Distinct changes in riparian sediment microbial communities with depth and time since dam removal

Abstract

Microbes drive biogeochemical changes in ecosystems, including carbon (C) and nitrogen (N) cycling. Dam construction has altered riparian ecosystems worldwide, yet we know little about microbial community composition in riparian sediments and how it changes following dam removal and sediment/soil drainage. Here, we evaluate how riparian microbial communities change with increasing depth in the sediment profile for existing dams and over time following dam removal/breach and assess how various physico-chemical sediment properties influence microbial community composition. We studied microbial community structure for 12 riparian sites over a chronosequence of 0-234 years since dam breach. Sediment was collected every 0.3 m to a depth of 4 m. Aerobic taxa involved with N cycling (e.g., Nitrospirota) were dominant in surficial sediments, and increased in deeper sediments as time since dam breach increased. Anaerobic taxa implicated in C cycling (e.g., Bathyarchaeia, Anaerolineaceae) and iron reduction (e.g., Sva0485) were dominant in deeper, anoxic sediments, but declined the fastest post dam breach. These microbial trends provide insights into how riparian biogeochemical functions are impacted by dam inundation and the recovery and restoration of these ecosystems following dam removals.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-37708-3.

Introduction

Microorganisms are key drivers of ecosystem processes and functions^1–3. Interfaces of terrestrial and aquatic ecosystems, such as riparian zones, stimulate and alter microbial community structure over short distances and time frames due to their dynamic moisture and redox regimes^4,5. Human activities have, however, significantly altered riparian zones and their environments^6–8. These impacts are especially pronounced with persistent, long-lasting anthropogenic structures such as dams, which can substantially alter riparian environments, affect water quality, and disrupt microbially-driven biogeochemical processes^9,10. Here, we investigate how microbial composition (bacteria and archaea) varies with depth in riparian zones upstream of dams, and how it changes over time following dam removal using a 0–234-year chronosequence. Given the large number of ongoing construction and removals of dams worldwide^11,12, this is a timely and unexplored opportunity to study anthropogenic alteration of microbiomes in riparian sediments where dams have existed for centuries, and where they have recently or long-since been breached.

While there is increasing appreciation and knowledge of how microbial taxa vary and affect biogeochemical processes in terrestrial environments, previous studies have been limited to shallow soil depths (~ 1–2 m; ^{3, 13}). These studies suggest that microbial biomass, diversity, and richness decline with increasing soil depth, due to progressively less favorable growth conditions and habitat availability^3,13. Key edaphic factors that affect microbial structure and functions include pH, soil oxygen and redox conditions, moisture, nutrient levels, and organic carbon (C) content and type^3,14. Nutrient (e.g., nitrogen and phosphorus) availability often declines with depth and organic matter becomes more recalcitrant¹³. Thus, copiotrophic taxa that thrive on nutrient-rich and labile C substrates flourish in surficial soils, while slow-growing, anaerobic, and oligotrophic microbial communities that process recalcitrant organic C are favored in deeper soil environments. Studies also suggest that deeper soil taxa are generally rarer, and novel compared to surficial communities¹⁵. Furthermore, the bacteria to archaea ratio declines with depth as obligate or facultative anaerobes, including many archaea sensitive to high-oxygen conditions, are better able to survive in nutrient poor conditions of deep soils/sediments^3,15.

Microbial communities in riparian zones upstream of dams could however be very different from upland locations and/or undammed fluvial soil profiles^9,10. This is because dams produce a complex depositional environment with layers of legacy sediments^16,17 of varying texture and bulk density underlain occasionally by the original, pre-dam floodplains^18,19 (Fig. S1). In many instances, the coarse-grained pre-dam floodplains are covered by finer grained sediments (primarily clays and silts) eroded from uplands with progressive coarsening towards the top as the dam reaches its sediment trapping and storage capacity^18,20. In addition to physical differences, legacy sediments and pre-dam floodplains also contain buried organic C layers and/or iron-oxide horizons, adding chemical complexity^21,22 (Fig. S1). Furthermore, when dams are in place, riparian sediments experience persistent saturation and reducing conditions for decades to centuries^23,24. In contrast, when dams are removed and streams incise, riparian terraces are hydrologically disconnected, drained, and oxidized^25,26 (e.g., Fig. S2), yielding a very different hydrologic environment. These contrasting and varying sediment and hydrologic conditions could have strong, complex, and occasionally contradictory influences on the vertical distribution of microbial communities and biogeochemical processes, but these have yet to be fully explained. Understanding and quantifying these microbiome shifts is not only important for advancing new paradigms and models of anthropogenic change, but also for guiding appropriate management strategies. As dams continue to be constructed and removed worldwide, environmental agencies and restoration managers are desperately seeking guidance on how these disturbances may affect current and future water quality and ecosystem health and the trajectory and time frames of ecosystem recovery post restoration.

Here, we investigated the changes in microbial communities with depth (up to ~ 4 m) and time since dam removal (0-234 years) for a chronosequence of three existing and nine breached/removed milldams located in the mid-Atlantic region of United States (US). At many of these sites, pre-dam floodplains with buried organic C and iron oxide horizons constituted the lower third or fourth of the riparian profile with upland legacy sediments deposited above them. Physical and chemical composition of these soils and sediments were described recently²⁷ and previously for existing and removed dams^{20,23,26,28–30} used in this study. These observations and previous studies indicated that: (a) riparian zones with existing dams had near-surface saturation year-round with highly reducing conditions that resulted in increasing concentrations of ammonium-N and Fe in soils and groundwaters; (b) where dams were recently removed concentrations of ammonium-N and Fe declined rapidly in the drained profile but were still elevated in the lower strata close to the new stream water level and with buried organic C and iron oxide horizons; and (c) many of these biogeochemical changes were attributed to microbial processes such as nitrification, denitrification, dissimilatory nitrate reduction to ammonium (DNRA), and dissimilatory iron reduction (DIR); thus, identification of microbial taxa associated with these sediments was key to confirming or rejecting our explanations. The novelty of these sites and settings is the large population of existing and breached dams spanning 200 + years in similar geologic, climate, and topographic settings, and the detailed physico-chemical characterization of the sediments to compare with the microbial character. To our knowledge, this is the first study to examine how microbial composition changes across a range of time (years to centuries) since dam removal, and through the entirety of the riparian sediment profile, from the modern surface to centuries-old buried organic soil below.

