Final Activity Report Summary - ISFREM (Improved spatial frameworks for river environment management)
This project undertook research of environmental classifications derived for use as spatial frameworks for river environmental management. Environmental classifications are spatially explicit (i.e. mapped) surrogates for patterns in ecological characteristics of interest to resource and conservation managers (such as environmental regimes and species composition).
The first project compared the performance (i.e. the degree to which classes discriminate areas with similar ecological character) of two classifications of the rivers of France; a conventional geographic regionalisation ('hydro-ecoregions') developed in accordance with the European Water Framework Directive and a numerical classification of segments of a digital river network. Both classifications were derived from a common dataset describing climate, topography and geology of France and were tested with independent water chemistry, invertebrate and hydrological data. The test results showed that there was little difference in the performance of the two classifications. The subsequent studies attempted to increase performance of network and other numerical classifications using statistical modelling. A classification of natural flow regimes of France was defined by first classifying gauging stations based on their flow records. Statistical modelling was used to predict class membership for all segments of a digital river network based on environmental variables. Tests of this spatial framework indicated that it had significantly better predictive performance than the hydro-ecoregions. However, despite strong scientific support for concept of the natural flow regime as the 'master variable' controlling ecological patterns in rivers, tests showed that flow regime classes had very weak relationships with fish assemblages observed at 297 sites throughout France. This result did not support the use of flow regime classifications as spatial frameworks for managing river flows.
Four subsequent projects applied various forms of statistical modelling to the definition of environmental classifications for terrestrial regions in Switzerland and rivers in Spain and New Zealand. These studies showed that statistical modelling (i.e. using a small dataset to 'train' a classification) can produce significant increases in performance of classifications. Thus, the research developed a progression from subjective to more objective techniques to define classifications for both river and terrestrial environments and demonstrated the potential performance gains in a variety of environmental contexts. Seven papers were generated and a software library comprising various functions to test and define classifications.
The first project compared the performance (i.e. the degree to which classes discriminate areas with similar ecological character) of two classifications of the rivers of France; a conventional geographic regionalisation ('hydro-ecoregions') developed in accordance with the European Water Framework Directive and a numerical classification of segments of a digital river network. Both classifications were derived from a common dataset describing climate, topography and geology of France and were tested with independent water chemistry, invertebrate and hydrological data. The test results showed that there was little difference in the performance of the two classifications. The subsequent studies attempted to increase performance of network and other numerical classifications using statistical modelling. A classification of natural flow regimes of France was defined by first classifying gauging stations based on their flow records. Statistical modelling was used to predict class membership for all segments of a digital river network based on environmental variables. Tests of this spatial framework indicated that it had significantly better predictive performance than the hydro-ecoregions. However, despite strong scientific support for concept of the natural flow regime as the 'master variable' controlling ecological patterns in rivers, tests showed that flow regime classes had very weak relationships with fish assemblages observed at 297 sites throughout France. This result did not support the use of flow regime classifications as spatial frameworks for managing river flows.
Four subsequent projects applied various forms of statistical modelling to the definition of environmental classifications for terrestrial regions in Switzerland and rivers in Spain and New Zealand. These studies showed that statistical modelling (i.e. using a small dataset to 'train' a classification) can produce significant increases in performance of classifications. Thus, the research developed a progression from subjective to more objective techniques to define classifications for both river and terrestrial environments and demonstrated the potential performance gains in a variety of environmental contexts. Seven papers were generated and a software library comprising various functions to test and define classifications.