Introduction

The Keelung River Basin, located in northern Taiwan near Taipei City, is a densely populated and economically significant region. The area frequently experiences intense rainfall events, with annual precipitation exceeding 3,600 mm. Typhoons and heavy rains frequently trigger severe flooding along the river. During Typhoon Nari in 2001, rainfall surpassed 1,000 mm within 2 days, resulting in severe levee breaches and flooding (Fig. 1). This prompted the Taiwanese government to initiate the Keelung River Regulation Project. Flood walls, dikes, and bank protections were upgraded from a 100-year peak flood capacity to a 200-year peak flood capacity, which are on the order of 10³–10⁴ m³/s (e.g., 3,910 and 4,180 m³/s at Guandu gage (river mouth)). However, early remediation methods devastated the surrounding environment. Over the succeeding 2 decades, as the importance of ecological protection became more widely recognized, environmentally friendly methods were increasingly employed for levee embankment construction, waterfront facilities, riparian green spaces, and recreational areas. Consequently, for the 42,535 m Keelung River embankment improvement project, although concrete has continued to be employed for some sections, the project has shifted toward application of ecologically sustainable methods, such as masonry, gabions, and reinforced revetments involving geosynthetics, with vegetation integrated on embankment surfaces (Taiwan Association of Hydraulic Engineer, 2006).

Fig. 1
figure 1

Destruction of the revetments of Keelung River after Typhoon Nari (a) Upstream reach; (b) Downstream reach.

Full size image

Bank vegetation strongly influences channel width (Bunn 1997, Gerstgraser 1999, Gurnell et al. 2002, Giupponi et al. 2019, Sundaravadivelu et al. 2004). In humid climates such as Taiwan’s, vegetative effects often dominate sedimentary factors, although interactions between bank vegetation and material can still be observed (Anderson et al. 2004, Arora et al. 2024, Puijalon et al. 2004). Thorne (James et al. 2004) demonstrated that vegetation considerably influences scour rates by affecting both force and resistance. Vegetation can slow the flow near a bank face by creating backwaters, weakening secondary circulation in bends. However, as the channel widens and deepens downstream, grass exerts less influence because the bank toe lies below the root zone. Consequently, replacing grass with forest does not cause widening due to bank erosion (Barman et al. 2023, Cheng and Nguyen 2011, Rutherfurd 2007, Thorne 1990, Zeng et al. 2002).

Vegetation enhances both riverbank stability and the environmental ecology. An eco-engineering approach, that is an “eco-friendly, engineering-based or conservation-oriented method for environmental and ecological stability, biodiversity conservation, and sustainable management,” can be employed to sustain the balance between human society and natural ecology. Natural-based solutions (NbSs) have substantial advantages in riverbank protection projects. Traditional methods, such as employing concrete structures, effectively prevent erosion but often impose long-term environmental burdens. By contrast, NbSs can be used to leverage natural ecosystems to achieve bank protection; they provide multiple benefits, including enhanced ecological functions and improved com-munity well-being (Sowińska-Świerkosz and García 2002, European Commission, 2015, IUCN, 2020, Johnson et al. 2022). Liu et al. (2022) demonstrated that NbSs not only improve the self-purification capacity of water bodies but also foster biodiversity. Additionally, Barton et al. (2023) emphasized that the root systems of riparian vegetation effectively stabilize soil, thereby reducing the rate and extent of erosion. NbSs are also more resilient and adaptable to varying environmental conditions than traditional structures are. Furthermore, Gordon et al. (2023) noted that NbSs typically incur lower maintenance costs and deliver additional ecological and social benefits. To investigate the effects of eco-friendly revetments on sustainable development, revetments in the Dingnei and Paifu flood protection sectors of the Keelung River embankment were selected for a comparison of the vegetation species changes and coverage rates before, during, and after an embankment reconstruction project (Smith et al. 2023, Zhang et al. 2020, Rey et al. 2019, Mickovski 2021, Wang et al. 2023).

