The evolution of dam induced river fragmentation in the United States

Rivers across the contiguous United States are heavily impacted by dams, which fragment networks and alter streamflow, affecting the transport of organisms, nutrients, sediments, and energy. Global-scale fragmentation assessments have relied on databases that include only large reservoirs (>100 million cubic meters), omitting most U.S. dams. In the U.S., there are over 50,000 medium to large dams that meet National Inventory of Dams (NID) criteria, but only 1,945 appear in global datasets. The study aims to quantify how river fragmentation and regulation have evolved over time in the U.S. by explicitly incorporating mid-sized dams, assessing trends in degree of regulation (DOR), fragment counts, and fragment lengths, and evaluating regional patterns across aridity gradients. The purpose is to reveal the full extent of anthropogenic fragmentation, especially the cumulative impacts of mid-sized dams that are underrepresented in global studies, and to understand how dam construction has reshaped natural connectivity patterns.

Prior global studies show that 63% of the world’s longest rivers are no longer free-flowing and 48% of river reaches have diminished connectivity, but acknowledge conservative results due to the focus on large dams (>100 MCM). The U.S. consistently appears as heavily regulated in these global assessments. However, global databases omit the majority of U.S. structures. Historical estimates suggest up to 2 million (mostly small) dams may have been built in the U.S., with state-level records indicating national datasets capture only a fraction of total dams (e.g., 17% in Wisconsin, 6% in Utah). Previous work using the National Anthropogenic Barrier Dataset (NABD) examined dam impacts on fish assemblages and computed fragmentation metrics, but often did not track temporal evolution or fully include mid-sized dams. This study builds on those foundations by integrating NABD with NHDPlusV2 to assess temporal and spatial changes in fragmentation and regulation, explicitly contrasting medium and large dams and highlighting how prior large-dam-only approaches underrepresent human impacts.

Study scope included 51,923 medium to large dams in the contiguous U.S. derived from the National Anthropogenic Barrier Dataset (NABD), itself a revision of the National Inventory of Dams (NID). Dams meeting NID thresholds but with storage <100 million m³ were classified as medium; ≥100 million m³ as large. NABD entries were cleaned in Python to remove duplicates, correct identifiers, and cross-checked against GRanD; missing large dams from GRanD were added. Approximately 6,300 dams lacking completion year were assigned to pre-1920 to enable temporal analyses, acknowledging spatially uneven impacts of this assumption. River networks and attributes were taken from NHDPlusV2 (≈2.6 million flowlines), using COMID to spatially link dams to flowlines. Key NHDPlusV2 attributes included segment length, hydrosequence (for upstream/downstream traversal), headwater flags, stream order, hydrologic unit codes (HUC), and estimated natural mean flow. Fragment identification algorithm: For a given analysis year (pre-development, 1920, 1950, 1980, 2010), dams completed after that year were excluded. The algorithm identified all headwater segments and traversed downstream for each, stopping when encountering (1) an already-processed fragment, (2) an anthropogenic barrier (dam), or (3) a natural terminal point (ocean, lake, internal sink). Fragments were labeled with unique IDs: fragments terminating at dams received IDs tied to the dam; natural fragments received incrementing IDs. This produced complete labeling of all network segments into natural or anthropogenic fragments for each time slice and basin. Metrics: Fragment density was computed as number of fragments divided by basin area (both by major river basins and HUC8). The relative contribution of medium vs large dams to fragmentation was estimated by comparing runs that included only large dams versus all dams. Degree of Regulation (DOR) quantified regulatory potential as reservoir normal storage divided by average annual natural flow, accumulated over upstream storage for each reach, following prior literature. DOR maps were produced for each time slice to show expansion of regulation into tributaries and headwaters. Analyses were summarized by major river basins (HUC2) and aridity gradients, including fragment length classes (<10 km, 10–100 km, 100–1,000 km, >1,000 km) and exceedance probabilities of fragment lengths for scenarios with no dams, large dams only, and all dams.

