Browsing by Author "Smith, Matthew J."

Now showing 1 - 2 of 2
  • No Thumbnail Available
    Item
    An ensemble of spatially explicit land-cover model projections : prospects and challenges to retrospectively evaluate deforestation policy.
    (2017) Bradley, Andrew V.; Rosa, Isabel Maria Duarte; Brandão Júnior, Amintas; Crema, Stefano; Dobler, Carlos; Moulds, Simon; Ahmed, Sadia E.; Carneiro, Tiago Garcia de Senna; Smith, Matthew J.; Ewers, Robert M.
    Ensemble techniques, common in many disciplines, have yet to be fully exploited with spatially explicit projections from land-change models. We trial a land-change model ensemble to assess the impact of policies designed to conserve tropical rainforest at the municipality scale in Brazil, noting the achievements made and challenges ahead. Four spatial model frameworks that were calibrated with the same predictor variables produced 21 counterfactual simulations of the actual landscape. Individual projections with a uniform calibration period gave estimates that between 29 and 68% of the simulated deforestation was saved, but lacked an uncertainty estimate, whilst batch projections from two different model frameworks provided a more dependable mean estimate that 38 and 49% deforestation was prevented with an uncertainty range of 1900 and 1000 km2. The consensus ensembles used agreement between the projections and found that the seven examples with a uniform calibration period produced an error margin of ±435.94 km2 and a prevented forest loss estimate of 50%. Using all 21 projections with diverse calibration periods improved these errors to ±179.26 km2 with a 53% estimate of prevented forest loss. Whilst we achieved a method of combining projections of different frameworks to reduce uncertainty of individual modelling frameworks, demonstrating a control model and accounting for non-linear conditions are challenges that will provide better confidence in this method as an operational tool. Such retrospective evidence could be used to make timely rewards for efforts of governments and municipalities to support tropical forest conservation and help mitigate deforestation.
  • No Thumbnail Available
    Item
    SimiVal, a multi-criteria map comparison tool for land-change model projections.
    (2016) Bradley, Andrew V.; Rosa, Isabel Maria Duarte; Pontius Junior, Robert G.; Ahmed, Sadia E.; Araújo, Miguel Bastos; Brown, Daniel G.; Brandão Júnior, Amintas; Câmara, Gilberto; Carneiro, Tiago Garcia de Senna; Hartley, Andrew J.; Smith, Matthew J.; Ewers, Robert M.
    The multiple uses of land-cover models have led to validation with choice metrics or an ad hoc choice of the validation metrics available. To address this, we have identified the major dimensions of land-cover maps that ought to be evaluated and devised a Similarity Validation (SimiVal) tool. SimiVal uses a linear regression to test a modelled projection against benchmark cases of, perfect, observed and systematicbias, calculated by rescaling the metrics from a random case relative to the observed, perfect case. The most informative regression coefficients, p-value and slope, are plot on a ternary graph of ‘similarity space’ whose extremes are the three benchmark cases. SimiVal is tested on projections of two deliberately contrasting land-cover models to show the similarity between intra- and inter-model parameterisations. We find metrics of landscape structure are important in distinguishing between different projections of the same model. Predictive and exploratory models can benefit from the tool.