Materials and methods

As observed prior to reconstruction, upstream pollution from Dingnei Reaches was due mainly to domestic waste water. The high embankments and river banks were mostly composed of rocks and only a few types of herbs. Ecological conservation of rivers is based on the maintenance of habitat biodiversity. The stretch between the Dingnei and Paifu Reach was relatively dense with shrubs and herbs, with a number of bamboo; i.e., greater biodiversity than that seen upstream. The reaches downstream from Paifu are near residential areas and the river basin is occupied by economic crops, resulting in a vegetation cover composed of few natural species.

After restructuring was completed, both the bank shapes and revetments were altered considerably. Riparian vegetation strips are frequently employed by river managers to enhance streambank stability. However, although such vegetation’s influence on bank stability is often discussed in the literature, this influence is rarely quantified, and the role of hydrologic processes—some of which may be harmful—is often underestimated (Simon and Collison 2002, Anderson et al. 2004, Bentrup and Hoag 1998). Engineering efforts, such as planning, design, construction, and maintenance, involve ecological evaluation at various stages to assess habitat restoration following water conservancy projects. The present study evaluated the adaptability of vegetation on the revetments and banks of the Keelung River and the extent of improvement in vegetation cover after the restructuring project. Archived satellite images and field surveys were used to identify plant species and calculate changes in vegetation coverage. Additionally, biodiversity analysis was conducted to obtain findings that could inform future projects on how they could sustainably conserve natural vegetation.

This study analyzed changes in vegetation coverage on various revetment surfaces and inside the dike of the Keelung River by using SPOT satellite images from 1999(SPOT-4), 2005(SPOT-4 & SPOT-5), 2010(SPOT-4), and 2020(SPOT-5) during the implementation of “The Keelung River Overall Master Plan”. The imagery was pre-processed to Level-3 to ensure both geometric and radiometric corrections, providing data suitable for quantitative analysis. To facilitate data analysis, only images with less than 10% cloud coverage that were captured during the same season were selected. The SPOT images contain four spectral bands (Red, Green, Blue, and Near-Infrared), which allowed for both visual interpretation and the derivation of the Normalized Difference Vegetation Index (NDVI), a widely used indicator of vegetation vigor, density, and spatial distribution. Remote sensing of visible light and near-infrared bands was applied to NDVI for automated classification, and high-resolution satellite images were used for classification correction and validation.

The range of each reach was defined by excluding the river, cloud-covered, and shaded areas. The remaining land was categorized on the basis of utilization as either river, bare ground, meadow, trees, embankment, or buildings (Fig. 2). Riverbanks on both sides were measured in 50-m increments, with classifications corrected using high-resolution satellite images captured during the same period. Finally, meadows and tree-covered areas were merged into zones marked as vegetation-covered. The comparison between 1999 and 2020 indicates a clear transition from larger areas of bare ground in 1999 (yellow) to substantially greater vegetation coverage in 2020 (light and dark green), demonstrating the effectiveness of eco-engineering interventions in restoring riparian vegetation.

Fig. 2
figure 2

Satellite images (a) 1999; (b) 2020. The satellite imagery was processed using ENVI software (version 5.6, NV5 Geospatial Software, https://www.nv5geospatialsoftware.com/Products/ENVI). The final maps and visualizations were created using ArcMap (version 10.8, Esri, https://www.esri.com/en-us/arcgis/products/arcmap/overview).

Full size image

The vegetation recovery was assessed separately using satellite imagery and field surveys. The hydraulic simulations were designed solely to evaluate hydraulic impacts of different revetment types and vegetation coverage scenarios. To evaluate the hydraulic performance of eco-engineered revetments, HEC-RAS was employed to simulate flow dynamics under different structural configurations. The hydraulic model was set up using a 2D unsteady flow computation framework in HEC-RAS (version 6.3.1), with governing equations based on the Saint-Venant shallow water equations (Eqs. (1)–(3)) for continuity and momentum conservation.

