Modelling

Oreskes, 1998

• Argues complex environmental models cannot be validated because natural systems are open, changing, and incompletely understood.

• Proposes evaluation, not validation, as the appropriate way to assess models by recognising their strengths and limitations.

• Identifies four key uncertainties: theoretical, empirical, parametric, and temporal.

• Uses the Limits to Growth model and lead exposure modelling to show why models cannot provide certainty.

• Concludes models should be question-driven, exploratory tools rather than sources of definitive predictions.

Oreskes, Shrader-Frechette and Belitz, 1994

• Argues numerical models of Earth systems can never be fully verified or validated; they can only be confirmed to varying degrees.

• Models are heuristic representations, useful for exploring hypotheses rather than proving reality.

• Highlights problems of auxiliary hypotheses and underdetermination (non-uniqueness), where different models produce identical outputs.

• Distinguishes benchmarking (testing numerical solutions) from calibration (adjusting parameters to fit observations).

• Models remain valuable for sensitivity analysis, identifying inconsistencies, and guiding future research.

Wainwright and Mulligan, 2013

• Defines models as simplified abstractions of reality, balancing realism with parsimony.

• Argues modelling is both a scientific and creative process rather than a purely objective exercise.

• Environmental models increasingly need to incorporate human decision-making, as society strongly shapes environmental systems.

• Reviews three approaches: scenario modelling, Integrated Assessment Models (IAMs), and dynamic human-environment interaction models.

• Models function as digital test beds for exploring policy options, while acknowledging unavoidable model discrepancies.

Cliff and Ord, 1975

• Argues modelling makes human geography more useful for policy and planning.

• Traces the rise of urban modelling from 1950s transportation planning.

• Advocates structured pluralism, combining quantitative and qualitative approaches.

• Demonstrates applications in healthcare, education, retail, finance, and public services.

• Models should be used to explore planning scenarios, not simply predict future outcomes.

Su, 1998

• Critiques 1990s GIS-based urban modelling as technology-driven rather than conceptually driven.

• Argues traditional GIS relies on overly simplistic Cartesian conceptions of space and linear time.

• Calls for models that incorporate formal, socio-economic, and subjective spaces.

• Shows globalisation and post-industrial economies undermine assumptions built into earlier urban models.

• Recommends more integrated spatial-temporal GIS capable of representing complex urban processes.

Clarke and Wilson, 1987

• Examines the decline of quantitative geography after the 1970s.

• Argues advances in computing allow increasingly sophisticated and complex modelling.

• Suggests technological development expands the potential applications of quantitative geography.

• Emphasises modelling's continued relevance despite changing disciplinary priorities.

• Highlights the growing applicability of quantitative methods.

Castelli et al., 2013

• Uses multi-level modelling to examine health inequalities across geographical scales.

• Models identify where variation occurs rather than directly predicting health outcomes.

• Demonstrates how modelling can inform English health policy by identifying priority intervention areas.

• Recognises individuals are embedded within wider administrative and political structures.

• Warns models cannot establish causality or replace political judgement in policymaking.

Ladet et al., 2010

• Uses the SMASH agent-based model to simulate future landscape change.

• Treats landscapes as coupled socio-ecological systems, integrating environmental and human processes.

• Simulates policy scenarios including CAP reform, urbanisation, and continuation of existing trends.

• Finds agricultural change drives reforestation more strongly than ecological factors alone.

• Demonstrates the importance of modelling diverse human behaviours while recognising simplification reduces realism.

Van Oel et al., 2010

• Models two-way feedbacks between water availability and water use in northeast Brazil's Jaguaribe Basin.

• Uses spatially explicit multi-agent modelling to capture upstream-downstream interactions.

• Incorporates behavioural rules derived from farmer interviews.

• Demonstrates that positive and negative feedbacks occur simultaneously depending on context.

• Acknowledges limited representation of local heterogeneity despite improved treatment of human behaviour.

Mayaud et al., 2017

• Models future landscape change in the Kalahari Desert under climate and land-use change.

• Uses a coupled vegetation and sediment transport model combining ecological and geomorphological processes.

• Adopts a social-ecological modelling framework.

• Ecological processes are parameterised, introducing uncertainty through simplified assumptions.

• Limited treatment of social complexity highlights the trade-off between realism and simplicity.

Manson, 2007

• Explores the difficulties of evaluating complex geographical models.

• Introduces equifinality, where different processes produce identical spatial patterns.

• Warns that models often mistake observed patterns for underlying causal processes.

• Highlights the science-policy gap: policymakers seek certainty that models cannot provide.

• Argues complex models should primarily be used for exploration rather than prediction.