieve that goal. The Paris Climate Agreement’s target of limiting global warming to 1.5 to 2 degrees Celsius exemplifies this approach. Researchers using target-seeking scenarios examine which combinations of policies, technologies, and behavioral changes could feasibly achieve this temperature target, systematically ruling out pathways that fail to meet the specified goal. This backward-looking approach provides crucial information for policy-makers about the feasibility and requirements of achieving specific sustainability targets.
Policy screening scenarios occupy a particularly important role in applied sustainability science, as they directly evaluate the potential impacts of specific policy interventions. Researchers develop a menu of policy options, such as carbon taxes, payments for ecosystem services, or regulations banning certain activities like coal extraction, and then model the outcomes of each policy under various conditions. This approach enables systematic comparison of policy effectiveness across chosen metrics, providing evidence-based guidance for policy selection and design. The ability to test policies in a modeling environment before real-world implementation represents a crucial advantage in addressing complex sustainability challenges.
Retrospective scenarios serve a unique validation and learning function by examining policies that have already been implemented. These scenarios model what analysts expected would happen when a policy was proposed, compare those projections to actual observed outcomes, and analyze the gaps between prediction and reality. This retrospective analysis serves two critical functions: it enables model calibration to improve future predictions, and it provides insights into why certain policies succeeded or failed relative to expectations. The learning generated from retrospective scenarios feeds back into improving all other types of scenario analysis.
Integration of SSPs with Representative Concentration Pathways
Understanding RCP-SSP Combinations
The integration of SSPs with RCPs creates a matrix of possible future scenarios that combine socioeconomic development pathways with climate forcing outcomes. RCPs are expressed numerically as watts per square meter of radiative forcing, with values ranging from 1.9 (very low forcing and minimal climate change) to 8.5 (high forcing and severe climate change). While not precisely equivalent to temperature increases, these values provide a useful approximation of climate change severity.
Theoretically, any SSP can be combined with any RCP, creating a vast space of possible scenarios. However, certain combinations are more plausible than others. For instance, SSP1’s sustainable development narrative aligns naturally with lower RCP values, while SSP3’s regional rivalry and limited cooperation makes higher RCP values more likely. Despite this logical alignment, the framework allows for exploration of seemingly contradictory combinations, such as sustainable development with high emissions or fossil-fueled development with low emissions, though achieving these combinations would require specific conditions or policies.
Practical constraints limit the number of SSP-RCP combinations that are fully developed and modeled. Computational limitations, resource constraints, and the need for tractable analysis lead to the selection of marker scenarios that represent specific SSP-RCP pairs. These marker scenarios, designated by combining the SSP and RCP numbers (e.g., SSP126 for SSP1 with RCP2.6, or SSP585 for SSP5 with RCP8.5), receive the most detailed analysis and are most commonly used in research and policy applications.
Implications for Energy and Land Use Projections
The SSP-RCP combinations generate specific projections for critical sustainability outcomes, particularly in energy production and land use patterns. Energy projections vary dramatically across scenarios, with different trajectories for the mix of oil and gas, coal, renewables, and nuclear power. The primary energy triangle visualization effectively captures these transitions, showing how energy sources shift over time under different scenarios. Historical patterns show transitions from biomass to coal during industrialization, then diversification to include oil and gas, with future scenarios diverging based on whether renewable energy expands rapidly (as in SSP1) or fossil fuels continue to dominate (as in SSP5).
Land use projections similarly vary across SSP-RCP combinations, reflecting different assumptions about agricultural intensification, dietary transitions, urbanization patterns, and forest conservation. For instance, scenarios with high crop yield improvements may require less agricultural land, potentially allowing for reforestation or habitat restoration. Conversely, scenarios with growing meat consumption and limited yield improvements may drive agricultural expansion into natural ecosystems. These land use trajectories have profound implications for biodiversity, ecosystem services, and the carbon cycle.
Integrated Assessment Models and the SSPs
Model Architecture and Capabilities
The SSP projections are generated using sophisticated Integrated Assessment Models (IAMs) that combine representations of human and natural systems. The primary models used with the SSPs include AIM (Asia-Pacific Integrated Model), GCAM (Global Change Assessment Model), IMAGE (Integrated Model to Assess the Global Environment), MESSAGE-GLOBIOM, and REMIND-MAgPIE. Each model has unique strengths and approaches, but all solve complex optimization problems to project future development pathways that meet specified objectives while respecting various constraints.