$$\frac{{\partial h}}{{\partial t}}{\text{+}}\frac{{\partial \left( {hu} \right)}}{{\partial x}}+\frac{{\partial \left( {hv} \right)}}{{\partial y}}=q$$
(1)
$$\frac{{\partial \left( {hu} \right)}}{{\partial t}}{\text{+}}\frac{\partial }{{\partial x}}\left( {h{u^2}+\frac{1}{2}g{h^2}} \right)+\frac{{\partial \left( {huv} \right)}}{{\partial y}}= - gh\frac{{\partial {z_b}}}{{\partial x}} - \frac{{{\tau _{b,x}}}}{\rho }+\frac{\partial }{{\partial x}}\left( {{v_t}h\frac{{\partial u}}{{\partial x}}} \right)+\frac{\partial }{{\partial y}}\left( {{v_t}h\frac{{\partial u}}{{\partial y}}} \right)$$
(2)
$$\frac{{\partial \left( {hv} \right)}}{{\partial t}}+\frac{{\partial \left( {huv} \right)}}{{\partial x}}{\text{+}}\frac{\partial }{{\partial y}}\left( {h{v^2}+\frac{1}{2}g{h^2}} \right)= - gh\frac{{\partial {z_b}}}{{\partial y}} - \frac{{{\tau _{b,y}}}}{\rho }+\frac{\partial }{{\partial x}}\left( {{v_t}h\frac{{\partial v}}{{\partial x}}} \right)+\frac{\partial }{{\partial y}}\left( {{v_t}h\frac{{\partial v}}{{\partial y}}} \right)$$
(3)

In the depth-averaged shallow-water equations used by HEC-RAS 2D, h (m) is water depth; u and v (m s−1) are the depth averaged velocity components in the x and y directions; zb (m) is bed elevation and zs = zb + h (m) is the water surface elevation (WSE); g (m s−2) is gravitational acceleration; ρ (kg m−3) is water density; τb, x and τb, y (N m−2) are the bed shear-stress components; νt (m2 s−1) is the eddy viscosity (representing sub-grid turbulent diffusion); and q (m s−1) is the lateral source/sink per unit plan area (positive for rainfall or tributary inflow, negative for infiltration or withdrawals). The bed friction source terms Sfx and Sfy (m s−2) are computed with Manning’s roughness n as Sfx = g·n2·u·|V|/h (1/3) and Sfy = g·n2·v·|V|/h (1/3), where |V| = sqrt(u2 + v2).

Two river cross-sections were selected for analysis: one in the Dingnei Reach and the other in the Paifu Reach. These two reaches were chosen because, among the 11 regulated sections of the Keelung River, the Dingnei Reach is characterized primarily by gabion revetments, while the Paifu Reach is mainly composed of reinforced earth revetments. In contrast, most of the other reaches contain a mixture of multiple revetment types, making them less suitable for a clear comparison. The study focused on the impact of roll gabion and reinforced earth revetments on flood conveyance and riverbank stabilization. Traditional reinforced concrete revetments, while structurally robust, often hinder ecological restoration and river naturalization. In contrast, gabion and reinforced earth revetments offer enhanced permeability and adaptability, potentially reducing flow-induced erosion while promoting vegetation growth.

Two hydrological scenarios were modeled1: a 25-year return period flood with a peak discharge of 789 m³/s and a corresponding water level of 12.5 m, representing a moderate event; and2 the Typhoon Nari scenario, with a peak discharge of 1,420 m³/s and a water level of 14.34 m, representing an extreme rainfall event. The second hydrological scenario represents Typhoon Nari (2001), which we classify as a near 200-year event for the upper Keelung River, compared with the WRA 200-year design discharge of 1,620 m³/s at the upstream control section. The 2001 scenario simulated conditions before eco-engineering interventions, while the 2010 scenario represented post installation of gabion and reinforced earth revetments (Fig. 3), which supported notable vegetative expansion and floodplain transformation.

Fig. 3
figure 3

Simulation of the study area (two river reaches) and HEC-RAS model setup. The satellite basemaps shown in these figures were generated directly within HEC-RAS (version 6.3.1, U.S. Army Corps of Engineers, Hydrologic Engineering Center, https://www.hec.usace.army.mil/software/hec-ras/).

