Protection of U.S. streams is insufficient to safeguard stream diversity and prevent habitat impairment

Despite covering a relatively small portion of the Earth’s surface, freshwater ecosystems offer habitat for a diverse range of species¹. However, these ecosystems are under distinct threats from human land use activities and climate change, more so than terrestrial ecosystems^2,3. Pressures including pollution, habitat modification and degradation, and overexploitation have been recognized as threats to freshwater systems for several decades², and newer stressors such as emerging pollutants and climate change only exacerbate the strain on these ecosystems⁴. Due to these cumulative threats, the abundance and distribution of 6600 populations and 1400 freshwater species have declined globally, on average, by 83% from 1970–2018⁵. Additionally, of the roughly 10,000 freshwater fish species assessed by the IUCN, 30% are at risk of extinction⁶.

The design and establishment of protected areas is a globally utilized conservation strategy aimed at sustainably managing human activity and preserving biodiversity and ecosystem services⁷. These protected areas can include hard, jurisdictional boundaries placed on lands and river corridors or restrictions on activities in and around streams through regulations, policies, or incentives. However, global ecosystem protection efforts have shown mixed success. As of 2021, approximately 17% of terrestrial areas and 8% of marine areas were protected for conservation, with protection defined as an area designated for nature conservation and implemented through legal or other means⁸. Despite this, area-based protection initiatives are not always effective⁹. In the U.S., for instance, current patterns of protected areas are misaligned with regions of high richness and endemism¹⁰. To combat these concerns, the 2022 Kunming-Montreal Global Biodiversity Framework (GBF) set international goals for conservation actions by 2030 and 2050. Specifically, Target 3 of the GBF sets an objective of protection for 30% of terrestrial and inland water areas of importance to biodiversity by 2030, hereafter referred to as 30 × 30¹¹. Previous national campaigns, such as the America the Beautiful and the Freshwater Challenge, set similar goals in the U.S. The importance of the biodiversity portion of 30 × 30 is essential because protected areas often are not located to maximize coverage of regions of particular biodiversity importance and irreplaceability^3,12. Thus, assessment of the current protection status of areas, both in terms of quantity and conservation efficacy, is vital to measure progress towards the 30 × 30 goal and to inform future protection efforts.

Protected areas are often created with terrestrial or marine ecosystems in mind, but they can confer varying benefits on stream habitats by limiting activities on land draining into a stream¹³. Terrestrial protection exists in different forms, ranging from the degree of allowable human activities to the extraction of natural resources. The U.S. Geological Survey (USGS) Protected Areas Database of the U.S (PAD-US) designates four different categories of land protection throughout the country, called GAP Statuses¹⁴. Land can be completely protected from land cover modification and either preserve a natural state, such as in wilderness areas, or allow management activities that might suppress natural ecosystem processes, like in wildlife refuges (GAP Statuses 1 and 2, respectively). Other protected lands may prevent land cover modification but allow for limited resource extraction, such as timbering and mining (GAP Status 3) or may be listed as protected simply because they are privately owned, even though there are no known restrictions on use (GAP Status 4). Hereafter, we do not consider lands with GAP Status 4 as providing protection¹⁵.

Because streams are complex ecosystems with spatiotemporally variant hydrology, connectivity, material transport, and biophysical properties, the protection of different components of the surrounding landscape (e.g., catchment/watershed, riparian zone/floodplain) offers unique benefits to stream habitat quality^16,17. For example, protection of the headwaters of a river system can stabilize hydrologic flows through the downstream network, thereby providing one form of protection further than the footprint of the protected area¹³. On the other hand, management of a buffer alongside a stream helps preserve the integrity of the riparian zone as a filter for all inputs to the stream as well as a rich ecosystem of its own¹⁸. Protection of the stream corridor itself is another strategy, such as through the National Wild and Scenic Rivers System (hereafter wild and scenic rivers) in the U.S., where extractive or disruptive activities along the stream corridor for designated streams are limited¹⁹. The disturbance of one stream component (e.g., headwaters) could offset the advantages gained from protection of another (e.g., riparian), which we term the stream protection-impairment paradox. For example, a study of 13 streams in an agriculturally dominated watershed before and after increased protection of the streams’ buffer found an increase in physical habitat quality but no improvement in indices of biotic integrity²⁰. Conversely, the benefit of pristine, intact forests within the larger stream watershed can be undermined by wastewater or agricultural discharge in the immediate vicinity of the stream¹³. Because it is well understood that protection alone does not safeguard biodiversity^9,21, this paper seeks to identify locations in the U.S. where this protection-impairment paradox exists to guide conservation efforts and decisions.

