Conservation science in rivers is still lagging behind its terrestrial and marine counterparts, despite increasing threats to freshwater biodiversity and extinction rates being estimated as five times higher than in terrestrial ecosystems. Internationally, most protected rivers have been assigned reserve status in the framework of terrestrial conservation plans, neglecting catchment effects of disturbance. While freshwater conservation tools are mainly index based (e.g. richness, rarity), modern terrestrial and marine conservation planning methods use complementarity-based algorithms - proven to be most efficient at protecting a large number of taxa for the least cost. The few complementarity-based lotic conservation efforts all use broad river classifications instead of biota as targets, a method heavily disputed in the literature. They also ignore current condition and future vulnerability. It was the aim of this thesis to develop a framework for conservation planning that: a) accounts for the connected nature of rivers b) is complementarity based and uses biota as targets c) integrates current status and future vulnerability I developed two different approaches using macroinvertebrate datasets from Australia, Canada and the USA. The first new method was a site/based two-tiered approach integrating condition and conservation value, based on RIVPACS/AUSRIVAS � a modelling technique that predicts macroinvertebrate composition. The condition stage assesses biodiversity loss by estimating a site-specific expected assemblage and comparing it to the actual observed assemblage. Sites with significant biodiversity loss are flagged for restoration, or other management actions. All other sites progress to the conservation stage, in which an index of site-specific taxonomic rarity is calculated. This second index (O/E BIODIV) assesses the number of rare taxa (as defined by &lt50% probability of occurrence). Using this approach on a dataset near Sydney, NSW, Australia, I was able to identify three regions: 1) an area in need of restoration; 2) a region of high conservation value and 3) an area that had high conservation potential if protection and restoration measures could counteract present disturbance. However, a second trial run with three datasets from the USA and Canada highlighted problems with O/E (BIODIV). If common taxa are predicted at lower probabilities of occurrence (p&lt50%) because of model error, they enter the index and change O/E (BIODIV). Therefore, despite an attractive theoretical grounding, the application of O/E (BIODIV) will be restricted to datasets where strong environmental gradients explain a large quantity of variation in the data and permit accurate predictions of rare taxa. It also requires extensive knowledge of regional species pools to ensure that introduced organisms are not counted in the index. The second approach was a proper adaptation of terrestrial complementarity algorithms and an extension to the Irreplaceability-Vulnerability framework by Margules and Pressey (2000). For this large-scale method, distributions for 400 invertebrate taxa were modeled across 1854 subcatchments in Victoria, Australia using Generalised Additive Models (GAMs). The best heuristic algorithm to estimate conservation value was determined by calculating the minimum area needed to cover all 400 taxa. Solutions were restricted to include rules for the protection of whole catchments upstream of a subcatchment that contained the target taxon. A summed rarity algorithm proved to be most efficient, beating the second best solution by 100 000 hectares. To protect 90% of the taxa, only 2% of the study area need to be protected. This increases to 10% of the study area when full representation of the targets is required. Irreplaceability was calculated by running the heuristic algorithm 1000 times with 90% of the catchments randomly removed. Two statistics were then estimated: f (the frequency of selection across 1000 runs) and average c (contribution to conservation targets). Four groups of catchments were identified: a) catchments that have high contributions and are always selected; b) catchments that have high contributions and are not always selected; c) catchments that are always chosen but do not contribute many taxa; d) catchments that are rarely chosen and did not contribute many taxa. Summed c, the sum of contributions over 1000 runs was chosen as an indicator of irreplaceability, integrating the frequency of selection and the number of taxa protected. Irreplaceability (I) was then linked to condition (C) and vulnerability (V) to create the ICVframework for river conservation planning. Condition was estimated using a stressor gradient approach (SGA), in which GIS layers of disturbance were summarised to three principal axes using principal components analysis (PCA). The main stressor gradient � agriculture � classified 75% of the study area as disturbed, a value consistent with existing assessments of river condition. Vulnerability was defined as the likelihood that land use in a catchment would intensify in the future. Hereby current tenure was compared to land capability. If a catchment would support a land use that would have a stronger effect on the rivers than its current tenure, it was classified as vulnerable. 79% of catchments contained more than 50% vulnerable land. When integrating the three estimators in the ICV-framework, seven percent of catchments were identified as highly irreplaceable but in degraded condition. These were flagged for urgent restoration. Unprotected, but highly irreplaceable and highly vulnerable catchments that were still in good condition made up 2.5% of the total area. These catchments are prime candidates for river reserves. The ICV framework developed here is the first method for systematic conservation planning in rivers that is complementarity-based, biota-driven but flexible to other conservation targets and accounts for catchment effects, thus fulfilling all the gaps outlined in the aims.