The challenge of modelling telecoupled land-use change

Modelling cross-scalar dynamics, including telecoupled interactions, influencing land-use change and food systems continue to challenge land system scientists (Verburg et al. 2019, Millington et al. 2017). Müller et al. (2020) likewise identifies the need to develop appropriate techniques for representing and analysing relationships and feedbacks across scales in relation to modelling food security. In addition, Müller et al. (2020) call for deeper integration of modelling approaches, as well as better representation of different agents in such models. 

In this blogpost, we introduce two new models, the TeleABM and the FLUTE, that are designed to study telecoupled land-use change but using different modelling approaches. Both models use soybean, one of the most traded commodities in the world, as an exemplary case to illustrate the ways in which they tackle the challenges of cross-scalar dynamics (Figure 1). The blogpost reports on a joint webinar hosted by the GLP Working Group on Telecoupling Research held in September 2020 and featuring presentations by the three first authors (view the recording and download the presentations here).

Figure 1. Soy land use in Mato Grosso, Brazil (top) as large-scale monocrop system, and in Heilongjiang, China (bottom) as small-scale mixed-crop system (with corn). (Source: James Millington (top), Yue Dou (bottom))

Using an agent-based modeling approach, the TeleABM model simulates how individual farmers’ behavior in an importing country aggregates and results in land-use changes in a telecoupled production country, and vice versa (Dou et al., 2019; Dou et al. 2020). In the model, the relationship between the sending and receiving systems is represented by a demand-supply function, which allows flexibility to simulate land use changes between the two systems and potential other places as well, and with other types of trade flows by calibrating the elasticity parameters based on literature. In a sample scenario, Yue Dou and colleagues used the TeleABM to demonstrate the land-use feedback between one municipality in Brazil and one county in China linked by the soybean trade flow (Dou et al. 2020). In the scenario, the authors simulated a “high-tariff” applied on soybeans exported from Brazil to China, and the results indicated that not only does agricultural intensification (from single cropping to double cropping cultivation) in Brazil slow down but it also affected the land-use changes in China, where fewer soybean areas were converted to corn and rice paddy. In this way, the model enables the representation and simulation of the feedback from distant land-use actors and processes in current land-use modeling practise, as a tool for policy-makers and researchers to evaluate potential externalities of land-use policies.

The Food and Land Use TElecoupling model (FLUTE), in turn, has been designed to investigate the dynamics and consequences of telecoupled interactions in the global food system, specifically the production and trade of important agricultural commodities. The FLUTE combines a spatially explicit land-use/land-cover (LUCC) model with a systems dynamics model representing trade-flows between global regions (Figure 2). Like the TeleABM, FLUTE has been developed to examine the case of soybean production and trade between Brazil and China. In their work, James Millington and colleagues use FLUTE to represent LUCC in Brazil based on a modified version of the agent-based CRAFTY modelling framework they call CRAFTY-BRAZIL. This element of the model simulates production of soybean, maize and beef as a result of spatially explicit LUCC due to biophysical, economic and technological drivers. Building on another existing model, a System Dynamics (SD) modelling framework is used to represent trade as flows of commodities between global regions. The two models are coupled such that the cumulative production of the spatially-explicit representation of LUCC in CRAFTY is passed to the SD model at each timestep, with demand for production passed back in return. 

Figure 2. The structure of the FLUTE model (Source: James Millington). Arrows indicate flows if information.

Testing of CRAFTY-BRAZIL shows that the model is more sensitive to estimates of scenarios of agricultural yield (e.g. due to fertiliser inputs) than climate and that confidence in outputs is greater for temporal trends than spatial patterns (a paper describing structure and results is in revision for publication in Environmental Modelling and Software). The coupled FLUTE model is currently being tested, but first results show that the model produces outputs that allow comparison of land use and flows of commodities between global regions. These initial results are promising, and indicate how FLUTE can be used in various applications for advancing understanding of telecoupled interactions in land systems. One such application could be the analysis of land-use impacts in Brazil (e.g. deforestation) under alternative scenarios of possible responses to the 2019 swine flu outbreak in China.

Both the TeleABM and FLUTE represent new types of models that integrate modeling techniques from different disciplines to address the challenge of cross-scalar interactions, yet in different ways. While TeleABM uses the supply-demand relationship to represent the trade flow between two specific countries (Brazil and China), FLUTE’s structure facilitates comprehensive representation of trade between all production regions around the world. However, the finest spatial scale of representation in FLUTE is 25 sq km, due to its reliance on CRAFTY-Brazil, which prevents consideration of telecoupling effects over all global-to-local scales. In contrast, TeleABM operates at a fine 30m spatial resolution within the two representative counties but can be scaled up to a larger geographic extent. In this way, TeleABM does enable examination of telecoupling dynamics across the full range of local to global levels but with land change represented for only two counties, whereas FLUTE aims for broad spatial coverage (but currently only for a single country - Brazil). When incorporating agency in the models, FLUTE focuses on multiple local land management types while TeleABM has a variety of agents from the supply chain from land users to government officials. Thus, FLUTE assumes homogeneity of decision-making in global trade (i.e. ‘the market decides’) whereas TeleABM can represent negotiations and relationships between individual traders based on specific characteristics. These characteristics illustrate the tradeoffs inherent in the different approaches to simulating telecoupling.

As has been argued elsewhere (Millington et al. 2017), the primary influence on the selection of modelling approaches should be the questions a researcher has about telecoupling dynamics (e.g. across what range of scales do questions apply, what particular aspects of the telecoupling framework are emphasised). A secondary issue in choosing modelling approaches is the available expertise, data and other resources. Comparing and contrasting TeleABM and FLUTE demonstrates these points further. It is hoped that the two models will continue to serve as a tool and a starting point for people who are interested in understanding and modelling telecoupled land-use changes in various locations and systems, with different flows, and for a range of different questions. Challenges remain, but progress can be made by both careful integration of existing models and exploration of novel approaches.