Degradation of floodplain integrity within the contiguous United States

Floodplains provide critical ecosystem functions and services, including flood reduction, groundwater storage, sediment regulation, organics and solute regulation, and habitat provisioning, as well as supporting agriculture, fisheries, recreation, and cultural values. However, they are among the most endangered ecosystems and have been heavily modified by human activities (e.g., dams, levees, channelization, dredging, agriculture, and urbanization). In the U.S., over 30% of floodplains have been cultivated or developed, and most rivers have been impacted by human activities. Despite their importance and extensive modification, there has been no comprehensive, national-scale assessment of human-caused floodplain degradation (loss of floodplain integrity) across the contiguous U.S. Prior efforts often focused on biological condition at catchment scales rather than holistic floodplain functionality. This study addresses that gap by assessing the influence of human stressors on five key floodplain functions—flood reduction, groundwater storage, sediment regulation, organics and solute regulation, and habitat provisioning—across the contiguous U.S., aiming to identify spatial patterns of degradation and inform restoration targeting.

The paper situates floodplain integrity within a body of research documenting floodplains' ecological and socio-economic benefits and the widespread human alterations that impair these systems. Prior work has characterized impacts from dams, levees, channel straightening, dredging, and land-use change, and documented extensive modification of U.S. rivers and floodplains. Existing assessments often focus on riparian biological condition or watershed integrity indices at catchment scales, with limited emphasis on comprehensive floodplain functional integrity across large geographies. While studies have evaluated specific functions or regional conditions (e.g., Puget Sound), a national, function-based integrity assessment for floodplains has been lacking. The authors leverage and extend prior integrity frameworks by applying an Index of Floodplain Integrity (IFI) nationally, using mechanistic, stressor-based geospatial indicators tied to specific functions, thereby enabling human-environment linkages relevant to management.

Study area and floodplain delineation: The authors assessed the contiguous United States (CONUS). Floodplain boundaries were obtained from a 30 m resolution shapefile of the 100-year (1% annual chance) undefended floodplain derived from 2D hydrodynamic modeling and regionalized flood frequency estimates. The dataset was processed by removing isolated pixel groups and filling small gaps (≤ 2700 m²), reducing the overall area from 740,967 km² to 721,799 km². Floodplains were subdivided by 12-digit hydrologic unit code (HUC12) boundaries within the U.S., producing 78,304 floodplain units totaling 662,566 km² (average unit area 8.46 km²). Maximum Strahler stream order per unit was determined using NHDPlusV2 flowlines; 2,855 units did not intersect flowlines due to dataset discrepancies. Stressor datasets and functional linkages: Publicly available, nationally consistent geospatial datasets representing anthropogenic stressors were compiled and assigned to five floodplain functions. Examples include buildings (reduced storage), leveed areas (disconnection), roads/railroads (overland flow interception), forest cover loss events (loss of wood/vegetation), developed area and impervious surface (land cover change and impermeability), canals/ditches (channelized overland flow), agricultural area (colmation and land cover change), groundwater wells (lowered water table), non-native introduced vegetation (species invasion), and hydrologic alteration (probability of change in peak flows; MH20 metric). Dataset sources included NLCD 2019, LANDFIRE EVT, OpenStreetMap, USACE National Levee Database, NHDPlusV2, USGS Groundwater Watch, and a national hydrologic alteration product. Stressor density calculations and rescaling: Within each floodplain unit, stressor prevalence was quantified as counts per km² (points), length per km² (lines), or area fraction (km² per km²; rasters). Raster-specific procedures included: agricultural area as % of NLCD classes 81 and 82; developed area as % of NLCD classes 22–24; forest cover loss as % of cells with loss events (2000–2020); percent imperviousness averaged over cells; non-native vegetation as % of relevant LANDFIRE EVT groups; hydrologic alteration as the average MH20 value for maximum-stream-order segments intersecting the unit. Stressor densities were rescaled to 0–1 where 0 indicates absence and 1 corresponds to the CONUS 90th percentile of stressor amount; for canals/ditches, leveed area, and groundwater wells, rescaling used the maximum observed value since the 90th percentile was zero. Correlation screening: Pearson correlations among stressor datasets were computed within each functional grouping. When two datasets within a function were correlated above 0.7, only one was retained to avoid overweighting. Functional IFI computation: For each floodplain unit and function, the functional IFI was computed as 1 minus an aggregate of the rescaled stressor densities, assuming a negative linear relationship between stressor prevalence and integrity and equal weight for each stressor included in that function. This yields functional IFI values between 0 (functional loss) and 1 (little degradation). Overall IFI: The overall IFI per unit was calculated as the geometric mean of the five functional IFIs, emphasizing that failure (zero) in any function drives the overall integrity to zero. Statistical and comparative analyses: Distributions were non-normal (Kolmogorov–Smirnov). Differences across categories were evaluated with non-parametric tests: Mann–Whitney (urban vs rural), Kruskal–Wallis with Dunn post hoc (stream order, ecoregions), with Holm/Bonferroni adjustments not affecting significance. IFI values were analyzed by urban-rural classification, maximum stream order, and aggregated Omernik Level III ecoregions. Comparisons to external datasets included the Index of Watershed Integrity (IWI) at national scale and functional classes from Konrad’s Puget Sound analysis; the latter required reprojecting, clipping, and aggregating class-based scores to 0–1 for comparison. All spatial processing used ArcMap and R (sf, terra, dplyr, etc.). Data and code are publicly available at the cited Dryad DOI.