Full size image

Riverbank geometry and cross-sectional profiles were derived from digital elevation models (DEMs) created using engineering design blueprints and processed with GIS tools (Fig. 4). The stepped gabion and reinforced earth revetment geometry from the design drawings was encoded in the DEM by enforcing breaklines along each bench crest and toe, with a 1-m step height per bench, so that the near-bank hydraulic gradients are preserved in the 2D mesh. The computational mesh consisted of 394,860 cells with variable resolution (2–5 m), refined along revetments and channel bends to capture detailed hydraulic process. The hydraulic resistance and infiltration capacity of different revetment types were parameterized using land-use-dependent Manning’s n values and infiltration rates which were assigned based on literature values: 0.025 for gabion revetments (Stordahl & Huffsmith, 2003; Plan, 2011; Shah et al. 2018, Merry 2017), 0.045 for reinforced earth revetments (Mohamed et al., 2006; Stanczak, 2007) and 0.012 for traditional concrete revetments (Table 1). Model outputs were analyzed by comparing flow velocities and water surface elevations across both scenarios. Particular attention was given to upstream and downstream variations to assess whether the gabion and reinforced earth revetments effectively mitigated flow intensity and enhanced bank stability without compromising flood conveyance.

Fig. 4
figure 4

DEM and mesh preparation with stepped gabion revetment representation. (a) Extracted design section showing 1-m bench steps; (b) DEM with enforced breaklines along bench crests and toes; (c) variable resolution unstructured mesh (~ 2–5 m; 394,860 cells) preserving the stepped bank geometry. The satellite basemaps shown in these figures were generated directly within HEC-RAS (version 6.3.1, U.S. Army Corps of Engineers, Hydrologic Engineering Center, https://www.hec.usace.army.mil/software/hec-ras/).

Full size image
Table 1 Infiltration, manning coefficient, and interception as functions of land use.
Full size table

Simulation results were analyzed by comparing flow velocities and water surface elevations across the two scenarios. Particular emphasis was placed on assessing upstream and downstream hydraulic responses to determine the effectiveness of gabion and reinforced earth revetments in mitigating flow intensity while preserving channel conveyance capacity.

Results and discussion

Dingnei Reach spans 5,668 m. The strength of the existing concrete revetment was considered sufficient, and consequently, no changes were made to the revetment, with the exception of the addition of a 1.5 m high reinforced concrete flood wall to ensure it could withstand 200-year frequency floods. Because of the limited hinterland on both sides of the riverbank and the proximity of roads and houses, gabions were used to green the concrete surface, and porous concrete was applied to recreational trails to create an ecologically friendly environment (Fig. 5). Additionally, gabions and riprap were installed at the bank toe to improve erosion resistance.

Fig. 5
figure 5

Left bank reconstruction of the Dingnei Reach (a) Disaster damage (2001, Sep.); (b) Design illustration of riverbank; (c) Early stage of the project (2007, Jun.); (d) Present situation (2023, Aug.).

Full size image

The Paifu Reach spans 8,600 m. The reach’s sharp bends and shallow depth previously led to severe erosion following typhoon-induced flooding, causing riverbank collapse. The construction of floodwalls and the dredging of 335,000 m³ were required to address this problem. After excavation was completed, stone gabions were placed on the foundation, and reinforced earth banks were constructed on top of the gabion foundation to a predetermined level. Geogrids 5 m in length and spaced 1.1 m apart were used to map the reinforced earth bank. An erosion control geomesh was applied to the surface of the reinforced earth bank, and riprap was installed to stabilize the toe and prevent erosion at the revetment base (Fig. 6).

Fig. 6
figure 6

Left bank reconstruction of the Paifu Reach (a) Disaster damage (2001, Sep.); (b) Design illustration of riverbank; (c) Early stage of the project (2007, Jun.); (d) Present situation (2023, Aug.).

Full size image

Vegetation cover rate

The results of satellite image analysis (Fig. 7) revealed that vegetation in the Dingnei Reach covered 70,721 m² in 1999, indicating a coverage rate of 47.61% over the total revetment area of 148,554 m². Reconstruction in 2007 considerably reduced vegetation to 43,725 m², with a coverage rate of 38.89%, that is, a decrease of 8.72% from 1999. However, by 2010, vegetation recovered to 104,423 m², reaching a 70.29% coverage rate, indicating an increase of 22.68% from before reconstruction. By 2020, vegetation coverage further increased to 112,457 m², with a coverage rate of 75.70%, reflecting an additional 5.41% increase from 2010. The use of gabion revetments successfully increased vegetation, contributing to the greening and sustainability of the revetment surface.

