Integrating Spatial & Temporal Dynamics in Ecological Modeling: Issues & Directions Michael Barton School of Human Evolution and Social Change Ecosystems are dynamic • Not just variation, but change across space and time • Dynamic modeling, such as GARP, offers new ways to study human ecology • Yet capturing the integrated dynamics of both space and time remains elusive Archaeology & paleoecology • Focus on long-term change • Also on human-environmental interaction • How to integrate these in research protocols? Space-time modeling: issues for current methods • Look at change over time at several localities • Look at change across space for several moments in time Space-time modeling: issues for current methods Space-time modeling: issues for current methods Epipaleo Early Neol Late Neol U. Paleo M. Paleo Space-time modeling: issues for current methods • High resolution diachronic record in one or a few • • sites unwarranted regional generalization across variable space High resolution spatial record for a few time periods past as a sequence of snapshots with change happening invisibly between (or stepwise change) Conceptualizing rich socioecological variation across space and complex dynamics through time desirable but difficult with current methods Future directions in space-time modeling • Goals for study of long-term dynamics of human ecology • High resolution models – spatial – chronological • Link dynamically so that – state of modeled ecosystem at any given time (t) is in part a function of ecosystem at prior time (t-1) and affects future state (t+1) – State of modeled ecosystem at any given locale is in part a function of state in surrounding locales • Test against archaeo-ecological record Future directions in space-time modeling: an example • Mediterranean Landscape • Dynamics Project Project locations at opposite ends of Mediterranean Basin – Encompasses wide range of ecological & social variation – Tracks initial spread of agriculture & replacement of foraging systems – Different trajectories to the appearance of social complexity and urbanism Meditereranean Landscape Dynamics Project • Agent-based simulation of human landuse: • • • • beginning of farming to beginning of urbanism Surface process models of ancient landscapes and landcover Synoptic climate models All linked within a GIS framework so that change in one module can affect state variables that serve as input to another Test and refine against rich archaeological and paleoecological record Dynamic space-time modeling Terrain modeling: multi-yr. steps • 3 interlinked modeling environments in GIS platform – Potential landscape model – Reference landscape chronosequence – Agropastoral socioecology model Climate model (temp&precip) Potential landscape model Vegetation modeling: multi-yr. steps Paleobotanical data Initial state Paleovegetation Veg. edaphic parameters Modern DEM Geological data Archeological data Paleoterrains (DEM’s) Prehistoric settlement & landuse Agent Modeling Climate model (temp&precip) Reference landscape chronoseq . Settlement & landuse modeling Initial state & validation at various stages Vegetation modeling: multi-yr. steps Terrain modeling: multi-yr. steps Agropasto ral socioecol ogy model Dynamic space-time modeling • Potential landscape model: surface processes and landcover Terrain modeling: multi-yr. steps Climate model (temp&precip) Potential landscape model Vegetation modeling: multi-yr. steps Dynamic space-time modeling • Reference landscape chronosequence: surface processes and landcover Paleobotanical data Paleovegetation Veg. edaphic parameters Modern DEM Geological data Archeological data Paleoterrains (DEM’s) Prehistoric settlement & landuse Reference landscape chronoseq. Dynamic space-time modeling • Human landuse: agent simulation Agent Modeling Information Inputs Temp. extremes Precip. amt & dist Terrain characteristics Current vegetation Decision Algorithms Settlement Establish Shrink/grow Abandon Location Landuse Location/extent Cultivation (which crops) Pasture (which animals) Settlement & landuse impacts on terrain & vegetation Dynamic space-time modeling • Climate • surface processes & landcover agent simulation Landscape socioecology Agent Modeling Settlement & landuse modeling Climate model (temp&precip) Vegetation modeling: multi-yr. steps Terrain modeling: multi-yr. steps Agropastoral socioecology model Polop Valley Landuse Intensity M. Paleolithicto Neolithic II Visualization & analyzing spacetime dynamics • As we begin to approach the goal of integrating space and time in ecological modeling, how can we visualize and analyze results? • How can we represent n-dimensional data in 2 dimensional media? • How can we compare and evaluate results quantitatively? Visualizing ecological dynamics • Animation Late Middle Upper Neolithic Paleolithic NeolithicIII Polop Valley Landuse Intensity Visualizing ecological dynamics • Animation – Looks dynamic – Realistically portrays change through time – Difficult to evaluate change, similarities, and differences—visually or quantitatively • Alternative methods of visualization and analysis to complement animation Visualizing ecological dynamics • Example from • the Penaguila Valley, Spain Terra ASTER imagery for 2000-2004 – Decorrelation stretch of bands 1-3 – Landcover change • Analysis with GRASS GIS Visualizing ecological dynamics • Stacked surfaces/landscapes – – – – Can evaluate change better than with animation Good for major changes Visual alignment (cell to cell) difficult for detailed evaluation Problematic with high temporal resolution--I.e., many stacked landscapes Visualizing ecological dynamics • Stacked surfaces with a cutting plane. Color blends from each surface into next – Solves alignment problem – Allows detailed visualization of change across multiple time intervals – Can only show a spacetime transect (i.e., one dimensional space) Visualizing ecological dynamics • Stacked surfaces with movable cutting plane • Permits interactive viewing of multiple transects Visualizing ecological dynamics • Space-time ) ce pa (s – 2D space (x,y) – Time (z) – Socioecological variable (w) Y • Z (time) • volumes (STVs) Produced through 3D interpolation 4D shapes that simultaneously show… X (space) Visualizing ecological dynamics • STVs showing time • • dynamics in component 1 of PCA (Terra ASTER vnir bands 1-3) Values ! 200 Proxy for barranco vegetation, mostly pine space Visualizing socioecological dynamics • STVs for landuse change in the Polop Alto Valley, Spain: M. Paleolithic - Late Neolithic + - landuse intensity Analysis of spatial-temporal dynamics • Robust suite of methods exist for time series analysis and spatial patterns • Comparative lack of quantitative protocols for characterizing and evaluating dynamics across space and time together • Brief survey of potential directions Analysis of spatial-temporal dynamics • Basic change analysis. Normally for surface pairs 2004 = 2000 red=positive, blue=negative Analysis of spatial-temporal dynamics • Time series analysis for landscapes Coefficient of variation (2000-2004). Terra ASTER PCA component 1. Red=high, blue=low Analysis of spatial-temporal dynamics • Change analysis for STVs red=2000-2001, green=2002-2004 Analysis of spatial-temporal dynamics • Change analysis for STVs (STV[2002-2004]STV[2000-2001]) 2002-2004 > 2000-2001 2002-2004 < 2000-2001 Analysis of spatial-temporal dynamics • Representing space-time dynamics as volumes offers potential for new methods of analysis – Extension of map algebra to volumes already available – Regression, trend analysis, and other methods now applied to surfaces and point distributions can potentially be applied to volumes • Example of ways to address new challenges posed by modeling human ecology across space and time GRASS as platform for ecocultural modeling • Open source – – – – Modifiable, scriptable Multi-platform GPL licensing for access and dissemination Internationalization • Rich suite of modeling tools (e.g., r.le, r.neighbors, r.mapcalc) • 3D volumetric tools • Visualization tools Support & collaboration • NSF: ERE Biocomplexity in the Environment Program, • • grant BCS-0410269 ASU: School of Human Evolution and Social Change, School of Earth and Space Exploration, School of Computing and Informatics, Geographical Sciences, International Institute for Sustainability Partners: Universitat de València, Universidad de Murcia, University of Jordan, North Carolina State University, University of Wisconsin, Hendrix College, Geoarchaeological Research Associates, GRASS GIS Development Team.