**4. Conclusions and implications for biodiversity conservation**

Recent developments in global freshwater habitat and biodiversity mapping products (and the rate at which they are updated) is encouraging for future conservation efforts. Assessing the conservation status of species and prioritizing areas of the globe for protection will continue to rely on spatially comprehensive and contiguous inventories of habitats, the biota they support, and evaluation of the degree of alteration at progressively higher spatial resolutions. Metrics are needed that translate anthropogenic stressors into meaningful measures of global habitat alterations in to freshwater systems. Depicting these relationships is challenging for freshwater ecosystems because they are inherently tied to upstream landscape processes. In turn, the field of trait biogeography shows promise in providing a predictive template to convert habitat alterations into specific biodiversity concerns.

While many nations have their own freshwater mapping initiatives conducted at relatively high resolutions (e.g., the US's NHD and NatureServe projects), many underdeveloped nations experiencing intense pressures from development (e.g., Brazil) are likely to rely on external globally-derived products to inform conservation efforts. Even so, local conservation efforts require more spatial fidelity to guide future development pathways. In particular, the Amazon basin is experiencing rapid hydropower development without proper knowledge of the full diversity and geography of fish, invertebrates, and amphibians, or the strategies needed to prevent extinction of these organisms during energy expansion [72]. The development and justification of global reserves for biodiversity conservation will also be contingent upon the accuracy and resolution of aquatic habitats and the organisms they support. New advances in our observation of earth (e.g. through remote sensing), provide opportunities for filling some of these gaps; however, understanding global biodiversity patterns at high resolutions will require exploring local knowledge bases and building predictive models before they disappear.