These models typically employ partial equilibrium approaches, meaning they represent certain sectors in great detail while treating others as exogenous or simplified. This contrasts with general equilibrium models that attempt to capture all economic interactions simultaneously. The partial equilibrium approach allows for much greater detail in key sectors like agriculture, energy, and land use, but may miss important feedback effects between sectors. For instance, a model might detail agricultural production and land use decisions while treating industrial development as an external input rather than an outcome influenced by agricultural patterns.
The models solve cost-minimization problems subject to various constraints, seeking efficient pathways to meet demands for food, energy, and other goods and services while respecting resource limitations and policy constraints. This optimization framework assumes rational economic actors and efficient markets, assumptions that may not hold in reality but provide useful benchmarks for analysis. The models can incorporate various policy instruments, technological options, and behavioral changes, allowing for extensive exploration of intervention strategies.
Limitations and Considerations
A critical limitation of partial equilibrium models lies in their treatment of key constraints and assumptions. These models often assume that certain fundamental challenges are resolved, then explore what additional actions are needed. For example, models might assume that agricultural yields will be sufficient to feed the global population, treating yield as an exogenous input rather than an endogenous outcome influenced by climate change, soil degradation, and other factors. If yields fall short of assumptions, the models may not capture resulting food security crises or land use pressures.
The assumption of economic optimization and rational behavior may not reflect real-world decision-making, which is influenced by cultural values, political constraints, institutional failures, and behavioral biases. Models may underestimate transaction costs, implementation challenges, and resistance to change that characterize real-world transitions. These limitations don’t invalidate the models but suggest caution in interpreting results and the need for complementary analyses that address these gaps.
Despite limitations, IAMs paired with SSPs provide invaluable insights into possible futures and the implications of different development pathways. They offer internally consistent, quantitative projections that can inform policy debates and identify key leverage points for intervention. The models help identify trade-offs between objectives, unexpected consequences of policies, and conditions necessary for achieving sustainability goals.
The Land Use Harmonization Project
Connecting Historical Data with Future Projections
The Land Use Harmonization Project (LUH2), described in Hurtt et al. 2020, represents a monumental effort to create consistent, spatially explicit land use datasets spanning from historical reconstructions to future projections. The project addresses a critical need in Earth system modeling for land use data that maintains consistency across the historical-future transition, enables model intercomparison, and provides the spatial and temporal resolution required for contemporary modeling applications.
The historical component of LUH2 extends back to 10,000 BCE, reconstructing land use transitions based on archaeological evidence, historical records, and modeling of agricultural expansion and intensification. This deep historical perspective provides context for understanding current land use patterns and rates of change, while also serving as validation data for models that must reproduce historical transitions before projecting futures. The historical reconstruction reveals the acceleration of land use change with industrialization and population growth, providing baselines for assessing future scenarios.
The future component incorporates the SSP-RCP scenarios, translating the aggregate land use projections from IAMs into spatially explicit, grid-cell level changes. This translation involves sophisticated downscaling algorithms that respect local constraints, historical patterns, and biophysical suitability while achieving aggregate targets. The resulting datasets provide annual, half-degree resolution global land use maps for multiple land use categories including cropland, pasture, forest, and urban areas.
Data Structure and Applications
The LUH2 dataset employs a state-and-transition framework that tracks both land use states (the amount of each land use type) and transitions between states (the annual flows between land use categories). This framework captures the dynamics of land use change, including gross transitions that may be masked in net change analyses. For instance, simultaneous forest loss in one region and forest gain in another might result in little net forest change, but the gross transitions have important implications for carbon cycling, biodiversity, and ecosystem services.
The dataset uses the NetCDF format, a standard in climate and Earth system modeling that enables efficient storage and access of multidimensional data. Each file contains multiple variables (different land use types), multiple time steps (annual data), and global coverage at half-degree resolution. This structure allows users to extract specific variables, time periods, or regions as needed for their applications. The standardized format ensures compatibility with major Earth system models and analysis tools.
Applications of LUH2 data span numerous research domains and practical applications. Climate modelers use the data to account for land use impacts on carbon cycling, surface energy balance, and atmospheric chemistry. Biodiversity researchers apply the data to assess habitat loss, fragmentation, and species exposure to land use change. Agricultural scientists use the projections to evaluate food security, water resources, and sustainable intensification strategies. Policy makers reference the scenarios to understand implications of different development pathways and identify intervention opportunities.