For this study, we focused particularly on bacterial and archaeal groups involved in C, N, and Fe cycling and taxa for deeper sediments including those associated with the pre-dam floodplains with buried organic C and iron oxide horizons. Specifically, we asked the following questions: (1) How does microbial abundance, diversity, and composition change with depth in the riparian zones with and without dams? (2) How do microbial characteristics and composition relate to riparian hydrology (groundwater) and physicochemical environment? (3) Are there unique microbial communities in the deeper, pre-dam floodplains underlying the legacy sediments, and can they explain the elevated concentrations of ammonium-N and Fe in these layers? 4) How does microbial abundance, diversity, and composition change with time in response to dam removal and what are the key drivers?

We hypothesized that milldams enhance the decline in microbial biomass and diversity with soil depth in riparian zones by maintaining a stable state of persistent saturation and reducing conditions for decades to centuries. On the other hand, we hypothesized that dam removal will translate into an overall increase in microbial diversity and richness in surficial sediments due to a shift to a more dynamic, drained environment. We expected relative abundances of anaerobic, oligotrophic, S/Fe reducing taxa to increase with riparian depth and be greatest in the lowest strata (Fig. 1). In contrast, we expected relative abundances of aerobic and copiotrophic taxa to be greatest in the upper sediment layers (Fig. 1). Microbial composition was expected to be influenced by groundwater depth, texture, organic matter content, and other geochemical characteristics of legacy sediments and buried, pre-dam soil floodplains. Given potential variations with depth, we expected that microbial changes post dam removal and drainage would be complex, but an increase in microbial diversity and richness could occur with time. We also predicted that relative abundances of aerobic, copiotrophic taxa would increase following dam removal and soil drainage, with higher rates of increase in surficial sediments (Fig. 1). In contrast, we expected relative abundances of anaerobic, oligotrophic taxa would decrease post-dam removal, with faster rates of change in the lowest strata.

Fig. 1.

Conceptual diagram indicating changes in microbial composition as riparian sediment depth increases (a) where dams are present, (b) where dams have been breached, and (c) with increasing time since dam breach and soil drainage.

Results

A total of 92 distinct phyla (80 bacteria and 12 archaea) and 1,341 genera (1,310 bacteria, 31 archaea) were found across watersheds. The most common phyla were Pseudomonadota (syn. Proteobacteria), Acidobacteriota, Chloroflexota, Bacillota (syn. Firmicutes), Actinomycetota, Thermoproteota (syn. Crenarchaeota), Planctomycetota, Gemmatimonadota, Methylomirabilota, Myxococcota, Patescibacteria, Verrucomicrobiota, Thermodesulfobacteriota (syn. Desulfobacterota/Deltaproteobacteria), Nitrospirota (syn. Nitrospirae), Sva0485, Bacteriodota, RCP2-54, Spirochaetota, Latescibacterota, and Thermoplasmatota, representing 93.4% of all sequences (Table S1). A total of 2,145,600 sequence reads were from the Chiques Creek watershed, while 2,011,500 reads were from the Christina – Big Elk watersheds.

Vertical distribution of microbial taxa in sediments

Depth-related changes in the relative abundances of phyla revealed close associations with hydrology, with many of the top 20 phyla sharply increasing or decreasing below groundwater (GW) level (Fig. 2). In the Chiques Creek watershed, relative abundances of Chloroflexota, Bacillota, Patescibacteria, Thermodesulfobacteriota, Sva0485, Bacteroidota, Spirochaetota, and Thermoplasmatota significantly increased with depth (P < 0.05), while Pseudomonadota, Acidobacteriota, Actinomycetota, Planctomycetota, Methylomirabilota, and Latescibacterota significantly declined (P < 0.05, Fig. S3). In the Christina – Big Elk watersheds, relative abundances of Thermoproteota, Patescibacteria, Thermodesulfobacteriota, Nitrospirota, Sva0485, Spirochaetota, and Thermoplasmatota significantly increased with depth (P < 0.05), while Pseudomonadota, Acidobacteriota, Actinomycetota, Planctomycetota, Methylomirabilota, and Latescibacterota significantly decreased (P < 0.05, Fig. S4). Additionally, phospholipid fatty acid (PLFA)-derived live bacterial biomass decreased with depth at many sites (Fig. 3). In the Chiques Creek watershed, total bacterial biomass decreased at the Roller, Siegrist, Krady, Heistand, and White Oak sites (see Fig. 3a for P and R² values). In the Christina – Big Elk watersheds, total bacterial biomass decreased with increasing depth at the Parks and Old Cooch sites (Fig. 3b for P and R² values).

Fig. 2.

Changes in relative abundances of the top 20 phyla with increasing sediment depth in the (a) Chiques and (b) Christina – Big Elk watersheds. Normalized Depth of 0 indicates the sediment surface, whereas 1 represents the deepest depth. The dashed blue lines represent ground water level at Normalized Depth, and the dashed black lines represent the top of the buried organic layer.

Fig. 3.

Changes in total bacterial phospholipid fatty acid (PLFA) concentration with increasing depth in the (a) Chiques and (b) Christina – Big Elk watersheds. Black trendlines with bold P and R² values indicate significance (P < 0.05), while gray trendlines and non-bold values indicate trends toward significance (P < 0.10). The dashed blue lines represents the depth to ground water level.

