Faculty of Mathematics and Natural Sciences - Biogeography

PASANOA - Work package Description


Work package 1: Understanding local land use decisions

Work package 2: Future land use pathways

Work package 3: Land use impacts on ES and biodiversity

Work package 4: Scaling up impacts of agriculture on ES and biodiversity

Work package 5: Trade-offs between agriculture, ES and biodiversity

Work package 6: Identifying sustainable solutions



Work package 1: Understanding local land use decisions

Lead: IAMO (Müller)

Contributors: UMdP, INTA-IRB

Rationale: Understanding land agents’ decision making is crucial for designing effective policies. This WP will explore land use decision making across land systems and actors in the Chaco under various policy options.

Work plan: We will select between 6 and 10 communities in the study area that are characterized by different conditions in terms of market accessibility, biophysical characteristics, land use strategies and human welfare. Within each community, we will conduct participatory mapping with groups of community members to delineate current land use and the major changes in land use for the last 30 years. We will then utilize qualitative interviewing techniques (cf. Müller et al. 2014, GEC) to disentangle the underlying drivers of local land use strategies and to discuss the changes therein. In addition, we will coordinate a structured survey of about 300 and 500 land-use agents (~50 per community) to obtain statistically representative community-level data. Based on these qualitative insights and quantitative data, we will construct and calibrate Bayesian Networks (BNs) to explore land use decision making at the plot and farm level (Celio et al. 2014; Frayer et al. 2014). BNs are ideally suited to combine qualitative expert and stakeholder knowledge with quantitative survey data, can handle nonlinear relationships and variable interactions, and are insensitive to multicollinearity. We will use the BNs to account for heuristic decision making of agents and allow modeling proximate and underlying drivers of land use change. The model results will be validated in the field in a second field campaign in mid-2016.

We will use the calibrated BN as the decision engine in an agent-based simulation model (ABM) at the community level (Sun and Müller 2013) to assess agent interactions (e.g., land rentals, sales), to test the impact of exogenous drivers (e.g., price changes) and to explore the effects of policy interventions (e.g., subsidy schemes, zoning laws) on local land-use strategies. The ABM model will serve as a decision support platform for local stakeholders to analyze the effect of the scenarios from WP2 on the local level. In WP3, the same decision platform will be used to simulate the impacts of scenario conditions on ecosystem services and biodiversity. We will present, discuss and validate the decision platform and the resulting outcomes with local stakeholders during the project dissemination workshop in 2017.

Main outputs:

(1) Deep understanding of land use decision-making

(2) Agent-based modeling framework to be used as an interactive decision-support platform for stakeholders


Work package 2: Future land use pathways

Lead: HUB (Kuemmerle)

Contributors: INTA-IRB (Gavier Pizarro), INTA-EEA (Volante), IAMO (Müller)

Rationale: Future agricultural expansion and intensification is likely in the Chaco, yet regionalized land-use scenarios are missing. This WP will develop scenario storylines and simulate associated land use patterns.

Work plan: We will define a set of 4-6 plausible, yet contrasting scenarios to explore future land use in the Chaco (Patela et al. 2007; Westhoek et al. 2006), building upon and interlinking with ongoing work on the future of agriculture in Argentina at INTA. Scenario development will be a two-stage process. First, we will explore the drivers of recent land-use change patterns, with a focus on understanding the factors driving the expansion of intensified ranching and soybean cultivation, currently the two main land-change processes in the Chaco. To do so, we will use a net returns model (Lubowski et al. 2008), parameterized using land-change datasets developed by consortium members and predictor variables capturing national and province-level economic factors (e.g., fuel and fertilizer costs, soybean price, export taxes), changes in infrastructure (e.g., accessibility, travel time to ports), and environmental conditions (e.g., soil quality, topography, rainfall patterns). This model will provide a powerful platform to explore the potential land-use change outcome of policy interventions with direct effects on net returns (e.g., changes in export taxes, subsidies, payment for ecosystem services).

Second, to draft scenario storylines, we will organize a workshop at the beginning of project year 2, bringing together PASANOA researchers and key stakeholders. Prior to the workshop, a screening of key potential future developments influencing land change in the Chaco will be carried out, including macro-economic trends (e.g., prices, trade agreements), shocks (e.g., devaluation), polices (e.g., export taxes, zoning), population trends (e.g., rural-urban migrations), technological innovation (e.g., new cultivars), infrastructure change (e.g., improved roads or railways), regional climate change (e.g. rainfall patterns), and conservation initiatives (e.g., new protected areas, REDD+). We will then distill a core set of factors, considering the outcomes from the drivers analyses, and develop scenario storylines using multi-criteria analyses (Gavier-Pizarro et al. 2015; Rounsevell et al. 2006; Xiang and Clarke 2003).These scenarios will consists of (a) a storyline describing developments resulting in the scenario, (b) short narratives describing scenario conditions and trends, as well as changes in location factors (e.g., infrastructure, zoning), (c) quantitative changes in land-use types (derived using the net returns model). Our scenarios will have a time horizon until 2030 and a spatial grain of 1 km.