Here, we quantify and assess the U.S.’s current state of land- and corridor-based protection for streams and rivers in relation to the GBF’s 30 × 30 goal. We differentiate between protection of the immediate local catchment of a stream and the protection of the larger stream watershed to the headwaters, hereafter referred to as network protection. We primarily focus on quantifying and assessing land-based protection of a stream’s catchment, though the degree of corridor protection is also assessed. Because studies have found protection of the habitat to be critical in the conservation of biodiversity²² and that stream habitat diversity strengthens aquatic communities against anthropogenic stressors²³, we also examine the habitat diversity of protected streams, in terms of physical configuration and biogeography. Furthermore, we examine the paradoxes of stream protection against human-induced impairment, such as localized stressors in streams with high network protection or local protection in streams sustaining heavy watershed stress.

We find that though current stream protection has met the 30 × 30 goal, large portions of protected streams are also impaired by human land use. When accounting for impairment in the form of stream impoundment and regulation, as well as developed or agricultural land use, up to 19% of streams designated as protected also fail thresholds of impairment. Almost 30% of unique stream habitats are protected, yet both the most common and the rarest stream types fall short of our target. This study offers a tool and dataset to support the optimization of siting future protection efforts to safeguard stream habitat health²⁴.

Total protected stream quantification

A total of 2,667,768 stream segments were assessed, representing 5,149,964 km of stream length. Of that length, 108,362 km was impounded, and an additional 2,663,240 km failed one or more of the impairment criteria, leaving a total of 2,378,362 km of free-flowing, unimpaired stream length in the conterminous U.S. (46% of total stream length). To achieve the 30×30 goal, 1,544,990 km of the total stream length must be protected by 2030. We designated streams as protected under six different scenarios of GAP Statuses and the spatial scale of protection considered. Exclusive scenarios consider GAP Statuses 1 and 2, while inclusive scenarios include GAP Statuses 1–3. We also differentiated between the protection of a stream’s local catchment, upstream watershed, or either spatial scale. See the Methods for a more detailed description of these scenarios.

Across all scenarios, protection ranged from 1,763,247 km or 34.2% of total stream length for the inclusive-either scenario to 1,236,413 km or 24% for the exclusive-local scenario with a 30% catchment threshold (Fig. 1, Supplementary Table 1). When including all GAP Statuses, every 20% catchment threshold scenario met the 30 × 30 goal; conversely, none of the exclusive scenarios met the target. The same pattern is seen for catchment and ARA protection, with catchment protection always having more stream length protected. While there are numerical differences between streams with local and network protection, those differences are less noticeable spatially (Fig. 2). The contrast between GAP Statuses 1 and 2 and GAP Status 3 is more tangible spatially (Fig. 1).

When removing streams with potential impairments from designation as being protected, protected stream length falls by 0.4–6.3% with the network protection scenarios being the most affected (Supplementary Table 2). Figure 3 highlights six scenarios notably affected by this paradox. Without considering impairment, the three inclusive scenarios pass the 30 × 30 threshold; with impairment, however, none of the scenarios met the target, and inclusive-either scenario is the most affected by impairment. Total agriculture was the largest source of impairment regardless of scale, followed by regulation in network catchments, although urban development and regulation were comparable in local catchments. Impoundment was consistently the second or third most common impairment and was relatively consistent across the catchment scale. Impairments were more prevalent in network catchments, both in magnitude and proportion. Figure 4 shows the stream length designated as protected that is potentially impaired by land cover, dam regulation, and impoundment in a stream’s local or network catchment (this data is also shown numerically in Supplementary Table 3). For example, the majority of streams with network protection by GAP Statuses 1–2 are locally impaired by agriculture (top right diagram), but that proportion is split with network impairments of regulation (bottom right diagram).

Figure 5 shows the spatial arrangement of impairment and protection, comparing streams with any catchment protection status to those that meet or exceed any of the impairment thresholds. For this comparison, 19% or 331,948 km of protected stream length is also impaired. Despite the seemingly low percentage, the spatial distribution is spread throughout the country. Much of the impairment is located from the Midwest to the East Coast, with most unimpaired protected streams lying in the West and Rocky Mountain regions. Urban impairment is concentrated along the East and California coast, agricultural impairment is prevalent throughout the Midwest and Chesapeake Bay region, and dam regulation extends into the Rocky Mountains.