At the genus level, microbial diversity and taxonomic richness varied with depth, and between breached and existing dams (Fig. 4). In the Chiques Creek watershed, Shannon diversity (P = 0.009, R² = 0.325), inverse Simpson diversity (P = 0.080, R² = 0.161), and taxonomic richness (P = 0.013, R² = 0.296) increased with depth for existing dam sites (Fig. 4a, e, i). However, at breached milldam sites, Shannon (P = 0.092, R² = 0.105) and inverse Simpson diversity (P = 0.047, R² = 0.144) decreased with depth (Fig. 4b, f), with no change (P > 0.10) in taxonomic richness (Fig. 4j). In the Christina – Big Elk watersheds, diversity and richness did not change with depth at the existing (Cooch) dam site (P > 0.10; Fig. 4c, g, k), whereas Shannon diversity (P = 0.047, R² = 0.114), inverse Simpson diversity (P = 0.021, R² = 0.151), and taxonomic richness (P = 0.061, R² = 0.102) decreased with depth (Fig. 4d, h, l).

Fig. 4.

Shannon Diversity (a-d), Inverse Simpson Diversity (e-h), and Taxonomic Richness (i-l) in riparian sediments at existing and breached dam sites in the Chiques and Christina – Big Elk watersheds. Red trendlines with bold P and R² values indicate significance (P < 0.05); gray trendlines and non-bold values indicate trends toward significance (P < 0.10). Dashed black lines represent smoothed means.

Across watersheds, sediment physico-chemical parameters such as pH, bulk density (BD), gravimetric water content (GWC), ammonium (NH₄⁺), nitrate (NO₃⁻), Mehlich-3 extractable iron (Fe_M3), oxalate extractable amorphous iron (Fe_o), and % sand and clay content varied with depth²⁷ and were significantly related to diversity metrics (Shannon diversity, inverse Simpson diversity, taxonomic richness) and relative abundances of top phyla (Figs. S3, S4). For example, NH₄⁺ and iron (Fe_M3, Fe_o) concentrations were negatively correlated with Pseudomonadota, Acidobacteria, and Myxococcota, yet positively correlated with Sva0485, Spirochaetota, and the Shannon and inverse Simpson diversity indices in both the Chiques (Fig. S3) and Christina – Big Elk watersheds (Fig. S4).

Changes in microbial composition of sediments over time

Across both watersheds, we observed shifts in microbial composition at the phylum level with increasing time since dam removal/breach (Fig. 5). The relative abundances of Acidobacteria, Actinomycetes, Planctomycetota, and Pseudomonadota increased over time, while Bacillota, Chloroflexota, Patescibacteria, Spirochaetota, Sva0485, and Thermoplasmatota decreased (Fig. 5). However, changes did not occur uniformly in the sediment profile, and responses were most evident when assessed separately in top, middle, and bottom strata (Fig. S5 for rates of change; see Figs. S6 and S7 for significant changes individually by phyla). In the Chiques Creek watershed, increases included Acidobacteriota in the top stratum (P = 0.099), Actinomycetota in the middle (P = 0.048) and bottom strata (P = 0.047), and Myxococcota in the bottom stratum (P = 0.012). Meanwhile, Sva0485 decreased in the middle stratum (P = 0.092), and Bacillota (P = 0.069) and Chloroflexota (P = 0.096) decreased in the bottom stratum (Figs. S5a, S6). Actinomycetota had the greatest rate of increase (middle: +9.1%/100 yrs., bottom: +10.2%/100 yrs.), while Chloroflexota and Bacillota had the greatest rates of decrease (bottom only, -10.0%/100 yrs. and − 8.6%/100 yrs.). In the Christina – Big Elk watersheds, Pseudomonadota (P = 0.042) and Planctomycetes (P = 0.051) increased in the top stratum, while Myxococcota increased in the middle (P = 0.089) and bottom (P = 0.022) strata, and Nitrospirota increased in the bottom stratum (P = 0.003). In contrast, Patescibacteria decreased in the top (P = 0.078) and middle (P = 0.068) strata, while Sva0485 decreased in the middle (P = 0.071) and bottom (P = 0.024) strata. Interestingly, Nitrospirota decreased in the top stratum (P = 0.049), yet increased in the bottom stratum (P = 0.003; Figs. S5b, S7).

Fig. 5.

Relative abundances of the top 20 phyla across all samples in the (a) Chiques and (b) Christina – Big Elk watersheds, with sites ordered by increasing time since dam removal/breach and soil drainage.

As time since dam breach and riparian drainage increased, genus-level diversity and taxonomic richness remained relatively stable, though subtle changes occurred in the top and middle strata of the soil profile (Fig. 6). In the Christina – Big Elk watersheds, the inverse Simpson diversity index significantly increased (P = 0.038, R² = 0.925; Fig. 6b), and there was a trend toward significant increase in the Shannon diversity index (P = 0.059, R² = 0.885; Fig. 6a) in the middle strata, yet there was no change in richness (Fig. 6c). In contrast, taxonomic richness decreased by approximately 22 and 25 taxa per decade in the top (P = 0.022, R² = 0.680) and middle (P = 0.022, R² = 0.685) strata of the Chiques Creek watershed (Fig. 6c), and there were no significant changes in Shannon or inverse Simpson diversity indices (Fig. 6a, b).

Fig. 6.

Changes in Shannon Diversity (a), Inverse Simpson Diversity (b), and Taxonomic Richness (c) by soil strata (Top, Middle, and Bottom) with increasing time since dam breach and soil drainage in the Chiques and Christina – Big Elk watersheds. Red trendlines with bold P and R² values indicate significance (P < 0.05), while gray trendlines and non-bold values indicate trends toward significance (P < 0.10). Dashed black lines represent smoothed means and are shown where P < 0.10.