To translate the storylines into future land use patterns, we will use the simulation model DynaCLUE that allocates land demand based on the suitability of a location for certain land use types (Verburg and Overmars 2009; Verburg et al. 2006). Land use types considered will include the main land systems in the Chaco: protected forests and grasslands, forest grazing linked to homesteads (i.e., puestos), silvopastoral grazing, intensive ranching, and agri-business agriculture (i.e., mainly soybean). Suitability for these land use types will be based on spatial regression models relating land-use change to predictor variables such as environmental factors (e.g., soil quality, topography, rainfall), accessibility (e.g., distance to ports) or population density (Müller et al. 2009; Pfaff 1999). Land use/change maps to parameterize these regression models are available, partly from the work of our consortium (Gasparri and Grau 2009; Hansen et al. 2013; Volante et al. 2012).

Main outputs:

(1) Insights into influential drivers of past land-use change patterns

(2) A set of plausible, yet contrasting future scenarios

(3) Future land-use simulations for each scenario


Work package 3: Land use impacts on ES and biodiversity

Lead: INTA-IRB (Gavier-Pizarro, Zaccagnini)

Contributors: UMdP, UTuc, HUB

Rationale: How agricultural production impacts ES provisioning and biodiversity in the Chaco remains unclear. This WP will assess the impact of land management on key ES and biodiversity indicators at the local and landscape scale.

Work plan: To assess agricultural production patterns, we will use meat and soybean production data from the literature and our own previous research (Macchi et al. 2013; Mastrangelo and Gavin 2012). Production estimates for grazing systems will be based on systematic forage harvest (dry matter / year), converted into kilograms of meat (Macchi et al. 2013). Protected forests will be assumed to having no food production (for regional or global markets).

Regarding ES, we will assess (a) carbon stocks, (b) soil salinization control, (c) water availability and (d) pest control. Field data for carbon stocks will come from existing survey plots and forest inventory plots (Gasparri and Baldi 2013; Gasparri et al. 2008). To assess soil salinization control, we will use data on soil water and chloride content collected in fields with different time lags since deforestation, as well as forest fragments. To assess water availability, we will use data on soil properties and water content collected in agricultural fields with different management practices. Data will come from fieldwork and our own previous work (Amdan et al. 2013).

To proxy pest control, we will use the richness and abundance of insectivorous birds and raptors. These bird species can maintain agriculture pest species (i.e., insects or rodents) below populations levels where they affect crop condition and harvests yields substantially (Sekercioglu 2006; Whelan et al. 2008). Since data on bird pest consumptions are rare and unavailable in Argentina, we will use an inferential approach (Gavier-Pizarro et al. 2012b).

Regarding biodiversity, we will use four indicators (a) diversity of bird species, (b) abundance of 10-15 focal bird species, (c) amphibian diversity and (d) landscape fragmentation. These indicators are highly complementary, sensitive to land management and relatively easy to survey across large areas. Our team has already collected much of these data (Gavier-Pizarro et al. 2012a; Macchi et al. 2013; Mastrangelo and Gavin 2012; Zaccagnini et al. 2010) and will collect more for PASANOA. Landscape fragmentation will be estimated using morphological image segmentation (Vogt et al. 2009; Vogt et al. 2007).

We will then employ regression analyses to link the three types of response variables (agricultural output, ecosystem services, and biodiversity indicators) to a set of land management factors (e.g., capital inputs, workforce, field size, farm size; collected in WP1) as well as a set of location factors that exert influence on our target variables (e.g., topography, soil quality, rainfall, accessibility). As a regression framework, we will use Boosted Regression Trees (BRTs), which are a non-parametric approach based on decision trees (Elith et al. 2008). BRTs are flexible regression models that can capture variable interactions and non-linear relationships, frequently outperform classic tools, and have recently been adapted to land-use applications (Levers et al. 2014; Müller et al. 2013). All models will be cross-validated and we will test the sensitivity of results to model parameterization.

Using the simulation platform developed in WP1, we will also assess how policy interventions would affect ES provisioning and biodiversity.

Main outputs:

(1) Insights on how land management affects ES provisioning and biodiversity

(2) Understanding of how policy interventions affect ES and biodiversity outcomes


Work package 4: Scaling up impacts of agriculture on ES and biodiversity

Lead: UMdP (Mastrangelo)

Contributors: INTA-EEA, INTA-IRB, HUB

Rationale: The spatial patterns of land management impacts on ES and biodiversity are not clearly understood. This WP will upscale the results from WP3 to map agricultural output, ES and biodiversity at the regional scale.