Values for each anthropogenic disturbance variable were used to represent impairment thresholds, beyond which ecosystem degradation is expected. The main results of this study are based on the mid-range values of thresholds reported in the literature; however, to evaluate the influence of different threshold values on the protection-impairment paradox, a sensitivity analysis was conducted. More conservative thresholds (lower thresholds) yield a higher percentage of paradox streams, whereas more lax thresholds (higher thresholds) yield a lower percentage. Impairments were sensitive to the threshold values (Supplementary Fig. 2, Supplementary Table 10), and the magnitude of impaired streams is dependent on the chosen percentage (Supplementary Table 11). In particular, the 10% threshold for agricultural land use might be overestimating impairment because it considers both row crop and pasture land cover (Table 1). However, more restrictive thresholds for agriculture did not dramatically decrease the percentage of paradox streams (Supplementary Table 10). Based on mid-range values of thresholds, 8–19% of protected streams are impaired, i.e., paradox streams, depending on the protection criteria, which translates to only 11–22% of streams, rather than 30%, being truly protected. When considering the most conservative thresholds, only 9–20% of streams are truly protected, whereas under more lax impairment thresholds, 21–28% of streams are protected. This suggests that any threshold of impairment indicates less than 30% protection of streams.

As a case study, we examined one watershed in Mississippi, in the Tallahala State Wildlife Management Area (Fig. 6). All streams in this watershed were designated as protected due to the GAP Status 1 and 2 lands throughout the management area, but we highlight two paradox streams. Stream A, the furthest downstream point in the watershed, faces local impairment primarily from agricultural use directly surrounding the stream corridor. However, the majority of its upstream network is protected lands. The two streams circled at B illustrate the reverse; they have almost 100% local protection, but there are agricultural impairments immediately upstream from their local catchment.

Stream habitat diversity protection

Stream diversity protection was assessed using the same protection criterion and GAP status categories as the total stream protection assessment, with results shown in detail in the SI (Supplementary Tables 4–9). Importantly, protection was only assessed for free-flowing streams (i.e., not impaired by impoundment or regulation). The most common stream types were unconfined headwater streams with moderate slopes, variably cool or warm temperatures, and intermittent or perennially flashy hydrology²⁴. The ten most prevalent stream types represent 693,928 km (14%) of classified stream length, yet only 100,626 km (15%) of that stream length is protected under GAP Statuses 1 and 2 (Fig. 7a–c). In fact, two thirds of the 100 most prevalent stream types do not have at least 30% protected length (Fig. 7d). Under inclusive scenarios, almost half of all classes have ≥30% of their stream length protected locally or network, and 29.6% of total classified stream length is protected, a deficit of 19,017 km (Fig. 7e, f). However, when removing impaired streams from consideration as protected, 24.9% of classified stream length is protected, 253,675 km under the 30% threshold. Protecting 30% of each stream type would require an additional 779,669 km of stream length, and raising the 10 most common stream types to 30% protection would add 134,950 km, while fully protecting the 4000 rarest classes would offer 72,721 km (Fig. 7f).

Of the 18 HUC 2 regions, six had at least 30% of their stream length protected: the California, Great Basin, Lower and Upper Colorado, Pacific Northwest, and Rio Grande regions. This reflects the regional bias of PAD-US lands to the western U.S. Several other regions would have met the 30% threshold if not for regulation and impoundment. Namely, the Mid Atlantic, Missouri, New England, and Tennessee regions had relatively large amounts of protected stream length, but many of those streams are not free-flowing.

Despite widespread acknowledgment of the necessity of protected area expansion for biodiversity conservation, the ability to achieve targets such as the 30 × 30 goal is limited by shifting legal policies²⁵ and rapid land use change²⁶. Identification of areas best suited for protection, particularly balancing critical habitat and offsetting or minimizing ongoing habitat impairment, poses an additional, complicated challenge^4,27. To help address these obstacles, this study cataloged current protection statuses for streams in the conterminous U.S. as measured by protection of stream catchments and determined how the physical and biogeographical diversity of streams is represented in protection schemes. Importantly, it also compared these schemes with indicators of stream habitat impairment to determine the efficacy of such protection.