Relationship of microbial communities with environmental parameters and indicator species

Non-metric multidimensional scaling (NMDS) revealed the separation of detailed microbial communities across watersheds and depths which was related to differences in environmental parameters (Fig. 7). In the Chiques Creek watershed, microbial communities in drier, surficial sediments (red color; negative direction [-] of NMDS axis 1) were significantly correlated with Bacterial PLFA, % sand, pH, NO₃⁻, and taxa classified at the genus (Gaiella, Hyphomicrobium), family (Gemmataceae, Nitrososphaeraceae, Vicinamibacteraceae), order (Rokubacteriales, Gaiellales), and class (KD4-96) taxonomic levels (P < 0.05; Fig. 7a, c). Communities in the wetter, deeper and anaerobic sediments (blue color; positive direction [+] of NMDS axis 1 and [-] NMDS axis 2) were significantly correlated with bioavailable iron forms (Fe_o, Fe_M3), GWC, soil Ripening Index, sulfur (S), and % clay, and anaerobic taxa belonging to the Sva0485 clade, the family Anaerolineaceae (phylum Chloroflexota), the Bathyarchaeia class of archaea, and the genus Niallia (phylum Bacillota) (P < 0.05; Fig. 7a, c). Sediments near GW level at breached dams ([+] NMDS axis 2) were significantly correlated with sodium (Na) and macroaggregates, while sediments near GW level at existing dams ([-] NMDS axis 2) were significantly correlated with Fe_c (Fig. 7a). In the Christina – Big Elk watersheds, drier surficial sediments ([-] NMDS axes 1 and 2) were correlated with Bacterial PLFA, NO₃⁻, the families Xanthobacteraceae and Gemmataceae, and the orders Vicinamibacterales and Gaiellales (P < 0.10; Fig. 7b, d). Sediments from mid-depths ([+] NMDS axis 2) were correlated with % clay content and the phylum RCP2-54 (P < 0.10). Additionally, drier mid-depth sediments ([-] NMDS axis 1, [+] NMDS axis 2) were significantly correlated with denitrification and various families (Gemmatimonadaceae, Nitrosotaleaceae), orders (Terriglobales, Elsterales, Rokubacteriales, Subgroup 2, Group1.1c), and the class AD3, while wetter mid-depth sediments ([+] NMDS axis 1 and 2) were correlated with the iron activity index (Fe_o/Fe_d) and the class Bathyarchaeia (Fig. 7b, d). Deep sediments ([+] NMDS axis 1, [-] NMDS axis 2) were correlated with soil Ripening Index, % OM, GWC, TN, the phylum Sva0485, and the genus Niallia (Fig. 7b, d).

Fig. 7.

Correlations of soil metrics (a, b) and dominant taxa (c, d) in the Chiques Creek and Christina – Big Elk watersheds with NMDS results showing separation of microbial communities at the genus level. Genera with relative abundance > 0.1% within the watershed were included in the analysis. Arrow overlays indicate significant correlations (P < 0.05). *indicates taxa not identified to genus level.

In the Chiques Creek watershed, distance-based redundancy analysis (dbRDA) showed that pH (P = 0.003), NO₃⁻ (P = 0.008), S (P = 0.013), total Bacterial PLFA (P = 0.021), GWC (P = 0.023), Na (P = 0.031), and NH₄⁺ (P = 0.032) significantly explained variation in microbial composition with depth (Fig. S8a). In the Christina – Big Elk watersheds, denitrification (P = 0.011), depth (P = 0.013), pH (P = 0.014), GWC (P = 0.024), % clay (P = 0.044), and NO₃⁻ (P = 0.031) were significantly related to microbial composition (Fig. S8b).

Indicator species analysis identified several taxa that were associated with buried organic soils in the deep, pre-dam floodplain (see Table S2 for relative abundances and full list including indicator taxa for Oxic and Saturated Zones). In the Chiques Creek watershed, the phyla Spirochaetota, Bacteroidota, and Chloroflexota were associated with buried organic soils, as were the archaean genera Marine Benthic Group D (MBG-D)/DHVEG-1 and SCGC-AAA011-D5, and bacterial genera SCGC-AB-539-J10, Spirochaeta, Candidatus Omnitrophus, Aggregatilinea, Streptomyces, and Aminicenans (all P < 0.05; Table S2). Additionally, the phyla Sva0485 (P = 0.056) and Thermoplasmatota (P = 0.056), the genus Desulfatiglans (P = 0.065), and the family Anaerolineaceae (P = 0.061) showed trends toward significance (Table S2). In the Christina – Big Elk watersheds, buried organic soils in the pre-dam floodplain were significantly associated with the phyla Sva0485, Thermodesulfobacteriota, Thermoproteota, Spirochaetota, genera MBG-D/DHVEG-1, Sideroxydans, SCGC-AB-539-J10, Candidatus Nitrosoarchaeum, Aminicenans, Desulfobacca, Candidatus Omnitrophus, and the class Bathyarchaeia (all P < 0.05; Table S2).

Discussion

Across the two watersheds and multiple sites, our data revealed clear and common patterns of stratification of microbial communities with increasing depth in riparian profiles. The relative abundance and diversity of these taxa reflect strong environmental filtering by gradients in moisture, oxidizing and reducing conditions or oxygen availability, sediment texture, nutrient availability and carbon lability^2,10. The sediment/soil profiles evaluated in this study differ markedly from those of upland soils and/or un-dammed floodplains². Unlike upland soils, which typically exhibit a unidirectional, monotonic gradient of soil development and weathering from the surface to the un-weathered parent material or bedrock, our profiles resembled a mosaic of soils and sediments of varying ages and degrees of weathering and development. The bottom was composed on Pleistocene-era gravels and sediments overlain by iron oxides likely deposited in the early Holocene as the climate warmed³¹. Above the Pleistocene-era deposits are organic horizons (now buried) that represented the Holocene-era wetlands and marshes that predominated the valley bottoms in this region^32–36. Together, these bottom layers constituted the pre-dam floodplains (~ 1–2 m thick) which evolved over hundreds to thousands of years, and likely had oligotrophic and facultative/obligate microbial communities representing the wetland/marsh environments of this period^8,35. These floodplains were subsequently rapidly (< 100–200 years; note ages of dams in Table 1) buried by large amounts of upland legacy sediments (thicknesses > 1 m depending on the heights of the dams) post dam construction. Thus, the buried floodplains and overlying legacy sediments represented very different habitat conditions and evolutionary time frames that shaped the microbial communities in these terraces.