Work plan: Levels of ES provision will be scaled up to the regional scale using the regression models produced in WP3 as well as an ES mapping platform developed by consortium members (ECOSER, Laterra et al 2012). The regression models from WP3 (Boosted Regression Trees) will be applied to spatially predict the levels of (a) agricultural production, (b) the  different ecosystem services, and (c) our biodiversity indicators for the entire study area. Regional-scale data regarding the necessary predictor variables will come from agricultural censuses (i.e. for land management factors) and GIS databases available at INTA and our other project partners (i.e., for location factors such as environmental data, infrastructure, zoning).

In addition to the ES modeled in WP3, we will also assess wind erosion control at the regional scale using the RUSLE model in ECOSER. The latter is a platform combining protocols for ES assessments, including the definition, mapping, valuation of ES, that has been developed and successfully applied by the UMdP team in other parts of Argentina (Laterra et al. 2012a). We will use the same database of location factors to ensure comparability of the ES mapped using the BRT models and soil erosion mapped using ECOSER.

We will then use the resulting regional maps of agricultural production, ecosystem service provisioning, and biodiversity to identify typical combinations of these target variables (Laterra et al. 2012b; Mastrangelo et al. 2014), To identify such bundles or portfolios of services (Raudsepp-Hearne et al. 2010), we will use Self-Organizing Maps (SOMs), a non-parametric spatial clustering techniques based on an unsupervised competitive learning algorithm (Agarwal and Skupin 2008; Kohonen 2001; Václavík et al. 2013). SOMs are a powerful tool to visualize, and reduce complexity in high-dimensional data, such as when overlaying multiple landscape functions, and have the advantage of preserving the typology of the input data by considering geographic closeness in the clustering process. In other words, SOMs identify typical combinations of ES while considering the spatial patterns of ES provisioning. SOMs result in two main outputs: (a) a cluster map, showing areas with similar ecosystems service provisioning, and (2) flower diagrams visualizing cluster characteristics (Raudsepp-Hearne et al. 2010; Václavík et al. 2013). The flower diagrams are a powerful tool to analyse and visualize trade-offs among agriculture, ES provisioning, and biodiversity (Foley et al. 2005; Raudsepp-Hearne et al. 2010).

Once future land-use maps are available (WP2), we will use our regression and ECOSER models for agricultural production, ecosystem service provisioning and biodiversity to assess how these target variables may change across future scenarios. Using the current ES bundles identified by the SOMs we will also assess how these bundles may change under alternative future scenarios.  Comparing these changes with the current bundles of services will allow to highlight hotspots of potential ES and biodiversity change in the Chaco, as well as local trade-offs and synergies between increased agricultural production and service provisioning.

Main outputs:

(1) Maps of how agricultural production, ecosystem services and biodiversity are distributed across the Chaco

(2) Maps and characteristics of ES bundles for the current and future scenarios


Work package 5: Trade-offs between agriculture, ES and biodiversity

Lead: UTuc (Grau)

Contributors: HUB, IAMO, INTA-IRB

Rationale: Existing trade-off analyses have mostly disregarded spatial heterogeneity and scale of ES. This work package will fill these gaps and systematically address the trade-offs between agriculture, ES and biodiversity.

Work plan: Two analyses will be carried out to analyze trade-offs between agricultural production, ecosystem services provisioning, and biodiversity. First, we will use the formalized relationships between target variables and predictors (i.e., regression models from WP3) to derive trade-off curves between each variable pair. Because these regressions models will rely on the same set of predictor variables, it is possible to predict target variables across the full range of environmental conditions, and thus to estimate pairwise efficiency frontiers (i.e., curves showing the co-provisioning of yields, ecosystem services, and biodiversity for each variable pair across all environmental conditions, Polasky et al. 2008; Polasky et al. 2005). The nature of these curves can provide deep insight into trade-off between variables and potential land-use strategies to mitigate these trade-offs depending on whether the identified efficiency frontiers are concave (i.e., suggesting a planning strategy targeted at integrating land use and conservation) or convex (i.e., suggesting a segregation of land use and conservation) (Butsic et al. 2012; Mastrangelo and Gavin 2012; Phalan et al. 2011).

Second, we will carry out multi-criteria optimization to identify ideal levels of agricultural production, ES and biodiversity that would maximize the total provisioning of these targets for the Argentine Chaco. To do so, we will use the newly developed software RobOff, which provides a rich toolkit to systematically analyze trade-offs between different land uses (Pouzols et al. 2012; Pouzols and Moilanen 2013). Using optimization routines, RobOff allocates area to these land uses in order to (a) maximize the provisioning of all targets (i.e., agricultural production, ES, and biodiversity levels) or (b) to maximize conservation gains given certain production targets (Moilanen et al. 2011). RobOff thus minimizes the trade-offs between target variables for the Chaco as a whole, while considering spatial heterogeneity in these variables. Moreover, RobOff is excellently suited to explore uncertainty (e.g., input variables, parameterization) and thus the robustness of identified optimal land uses levels (Pouzols and Moilanen 2013).