A central question of this study is, “What does terrestrial protection entail for stream ecosystems?”. Even though most of the protected areas considered were not created with streams in mind, how do U.S. streams benefit from the restriction of activity in their watersheds? Our study, through its focus on land use/land cover in stream catchments, does not directly measure the effectiveness of protected area management on stream habitats. Rather, it highlights that meeting percentage-based indicators of protection does not safeguard a stream from impairment²¹; in fact, designating an area as protected does not guarantee that the land has not already been anthropogenically modified. Land cover maps show areas where potentially harmful land use overlaps with designated protected areas (Fig. 6), and even watersheds that contain large amounts of protected land may also allow disruptive activities elsewhere. This paradox of protection versus impairment is illustrated by the approximately 19% of stream length categorized as protected that also faced potential impairment from anthropogenic activity in stream catchments, and the Tallahala State Wildlife Management Area case study underscores this paradox, where protection at one scale is juxtaposed with impairment on another. As such, our findings suggest that the sole existence of protected lands within stream watersheds will not safeguard biodiversity, particularly from legacy effects²⁸. This is especially true considering that not all protected lands offer the same potential benefits due to differences in allowed activities (such as between GAP Statuses 1 & 2 and 3) and the location, design, and management of a given area^29,30. Consequently, conservation efforts must focus on the ecological effectiveness of protection rather than percentage- or threshold-based targets alone¹¹.

Streams are important targets for conservation because they are physically diverse ecosystems that host great biodiversity, and protection of these habitats, particularly the rarest, is vital to safeguarding the species within³¹. As could be expected given the terrestrial intent of most protected areas, current protection efforts poorly represent the diversity of the streams; however, the rarest 2000 stream types could be fully protected by adding 30,600 km of protection nationwide (Fig. 7f). One challenge is the likely uneven distribution of these streams across the country which would require coordination across multiple states and agencies to protect—a serious hurdle. Expansion of protection where it already exists or increasing land use restrictions, especially in the hotspots of biodiversity such as the Southeastern and central Californian mountains¹⁰, is one strategy for achieving protection targets. The creation of new protected lands or expansion of already existing areas often runs counter to socioeconomic and political goals²⁵, and actions without targeted and cohesive planning will likely fail to maximize ecological benefits³².

Unfortunately, the practicality of expanding conservation lands sufficient to ‘protect’ the breadth of stream diversity in the U.S. (i.e., increasing protection by 50%) is questionable. In part, this stems from this paper’s perspective of mutual exclusion of ‘protection’ as land used strictly for biodiversity conservation versus that used for meeting human needs, i.e., food production and resource extraction, presumably leading to stream impairment. For over three decades, scientific debate has transpired on how to use land to support both biodiversity and human needs, particularly in agricultural landscapes³³. These two opposing endpoints have since been formally reframed as alternative strategies, i.e., land sparing versus land sharing, only intensifying the debate on conservation approaches^34,35,36. More recently, efforts have aimed to reduce the extreme dichotomy of the alternative extremes by making use of both strategies as needed³⁷. However, far less attention has been devoted to extending the land-sharing/sparing concept to river systems^38,39. One limitation of land sharing schemes is that thresholds for terrestrial ecosystem health may be very different from thresholds of impairment to aquatic systems. Congruent with most of the land sharing/sparing debate over wildlife and land use, agriculture was also the largest source of stream impairment paradox in the present analysis; hence, novel strategies to identify effective and targeted protection measures for streams are needed beyond land sparing, such as floodplain protection³⁸ or increasing watershed buffering capacity⁴⁰.

This work highlights clear needs for further research in optimal design and siting of future protection efforts, illustrated by the coincidence of protection efforts with sources of impairment at different landscape scales. Even if the quantity portion of the 30×30 goal has been met, it can be argued that protection does not yet account for areas of ecological significance or provide effective conservation benefit. Thus, a prioritization of stream protection to account for ecological significance and impairment avoidance or mitigation is necessary. Schema using measures of the biological, social, and physicochemical resilience of an ecosystem⁴¹ as well as region-specific landscape health indicators^42,43 can inform protection and restoration efforts. For example, our results can be joined with existing data on riparian connectivity networks⁴⁴ and the Freshwater Resilient and Connected Network from the Nature Conservancy⁴⁵ to identify, in a spatial prioritization model, streams/catchments that would be optimal for habitat restoration/protection given the existing impairment or resilience of the system. Species distributions for target taxa, such as those on the IUCN Red List or overlaying maps of land identified as critical for threatened species^26,46 could also be added to identify streams that would most benefit from conservation efforts. This study offers insights into watershed protection and impairment that can contribute to the wealth of existing freshwater biodiversity and condition data to better inform future conservation efforts.