Table 1.

Watershed, milldam name, location (State), latitude (Lat.), longitude (Long.), maximum sampling depth, date breached and years since breach (Breach year & age), drainage area, and SSURGO soil type of sampled milldams.

Watershed	Milldam	State	Lat.	Long.	Stream	Depth (m)	Breach Year & age	Drainage area (km2)	SSURGO Soil Type
Chiques	Heistand	PA	40.0553	-76.5268	Chiques	2.13	2015 (9)	326.34	Lindside-Linden complex
	Johnston	PA	40.0634	-76.5154	Chiques	1.22	1926 (98)	279.72	Lindside silt loam
	Krady	PA	40.0689	-76.4998	Chiques	2.74	2018 (6)	162.13	Lindside-Linden complex
	Siegrist	PA	40.0749	-76.4685	Chiques	2.74	Existing	153.85	Hagerstown silty clay loam
	Roller	PA	40.1083	-76.4431	Chiques	3.35	Existing	127.43	Hagerstown silt loam
	Shenck	PA	40.1165	-76.4239	Chiques	1.52	1954* (70)	108.26	Hagerstown silt loam
	White Oak	PA	40.2060	-76.3943	Chiques	2.13	1972 (52)	47.40	†Bowmansville silt loam
Christina – Big Elk	Parks	MD	39.6751	-75.8288	Big Elk	3.96	1935 (89)	125.36	Hatboro-Codorus complex
	Scott	MD	39.6889	-75.8272	Big Elk	3.96	1935 (89)	122.25	Hatboro-Codorus complex
	Springlawn	PA	39.7389	-75.8671	Big Elk	3.96	1958* (66)	81.58	Codorus silt loam
	Cooch	DE	39.6456	-75.7424	Christina	3.35	existing	49.73	Hatboro-Codorus complex
	Old Cooch	DE	39.6415	-75.7371	Christina	2.13	1790 (234)	49.99	Hatboro-Codorus complex

* Estimated based on last appearance of milldam in historical aerial imagery.

† Location is classified as “Water” due to outdated mapping; thus, soil type of the nearest land area is shown.

While total bacterial biomass consistently declined with depth across nearly all sites, patterns in diversity and richness were more variable. At the existing Chiques Dam sites (Roller and Siegrist), diversity and richness increased with depth, yet both metrics declined at breached dam locations. In contrast, a previous study at the Roller site⁸ (at the OTU level versus the genus level used here) reported a consistent decline in Shannon diversity and Faith’s phylogenetic diversity (PD), but a mixed response in Chao1 richness. The observed decline in microbial biomass, diversity, and richness with depth aligns with numerous studies attributing these patterns to reductions in oxygen, nutrients, and labile carbon availability^13,15. The increase in diversity and richness at the existing dam sites was primarily driven by measurements at the base of the Roller profile (Fig. 4), which likely represent microbial communities preserved in pre-dam, buried floodplain sediments. These anaerobic, oligotrophic taxa capable of utilizing recalcitrant C are not as present or abundant in the aerobic, drained profiles of breached dam locations, likely contributing to the decrease in diversity. Prior to dam construction, these floodplains were likely more dynamic in terms of aeration and hydrology, creating conditions that supported more diverse microbial populations. However, the extent to which this historic microbial community remains viable is unknown. It is important to note that the DNA-based sequencing methods used in this study capture the total potential microbial community and do not distinguish between living, dead, or dormant microbes³⁷. While many functions/processes are known to occur within taxa in our study, hereafter we speculate on their potential roles and not confirmed processes based on metabolically active community.

The bacteria that were dominant in surficial soils across the sites but decreased with depth included Pseudomonadota, Acidobacteriota, and Actinomycetota. These phyla have previously been reported for surficial soils because of their low tolerance for saturated and anoxic conditions, as well as their preference for labile C sources^2,13. Groundwater levels were clearly an important control on microbial abundance at our sites - Acidobacteria abundance decreased below groundwater level and was lower at existing and recently breached sites (e.g., Roller, Krady, and Cooch; Fig. 2) than sites where dams were breached long ago (e.g., Cooch versus Springlawn, Scott, and Parks; Fig. 2). Correlation analyses (Figs. 7, S3, S4) also revealed that surficial taxa were related to % sand, pH, and NO₃⁻, indicating their preference for more porous, acidic, and N-rich oxidized environments.

In contrast to surficial taxa, Chloroflexota, Bacillota, Patescibacteria, Thermodesulfobacteriota, and Sva0485 were identified as dominant phyla in lower sediment strata, while Bathyarchaeia and Niallia were the dominant class and genus, respectively (Fig. 7). The increase of a few taxa (e.g., Sva0485, Patescibacteria, Spirochaetota) below the groundwater levels was especially apparent (e.g., Roller, Siegrist, Cooch, and Krady sites in Fig. 2) indicating strong controls of groundwater and reducing conditions. Indeed, many of these taxa have been reported to be oligotrophic and obligate anaerobes that can survive on recalcitrant C^2,14. Bathyarchaeia, a key taxon in our data and one of the largest archaeal classes, is capable of surviving in energy limited environments because of its metabolic plasticity³⁸. Like our observations, Zhou¹⁰ found Dehalococcoidia (phylum Chloroflexota) and Patescibacteria were key taxa in buried sediments associated with a dam in China. They attributed the presence of Dehalococcoidia in deep sediments to anaerobic organohalide respiration, and Patescibacteria to its minimalist lifestyle and adaptability to nutrient-limiting deep sediments.

