Dynamic interactions between vegetation and land use in semi-arid Morocco: Using a Markov process for modeling rangelands under climate change

https://doi.org/10.1016/j.agee.2011.01.011Get rights and content

Abstract

Integrated scientific assessments of semi-arid agroecosystems with mathematical models are challenging because of computational constraints. These constraints arise from exponentially increasing decision options due to dynamic interactions between the biophysical states of rangeland vegetation and farsighted decisions taken by pastoral stakeholders. This study applies a methodology that integrates these interactions in a computationally feasible manner. We equip a dynamic land use decision model with a detailed representation of biophysical processes by using a Markov chain meta-model of EPIC (Environmental Policy Impact Calculator). Using separate Markov chains for different weather scenarios, we investigate the economic and ecological impacts of droughts on rangeland management in southern Morocco. The drought simulations (2 years with 33% less precipitation) show a decrease in profits from pastoralism by up to 57%. Pastoral land use of the rangeland in our model increases surface runoff by 20%, doubles infiltration, and thus influences irrigation agriculture. The economic and ecological impacts of drought in our simulation go substantially beyond its meteorological time horizon.

Research highlights

► Coupling biophysical vegetation dynamics to farsighted land use decisions using Markov chains. ► Economic and ecological impacts of drought go substantially beyond its meteorological time horizon. ► Intense grazing of rangelands enhances water availability for down-stream irrigation agriculture.

Introduction

Pastoralism is the dominant land use in semi-arid and arid areas. These areas occupy 41% of the world's land surface and are inhabited by more than two billion people (Millennium Ecosystem Assessment, 2005). However, in several large-scale economic assessments of global change, semi-arid areas have been found to not play an important role because the overall impacts of climate change are accounted for mostly in terms of the percentage of global gross domestic product (GDP) (e.g. Tol, 2009), and drylands have the lowest GDP per capita (UNCCD, 2007). Hence, the socio-economic effects of climate change in these areas are at present of no great influence in large-scale economic models and the resolution of system properties is low. Furthermore, grazing is not generally considered as part of dynamic global vegetation models (Diaz et al., 2007). Nevertheless, especially in developing countries, pastoralism is a major source of income for large parts of the population (Gertel and Breuer, 2007). At the same time, the social impact of climate change for people living in semi-arid areas has the potential to be quite substantial, since 90% of the affected areas are located in developing countries (Millennium Ecosystem Assessment, 2005) and drylands have the highest infant mortality rates compared to other land use types (UNCCD, 2007). The conflict in the Sudanese Dafur, which can be traced back in part to changes in a pastoral agroecosystem, exemplifies the impacts climate change can have on society (Prunier, 2005). Hence, investigating the effects of climate change in these areas is of great importance.

In order to adequately assess the influence of climate change on large-scale agroecosystems and society, mathematical models can be used. However, these often require very high levels of computational effort, in particular for the integration of vegetation dynamics in combination with decision-making. If human decision-making is farsighted, the number of possible land management plans and related vegetation states can quickly lead to the so-called “curse of dimensionality” (Bellman, 1961), where the computational effort is exponentially related to the number of considered time periods. To overcome this, large-scale land use models use either a static representation of biophysical properties, such as biomass growth, or myopic decision-making, such as prescribed scenarios or exclusion of inter-temporal planning (Lambin et al., 2000, Schaldach and Priess, 2008).

The aim of this study is to quantify the implications of droughts in a medium-scale Moroccan pastoral agroecosystem, and we approach this by estimating the changes in profits from pastoral activities of rural households. In addition, we also assess the relationships between land use intensity and local hydrological and biophysical properties, including the infiltration of water into the groundwater, surface runoff, evapotranspiration (ETP), and albedo. These biophysical parameters might have further implications for land use decisions since, for example, altered hydrological properties can affect downstream oasis agriculture.

To address these aims, we develop an augmented mathematical land use decision model (LDM) that combines a dynamic representation of vegetation with farsighted, profit-oriented decision-making. In this way, it is hoped that our research will help bridge the scientific gap between those existing models that address either the detailed representation of farsighted decision-making on the one hand, or concentrate on describing accurately the biophysical vegetation dynamics on the other. We use a Markov chain to integrate into our LDM the results of an elaborated biophysical soil-vegetation model, as well as parts of its dynamic properties. The applied method is suitable for use at large scales, i.e. for a more adequate representation of dryland agroecosystems in global LDMs, such as GLOBIOM (Havlík et al., in press).

Section snippets

Study site and setup

The study site is located in the Moroccan province of Ouarzazate, on the southern slopes of the High Atlas mountain range (Fig. 1). The region is characterized by a semi-arid to arid climate and a strong precipitation gradient (200 mm to more than 700 mm per year), which exists because of a similarly steep altitudinal gradient (Schulz and Judex, 2008). Climate projections for this region differ greatly and indicate large uncertainty in the direction of precipitation development (Sillmann and

Simulation results

To investigate the effects of drought in our agroecosystem of interest, we simulate 2 years with 33% less precipitation. The overall time horizon is 10 years, with years 4 and 5 experiencing the drought. The remainder years are simulated with average weather. We use an ensemble of six model runs to separate the effects of droughts from the effects of the model's initial state. The 90% confidence intervals of the individual runs are shown by error bars in Fig. 10, Fig. 11, Fig. 12.