Practical Implementation with QGIS
Working with Land Use/Land Cover Data
The practical component of the course introduces students to working with actual land use data using QGIS, a free and open-source geographic information system that has become increasingly popular in academic and professional settings. QGIS provides capabilities comparable to commercial alternatives like ArcGIS while offering the advantages of open-source software including free access, community support, and extensibility. The software handles various data formats common in land use analysis, including raster data (like TIFF files) and multidimensional datasets (like NetCDF files).
The LULC_CurrentWillamette.tiff file serves as an introduction to raster land use data. This file contains a two-dimensional matrix where each cell holds a numerical value representing a specific land use class. The classification system assigns unique numbers to different land use types: 90 for woody wetlands, 82 for cultivated crops, and so forth. This numerical encoding allows for efficient storage and processing while maintaining categorical information about land use types. The 30-meter resolution of this dataset provides sufficient detail for landscape-level analysis while remaining computationally manageable.
Loading and visualizing this data in QGIS demonstrates fundamental GIS operations that students will use throughout the course. The drag-and-drop interface simplifies data loading, while the symbology tools enable customization of visualization to highlight patterns of interest. Converting from the default grayscale display to a pseudocolor representation using appropriate color ramps makes land use patterns immediately visible and interpretable. The ability to zoom, pan, and query individual pixels provides interactive exploration of the data, helping students develop intuition about land use patterns and spatial relationships.
Exploring Temporal Dynamics with LUH2 Data
The LUH2 NetCDF files introduce additional complexity through their temporal dimension, containing annual land use data spanning historical reconstructions and future projections. This multidimensional structure requires more sophisticated handling than simple raster files, as users must specify which variables and time steps to visualize. QGIS handles this complexity through a selection interface that allows users to choose specific variables (such as cropland, rangeland, or forest) and specific years from the time series.
The rangeland variable provides an illustrative example of the data’s content and utility. Each pixel contains the proportion of that grid cell covered by rangeland, ranging from 0 (no rangeland) to 1 (complete rangeland coverage). This proportional representation acknowledges that half-degree grid cells (approximately 55 km at the equator) contain mixtures of land use types rather than homogeneous coverage. Visualizing rangeland distribution reveals global patterns such as the extensive rangelands of central Asia, Australia, and parts of Africa, contrasting with the minimal rangeland in tropical forests or intensive agricultural regions like the U.S. Corn Belt.
The temporal dimension enables analysis of land use change over time, a critical capability for understanding trends and projecting futures. By selecting different time bands within the same dataset, students can observe how land use patterns evolve under different scenarios. For instance, comparing rangeland extent in 2020 versus 2100 under different SSPs reveals dramatically different trajectories depending on assumptions about agricultural intensification, dietary choices, and conservation policies. This temporal analysis provides concrete illustrations of how different development pathways lead to different environmental outcomes.
Synthesis and Future Applications
Connecting Concepts to Applications
The integration of conceptual frameworks, quantitative scenarios, and practical data analysis tools provides students with a comprehensive toolkit for engaging with land use and sustainability challenges. Understanding the IPBES scenario typology enables appropriate selection of analytical approaches for different questions. Familiarity with the SSPs and their underlying assumptions allows critical evaluation of projections and their applicability to specific contexts. Hands-on experience with actual datasets and GIS tools transforms abstract concepts into concrete analytical capabilities.
This foundation prepares students for increasingly sophisticated analyses in subsequent course modules. Future assignments will build on these basics to explore topics such as ecosystem service modeling, where land use scenarios drive changes in carbon storage, water provision, and biodiversity habitat. Students will learn to couple land use projections with process models to assess impacts on specific sustainability outcomes. The combination of scenario analysis and spatial modeling provides powerful capabilities for evaluating policy options and identifying sustainable development pathways.
Implications for Sustainability Science and Practice
The tools and concepts covered in this lecture have immediate relevance for contemporary sustainability challenges. The Paris Agreement’s global stocktake process relies on scenarios like the SSPs to assess collective progress toward climate goals. National and subnational governments use land use scenarios to develop territorial plans that balance development needs with environmental protection. Conservation organizations apply these approaches to identify priority areas for protection under different development scenarios.