Fig. 7
figure 7

The influence of satellite video images on Dingnei Reach (a) 1999; (b) 2020. The satellite imagery was processed using ENVI software (version 5.6, NV5 Geospatial Software, https://www.nv5geospatialsoftware.com/Products/ENVI). The final maps and visualizations were created using ArcMap (version 10.8, Esri, https://www.esri.com/en-us/arcgis/products/arcmap/overview).

Full size image

In the Paifu Reach, vegetation coverage in 1999 was 247,392 m², with a coverage rate of 45.51%. Dredging and disturbances in 2005 reduced vegetation to 152,969 m², reducing the coverage rate to 28.14%. By 2010, after remediation, vegetation dramatically increased to 335,466 m², with a coverage rate of 79.74%, a 34.23% increase from 1999. By 2020, vegetation expanded further to 480,596 m², with an 88.41% coverage rate, indicating that the reinforced earth banks substantially encouraged plant growth and increased the vegetative coverage rate of the revetment (Fig. 8).

Fig. 8
figure 8

The influence of satellite images on Paifu Reach (a) 1999; (b) 2020. The satellite imagery was processed using ENVI software (version 5.6, NV5 Geospatial Software, https://www.nv5geospatialsoftware.com/Products/ENVI). The final maps and visualizations were created using ArcMap (version 10.8, Esri, https://www.esri.com/en-us/arcgis/products/arcmap/overview).

Full size image

Plant adaptation

Plants are crucial to river revetment ecology. Their functions comprise preventing riparian surface erosion, consolidating soil, delaying flood peaks, reducing floods, conserving water, adsorbing pollution, improving water quality, providing landscaping, and providing food sources and habitats for wildlife. However, basal characteristics, climatic conditions, and human factors cause variations in plant species within specific areas. In addition to analyzing the revetment surface’s vegetation coverage, this study examined the species and quantities of the plants in this coverage to understand the vegetation adaptability, species diversity, and other ecological benefits of embankment reconstruction. The survey was conducted over 2 years, covering eight seasons from 2019 to 2020.

The plant species growing on Keelung River revetments comprised ferns, dicots, and monocots. The results of the plant family and species surveys for the Dingnei and Paifu Reaches are presented in Table 2. The following paragraphs summarize the major vegetation species in each reach.

  1. 1.

    Pteridophytes: A total of 8 fern families and 14 species were surveyed, including Christella acuminata (Houtt.) Lev., Cyclosorus parasiticus (L.) Farw., Nephrolepis auriculata (L.) Trimen, Pteris fauriei Hieron., Pteris multifida Poir., Microlepia strigosa (Thunb.) C. Presl, Diplazium dilatatum Blume, Anisogonium esculentum (Retz.) Presl., Athyrium japonicum (Thunb.) Copel., Lemmaphyllum microphyllum C. Presl, Colysis pothifolia (Don) Presl, Microsorum pteropus (Blume) Copel., Osmunda banksiaeifolia (C. Presl) Kuhn, and Lygodium japonicum (Thunb.) Sweet. The Paifu Reach contained 8 families and 13 species, whereas the Dingnei Reach contained only 1 family and 1 species.

  2. 2.