Indicator species analysis demonstrates that the lower, pre-dam horizons with buried organic soils and iron oxides contained the phyla Sva0485, Thermodesulfobacteriota, Thermoproteota, Spirochaetota, Chloroflexota, and Bacillota, the class Bathyarchaeia, and the family Anaerolineaceae. Many of these taxa possess the metabolic capacity to degrade recalcitrant organic matter. For example, the class Anaerolineae, which includes the family Anaerolineaceae, is known for its role in degrading complex organic compounds³⁹ and at least one subgroup of Bathyarchaeia can degrade lignin⁴⁰. Correlation analyses from the current study also revealed that many of these deep taxa were strongly correlated with bioavailable iron forms (Fe_o, Fe_M3), GWC, sulfur (S), and % clay (Figs. 7, S3, S4). We speculate that highly reducing conditions coupled with the availability of recalcitrant organic carbon and iron oxides (Fig. S3; including elevated dissolved and soil bound amorphous iron, Fe_o) in these soils could be providing a suitable environment for these microbial communities to survive^28,30. The buried organic C and iron oxides may serve as valuable electron donors and acceptors to meet the energy needs of these taxa⁴¹ in an otherwise low energy environment. For example, the facultative anaerobic autotroph Sva0485 can perform dissimilatory iron and sulfate reduction (DIR, DSR) under anoxic conditions, can produce ammonium-N, and has been reported in the presence of ferrous iron⁴². The presence of Sva0485 in the pre-dam horizons, above the current groundwater levels at breached dam locations (especially Parks and Springlawn, Fig. 2), further supports this premise. Similarly, other taxa identified in these horizons - Bathyarchaeia and Thermodesulfobacteriota - are capable of utilizing recalcitrant C and performing DNRA, another key mechanism in addition to DIR^40–44. Together, DNRA and DIR may help explain the elevated iron and ammonium-N concentrations observed, particularly their increase with depth near pre-dam horizons^23,28–30. Thus, the stratification of microbial communities in our dam-affected riparian profiles may not only reflect historic and anthropogenic legacies but suggests they could be shaping contemporary soil and water chemistry if metabolically active. Future research should focus on the live, functionally active portion of the community rather than describing the potential community in its entirety.

In addition to depth patterns, our analyses revealed significant temporal shifts in microbial community composition across the 0–234-year dam removal chronosequence. Comparing sites with different times since dam breach provides evidence that drained sites have distinctly different microbial communities than riparian sediments near existing dams. In the Christina – Big Elk watershed, microbial species diversity increased in post-removal sediments relative to the dam-influenced site, indicating an initial surge in diversity following the disturbance of dam removal; however, diversity decreased over time in the Chiques watershed. This aligns with recent studies that the elimination of long-term flooding can increase microbial diversity and functional potential as aerobic niches are restored^8,15. The abrupt transition from saturated, anoxic conditions to a drained, oxic environment after dam removal²⁶ represents a major perturbation that likely disrupts competition and creates new microhabitats, thereby fostering greater diversity. Such increases in microbial diversity with disturbance support an intermediate disturbance hypothesis in these sediments, where the newly introduced fluctuations in moisture and oxygen allow coexistence of both pioneer opportunists and persisting anaerobic specialists during the early stages of recovery.

Community composition also shifted toward a more terrestrial soil-like microbiome as time since removal increased. Likewise, nitrifying taxa (e.g., Nitrososphaeraceae, MND1) and denitrifying taxa (e.g., Hyphomicrobium) were more prominent in sites that had been oxic for longer, suggesting a resurgence of aerobic activity as the habitat transitions away from anaerobic conditions. In contrast, sediments at existing or recently breached dams showed greater proportions of obligate anaerobes, including iron/sulfate reducers (e.g., Sva0485, Thermodesulfovibrionia), taxa that degrade complex recalcitrant C (e.g., Bathyarchaeia, KD4-96) and syntrophic taxa that potentially contribute to methanogenesis (e.g., Anaerolineaceae) or DNRA (e.g., Aminicenans, Rokubacteriales), reflecting the legacy of prolonged anoxia prior to removal⁴⁵. While the relative abundances of Sva0485 were comparatively low, this phylum should not be overlooked as it may signify that important changes are occurring in previously anoxic soils. Members of Sva0485 express the genes necessary for N fixation (nifHDK) and denitrification (norB, nosZ)⁴², confirming the potential to both fix atmospheric N₂ and remove N from the system even in deep sediments. However, its relative abundance in mid and lower depths significantly declined over time since dam removal, reflecting the potential shift toward a more aerobic and oxidizing microbial community. Again, we caution that these are not confirmed processes in our system but potential roles these taxa may play if metabolically active.

Over time, as riparian sediments dry and connections with the atmosphere and surrounding soils are re-established, the microbial community becomes more aerobic and similar to communities of upland terrestrial soils^1,36,46. This trajectory of community change over the dam removal chronosequence is consistent with broader patterns of microbial succession observed in other ecosystem restoration contexts, where early post-disturbance communities gradually give way to stable, reference-like communities as environmental conditions stabilize^47–49. We do recognize though that the temporal shifts in microbial taxa varied with depth (top, middle, and bottom) in the riparian profile. This could be due to the complex and varying physico-chemical environment and its control on microbial communities highlighted in the previous section. While hydrology was a key driver of microbial change and responded most rapidly following drainage and dam removal, other controlling factors, especially sediment characteristics such as texture, pH, organic C, and iron oxides have greater inertia and longer lag periods in their response. Thus, the net temporal changes in microbial communities could be complex and vary with depth, as observed in this study.