The limitations and uncertainties inherent in scenario analysis also carry important lessons. No model or scenario can perfectly predict the future, and all projections embed assumptions that may prove incorrect. However, the value of scenarios lies not in prediction but in systematic exploration of possibilities, identification of robust strategies that perform well across multiple futures, and recognition of key uncertainties that require monitoring and adaptive management. By engaging with both the power and limitations of these tools, students develop the critical thinking necessary for effective sustainability practice.
The increasing availability of open-source tools like QGIS and freely accessible datasets like LUH2 democratizes access to sophisticated analytical capabilities that were once restricted to well-resourced institutions. This democratization enables broader participation in sustainability analysis and decision-making, potentially leading to more inclusive and locally relevant solutions. As students develop proficiency with these tools, they join a global community of researchers and practitioners working toward sustainable futures, equipped with shared frameworks, data, and analytical approaches that enable collaboration across disciplines and borders.
Transcript
All right, everyone, let’s get started. Welcome to Lecture 4: Scenarios for the Future and Land Use Change Modeling.
I want to quickly go over the agenda and take stock of where we are. Today, we’re going to talk about scenarios, building on the framework from last class, but this time we’ll discuss them much more specifically—not just scenarios in general, but specific datasets, particularly for land use and land cover. I assigned two key readings: Pop et al. 2017 and Hurtt et al. 2020. I’ll also mention another paper, Riahi 2017, which I didn’t require but is relevant. These will catch you up to the state of the art on scenarios and land use within those scenarios.
After discussing these topics in general, we’ll get hands-on with QGIS and some of the LUH2 data. Hopefully, everyone saw my message to have QGIS installed on your computer. Any issues with that? Good, excellent.
That’s our plan for today. I also want to look ahead at the calendar and review where we’ve come from. We started with an introduction to earth economy modeling, looked at its origins in national sustainability models and general equilibrium approaches, and today we’re focusing on the scenarios that go into these models.
Looking ahead, we’ll shift gears a bit for Tuesday’s class and take a hands-on approach to understanding integrated assessment models for climate change. Who has heard of William Nordhaus or the DICE model? It’s a pretty famous model, as I mentioned before. It’s been used by climate skeptics to argue that climate change won’t be that bad, but it’s still a foundational model. We’ll also look at an update called the Green DICE Model from Frances Moore. I’ll send instructions for installing that on your computer. Starting today, and especially on Tuesday, we’ll be more hands-on and really embrace the applied nature of this course. We want to not just learn about these concepts, but use the tools.
From there, we’ll have a lecture on inclusive wealth, and then move on to another hands-on session with ecosystem services. That’s the roadmap for the next few classes.
Before diving into the lecture, let’s start your data downloading so it’s ready for the hands-on part at the end of class. But first, I need to explain a bit about the context and how we’ll organize our files.
We’ll return to this in a moment, but please check out the Google Drive link I shared with you. I’ve been writing it as “ya’ll,” but I was recently informed by a classmate from the South that the correct spelling is “y’all.” Apparently, that’s important. I like it because it’s a gender-inclusive word and sounds friendly. Anyway, I shared a Google Drive link with you all. Did everyone get that? Excellent. That’s the link to the NATCAP team’s base data—many terabytes of data, though not the full 50 terabytes.
Ideally, you’d use Google Drive Desktop Sync or something similar to get the data on your computer, but I don’t require you to install that. That’s how I do it, and it’s much faster than downloading individual files. The backup method, which we’ll use today, is to manually download the files if you don’t have the sync app. Everything I say today can also be adapted to work with Google Drive Sync, which is still the preferred method.
Let’s take a look at what I sent you. Did everyone get the link? Here’s the critical part, so I’m going to share it with you right now. Did everyone else get it? Okay, I must have typed in a wrong name or something. This won’t go on the recording video. Is that the email you prefer? Let me send it again. Yes, I sent it to your UMN address. Maybe I hit the wrong one. Anyway, that’s the link we’ll use throughout, so it’s good to get that right.
When I shared that with you, I shared something in our shared drive. You don’t have access to the whole drive, but in a folder called Files, we have this base data, and that’s where we’ll be using all the data. In the screenshot, I’m pointing to a specific subset of the base data: the invest sample data, and specifically the carbon model. Let’s take a look at this live, so I’m going to unshare my screen for a moment.