    Dicotyledons: A total of 15 dicot families and 26 species were identified, including Boehmeria densiflora Hook. & Arn., Boehmeria nivea (L.) Gaudich. var. tenacissima (Gaudich.) Miq., Elatostema lineolatum Forst. var. majus Thwait., Debregeasia orientalis C. J. Chen, Ipomoea cairica (L.) Sweet, Ficus fistulosa Reinw. ex Blume., Morus australis Poir., Humulus scandens (Lour.) Merr., Ficus ampelas Burm. f., Wedelia trilobata L., Bidens pilosa L. var. radiata Sch. Bip., Emilia praetermissa Milne-Redh., Sonchus arvensis Linn., Ageratum houstonianum, Erigeron annuus (L.) Pers., Drymaria cordata L., Stephania japonica (Thunb. ex Murray) Miers, Polygonum multiflorum Thunb. ex Murry var. hypoleucum (Ohwi) Liu, Ying & Lai, Polygonum lapathifolium L., Rumex japonicus Houtt., Alternanthera philoxeroides (Moq.) Griseb., Alternanthera bettzickiana (Regel) Nicholson, Amaranthus viridis L., Phyllanthus urinaria L., Leucaena leucocephala (Lam.) de Wit, Pueraria montana (Lour.) Merr., Salix warburgii Seem., Dicliptera chinensis (L.) Juss., Hydrocotyle sibthorpioides Lam., Litsea hypophaea Hayata, Trema orientalis L. Blume, and Cuphea ignea. This study identified 14 families and 27 species in the Paifu Reach, whereas 6 families and 10 species were located in the Dingnei Reach.

  3. 3.

    Monocotyledonous: The identified monocots species included Ischaemum floridulum (Labill.) Warb., Paspalum conjugatum Berg., Cynodon dactylon L. Pers., Eremochloa ophiuroides (Munro) Hack., Imperata cylindrica (L.) Beauv. var. major (Nees) Hubb. ex Hubb. & Vaughan, Alocasia macrorrhiza (L.) Schott & Endl., Cyperus eragrostis Lam., Torulinium odoratum (L.) S. Hooper, and Canna indica L. var. orientalis (Rosc.) Hook. f. Seven species from four families were recorded in the Paifu Reach, and three species from one family were located in the Dingnei Reach.

Table 2 Distribution of vegetation families and species over Revetments.
Full size table

λ: Found species; ν: Dominant species; τ: Endemic species.

In the Paifu Reach, the number of fern species and their overall quantity exceeded those in the Dingnei Reach. The most common fern was N. auriculata (L.) Trimen (Oleandraceae), a lithophyte or terrestrial fern found at lower altitudes, forest clearings, ridges, and masonry walls. The dominant dicot species in Paifu Reach was B. densiflora Hook. & Arn. Additionally, three endemic species were identified: P. multiflorum Thunb. ex Murry var. hypoleucum (Ohwi) Liu, Ying & Lai, S. warburgii Seem., and L. hypophaea Hayata (Fig. 9). The dominant species in the Dingnei Reach was M. floridulus (Labill.) Warb.

Fig. 9
figure 9

Left bank reconstruction of the Paifu Reach (a) P. multiflorum Thunb. ex Murry var. hypoleucum (Ohwi) Liu, Ying & Lai; (b) S. warburgii Seem.; (c) L. hypophaea Hayata.

Full size image

Eco-performance of revetments

The primary revetment type in the Dingnei Reach is gabions, whereas that in the Paifu Reach is reinforced earth revetments. As presented in Fig. 10, both gabion and reinforced earth revetments have effectively restored and enhanced vegetation coverage in the regions, with vegetation in each revetment exhibiting a consistent increase over time. However, the vegetation coverage for reinforced earth revetments is substantially higher than that of gabion revetments. This difference can be attributed not only to the geomesh covering on the reinforced earth revetments, which protects the soil from erosion, but also to the inherent hydrological setting of the Paifu Reach. Located in the middle section of the river, the Paifu Reach is characterized by naturally slower flow velocity, which favors sediment deposition and provides a stable substrate for vegetation. In addition, the rougher surface texture of reinforced earth revetments further reduces near-bank velocity, enhancing sediment retention and supporting a richer plant community.

Fig. 10
figure 10

Variation in revetment vegetation coverage rate.

Full size image

The aforementioned two types of revetments were used to improve vegetation coverage; however, their effects on species diversity differed considerably. Figure 11 llustrates the distribution of each vegetation type across the two river reaches. Dicots are the most numerous and dominant species. Vegetation adaptability in the Paifu Reach surpasses that in the Dingnei Reach, primarily because the Paifu Reach is located in the middle section of the river, where slower flow velocity and the sediment-accumulating surface of the reinforced earth revetment result in a thicker coverage layer. By contrast, the gentler slope of the gabion revetments in the Dingnei Reach provides a more favorable environment for climbing plants.

