4.03.01 – Uncertainty analyses of spatial and non-spatial models

Mathematical models are widely used for assessing sustainability.  They are used to predict the future state of forest systems and for testing hypotheses. Models are now frequently used to answer many questions. Is a desired forested landscape sustainable today and in the foreseeable future? What is the influence of climate on sustainability of tree species in a mountainous environment? What is the current size and health of forest resources of a nation or nations? What is the size and long term viability of certain wildlife populations that reside in forests fragmented by urban sprawl? Models are being used by everyone, from small landowners in People’s Republic of China using a growth and yield model to manage their woodlots based on standards of sustainability needed for forest certification, to the Intergovernmental Panel on Climate Change (IPCC) decision makers determining environmental policy relating to global warming using complex ecosystem models.

Needless to say, for real world systems, projections made with the simplest to the most complicated model have statistical errors and uncertainties.  For many ecological and environmental models, there can be hundreds of sources of uncertainties due to measurements, sampling, knowledge gaps, parameter estimates, multiple temporal and spatial scales, stochasticity, etc. If one doesn't account for the uncertainties, the outcomes from models have little or no value. Assessing consequences of the propagation of uncertainties becomes particularly complex as scientists make spatially explicit projections forward in time, or when they test complex hypotheses based on models.

The Working Party goals have been to develop a comprehensive framework to statistically identify and manage error and uncertainty for both non-spatial and geospatial large scale natural resource monitoring and projection systems. Emphasis has been given to geospatial systems, where both ground and remote sensed monitoring systems have been used for inputs of large scale landscape natural resource modeling systems.

Recent emphasis has been given to geospatial systems, where both ground and remote sensed monitoring systems have been used for inputs for large scale landscape natural resource modeling systems. Spatially identifying the sources of uncertainties, modeling their accumulation and propagation, and finally, quantifying them locally and globally for current and future landscape maps has been the focus of the research. A variety of non-spatial and geospatial natural resource systems are being considered including global change models, avalanche hazard models, carbon sequestration models, fragmentation assessment models, forest growth process models, land disturbance and restoration models, plant biodiversity models, spatially explicit soil erosion models, sustainability models, topographic models, urban sprawl models, etc.