That’s the web version, but what we’re going to do is download the files in such a way that they maintain the exact file structure. I want to make a quick side note: I spend a lot of time programming, and the majority of coding bugs you’ll run into are almost always just pointing to a file that doesn’t exist. I don’t have a source for that statistic, but that’s how it feels to me. Having taught many courses, almost all bugs you get until you reach an advanced stage in programming boil down to downloading a file, putting it in the wrong place, and then telling your program to look for it in the wrong location. To avoid this, we’re all going to use the same file organization, so the sample code I provide will work for everyone.
To follow best practices among programmers, we’re going to use your user directory. If you already do this, great. If not, you might have to get used to it. We’ll organize things in a way that’s the same for all of us. In particular, let me show you: I have a shortcut here, but on Windows, the C drive has a Users directory, and then your username (for example, J.A. Johns). This is important because it allows your code to work across Windows, Mac, and Linux. Many people store everything in the user directory, but I don’t like that because it gets messy with configuration files from different programs. For convenience, I have us create a folder called Files, and that’s where we’ll keep everything we control.
You can see my full organization here. Create a directory called BaseData. From this directory, we’re going to mimic the structure from all the other files. You can see in the shared drive, the path is NetCap Teams > Files > Base Data. All the folders after base data, we’re going to replicate in our user directory. Whenever I say to download a file, like the land use land cover map we’ll get to in a minute, you’ll replicate that folder structure. You wouldn’t just download it into base data; you’d create a folder called Invest Sample Data, then a folder called Carbon, and then download the file into that.
If you have the Google Drive Sync application, it will do that automatically for you. But we’re not just going to download random files; we’re going to keep everything in this directory structure.
Because I’d like to start your downloads now, go ahead and do that with the LULC_CurrentWillamette.tiff file. That’s in the Invest Sample Data > Carbon folder. Is everyone with me? This is one of the most boring points of class, but we have to get it right so we don’t spend a lot of time later not finding files.
Just to show it from the Google Drive interface: the internet here isn’t doing well today, probably because we’re all downloading files at the same time and overwhelming the network. Has anyone besides you (since you don’t have your computer) not gotten the files downloaded, or is there anything about the organization that doesn’t make sense? I shared this base data folder with you. In there, go to Invest Sample Data, then the Carbon folder. I’d like you to download the LULC_CurrentWillamette.tiff file. There are lots of other files that look similar, so make sure you get the exact right one. We don’t want the .tiff.ox.xml file; we want the one that is just .tiff. If you’re using the web interface, you can download it, but it’s tempting to put it in the wrong place. When the download screen pops up on Windows, don’t save it in the Downloads folder. If you do that by mistake, move it afterward. Instead, navigate to your user directory, then Files, then Base Data, and create all the necessary folders if you haven’t already.
We’re going to copy that structure: Invest Sample Data > Carbon. I’ve downloaded everything because I sync the whole drive, but this is where you should have downloaded the LULC_CurrentWillamette.tiff file.
One other note: this course will be increasingly hands-on, with you doing things on your computers. Please, when I ask a question like “Does everybody have it?” that’s not hypothetical—I actually want to see heads nod or thumbs up. It’s hard for me to keep everyone on the same page, so please respond actively if you’re up to speed, or raise your hand if you’re not. It’s much easier to stop and fix a question than to have everyone get out of sync. So, does everyone have the LULC_CurrentWillamette.tiff downloaded in the right location?
It should be in your own user path: user > Files > Base Data > Invest Sample Data > Carbon. Any questions about that? Okay, back at it. That’s the first file I wanted you to download. That one should have been relatively quick, since it’s not a huge file. I also want you to download another file, which is much larger. This is the land use harmonization data that we’ll use later today. For that, go into the Google Drive interface on the web, navigate to base data again, but this time, instead of “Invest Sample Data,” go to “LUH2.”
From there, go to “raw data.” Let me show you on my screen. Let’s go up a few levels—here’s the base data, LUH2, and then raw data. We’re going to learn what all these numbers mean in a moment. Pick one of the files at random, besides the historical one. For example, I’m going with RCP4.5SSP2, but you can choose a different one if you want some variety. In this folder, you’ll find a few files. The file names are long, so make sure you can see the end of the file name to get the right one.
