Modeling the Scale Dependent Drivers of LCLU Dynamics in Northeastern Ecuador: Simulating Patterns of Landscape Change and Assessing their Cause and Consequence through Multi-Level Models and Cellular Automata

Project Overview

Funding Agency: NASA
Begin Date: March 1, 2003
End Date: April 30, 2006

Research Symposium in Quito, Ecuador – June 10, 2004
Click here to view a description of the symposium conducted by the Ecuador Project team in Quito on June 10, 2004.

As a continuation of the Agricultural Colonization in the Ecuadorian Amazon project, this project attempts to further delve into the factors related to changing land cover and land use (LCLU) in the northern Oriente of Ecuador. Exciting new methods are tested that provide greater insight into the various LCLU dynamics in the Amazonian region.

Using longitudinal household survey data collected in 1990 and 1999, a 2000 community survey, a multi-resolution remote sensing time-series, GIS coverages of resource potentials and endowments, and field verification and geodetic control data, we analyze the determinants of changes in LCLU at the plot, sector, and regional levels, and for annual and decadal periods. The fundamental research questions revolve around (a) the rates, patterns, and mechanisms of forest conversion to agricultural and urban uses; (b) the relative importance of exogenous and endogenous variables on these land uses; (c) the associated scale dependent drivers of LCLU dynamics and patterns operating across socio-economic and demographic, biophysical, and geographical domains; (d) rate and pattern of land conversion from forest to agricultural crops, pasture, secondary plant succession, and urbanization, as well as the rate and pattern of land abandonment at the farm level; and (e) plausible scenarios of future land cover change and their policy implications as assessed through multi-level models that are responsive to multi-scale effects as well as spatial simulations of LCLU dynamics through a cellular automata (CA) approach.

The survey periods and the assembled satellite time-series images serve as our reference dates that are integrated to define relationships through (1) multivariate logit models of LCLU for 1990, 1999, and for changes between those two survey periods; (2) satellite image classifications and change-detections of LULC dynamics, space-time trajectories of pixel histories, and pattern metrics of landscape organization to define LCLU composition and spatial structure; (3) LCLU simulation through cellular automata, informed by the satellite and multivariate models of LCLU change, to create spatial simulations of LULC dynamics; and (4) multi-level models to integrate variables and effects from multiple scales into an integrated model of LCLU dynamics to assess the scale dependence of variable interactions on LCLU patterns.

The analysis is framed within a dynamic systems approach that emphasizes non-linear relationships, feedback mechanisms, and critical thresholds in population-environment interactions. Theoretical foundations include principles involving the interplay of political ecology, human ecology, landscape ecology, and complexity theory.

The multi-level models are used to integrate household, community, and regional variables that impact household decision-making and hence the mapped LCLU patterns at the farm level. The spatial simulations are developed through CA approaches at the annual and decadal scales. Linear and non-linear responses or “critical landscapes” are studied to model the ecological responses to a range of spatial patterns of LCLU derived through hypothetical, modeled, and observed conditions. The multi-level models are assessed through statistical measures of model performance, whereas the derived CA patterns are compared to the actual patterns represented in the satellite time-series and assessed through image change-detections, change trajectories or pixel histories, and summary correlations and pattern metrics for comparisons of expected vs. observed LCLU patterns. System behaviors are interpreted within a policy-relevant context by comparing simulated LCLU scenarios to targeted land management outcomes. Multiple LCLU change scenarios are developed around defined policy goals. Model convergence and variable sensitivity are examined relative to the LCLU patterns, model variables, and policy goals and expectations.

The basic intent of this research is to assess the rate, pattern, and mechanisms of forest conversion to agricultural and urban land uses by examining the scale dependent drivers of LCLU dynamics of the Ecuadorian Oriente so that models of household decision-making regarding LCLU dynamics can be derived through the integration of exogenous and endogenous variables that operate within space-time scales and represent socio-economic and demographic, biophysical, and geographical domains, linked to a satellite time-series view of landscape change. Using such models and satellite views of LCLU change, multilevel models will integrate space-time scales as well as global, regional, and local effects, and spatially-explicit simulations of LCLU change will be derived using rules and weights of variable behavior and interactions derived through the empirical models, neighborhood conditions, feedback and threshold relationships, and initial conditions. The spatial simulations will be space-time sensitive and policy relevant. To accomplish the above set of goals, the following research aims will rely upon our previous data collection and preliminary research findings to bridge to the more integrated modeling that will consider population-environment interactions as cause and consequence of LCLU dynamics.

There are five primary research aims that we hope to address in this project.