The augmented

Discussion

The Markov-chain-based integration of EPIC with an economic LDM has succeeded in providing a more accurate picture of the complex land use system dynamics than the two modeling components alone. However, the reduction of EPIC to a single-state-variable-based Markov chain is not cost-free. We assume that the entire land management history of a certain site is fully captured by the value of a single state variable, namely AGPM. In reality, the grazing and precipitation history affects many other

Conclusion

The methodology presented in this paper allows a computationally feasible integration of a complex biophysical model into an economic LDM. The dynamic interactions between land management and vegetation in semi-arid areas can be more accurately depicted. The Markov-chain-based approximation of the biophysical impacts reduces drastically the computational requirements compared to a direct coupling with explicit land use management trajectories. By using separate Markov chains for different

Acknowledgements

The authors wish to thank the BIOTA Maroc project team, whose expertise, data, and infrastructure greatly benefitted this study (German Federal Ministry of Education and Research, Förderkennzeichen 01 LC 0601A). Furthermore, thanks also to the IMPETUS research project team (Universities Cologne and Bonn), who kindly permitted us access to their invaluable empirical datasets. Particular thanks to Claudia Heidecke, Anna Klose and Oliver Schulz, as well as Zakia Akasbi for her helpful

References (49)

  • Baumann, G., 2009. How to assess rangeland condition in a semiarid ecosystem? Dissertation, University of Cologne,...
  • Barbier, B., Bergeron, G., 2001. Natural Resource Management in the Hillsides of Honduras – Bioeconomic Modeling at the...
  • R. Bellman

    Adaptive Control Processes: A Guided Tour

    (1961)
  • BIOTA, 2010. http://www.biota-africa.org/...
  • K. Born et al.

    Shifting climate zones in the Northwestern Maghreb

  • Boudiab, A., 1981. Contribution à l’étude du système de production animale sur les parcours dans la région de...
  • I. Breuer

    Marketing from the margins: the Ilmichan pastoralists of the Pre-Saharra

  • S. Bryl

    Untersuchungen ausgewählter primärer Pflanzeninhaltsstoffe von weiderelevanten Pflanzenarten in südmarokkanischen Trockensteppen. Staatsexamensarbeit

    (2009)
  • D.K. Davis

    Neoliberalism, environmentalism, and agricultural restructuring in Morocco

    Geographical Journal

    (2006)
  • S. Diaz et al.

    Plant trait responses to grazing—a global synthesis

    Global Change Biology

    (2007)
  • M. Finckh et al.

    Monitoring biodiversity on the Saharan slopes of the High Atlas, Morocco

    Mountain Forum Bulletin

    (2007)
  • J. Gertel et al.

    Pastoral morocco – introduction

  • F. Guessous et al.

    Sheep in Morocco

    FAO Animal Production and Health Paper

    (1989)
  • Havlík, P., Schneider U.A., Schmid E., Böttcher H., Fritz S., Skalský R., Kentaro A., de Cara S., Kindermann G.,...
  • Cited by (35)

    • Assessing drought risk in Mediterranean Dehesa grazing lands

      2016, Agricultural Systems
      Citation Excerpt :

      An innovative approach to risk valuation is the work of Lybbert et al. (2010), who explore the potential of field experiments to better understand how poor valuate drought risk mitigation options in a dynamic context. Linking famers' decision-making and biophysical models in a stochastic context is a computational challenge highlighted by Freier et al. (2011). These authors adopt a Markovian approach and develop an optimization decision-making model in order to identify economic and environmental effects of long persistent drought on extensive livestock systems.

    • The dynamic soil organic carbon mitigation potential of European cropland

      2015, Global Environmental Change
      Citation Excerpt :

      Using Markov chains enables us to account for biophysical SOC dynamics in an economic partial equilibrium model. However, the applied approach is not limited to soil organic carbon representation only but could be applied in various mathematical models to dynamically represent different biophysical processes i.e. pasture dynamics (Freier et al., 2011). Besides biophysical dynamics and relationships, we are also able to capture trade-offs and feedbacks between the different land use sectors, markets and regions in our bottom-up modeling framework and thereby to come up with a more robust quantification of the economic SOC mitigation potential for European cropland.

    • How to avoid unsustainable side effects of managing climate risk in drylands - The supplementary feeding controversy

      2015, Agricultural Systems
      Citation Excerpt :

      Drought-induced economic losses are seen as a major threat to global food security, particularly in drylands (e.g., Haile, 2005; Fraser et al., 2011; Sietz et al., 2011). The extent of this problem is predicted to increase with future climate change (Reynolds et al., 2010; Freier et al., 2011; Rufino et al., 2013). Livestock systems are of particular concern in this regard (FAO, 2009).

    View all citing articles on Scopus
    View full text