I want you to download the file that starts with “multiple states” and ends with “.nc”—not “.NCOUX.” Download this file, but again, make sure you keep the same folder structure as before.
This means that in your file structure, under your user directory, go to Files > Base Data, and create a folder called LUH2, then inside that, “raw data,” and then have a folder name that exactly matches the one you chose. I know the screen is a little small, but try to match the folder names exactly.
You only need to create the folders that contain the file I’m asking you to download. The reason this is important is that syncing base data sets can be a nightmare, especially since most of you probably don’t have 10 terabytes of storage on your hard drive. If any of you do, I’d be very impressed. I pay a lot just to have 4TB on my machine, and I’ve never heard of anyone having 10TB locally.
So, if you’re downloading this file, just create the folders that contain it. For example, I have base data > LUH2 > raw data, and I chose RCP26_SSP1. Into that folder, I pasted this large file. It’s about 612 megabytes, so it’s large enough to potentially overwhelm our Wi-Fi for a bit.
Is anyone having trouble getting that download started? If so, please wave your hand at me, because I’m going to assume these files are on your computer after this.
Great. While that’s downloading, let’s jump back to the slides and talk about what we’ve just started to download.
Fantastic. Thank you, Geronimo. Sorry, everyone. Sorry, I’m just reading the long series of texts. I did feel my phone vibrate, but I don’t typically answer my phone when I’m lecturing. I’m going to assume you can see the screen now—is that true? Nobody’s answering. Let’s move on for the people in person. If we don’t hear from you, you’ll be voted off the island. Yes, thank you, Haku! Now let’s go forward.
Hopefully, everyone online was able to get to the point where you’re downloading the base data. Let’s jump back into the discussion of scenarios and the data we’re using.
Last class, we talked about scenarios in general, and now we’re going to talk about very specific scenarios and data. The context for this originates largely from an international body called IPBES, the Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services. This organization is dedicated to summarizing the current state of our understanding of biodiversity and ecosystem services. They combine, fund, and support science in this area, addressing the fact that biodiversity has often been ignored compared to issues like climate change. You may have heard of the IPCC, the Intergovernmental Panel on Climate Change. IPBES is essentially the equivalent for biodiversity.
IPBES became well-known through a publication by Sandra Diaz et al. in 2015, which presented their conceptual framework. We won’t go into the details of this framework, as it’s quite theoretical, but one of the most practical things they did was to identify a useful typology for scenarios. I want to talk through these.
The four scenario types in the IPBES conceptual framework are: exploratory, target-seeking, policy screening, and retrospective.
First, exploratory scenarios are what we usually think of when we talk about scenarios. We’re at a point in time, we’ve seen what’s happened in the past, and we want to explore the space of possible futures. For example, if nature’s benefits have been declining, we want to know how we might reverse that trend. These scenarios explore possibilities without being specific about how to achieve them.
Second, target-seeking scenarios are about achieving a specific goal, such as the Paris Climate Agreement’s target of limiting warming to 1.5 to 2 degrees. Here, we look at what pathways could feasibly get us to that target and rule out those that don’t.
Third, policy screening scenarios are perhaps the most common in sustainability science. We might have a menu of different policy options—like a carbon tax, payments for ecosystem services, or a law banning coal—and we try out these scenarios to see how they perform. This helps us identify which policies are most effective on our chosen metrics.
The last type is retrospective scenarios. These are useful for assessing how a policy that was actually implemented performed. We model what we thought would happen, compare it to what actually did happen, and analyze the gap. This is also useful for calibrating our models to reality.
In this context, IPBES and other communities, in coordination with the IPCC, decided to develop specific versions of these scenarios, which led to the creation of the SSPs—Shared Socioeconomic Pathways. The SSPs come from a major paper by Riahi et al. (2017), which involved multiple teams and stakeholders to define a comprehensive set of scenarios for assessing biodiversity and nature in the future.
The SSPs were designed to provide the same benefits for broader sustainability questions as the RCPs—Representative Concentration Pathways—did for climate. RCPs are specific to climate change and describe trajectories of greenhouse gas concentrations, from which we can derive climate change impacts. The SSPs, in contrast, are broader and include factors like population and productivity, making them more complex.
Developing the SSPs involved stakeholder engagement to define different storylines or narratives about the future. These storylines are then translated into specific drivers, such as population, urbanization, and productivity. Policies are also considered, allowing for policy screening within the scenarios. The result is a comprehensive set of country-level projections for inputs like GDP and population, as well as outputs like energy supply, demand, and land use change.