Exogenous Impacts & Relationships: Does LCLUC (land use and land cover change) respond in a linear or non-linear way to changes in commodity prices, new or improved infrastructure, in-migration and population growth, institutional policies, or extensions of the road network? Feedbacks may produce a system with a critical point, subject to small or large periods of incremental change in such exogenous factors functioning with time lags. Determining what are the key external factors, examining feedbacks and lags, and identifying thresholds of change are critical for the analysis of system behavior and the effects of context on LCLUC. This can lead to the re-formulation of social and environmental policies that are otherwise usually assumed to have linear effects, and exogenous effects might also serve as shocks that may destabilize the region and introduce a dynamic equilibrium condition to the changing LULC pattern.

Migrant Colonists & LCLU Dynamics: What are the rates and scale dependencies of different types of land conversions at the finca, sector and regional levels? Typical land conversion types include forest-to-crop, forest-to-pasture, forest-to-urban, forest-to-barren, crop-to-pasture, crops-to-urban, pasture-to-secondary forest, pasture-to-urban, barren-to-secondary forest, and barren-to-crops. It is likely that conversion rates vary across scales from the finca to the region. It is also likely that conversion rates change over time and that the spatial and temporal considerations are conversion-type specific. What are the socio-economic and demographic, biophysical, and geographical factors and processes that drive land-use/land-cover conversion characteristics? Socio-economic and demographic factors include family size and demographics, educational level, past agricultural experience, family life cycle, duration of residence, economic status; biophysical factors include soil conditions, topography, water supply, surrounding land uses, existing LCLU characteristics at time of settlement; and geographical factors include proximity to a road, proximity to a town, proximity to a school, size of local market, and access to banks (credit and loans).

Multivariate Models of LCLU & LCLUC: What (parsimonious) set of socio-economic and demographic, biophysical, and geographical variables best explain LCLU, LCLUC, and plant biomass variation? What is the nature of the spatial autocorrelation (or degree of randomness) across spatial scales, including cell resolutions? How stable are multivariate models developed for the current landscape for other time periods and for other regions? What land units and spatial aggregations are the most useful for characterizing the composition and spatial organization of LCLU types within the Oriente? What are the interrelationships between system stakeholders in terms of LCLU dynamics in a context where deforestation, agricultural extensification, secondary plant succession, and urbanization are all occurring in response to population in-migration, road building, and the expansion of the market economy? How does development and the emergence of and growth of towns in a central place hierarchy affect LCLUC? What is the form of multi-level models that seek to integrate multi-thematic variables across space and/or time scales, and what are the structure of the spatial autocorrelation and the nature of the autoregressive terms in the models to account for this effect of location on the ordering of data?

Spatially-Explicit, Dynamic Simulations of LCLUC: Develop a cellular automata (CA) system to generate LCLU simulations based upon actual conditions observed through the satellite image time-series and extended in time and space through observed rules and interactions between neighbors. The empirical models of LCLUC from the longitudinal survey data will be used to inform the CA models by generating weights and rules of behavior in the model. As well, the satellite time-series and the image change-detections and pixel history or trajectory work will also inform the CA models by imposing preferred pattern, and settlement trajectories, deforestation and reforestation pathways, and in short, landscape behavior to socio-economic and demographic, biophysical, and geographical conditions and characteristics for the simulation of LULC dynamics over time and throughout the region. Among the research questions to be considered include the following: are the rates and patterns of LCLUC non-linear due to feedbacks between the process of change and the existing patterns of land use and land cover? Do spatial patterns of LCLU cause a change in the fitness of the weighting of the transition probabilities? What are the LCLU patterns at year 2025 and at other selected time frames when LCLU dynamics are simulated using variables representing the socio-economic and demographic, biophysical, and the geographical domains? Are feedbacks between human activities and ecological dynamics non-linear? How can the science of complexity be used to understand how simple, fundamental processes combine to produce complex holistic systems? How can non-equilibrium systems, with feedbacks leading to non-linearity, evolve into systems that exhibit criticality, and thereby capture key elements of the dynamics of LCLU in the Amazon? What are the emerging patterns or trajectories of LCLUC across the satellite time-series, and do they represent LCLU patterns indicative of agricultural extensification or intensification, and how much secondary plant succession is occurring and why?

Policy-Relevant Scenarios & LCLUC Simulations: What are the effects of continuing petroleum exploration and extraction and road building on future in-migration of spontaneous colonists to both farm plots and towns? What are their effects on deforestation and land conversions at the finca, sector, and regional scales? How do land management-targeted outcomes of LCLU by strata compare to the LCLUC simulations? What are the implications for LCLUC in terms of government services (e.g., land titles, credit, agricultural extension), infrastructure (especially road development but also schools, health clinics), and institutional development (markets, community organizations, local leadership and policies) in the region? What have been the effects of the creation of the large Yasuni National Parks and the Cuyabeno Nature Reserve within the region on LCLU dynamics within the Oriente and lands adjacent to these conservation areas? How can information on human dynamics and LCLU dynamics be related to policy-relevant research by having the policy analysis contextually inform the research and contribute to policy formation at the national, regional, and local levels?