Conclusions and environmental implications

This study highlights that anthropogenic disturbances related to dam construction and removal alter the vertical and temporal dimensions of riparian microbial communities. Vertical stratification of physico-chemical conditions in legacy sediments and underlying pre-dam floodplains shapes microbial habitat and structure, which can subsequently alter water and soil chemistry. Dam-induced stratification of microbes, formed by decades of reducing conditions, begins to break down after breaching as surface and subsurface communities converge towards aerobic metabolism and diversity. This chronosequence reveals a successional dynamic wherein microbial communities rapidly respond to dam breaching, then change in composition as the ecosystem moves toward a new equilibrium state. Together, these findings underscore the resilience and adaptability of sediment microbial communities in the face of large-scale environmental change and carry important practical implications: as managers consider dam removals, understanding the likely trajectories of microbial community shifts can inform predictions of biogeochemical and water quality outcomes such as nutrient release or retention. For example, the anaerobic microbial taxa observed for existing dams in this study matched the elevated ammonium-N concentrations for these sites, whereas at breached dam sites the microbial taxa were aerobic and nitrifying with low N concentrations. Understanding the shifts in microbial communities could thus provide important insights into the potential mechanisms and fate of N associated with dam removals. Long-term monitoring of recently breached or planned removal sites would confirm whether microbial assemblages eventually converge with those of older breach sites, but our data suggest a trajectory toward microbial recovery within the first decade post-removal. These assessments could help design appropriate riparian and stream restoration and mitigation strategies.

Methods

Study design

Twelve riparian sites located upstream of current (n = 3), and former (n = 9) milldams were selected for sampling. Streams sampled included the Chiques Creek, Big Elk Creek, and Christina River²⁷. Mean annual temperatures in the region range from 10.6 to 13.2 °C, and mean annual precipitation ranges from 115 to 121 cm. Milldam breach dates ranged from recent (2018) to long ago (1790); exact locations, sampling depths, breach dates, and drainage areas of milldam sites are reported in Table 1.

The Chiques Creek watershed is in the Piedmont region of central Pennsylvania, and watershed land use/land cover is predominantly agriculture (54%), followed by forest (22%), and urban/developed land (20%)²⁸. The Christina River watershed and Big Elk Creek sub-watershed span the Piedmont/Coastal plain transition region along the Pennsylvania, Maryland, and Delaware. Land use/land cover is predominantly urban/developed (47%), forested (30%), and agricultural (23%)²⁸. Chiques Creek and Big Elk Creek are tributaries of the Susquehanna and Elk Rivers, respectively, with outflows to the Chesapeake Bay. The Christina River is a tributary of the Delaware River, which has its outflow to Delaware Bay. The Big Elk Creek and Christina River are adjacent to one another and are considered as one large drainage system (the Christina-Big Elk watershed) for this study, though they are part of different drainage basins. The Chiques Creek watershed was considered separately due to differences in location, geography, and soil texture/type.

Sediment sampling was conducted between April and May 2024. A single sediment core was collected 100–300 m upstream from known milldam locations using a 3.5 cm diameter manual auger, approximately 2–5 m from the stream bank. Sediment cores were collected to refusal when possible, and depth varied from 1.2 to 4 m. A shovel was used to collect surface soils at two sites (Cooch, Old Cooch) due to high sand content, and sediment samples were taken from the stream bank face after scraping away exposed soil with a hand trowel at one site (Springlawn) due to difficulty with augering. Samples were separated into equal ~ 0.3 m intervals, placed into labeled Ziplock bags, and kept on ice until they were delivered to the lab. Subsamples were collected and stored in a refrigerator or freezer depending on the analysis (below).

Due to site-specific variability, sampling depth was normalized in each site (rescaled from 0 to 1, with 0 representing the top 0.3 m of surface sediment and 1 indicating the deepest 0.3 m at the site) using the equation where x_i is a given foot of depth, min(x) represents the uppermost depth, and max(x) the lowest depths. This allowed us to better compare sites that were sampled to different maximum depths (since riparian terrace height and sediment depth is largely determined by the height of the dam, and dam heights differed), while still accounting for groundwater effects. Depth to stream water (in meters) from the top of the riparian terrace was noted at each location during sampling and was assumed to be the groundwater (GW) depth. Normalized GW depth was calculated by replacing x_i in the above equation with the GW depth at each site. Lastly, Normalized Depth-to-GW was computed for each sample depth by subtracting the Normalized Sample Depth from Normalized GW Depth. Values ranged from approximately − 1 to + 1.25, with 0 indicating GW level, and increasingly negative or positive depths lower or higher above GW, respectively.

Sample processing and analysis

Approximately 200 g of refrigerated sediment from each 0.3 m depth interval was shipped to Regen Ag Lab (Nebraska, USA) for determination of Mehlich-3 extractable nutrients (P, S, K, Ca, Mg, Na, Zn, Fe, Mn, Cu, B, Mo, Co), KCl extractable nitrate (NO₃⁻) and ammonium (NH₄⁺) -N, percent total organic carbon (% TOC) and total nitrogen (% TN), wet aggregate stability, particle size/texture (i.e., sand, silt, and clay), and phospholipid fatty acid (PLFA) content. Briefly, PLFAs are a key component of most microbial cellular membranes (though they are absent in the Archaea), and quickly decompose following cell death⁵⁰. Thus, PLFA analysis allows quantification of living biomass among different types of microbes within a sample.

