Keywords
ecosystem service flows, probabilistic modelling, span, service path attribution networks, uncertainty propagation, bayesian networks
Start Date
1-7-2012 12:00 AM
Abstract
Ecosystem service models are increasingly in demand for decision making.However, the data required to run these models are often patchy, missing, outdated, oruntrustworthy. Further, communication of data and model uncertainty to decision makersis often either absent or unintuitive. In this work, we introduce a systematic approach toaddressing both the data gap and the difficulty in communicating uncertainty through astochastic adaptation of the Service Path Attribution Networks (SPAN) framework. TheSPAN formalism assesses ecosystem services through a set of up to 16 maps, whichcharacterize the services in a study area in terms of flow pathways between ecosystemsand human beneficiaries. Although the SPAN algorithms were originally defineddeterministically, we present them here in a stochastic framework which combines probabilisticinput data with a stochastic transport model in order to generate probabilisticspatial outputs. This enables a novel feature among ecosystem service models: the abilityto spatially visualize uncertainty in the model results. The stochastic SPAN model cananalyze areas where data limitations are prohibitive for deterministic models. Greateruncertainty in the model inputs (including missing data) should lead to greater uncertaintyexpressed in the model’s output distributions. By using Bayesian belief networksto fill data gaps and expert-provided trust assignments to augment untrustworthy or outdatedinformation, we can account for uncertainty in input data, producing a model thatis still able to run and provide information where strictly deterministic models could not.Taken together, these attributes enable more robust and intuitive modelling of ecosystemservices under uncertainty.
Modelling Ecosystem Service Flows under Uncertainty with Stochastic SPAN
Ecosystem service models are increasingly in demand for decision making.However, the data required to run these models are often patchy, missing, outdated, oruntrustworthy. Further, communication of data and model uncertainty to decision makersis often either absent or unintuitive. In this work, we introduce a systematic approach toaddressing both the data gap and the difficulty in communicating uncertainty through astochastic adaptation of the Service Path Attribution Networks (SPAN) framework. TheSPAN formalism assesses ecosystem services through a set of up to 16 maps, whichcharacterize the services in a study area in terms of flow pathways between ecosystemsand human beneficiaries. Although the SPAN algorithms were originally defineddeterministically, we present them here in a stochastic framework which combines probabilisticinput data with a stochastic transport model in order to generate probabilisticspatial outputs. This enables a novel feature among ecosystem service models: the abilityto spatially visualize uncertainty in the model results. The stochastic SPAN model cananalyze areas where data limitations are prohibitive for deterministic models. Greateruncertainty in the model inputs (including missing data) should lead to greater uncertaintyexpressed in the model’s output distributions. By using Bayesian belief networksto fill data gaps and expert-provided trust assignments to augment untrustworthy or outdatedinformation, we can account for uncertainty in input data, producing a model thatis still able to run and provide information where strictly deterministic models could not.Taken together, these attributes enable more robust and intuitive modelling of ecosystemservices under uncertainty.