Fig. 11
figure 11

The concentration of each vegetation type.

Full size image

On the basis of these observations, this study concluded that reinforced earth revetments outperform gabion revetments in improving both vegetation coverage and species richness. Sediment on gabion revetments is prone to erosion due to water flow, resulting in a thin coverage layer that cannot support plants requiring a thick substrate and resulting in a less diverse species distribution. Additionally, gabion revetments exhibit poor water retention, which reduces vegetation coverage on higher sections compared with those closer to the water level. By contrast, the geomesh on reinforced earth revetments traps and deposits upstream sediment, forming a thick vegetative layer. The excellent water absorption properties of nonwoven geotextiles also provide excellent water retention, promoting the proliferation of diverse plant species. Although gabion revetments offer strong erosion resistance and create favorable conditions for fish habitats, they are more suitable for upstream sections with higher flow rates and greater species diversity.

Dense vegetation and plant variety provide ample hiding and nesting spaces for mammals and birds. Seasonal surveys over the course of 1 year recorded 176 birds and 23 small mammals in the Dingnei Reach, whereas the Paifu Reach hosted 337 birds, 51 small mammals, and 8 reptiles. These findings indicate that reinforced earth revetments foster a more diverse ecological habitat than gabion revetments do. Additionally, the presence of numerous frugivorous birds in the Paifu Reach aids in seed dispersal, a key factor in promoting greater diversity of dicots and monocots on the reinforced earth revetments.

Hydraulic analysis of Gabion revetments for riverbank safety

Following the installation of gabion and reinforced earth revetments in 2010, vegetation coverage in the two reaches increased as observed in satellite imagery (as shown in Figs. 7 and 8). In this section, we focus on HEC-RAS simulations to evaluate how such vegetation coverage changes and revetment structure properties influence flood behavior. To evaluate the hydraulic performance of gabion and reinforced earth revetments in enhancing riverbank safety and mitigating flood risks, two cross-sections were analyzed which are Dingnei Reach (CS-1) and Paifu Reach (CS-2). Hydrologic modeling was performed under two scenarios: a 25-year return period rainfall event and an extreme rainfall event corresponding to typhoon Nari in 2001. Nevertheless, the Manning’s roughness coefficient of the revetment structure changed from 0.012 (pre-installation) to 0.025 (gabion) and 0.045 (reinforced earth) due to the porous nature of the revetment structure.

HEC-RAS simulation results demonstrated a consistent reduction in water depth across both scenarios when gabion and reinforced earth revetments were present (Table 3), confirming their beneficial role in flood mitigation. The water depth time series data shows maximum water depth under the typhoon Nari and 25-year rainfall event scenario at Sect. 1 decreased, indicating improved conveyance capacity due to the gabion and reinforced earth installation (Fig. 12).

Table 3 Different hydrodynamic simulation scenarios and the corresponding maximum water depths.
Full size table
Fig. 12
figure 12

Impact of gabion and reinforced earth revetment installation on the water depth time series in CS-1 under Typhoon Nari scenario.

Full size image

To further explore the spatial impact of gabion and reinforced earth revetments, cross-sectional water depth distributions were analyzed across both rainfall scenarios (Figs. 13). Under the typhoon Nari scenario, peak water depth (Table 3) decreased at both cross-sections in 2010 relative to the 2001 concrete baseline (Fig. 13b and d). At CS-1 (upstream), depth decreased from 6.02 m to 5.35 m (gabion) and 5.47 m (reinforced earth). At CS-2 (downstream), depth declined from 16.45 m (2001) to 11.31 m (2010 gabion) and 11.32 m (2010 reinforced earth). The stronger attenuation downstream reflects cumulative effects of the stepped near-bank geometry and lateral losses, with gabion yielding a slightly larger reduction than reinforced earth.