Bulk density (BD) and gravimetric water content (GWC) were determined by oven drying a wet sediment subsample of known mass and volume at 65 °C for 72 h. In some cases, compaction, saturation, and/or sampling method (i.e., shovel or bank-face collection) prevented BD measurement, and only GWC was recorded. Potential denitrification rates were determined by denitrification enzyme assay²⁹. All remaining sediments were air-dried, ground, and sieved through a 2 mm sieve prior to analysis. TOC and TN were determined by were determined by combustion on a LECO TOC/TN analyzer. Percent sand (2,000–63 μm), silt (63–2 μm), and clay (< 2 μm) content were determined by standard hydrometer method. Dried sediments were homogenized via ball mill (Mixer Mill MM200, Retsch, Haan, Germany) for extraction and determination of amorphous and total free iron oxide content. Amorphous iron content (Fe_o) was determined via ammonium oxalate extraction, while free iron oxide (Fe_d) content was determined via dithionite-citrate-bicarbonate (DCB) extraction⁵¹. Crystalline iron content (Fe_c) was calculated as Fe_d - Fe_o, and the iron activity index, a proxy for soil age⁵², was calculated as Fe_o / Fe_d.

DNA extraction and high-throughput sequencing

Subsamples (~ 2 g) were collected from intact sediment/soil cores immediately after augering and placed into 2.0 mL microcentrifuge vials using a micro-spatula, sterilized with 75% ethanol between samples. Vials were placed on ice for transport to the lab, then stored at -80 °C until DNA extraction. Total genomic DNA from each sample was extracted using DNeasy PowerSoil Pro Kits (Qiagen, Germany) following the manufacturer’s protocols. DNA quantity and quality were checked with an ND-2000 NanoDrop spectrometer (Thermo Fisher Scientific, USA) before high throughput sequencing prep. The V4 variable region of 16S rRNA genes were amplified and sequenced on an Illumina high throughput sequencing platform (MagiGene Biotechnology Co. Ltd., Guangzhou China) following the manufacturer’s recommendations and protocol.

Raw Illumina sequences were processed with the QIIME2 software package (version 2024.2). After demultiplexing, raw sequence reads were carried out with quality control, denoising, filtering (to remove unknown Kingdom, mitochondria, chloroplast, and singleton DNA fragments), merging, and chimera removal through q2-DADA2. Amplicon sequence variants (ASVs) were generated, and sequences were classified to taxa using the SILVA 138.2 database for 16S rRNA. To minimize sampling effects, the original ASV tables were rarified to a depth of 44,700 sequences per sample, with coverages over 99.5% of the total diversity, resulting in 104,232 unique ASVs. Relative abundances (rather than raw counts) were calculated as the proportion of each ASV in the soil sample after rarefication, and ASVs were then aggregated at the phylum and genus levels for all downstream analyses.

Data and statistical analysis

All analyses were performed using R version 4.4.3 (R Core Team 2025) and JMP Pro 17.2.0 (SAS Institute). Statistical analyses were performed separately for the Chiques Creek watershed and Christina-Big Elk watersheds. Statistical significance was considered at the α = 0.05 level, though trends toward significance are reported at the α = 0.10 level in some cases.

For each soil sample, alpha diversity metrics (i.e., Shannon diversity, inverse Simpson diversity, and taxonomic richness) were calculated at the genus level using the diversity function from the ‘vegan’ R package. Vertical profiles were generated for each site to investigate how microbial composition at the phylum level changes as sediment depth increases. Linear regression was performed to establish how microbial diversity and bacterial PLFA change as sediment depth increases. Spearman rank correlations were used to evaluate relationships between microbial taxa (diversity and relative abundances) and soil physico-chemical parameters, which often vary with depth²⁷. To discern patterns of change post dam removal with sediment depth, we further grouped sediment samples into 3 depth strata: top, middle, and bottom third of the profile. Average diversity, richness, and relative abundances of top phyla were calculated for each stratum, and linear regressions were performed to assess how these metrics changed with increasing time since milldam breach and soil drainage occurred. We assumed that pre-dam floodplains were likely located in the bottom strata.

To understand how microbial communities differ across sites and sampling depths, we performed non-metric multidimensional scaling (NMDS) on relative abundances using the metaMDS function in the ‘vegan’ R package, with Bray-Curtis dissimilarity as the distance measure. Only genera with relative abundances greater than 0.1% across samples within each watershed were used in the analysis. NMDS stress values < 0.2 indicated a two-dimensional NMDS solution was acceptable for each ordination. Pearson correlations between NMDS scores and soil physicochemical parameters²⁷and microbial taxa were calculated via the envfit function in the ‘vegan’ R package.

To further establish which soil parameters significantly explain variations in microbial composition and structure, distance-based redundancy analysis (dbRDA) was performed using the ‘vegan’ function dbrda with Bray-Curtis dissimilarity as the distance measure. Only genera with relative abundances greater than 0.1% across samples within each watershed were used in the analysis. To reduce the initial number of variables in the model, only soil parameters which were significantly correlated with NMDS community structure were selected as environmental constraints in the ordination. Variables were first standardized, then stepwise (forwards and backwards) regression was used to obtain the terms with significant explanatory relationships in the model. Lastly, indicator species analysis was performed using the multipatt function from the ‘indicspecies’ R package to determine microbial taxa that were associated with 3 distinct zones: oxic sediments (above GW), saturated sediments (below GW), and buried organic soils (in the pre-dam floodplain).

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Eric R. Moore (EM) and Shreeram Inamdar (SI) wrote the article. EM, Md. Moklesur Rahman (MR), Matthew Sena (MS), Joseph G. Galella (JG), Bisesh Joshi (BJ), and Alexis Yaculak (AY) collected the data. Marc Peipoch (MP) conducted denitrification enzyme assays and Jinjun Kan (JK) performed high throughput sequencing steps. EM did the data analysis and created the tables and figures. All authors reviewed the manuscript and suggested edits.

Data availability

The datasets generated and/or analyzed during the current study are available in the GenBank repository, [accession number PRJNA1346466].

Declarations

Competing interests

The authors declare no competing interests.

Informed consent

Informed consent from all subjects/participants and/or their legal guardian(s) for publication of identifying information/images in an online open-access publication.