How to include Human Actions in Earth System Science Modelling? Gilberto Câmara Earth System Science Centre, INPE Workshop on Earth System Science Models, São José dos Campos, 2009

Global Change Where are changes taking place? How much change is happening? Who is being impacted by the change? Human actions and global change photo: A. Reenberg photo: C. Nobre

Are targets of deforestation possible for the Brazilian Amazon? National Plan for Climatic Change (Brazil, 2008)

Earth system science needs to model the interactions between nature and society Nature: Physical equations Describe processes Society: Decisions on how to Use Earth´s resources

Slides from LANDSAT Aral Sea Bolivia images: USGS Modelling Nature-Society Interactions How do humans use space? How to describe and predict changes resulting from human actions? What computational tools are needed to model Modeling the interaction nature-society interactions?

land_cover cells Question #1 for Nature-Society models Temperature What ontological kinds (data types) are required for nature-society models?

Climate Change models deals with ST fields A field is a spacetime continuum Field: T  S  A

land_cover cells in 1985 Societal data are modelled as ST objects land_cover cells in 2000 An object is an individual that exists in space and time (Object: ID  T  [S,A])

land_cover cells (objects) Requirement #1 for Nature-Society models Temperature (fields) Nature-society models need to describe fields and objects (and store their attributes in a database)

Question #2 for Nature-Society models What models are needed to describe human actions?

Modelling Human Actions Models based on global factors  Explanation based on causal models  “For everything, there is a cause”  Human_actions = f (factors,....) Emergent models  Local actions lead to global patterns  Simple interactions between individuals lead to complex behaviour  “More is different”  “The organism is intelligent, its parts are simple-minded”

Statistics: Humans as clouds Establishes statistical relationship with variables that are related to the phenomena under study Basic hypothesis: stationary processes Exemples: CLUE Model (University of Wageningen) y=a 0 + a 1 x 1 + a 2 x a i x i +E

Statistics: Humans as clouds Statistical analysis of deforestation [Aguiar et al, 2007]

Agents as basis for complex systems Agent: flexible, interacting and autonomous An agent is any actor within an environment, any entity that can affect itself, the environment and other agents.

Agent-Based Modelling Goal Environment Representations Communication Action Perception Communication source: Nigel Gilbert

Agents: autonomy, flexibility, interaction Synchronization of fireflies

Bird Flocking No central authority: Each bird reacts to its neighbor Bottom-up: not possible to model the flock in a global manner. It is necessary to simulate the INTERACTION between the individuals

Requirement #2 for Nature-Society models Models need to support both statistical relations (clouds) and agents (ants) [Andrade-Neto et al., 2008]

Question #3 for Nature-Society models What types of spatial relations exist in nature-society models?

Rondonia Natural space is (usually) isotropic Societal space is mostly anisotropic

Which spatial objects are closer? Societal spaces are anisotropic Which cells are closer? [Aguiar et al., 2003]

Euclidean space Open network Closed network D2 D1 Requirement #3 for Nature-Society models: express anisotropy explicitly [Aguiar et al., 2003]

Question #4 for Nature-Society models How do we combine independent multi-scale models with feedback?

Models: From Global to Local Athmosphere, ocean, chemistry climate model (resolution 200 x 200 km) Atmosphere only global climate model (resolution 50 x 50 km) Regional climate model (resolution 10 x 10 km) Hydrology, Vegetation Soil Topography (e.g, 1 x 1 km) Regional land use change Socio-economic changes Adaptation (e.g., 100 x 100 m)

National level - the main markets for Amazonia products (Northeast and São Paulo) and the roads infrastructure network; Regional level - for the whole Brazilian Amazonia, 4 million km2; Local level - for a hot-spot of deforestation in Central Amazonia, the Iriri region, in São Felix do Xingu, Pará State grid of 25 x 25 km2 grid of 1 x 1 km2 Nature-Society models need multi-scale modelling [Moreira et al., 2008]

Not all multiscale models have nested grids Environmental Modeler [Engelen, White and Nijs, 2003] CLUE model [Veldkamp and Fresco, 1996] Multi-scale modelling: hierarchical relations need to be described

Network-based relations Flow of timber from Amazonia Multi-scale modelling includes networks National market chains in Brazil [Moreira et al., 2008] (source: )

Requirement #4 for Nature-Society models: support multi-scale modelling using explicit relationships Express explicit spatial relationships between individual objects in different scales [Moreira et al., 2008] [Carneiro et al., 2008]

Question #5 for Nature-Society models Small Farmers Medium-Sized Farmers photos: Isabel Escada How can we express behavioural changes in human societies? When a small farmer becomes a medium-sized one, his behaviour changes

Old Settlements (more than 20 years) Recent Settlements (less than 4 years) Farms Settlements 10 to 20 anos Societal systems undergo phase transitions Isabel Escada, 2003 [Escada, 2003]

Requirement #5 for Nature-Society models: Capture phase transitions Newly implanted Deforesting Slowing down latency > 6 years Deforestation > 80% Small Farmers Iddle Year of creation Deforestation = 100% Deforesting Slowing down Iddle Year of creation Deforestation = 100% Deforestation > 60% Medium-Sized Farmers photos: Isabel Escada