• Mid-sized dams dominate structure counts: They represent 96% of major inventoried structures nationally and at least 80% of structures in every major basin; in the Mississippi basin they comprise 97% of structures. - Storage share: Mid-sized dams account for 48% of total reservoir storage nationally; their contribution is as high as ~80% in the Rio Grande and ~50% in the Columbia. - Regulation increased steadily: DOR rose from mostly <0.2 pre-1920 to widespread >0.4 by 1980, with extensive reaches >1 by 2010, reflecting marked growth in regulatory potential across mainstems, tributaries, and headwaters. - Medium dams’ regulatory role: Medium dams contribute at least 10% of total DOR in every region, often ≥40% along the East Coast, and >80% of headwater regulation in systems such as the Colorado, Columbia, and Missouri. Their relative contribution generally decreases downstream and westward. - Fragmentation patterns reversed: Pre-development, arid and coastal regions had the highest natural fragment densities due to ephemeral flows and terminal sinks. By 2010, humid basins (e.g., North/South Atlantic, Gulf Coast, Mississippi) became among the most fragmented, reflecting dam construction and flood-control needs; arid basins show increases but less dramatic due to fewer, larger reservoirs. - Medium dams drive fragmentation: They accounted for ~40% of fragments in 1920, increasing to ~73% by 2010. Nationally, mid-sized dams are responsible for 73% of anthropogenically created fragments and disproportionately create short fragments (<10 km). - Short fragments proliferated: Fragment densities increased across all length classes, with the largest increases for fragments <10 km, especially in humid basins. Medium dams account for >70% of fragments <100 km nationally; in Gulf Coast and Mississippi basins, medium dams contribute ~80% of these short fragments, whereas in arid basins like California, Great Basin, and Rio Grande their contribution to <100 km fragments can be ~20%. - Loss of long fragments: Exceedance analyses show a systematic shift away from long river fragments (connected habitats), more pronounced when including medium dams than when considering large dams alone. - Regional nuances: In some southwestern mainstems (lower Colorado and Rio Grande), medium dam impacts increase downstream due to tributary regulation feeding into downstream reaches; aridity allows medium dams to operate over larger drainage areas due to lower flows.

By explicitly including mid-sized dams, the study reveals substantially greater anthropogenic fragmentation and regulation than suggested by large-dam-only global assessments. The findings address the research question by demonstrating that mid-sized dams are the principal drivers of increased fragment counts and shortening of fragment lengths, especially in headwaters and tributaries that are critical for biodiversity and ecosystem function. The reversal of natural fragmentation patterns—where humid basins now exhibit the highest fragment densities—highlights the interplay between water availability, infrastructure purposes (flood control, hydropower, water supply), and human settlement. Ecologically, the proliferation of short fragments and loss of long, connected reaches indicate reduced habitat connectivity and potential declines in freshwater biodiversity, reinforcing concerns about migratory species and headwater–mainstem linkages. The growth in DOR across networks underscores the broad potential for altered flow regimes, though actual ecological outcomes depend on dam operations and mitigation measures. Overall, accounting for mid-sized dams is essential for accurate national-scale assessment of river network health and for informing restoration, dam removal, and management strategies.

This national-scale evaluation incorporating more than 50,000 dams shows that mid-sized dams, though often omitted from global datasets, are the dominant contributors to river fragmentation and a major source of regulatory potential in U.S. rivers. Dam construction has fundamentally reshaped connectivity, reversing natural patterns such that humid basins are now the most fragmented, and greatly increasing the prevalence of short fragments while reducing long connected reaches. The work underscores the need to integrate mid-sized structures into policy, conservation, and restoration planning. Future research should develop more comprehensive inventories that include small and very small barriers, compile accessible national datasets on dam purposes and operations, and integrate information on fish passage and other mitigation structures to better quantify actual ecological impacts and guide effective management and removal prioritization.

• Database coverage: NABD/NID focus on structures with higher hazard or size thresholds; millions of small and very small barriers are likely unmapped, making results conservative. - Missing metadata: Roughly 6,300 dams lacked construction years and were assigned pre-1920 dates; uneven spatial distribution of these records may bias temporal fragmentation patterns locally. - Operations unknown: DOR reflects potential regulation based on storage-to-flow ratios; actual impacts depend on operational policies, reservoir purposes (e.g., flood control vs water supply), and interannual variability. - Ecological passage not resolved: Presence/efficacy of fish ladders and other passage structures are not comprehensively captured, potentially moderating realized connectivity impacts. - Regional data heterogeneity: State-level inventories may contain additional dams not present in national datasets, and coastlines were filtered to avoid flowline chaining artifacts, which may affect local fragment delineations.