Under the 25-year scenario, water depth changes are smaller. At CS-1, peak depth reduced from 2.98 m (2001) to 2.65 m (gabion) and 2.74 m (reinforced earth). At CS-2, peak depth showed a slight decrease from 6.79 m (2001) to 6.66 m (gabion) and 6.63 m (reinforced earth). Cross-sectional profiles (Fig. 13a and c) visualize these differences and indicate that near-bank attenuation is achieved while overall conveyance along the thalweg is maintained. It shows that eco-engineered revetments substantially lower peak stage under extreme conditions especially at CS-2, consistent with the goal of reducing overtopping risk while preserving channel conveyance.

Fig. 13
figure 13

Cross-sectional depth profiles at peak flow under Typhoon Nari and 25-year scenarios: (a) CS-1, 25-year, (b) CS-1, Nari, (c) CS-2, 25-year, (d) CS-2, Nari.

Full size image

Flow velocity analysis focused on the 25-year return period scenario, were conducted to assess long-term erosion potential. While the flow velocity slightly increased along the centerline, the velocities near the banks generally declined in Sect. 1 (Fig. 14a). This suggests reduced erosive forces along the revetment toe and slopes, thereby contributing to enhanced riverbank stability. Additionally, variations in flow velocity were influenced by the extent of vegetative cover and surface roughness, indicating that post-installation vegetation growth further supports hydraulic performance. At the downstream cross-section (CS-2), the reinforced earth revetment produces a stronger reduction of near-bank (bank-strip) velocity than the gabion alternative (Fig. 14b). Because peak stages are nearly identical between the two eco-revetments under the extreme event (Typhoon Nari) and changes under the 25-year event are modest, revetment selection at this reach should be driven by erosion control, which shows reinforced earth is preferable at CS-2 where reducing toe shear and protecting the bank are the primary objectives, whereas gabions remain suitable where marginally greater stage drawdown, constructability, or maintenance considerations are prioritized. This trend highlights the dual benefit of gabion and reinforced earth revetments, not only do they strengthen structural stability, but they also reduce erosion risk and enhance slope integrity, contributing positively to sustainable river management.

Fig. 14
figure 14

Cross-sectional velocity profiles under 25-year scenarios: (a) CS-1, (b) CS-2.

Full size image

Conclusions

The selection of appropriate revetment types following flood-induced erosion is critical for restoring riverbank stability and promoting ecological resilience in line with sustainable development goals. This study assessed the long-term ecological and hydraulic performance of gabion and reinforced earth revetments along the Keelung River, using satellite imagery and field data from 1999 to 2020.

Although vegetation coverage initially declined due to construction disturbances, both revetment types demonstrated favorable conditions for vegetation recovery post-installation. In fact, vegetation cover not only rebounded but surpassed pre-construction levels. Reinforced earth revetments exhibited approximately 13% greater vegetation coverage than gabion revetments. This enhancement is primarily attributed to the geomesh layer, which effectively captures upstream sediment and, in combination with geotextiles, provides superior water retention—creating a stable substrate for plant colonization and growth. While gabion revetments also supported vegetative development, their coarser substrate and lower moisture retention limited species richness and ecological diversity compared to reinforced earth structures.

In terms of spatial application, gabion revetments are most advantageous in areas where slightly greater stage drawdown, constructability, or ease of maintenance is prioritized. This is particularly true for reaches where reducing peak water levels is critical for flood mitigation. Reinforced earth revetments are preferable where reducing toe shear stress and controlling near-bank erosion are the main objectives, given their stronger effect on lowering bank velocities. Hydraulic simulations under both 25-year return period and Typhoon Nari scenarios showed nearly identical peak stages between the two eco-revetments during extreme events, with modest channel-wide velocity changes. However, both consistently reduced near-bank velocities, lowering erosion risk and enhancing slope stability.

For future river management, results suggest that revetment type selection should be guided by site-specific hydraulic conditions, ecological goals, and maintenance constraints. In reaches with high erosive forces, reinforced earth revetments should be prioritized; in flood-prone reaches where constructability and rapid installation are key, gabions remain effective. Long-term monitoring using UAV and satellite imagery should be institutionalized to track vegetation recovery and structural stability, informing adaptive maintenance. Finally, integrating nature-based engineering solutions into climate adaptation strategies can ensure that flood resilience and ecological restoration are achieved concurrently, providing a sustainable pathway for river corridor management.