TerraME: Computational environment for developing nature-society models Cell Spaces Support for cellular automata and agents TerraME: Modular modelling tool [Carneiro, 2006]

TerraME´s way: Modular components Describe spatial structure 1:32:00Mens :32:10Mens :38:07Mens :42:00Mens return value true 1. Get first pair 2. Execute the ACTION 3. Timer =EVENT 4. timeToHappen += period Describe temporal structure Newly implanted Deforesting Slowing down latency > 6 years Iddle Year of creation Deforestation = 100% Describe rules of behaviourDescribe spatial relations [Carneiro, 2006]

Spatial structure in TerraME: Cell Spaces integrated with databases

Spatial Relations in TerraME Spatial relations between entities in a nature-societal model are expressed by a generalized proximity matrix (GPM) [Moreira et al., 2008]

TerraME: multi-scale modelling using explicit relationships Generalized proximity matrices express explicit spatial relationships between individual objects in different scales up-scaling Scale 1 Scale 2 father children [Moreira et al., 2008] [Carneiro et al., 2008]

To Agent Cell a b a b c c Cell Agent From GPM: Relations between cells and agents [Andrade-Neto et al., 2008]

TerraME uses hybrid automata to represent phase transitions State A Flow Condition State B Flow Condition Jump condition A hybrid automaton is a formal model for a mixed discrete continuous system (Henzinger, 1996) Hybrid Automata = state machine + dynamical systems

Hybrid automata: simple land tenure model STATEFlow ConditionJump ConditionTransition SUBSISTENCEDeforest 10% of land/yearDeforest > 60%CATTLE Extensive cattle raisingLand exhaustionABANDONMENT Forest regrowthLand revisionRECLAIM Public repossessionLand registrationLAND REFORM Land distributionFarmer gets parcels SUBSISTENCE Deforest 20%/year Farmer gets parcel deforest>=60% Land exhaustion CATTLE Extensive cattle raising ABANDONMENT Regrowth RECLAIM Public repossession Land revision LAND REFORM redistribution Land registration

TerraME Software Architecture TerraLib TerraME Framework C++ Signal Processing librarys C++ Mathematical librarys C++ Statistical librarys TerraME Virtual Machine TerraME Compiler TerraME Language RondôniaModelSão Felix Model Amazon ModelHydro Model [Carneiro, 2006]

Lua and the Web Where is Lua? Inside Brazil  Petrobras, the Brazilian Oil Company  Embratel (the main telecommunication company in Brazil)  many other companies Outside Brazil  Lua is used in hundreds of projects, both commercial and academic  CGILua still in restricted use until recently all documentation was in Portuguese TerraME Programming Language: Extension of LUA LUA is the language of choice for computer games [Ierusalimschy et al, 1996] source: the LUA team

TerraME programming environment [Carneiro, 2006]

Multi-scale, multi-locality land change scenarios Current studies: Macro Amazonia PA 279/SFX e BR 163/Santarém Planned studies: North of MT, South of Amazonas Deforestation Forest Non-forest Clouds/no data INPE/PRODES 2003/2004:

Paved roads in 2010 Unpaved roads Main rivers 0.0 – – – – – – – – – – 1.0 % change 1997 a 2020: Baseline Scenario A – Hot spots of deflorestation from 1997 a 2020 São Felix/Iriri (Terra do Meio) BR 163 (Cuiabá-Santarém) South of Amazonas BR 319 (Porto Velho-Manaus) New frontiers in Central Amazonia: Source: Aguiar, 2006

2) São Felix do Xingu study: multiscale analysis of the coevolution of land use dynamics and beef and milk market chains Current studies: Macro Amazonia PA 279/SFX e BR 163/Santarém Planned studies: North of MT, South of Amazonas Deforestation Forest Non-forest Clouds/no data INPE/PRODES 2003/2004:

Forest Not Forest Deforest River Observed deforestation from 1997 to 2006: leading deforestation rates and cattle industry organization

Land use Change model Beef and milk market chain model Small farmers agents Medium and large farmers agents Land use Change model Small farmers agents Medium and large farmers agents Landscape metrics model Pasture degradation model Several workshops in 2007 to define model rules and variables Landscape model: different rules for two main types of actors

Landscape model: different rules of behavior at different partitions Forest Not Forest Deforest River FRONT MIDDLE BACK SÃO FÉLIX DO XINGU

Landscape model: different rules of behavior at different partitions which also change in time FRENTE MEIO RETAGUARDA Forest Not Forest Deforest River FRONT MIDDLE BACK SÃO FÉLIX DO XINGU

Modeling results 97 to 2006 Observed 97 to 2006

Acknowledgments and thanks to Tiago Carneiro (UFOP): TerraME architect and chief programmer Ana Aguiar (INPE): Land use models and new concepts in TerraME Miguel Monteiro (INPE): Agent models in TerraME Sergio Costa, Pedro Andrade-Neto, Karine Ferreira, Gilberto Ribeiro, Eva Moreira, Giovana Espinola: PhD Students The LUA team at PUC-RIO