Modelling future land use scenarios based on farmers’ intentions and a cellular automata approach

Highlights

• We coupled cellular automata-Markov chain and farmers’ intentions to simulate LUCC.

• We measured LUCC consequences under four explorative scenarios.

• Recommendations to decision-makers to monitor LUCC were provided.

Abstract

Different mechanisms drive land use and land cover changes (LUCC). This paper presents an exploratory analysis aimed at understanding the complex dynamics of LUCC based on farmers’ intentions when they are faced with four scenarios with the time horizon of 2025: (1) A0 – current social and economic trend; (2) A1 – intensified agricultural production; (3) A2 – reduced agricultural production; and (4) B0 - increasing demand for urban development. LUCC models are applied to a Torres Vedras (Portugal) case study. This territory is located in a peri-urban area near Lisbon dominated by forest and agricultural land, which has been suffering considerable urban pressure in the last decades. Farmers — major agents of agricultural land use change — were interviewed to obtain their LUCC intentions according to the scenarios studied. To model LUCC a Cellular automata-Markov chain approach was applied. Our results suggest that significant LUCC will occur depending on their intentions in the different scenarios. The highlights are: (1) the highest growth in permanently irrigated land in the A1 scenario; (2) the biggest drop in non-irrigated arable land, and the highest growth in forest in the A2 scenario; and (3) the greatest urban growth was recognized in the B0 scenario. To verify if the fitting simulations performed well, techniques to measure agreement and quantity-allocation disagreements were applied.These outcomes could provide decision-makers with the capacity to observe different possible futures in ‘what if’ scenarios, allowing them to anticipate future uncertainties, and consequently allowing them the possibility to choose the more desirable future.

Introduction

Land use and land cover is the result of a set of complex systems linked by the interaction of environmental, social, and human activities and economic factors (Turner et al., 2007; Robson and Berkes, 2011). Land use and land cover changes (LUCC) result from the interaction between these different factors (Geist and Lambin, 2002) in non-linear relationships, and they can take place globally or locally. In peri-urban areas, the main concern about land use/cover is related mostly to the conversion from agricultural land to urban development. The consequences can range from endangered food security (Abrantes et al., 2016; Gomes et al., 2018; Spilková and Vágner, 2016) to negative impacts on the economy (Hein et al., 2008), ecosystems (Seki et al., 2017), and climate variability (Li et al., 2009).

These areas are dynamic, and characterised by commutes to work, and heterogeneous activities (Lambin et al., 2003). The proximity to urban settlements and the urban pressure felt in these places present farmers with new challenges for the future. The literature focuses on three main challenges: (1) maintaining their farmland (Malan, 2015); (2) expanding their farmland (Deininger and Byerlee, 2011); and/or (3) selling their farmland for urban development (Curran-Cournane et al., 2016; Satterthwaite et al., 2010a, b). Analysing and understanding how farmers would change the territory are of great importance to anticipate the uncertainties of the future.

LUCC models have been developed since the 1950s and 1960s (Yu et al., 2011). They have been performed by a multidisciplinary assessment (Agarwal et al., 2002; Verburg et al., 2006) analysing the relationship between different types of behaviour to understand complex dynamics, and artificially recognise what can happen in the real world (Macal, 2016). LUCC methods have evolved to integrate a variety of methods, coupling artificial neural networks, cellular automata, agent-based models, or multiple regressions.

Several works have been published using different methods and applied to empirical case studies, increasingly stirring interest among policymakers. For instance, Agarwal et al. (2002) present an analysis of different types of models, Boavida-Portugal et al. (2016) assess the impacts of tourism on built-up areas, Lambin et al. (2003) explore LUCC in Tropical Regions, Morgado et al. (2014) analyse contested land use visions, and Gomes et al. (2019a) and Puertas et al. (2014) simulate urban growth in a metropolitan area context.

Cellular automata (CA) are one of the most widely used methods (Macal and North, 2010). It is a powerful method for studying complex systems and exploring principles of system evolution and self-organisation (Mitchell, 1998). CA became more common and popular when, in 1970, John Conway designed the Game of Life (Huang et al., 2009), the most popular 2-D binary CA. Following successive theoretical structure improvements, CA have become a well-established method for modelling LUCC in recent decades, and have been widely used since they were introduced by Tobler (1979). CA have the ability to simulate dynamic development from a bottom-up perspective (Liu et al., 2008a,b) based on complex spatial forms (e.g., Agarwal et al., 2002; de Almeida et al., 2003; Wang and Li, 2011). CA based on Markov chains are increasingly employed in LUCC (Dezhkam et al., 2017; Meneses et al., 2018; Sang et al., 2011), incorporating the relationships between land use and driving forces.

LUCC models are likely to become a significant tool for spatial planning (Herold et al., 2005). A better LUCC analysis can support better planning practices (Yirsaw et al., 2017), and identify the valuation of different land use options and socioeconomic settings in order to recognize desirable land uses (FAO, 1993). Land use planning can shape policies to promote regulatory land use implemented by decision-makers. These policies intend to control land use activities in the future, aiming to preserve open landscapes for agriculture and nature, and encourage sustainable development. But these processes are too rigid, particularly when applied to peri-urban areas where land conversion is very fast. They are one of the major policy challenges at the moment, and more studies are needed (Abrantes et al., 2016; Gomes et al., 2019a), namely engaging stakeholders to recognize their intentions in the LUCC process. Scenarios are a good way to identify future uncertainties in endogenous and exogenous developments (Rounsevell et al., 2006) to prepare for the needs of the future (Cork et al., 2000).

This paper proposes a LUCC model incorporating farmers’ intentions and assessing the impacts of their intentions in agricultural, forest, and urban land. We identified how farmers’ intentions may affect future land use whenever they are faced with four different scenarios, such as (a) A0 - current social and economic trend; (b) A1 - intensified agricultural production; (c) A2 - reduced agricultural production; and (d) B0 - increasing demand for urban development.

The methodology used in this study included interviews with farmers (to capture farmers’ LUCC intentions), and then a combination of geographic information systems (GIS), and CA – Markov Chain – developing step-by-step guidelines towards the creation of a simulation model to predict LUCC. The main contributions of this paper are: (a) to spatially analyse and model future LUCC and their impacts on the territory, and (b) to help understand how farmers’ decisions can affect the decline, preservation, or maintenance of agricultural land in peri-urban regions.