Using the RobOff framework, we will implement two analyses. First, we will identify opportunities to increase ES provisioning and biodiversity given current levels of agricultural production. We will test alternative land-use strategies, such as such land sparing (agri-business production on less land) vs. a land sharing (i.e., favoring less intensive agriculture on more land) to reach the same production targets (Phalan et al. 2011; Tscharntke et al. 2012; Wright et al. 2012). Moreover, we will carry out optimization at the province and study region level to reveal how optimal solutions may change across planning scales. Second, we will use RobOff to derive how optimal levels of land uses in the Chaco would change across future scenarios (WP3). We will compare those levels with current levels (above) and those envisioned in the new zoning plan that Argentina implemented recently (i.e., the Ley de Bosques). This will highlight opportunities to mitigate trade-offs potentially entailed in the current zoning. As in the above analyses, we will consider alternative land management strategies to reach production targets and carry out land-use optimization at province and study region scale to explore scale sensitivity.

Main outputs:

(1) Pairwise trade-off curves between agricultural production, ES and biodiversity

(2) Optimal target levels of land uses that would mitigate these trade-offs – now and across future scenarios


Work package 6: Identifying sustainable solutions

Lead: HUB (Kuemmerle)

Contributors: INTA-IRB, INTA-EEA, UMdP, UTuc, IAMO

Rationale: Smart landscape planning can mitigate land use conflicts substantially. This work package will use spatial priorization tools to identify landscape patterns that correspond to the optimal target levels obtained in WP5.

Work plan: We will use the novel spatial priorization tool Zonation (Moilanen et al. 2009; Moilanen et al. 2011b) as well as the maps of target variables (WP2 and WP4) to translate the optimal target levels (WP5) into landscape patterns. Zonation is a tool developed to balance habitat requirements of different species across a landscape, but it can equally well be used to balance different land uses, including both production and conservation uses. The Zonation algorithm (Moilanen et al. 2011b) produces a hierarchical prioritization of the relative value of each landscape element for reaching a certain target across the entire landscape. Zonation starts from the assumption that protecting everything would be best for conservation, and then iteratively assigns cells of the landscape to production uses, using minimization of marginal loss as the criterion to decide which cell to remove next. Simultaneously, information is collected about the decline of ecosystem service and biodiversity levels as agricultural production increases. This information is used to derive performance curves, which quantify how each target function (agricultural production, ecosystem service levels and biodiversity) varies as cells are replaced (Moilanen et al. 2011b). This allows deriving land use patterns for the study region as a whole that correspond to the optimal target levels (WP5) as well as testing how sensitive these land use patterns are to changes in target levels.

In analogue to the procedure in WP5, we will carry out two analyses. First, we will identify optimal land use patters given current agricultural production levels. These optimal landscapes can then be compared to the actual Chaco landscape to identify how land-use and conservation planning could mitigate current trade-offs. Second, we will derive optimal land-use patters for all future scenarios (WP4) and compare those to the current situation, thus highlighting areas that planning should focus on to avoid unwanted and avoidable trade-offs. We will also compare the resulting optimal landscapes for current production levels and our future scenarios to the recently implemented zoning plan (Ley de Bosques) to highlight where revisions of this zoning plan would be useful in order to make use of synergies between agricultural production, ecosystem service provisioning and conservation.

As in WP5, we will explore how optimal landscapes patters change if alternative land-use planning and conservation strategies are adopted (e.g., land sparing vs. land sharing paradigms) and how sensitive identified landscapes are to the extent over which optimization is carried out (e.g., province vs. entire study region). The latter is particularly important, given that the current zoning plan was drafted in a decentralized way, thereby potentially missing opportunities to mitigate trade-offs.

Once optimal land-use patterns for current and future land use scenarios, for alternative management paradigms, and for different scales are derived, we will discuss these maps with relevant stakeholders in a workshop. We will document trade-offs and potential pathways to mitigate them connected to each scenario, and discuss fine-scale (WP1) and regional-scale (WP3) policy levers to implementing desirable land-use and conservation landscapes in the Chaco. We will also present and document the entire methodology used in PASANOA to allow interested stakeholder to implement these tools.

Main outputs:

(1) Maps of land use patterns that balance agricultural production, ES provisioning, and biodiversity for the Argentine Chaco

(2) Provision of sustainable solutions for stakeholders at local to regional scale