These scenarios are intended to help communities plan for sustainability in a complex world. The SSPs define not just climate variables, but also socioeconomic challenges related to climate change.
To visualize the SSPs, imagine a space with two axes: mitigation challenges (vertical) and adaptation challenges (horizontal). Mitigation refers to preventing climate change, such as through reducing emissions or investing in renewable energy. Low mitigation challenges mean it’s easy to reduce emissions, while high challenges mean it’s difficult.
Adaptation refers to dealing with the impacts of climate change, such as through air conditioning or health infrastructure. Low adaptation challenges mean it’s easy and affordable to adapt, while high challenges mean it’s difficult, often due to factors like income inequality.
In this space, the “good” scenarios are those with low challenges for both mitigation and adaptation. For example, SSP1 represents low challenges on both axes. SSP3, on the other hand, represents high challenges for both mitigation and adaptation—the worst-case scenario. The off-diagonal scenarios are also interesting: for example, high mitigation challenges but low adaptation challenges might represent a world where climate change is severe, but societies are resilient and able to cope, perhaps through widespread use of air conditioning. The opposite—low mitigation challenges but high adaptation challenges—might occur if we can reduce emissions but still face significant inequality or lack of infrastructure.
SSP2 is the middle scenario, often called the baseline or business-as-usual, and refers back to what we discussed last lecture. Any questions on the overall framing of that? Excellent. Unlike what we discussed in the previous lecture, where we talked about general scenarios, these are very specific. This table gives you a sense of the differences: for each scenario, there are different assumptions about the percentage change in cropland, maximum population, the percentage of kilocalories from beef and other sources, GDP, fertilizer, irrigation, and many other factors.
One of the most confusing aspects of the SSPs is how they combine with the RCPs. As I mentioned, RCPs are Representative Concentration Pathways. These are always expressed with a number, which refers to the watts per meter squared of radiative forcing. For example, if 1.9 watts per meter squared are hitting the Earth, that will result in very little climate change. If 8.5 watts per meter squared are hitting, that will result in severe climate change. You can almost think of these as the number of degrees Celsius hotter it will be, though that’s not exactly what they define, but it’s a useful approximation.
It’s possible to pair any SSP with any RCP. In theory, you could have a “good” SSP but still have high emissions, though in practice that combination is unlikely. In reality, there are a set of marker scenarios that are specific combinations of an SSP and an RCP. For example, SSP1 paired with RCP2.6 is referred to as SSP126. Similarly, SSP5 with RCP8.5 is SSP585. While you could do all possible combinations, for computational reasons, only a subset are typically used. This is a limitation that can introduce biases in analyses, but it’s the current practice.
With these combinations in place, we can start to make projections. For each scenario, we define assumptions about population, education, GDP, and other factors. These assumptions are the foundation for the different scenarios.
From there, we can calculate projections for critical outcomes, such as energy production and the mix of energy sources. I want to highlight a figure called the primary energy triangle, which shows the different sources of energy: oil and gas, coal, renewables, and nuclear. For example, in 1858, most energy came from burning wood, which is considered renewable. With the Industrial Revolution, coal became dominant, and later, oil and gas. By 2010, the energy mix had shifted again. The different SSPs have different trajectories for the mix of energy sources. SSP1, for example, moves toward renewables.
You might wonder how the input parameters, like population and GDP, are chosen. The Riahi et al. and Pop et al. papers discuss this in detail. Each SSP has a descriptive name: SSP1 is “Sustainability, Taking the Green Road,” SSP2 is “Middle of the Road,” SSP3 is “A Rocky Road,” SSP4 is “A Road Divided” (inequality), and SSP5 is “Fossil-fueled Development.” For example, SSP4 represents a world where some countries cut themselves off from international collaboration, leading to greater inequality. SSP5 is interesting because, while it involves high fossil fuel use, it also assumes successful economic growth and adaptation, creating a kind of techno-utopia where climate change occurs but societies adapt, such as in wealthy cities with extensive air conditioning.
Game theory can be relevant here, especially when countries are competing and facing the free rider dilemma. For example, if one country doesn’t invest in climate change mitigation because it expects others not to, this can lead to suboptimal outcomes.