Due to the national scale of this work, the underlying data for attributes such as stream discharge and the physical classifications are often estimated based on smaller samples, giving rise to variance and error that may not be representative of actual conditions in every case. The chosen indicators of both ecological stream diversity and stream impairment are by no means exhaustive, likely contributing to an overestimate of effective stream protection in the conterminous U.S. Additionally, the thresholds used to estimate impairment cannot be nuanced enough to accurately capture the dynamics of each stream ecosystem. The selection of impairment thresholds naturally influenced our results, which ideally provide a foundation for future interrogation. Efforts to create more comprehensive stream habitat data on a nationwide scale will allow for similar exercises to this one and give a clearer picture of the status and needs for biodiversity protection in stream ecosystems.

Overview of approach

Our approach layers attributes of protection, diversity, and impairment for each stream segment, resulting in a database used to label and filter streams according to their unique characteristics. We relied on the NHDPlusV2 hydrography dataset as the geometry for every stream segment in the conterminous U.S.⁴⁷. Our analysis considers 2,667,768 stream segments, each with a unique common identifier that was used to jointly summarize and evaluate protection, human disturbances, and habitat diversity for streams. Generally, we report the frequency of stream protection, disturbance, and diversity in terms of total stream length (km).

Catchment area protection

We calculated the protected catchment area via geospatial summarization of the area and percentage of protected lands within each stream’s local and network catchment and floodplain, termed the Active River Area (ARA). To quantify land protection for the local and cumulative upstream landscape, we first intersected the NHDPlusV2 catchments with previously delineated unconstrained valley bottoms³¹ to get local ARA polygons for each local catchment. We then intersected the PAD-US Version 2.0 Database⁴⁸ with both the local catchments and local ARAs to quantify the area covered by each GAP Status Code. Next, we set up the NHDPlusV2 national database with the ArcHydro (ArcHydro for ArcMap 10.3) database schema required to run the ArcHydro Accumulate Attributes tool and joined the local catchment and ARA protected areas via the NHDPlusV2 COMID field to the NHD flowlines stream network layer. Then we ran the Accumulate Attributes tool to get the accumulated upstream catchment and ARA area for each GAP Status Code. Figure 8 shows the spatial scales considered. Wild and scenic rivers^19,49 were considered protected along their total length as a separate metric from PAD-US protection.

Using the proportion of local and accumulative catchment area protected for each segment¹⁴, we designated streams as protected under twelve different configurations according to the GAP Status (Statuses 1 and 2 vs. Statuses 1–3), threshold of catchment area protected (20% vs. 30%), and spatial scale of protection (local, network, or either). Any number of catchment protection thresholds could be used, but for simplicity, we only chose 30% in reference to the 30×30 target and 20% as a bridge between the GBF goal and the earlier Aichi Target of 17%⁵⁰. We report results for only the 20% catchment protection threshold here (Table 2), but results for all scenarios can be found in the Supporting Information and the data release²⁴. Percentage-based designations for protection are imperfect metrics for understanding effective biodiversity conservation, especially in isolation. However, they offer tangible and oftentimes ambitious targets for decision makers to be used in conjunction with other conservation actions⁵¹. Thus, we offer the two thresholds of catchment protection as a first step in understanding and prioritizing stream habitat conservation.

Stream impairment

We considered three categories of stream impairment: impoundment, regulation, and land use disturbance in the drainage area. To identify streams that are impounded and no longer free-flowing (and thus cannot be considered protected), we overlaid NHDPlusv2 flowlines with fish barriers from the National Anthropogenic Barrier dataset⁵² as points. Any flowlines that intersected with these points (within a 100 m buffer) were selected. We used the selected streams to identify intersecting NHD Waterbodies and Areas. Only NHD Areas that represented inundated or impounded water (inundation area, canal ditch, dam weir, lock chamber, special use zone, spillway, submerged stream) were kept for this selection. Sections of NHD flowlines overlapping selected waterbodies and inundated areas were then erased from maps (see Supplementary Fig 2). The new length for each stream segment was captured as a new variable, and streams were flagged if their entire length was impounded.