- National IFI patterns: Floodplain degradation is concentrated in the southeastern U.S., along major river corridors (e.g., Mississippi River), and in California’s Central Valley. The IFI distribution indicates that 68% of total floodplain area in the CONUS is in poor condition based on an IFI threshold of 0.7. - Overall statistics (CONUS): Median overall IFI = 0.762; mean = 0.740; standard deviation = 0.152. Functional IFIs are highly correlated (0.7–0.91). Among functions, sediment regulation has the lowest median IFI (0.697), while flood reduction has the highest median IFI (0.893). Functional means: flood reduction 0.831; groundwater storage 0.862; sediment regulation 0.676; organics/solutes regulation 0.675; habitat provision 0.699. - Subregional assessments: • San Joaquin subregion: Floodplain area 8,248 km² across 435 units; average overall IFI = 0.691. Lower IFI concentrated in western, densely populated and agriculturally intensive areas (e.g., Salinas Valley). • Missouri-White subregion: Floodplain area 1,740 km² across 594 units; average overall IFI = 0.860, reflecting predominantly grassland/herbaceous cover and fewer anthropogenic alterations. • Lower Mississippi–Yazoo subregion: Floodplain area 17,969 km² across 391 units; average overall IFI = 0.561. Over 68% of floodplain intersects levees, impairing connectivity, flood reduction, and sediment regulation; functional IFIs show spikes at lower values for flood reduction and sediment regulation. - Urban vs rural: Floodplains in urban areas (n=9,352) are significantly more degraded than rural (n=58,565) (P<0.0001). Urban IFI: median 0.557, mean 0.561, SD 0.142; Non-urban IFI: median 0.790, mean 0.769, SD 0.127. - Stream order: No significant differences in mean IFI between most adjacent stream orders except between 3rd and 4th; general trend of decreasing IFI with increasing stream order, especially >8, though some orders with small sample sizes deviate. - Ecoregions: Significant differences among nine aggregated ecoregions. Highest IFI in Northern Plains (median 0.878; mean 0.855; SD 0.079); lowest in Coastal Plain (median 0.625; mean 0.607; SD 0.142). Regions with high IFI align with high grassland/shrub or forest cover; regions with low IFI align with high agricultural development. - External comparisons: No meaningful relationship between overall IFI and IWI at national scale. Compared to Konrad’s Puget Sound analysis, IFI estimates were generally higher for flood reduction and sediment regulation and similar for habitat provisioning across overlapping floodplain areas (~2,338 km²).

Applying the IFI nationally provides a quantitative, spatially explicit assessment of floodplain functional integrity, revealing heterogeneous degradation patterns and enabling identification of areas most in need of restoration. Lower IFI values in higher-order streams and urban areas are consistent with historical concentration of human development along larger rivers and within cities, reinforcing the negative relationship between anthropogenic modification and floodplain integrity. The congruence between IFI patterns and watershed integrity patterns by ecoregion suggests IFI captures broader river corridor condition gradients, although watershed integrity is not a direct proxy for floodplain function. The methodology’s flexibility allows replication at finer spatial scales to guide local restoration planning. Comparative analysis indicates potential overprediction of integrity for certain functions relative to regional, classification-based studies, implying that finer-resolution stressor data may be needed in areas where IFI predicts high integrity. Overall, the IFI framework demonstrates utility for strategic prioritization and for understanding linkages between stressors and function-specific degradation across large geographies.

This study delivers the first CONUS-wide, function-based Index of Floodplain Integrity, demonstrating pervasive human-driven degradation, with approximately two-thirds of floodplain area exhibiting poor integrity by the chosen threshold. The analysis identifies spatial hotspots of impairment (e.g., southeastern U.S., Mississippi corridor, Central Valley) and clarifies how urbanization, higher stream order, and ecoregional land use relate to reduced integrity. The approach and datasets provide a replicable, adaptable framework to prioritize restoration and conservation by function and location. Future research should focus on: integrating additional or finer-resolution stressor data (e.g., bank stabilization, pesticides, legacy wood removal, beaver extirpation), incorporating off-floodplain stressors, developing empirically supported non-linear stressor–function relationships and thresholds, and conducting localized assessments to validate and refine the index for site-scale decision-making.

- Data availability and representation: Some relevant stressors are missing or coarsely represented due to lack of national datasets or legacy effects (e.g., pesticides, non-native vegetation in some contexts, beaver extirpation, historical removal of large wood, bank stabilization, watershed land cover history). - Spatial scope of stressors: IFI accounts only for stressors within floodplain boundaries, not adjacent or upstream stressors that can influence floodplain functions. - Assumed linearity and equal weighting: The negative linear relationship between stressor density and function, and equal weighting among stressors, oversimplify complex, potentially non-linear, threshold-governed processes; this may affect magnitude and range of IFI values. - Validation constraints: Scarcity of quantitative, spatially comparable floodplain condition datasets and scale incongruities impede robust validation; comparison with Konrad suggests possible overprediction for certain functions. - Resolution and generalization: National-scale datasets may lack the spatial granularity needed to capture local conditions; IFI results are best interpreted for broad patterns and prioritization rather than local diagnostics without further, finer-scale analysis.