ABSTRACT
Environmental policy and management is increasingly required to optimise both the use of ecosystems for commodity supply and the maintenance of ecosystem function and viability to ensure continuity of life supporting services. This optimisation can not be achieved without access to improved information on ecosystem state and performance at landscape scales. We consider here approaches to delivering enhanced information by combining remote sensing with ecological modelling and prediction. The advantages and limitations of statistical, process-based, and knowledge-based models are discussed, using examples from two important issues of international environmental concern: forest biodiversity, and carbon sequestration. Current limitations on the availability and accuracy of spatial information to address these and similar issues are identified. Possibilities for removing these limitations through advances in optical and radar remote-sensing technology are presented, and also reviewed are recent advances in the processing of remotely sensed data to deliver more accurate biophysical information. We conclude that although much progress can yet be made in ecosystem classification, modelling, and forecasting, more attention must be given to removing the shortcomings in the scale and accuracy of spatial data, and also in the specification of spatial error-particularly for the climate and soils spatial datasets that underpin ecological analysis at landscape scales. Also required is more investigation of the advantages and limitations of the various analysis and modelling approaches applied to ecological datasets, given the apparent divergence in results when differing analysis methods are applied to the same dataset.