Dam regulation was also considered as an impairment using the Degree of Regulation (DOR), which represents a percentage of total cumulative reservoir storage volume in each stream’s upstream network divided by the annual discharge volume^53,54. Generally, streams with DOR ≥ 4% have been designated as impaired, i.e., showing signs of notable departure from natural flows⁵⁴. Dam storage density, measured as dam storage per unit area of a stream’s catchment (m³/km²), was obtained from StreamCat⁵⁵ and was multiplied by each stream’s cumulative upstream drainage area to calculate raw dam storage. Annual daily average stream discharge was obtained from NHDPlus Enhanced Unit Runoff Method flow estimates⁴⁷ and converted into annual discharge volume to calculate DOR.

Alongside impoundment and regulation, there are numerous indicators of stream ecosystem impairment across spatial scales. Among these, anthropogenic land cover in the stream floodplain or watershed is well-studied as an impediment to stream health^56,57,58,59, and several studies have used land cover and other anthropogenic disturbances to create indices of stream, floodplain, and watershed health^60,61. To isolate land cover as an impairment and to leverage spatial flexibility, we used StreamCat⁵⁵ variables for urban development (open space, low-, medium-, and high-intensity), impervious areas, and row crops/pasturelands within catchments and watersheds to identify land use disturbance within stream drainage areas. We used a threshold approach to define percentages of land use within stream catchments as quantitative boundaries where biological integrity (i.e., species richness, water quality) either rapidly declined or became noticeably lower. Streams were identified as impaired by land use if the local and/or network catchments exceeded 15% urban development area (NLCD developed classes)⁶², 6% total impervious area⁶³, and/or 10% total agriculture area (Table 1)⁶⁴. Despite the known influence of land cover on stream health⁵⁹, these particular thresholds may not reflect abrupt non-linear responses indicative of ecological impairment or shifts in alternative community states. Indeed, a range of land use thresholds have been reported for inducing stream impairment, depending on the metric and context (Table 1). To consider the potential effects of various threshold values, a sensitivity analysis was conducted to examine the influence on the protection-impairment paradox. Sensitivity results of stream impairment frequencies relative to land cover thresholds are shown in Supplementary Fig. 2 and Supplementary Tables 10 and 11. Each stream was given a binary flag to indicate whether it failed any impairment thresholds at a given spatial scale, which we used to identify streams marked as protected yet also impaired. Frequency distributions of stream impairment, quantified as the % of total stream length impaired and % of protected streams falling under impairment, were compared to threshold values to visually examine the occurrence breakpoints where shifts in the distribution of impairment were obvious (Supplementary Fig. 2). Thresholds used in our study are reflective of these breakpoints, indicative of major shifts in impairment frequency. These values tend to agree well with mid-range values reported in literature (Table 1). Again, while this does not insinuate ecologically meaningful thresholds, these mid-range values are conservative and sufficient to communicate the protection-impairment paradox.

Stream diversity

We used the catchment protection data in conjunction with geophysical stream classification³¹ to understand the diversity of physical stream habitats protected. The geophysical classification included five layers of sub-typologies, including stream hydrology, size, gradient, temperature, and valley confinement (Supplementary Table 11). Each sub-typology had multiple classes, which we grouped together to create a final physical classification. This theoretically could yield 435,897 different types of streams. However, due to association among sub-typology classes, the resulting physical types of streams in the conterminous U.S. numbered 5948, representing 4,963,171 km of stream length. We then ranked the physical classes according to the total stream length represented by each class, with the most prevalent stream type receiving a rank of 1. These rankings were plotted against each class’s total and protected stream length to examine the percentage of protection relative to the prevalence of stream types in order to quantify the deficit of protection for each stream type, assuming each stream type should receive 30% protection. We also classified streams according to the HUC 2 region¹ they fell within to understand how biogeographical regions were protected.

All analyses were conducted within ESRI ArcMap 10.3, ArcGIS Pro 3.1.3, and within R programming environment²⁴, and in-depth discussion of the methods can be found in the Supporting Information. This effort provided a dataset used to quantify protection of streams in the U.S., identify underrepresented or under-protected stream types, and isolate streams with contradictory protection and impairment statuses.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.