The Pop et al. paper, which you read, dives deeper into land use in the different SSPs. Just as you can input SSP assumptions into an energy model to project future energy mixes, you can do the same with land use models. For example, if crop yields increase significantly, less land may be needed for agriculture. The SSPs enable nuanced, quantitative exploration of a wide range of potential futures.
The SSP database provides detailed, downloadable information on all these metrics, and we’ll use it in future assignments.
All of these projections are generated by models. In the first paper of the semester, we saw a plot of earth economy models, categorized by economic and spatial detail. The main integrated assessment models (IAMs) used with the SSPs are AIM, GCAM, IMAGE, MESSAGE-GLOBIOM, and REMIND-MAgPIE. These are complex models that solve cost-minimization problems to achieve objectives like food production and minimizing environmental impact, given certain assumptions.
Most of these models are partial equilibrium models, meaning they represent some sectors in detail while holding others fixed. This contrasts with general equilibrium models, which allow all sectors to interact. For example, the DICE model by Nordhaus is a general equilibrium model focused on climate, while the IAMs paired with the SSPs are partial equilibrium and focus on sectors like agriculture and energy.
A key limitation of partial equilibrium models is that they may assume away critical constraints. For example, they might assume agricultural yields will be sufficient to feed the world, rather than modeling yield as an endogenous outcome. If yields are lower than expected, the models may not capture the resulting food shortages. Thus, these models often assume the most important challenges are solved, and then ask what should be done next.
For the last part of class, we’ll focus on key outputs from these models related to land use and land cover scenarios, as reported in Pop et al. Their task was to take the SSPs and project what will happen to land use, using the best comprehensive IAMs. They produced projections of land use activities, their effects on the climate system, and how to generate spatially and temporally detailed projections consistent with historical data.
A major contribution of their work is making the data usable by Earth system models. For example, they report changes in crop demand, cropland and pasture, and forest cover under different scenarios. These outputs were combined into the Land Use Harmonization Project, as described in Hurtt et al. (2020). This project connects historical land use data (going back to 10,000 BCE) with future projections under the SSPs and RCPs, creating consistent time series of land use transitions.
Now, let’s dive into the data. Hopefully your downloads have finished. We’ll look at two maps: a specific land use/land cover map, and the LUH2 time series.
First, you downloaded a TIFF file. A TIFF is just a two-dimensional matrix of numbers, where each number represents a land use class (e.g., 90 for woody wetlands, 82 for cultivated crops). This is what you found in the invest sample data, in the carbon directory: the WillametteCurrentLULC.tiff file.
Let’s load this into QGIS. QGIS is free and open source, and has become a superior alternative to ArcGIS for many users. To add data, the easiest way is to drag and drop the file into QGIS. Make sure you select the correct file extension. Once loaded, you can view the legend and see the different land use classes.
If your map appears in black and white, you can change the display settings. Double-click the layer, and in the symbology tab, select “single-band pseudocolor” to apply a color ramp. This will color the map according to the land use classes. The legend will update accordingly. Each pixel represents 30 meters, so you can estimate distances and areas.
This is just an introduction; we’ll use this data more in future assignments.
Next, let’s load the LUH2 dataset. This is a NetCDF file, which is a time series of rasters (essentially, a stack of TIFFs over time). Drag the NetCDF file into QGIS. It won’t display immediately because NetCDFs can have multiple dimensions. In this case, there are layers for different variables and years.
Select the variable “rangeland” and add the layer. If you don’t see anything, right-click and choose “zoom to layer” to adjust the view. This map shows the proportion of each grid cell covered by rangeland, as calculated by Pop et al. and the Land Use Harmonization Project.
To improve the display, double-click the layer, go to the symbology tab, and select “single-band pseudocolor.” You can choose which year to display by selecting the appropriate band (e.g., time equals 11 corresponds to 2026). Apply a color ramp to visualize the data.
This map shows the distribution of rangeland globally. For example, in some regions, nearly all land is rangeland, while in others, such as the U.S. crop belt, there is little rangeland.
We’ll pause here, as we’re out of time. Assignment 2 will have you explore these datasets further, looking at changes over time. Are there any questions? Did everyone get both datasets loaded into QGIS?
Excellent. You are now QGIS specialists and will use these skills throughout the course.
Cool. Alright, we’re done for now. Thanks, everybody.