Reservoir sedimentation · cross-country cost assessment
Every year, rivers carry eroded soil into the world's reservoirs, permanently burying water-storage capacity that society depends on for drinking water, irrigation, hydropower, and flood protection. Panagos et al. (2024) estimated that removing the sediment delivered by water erosion alone to EU + UK reservoirs would cost €2.3 billion per year (≈135 million m³), rising to €5–8 billion per year once all soil-loss processes (gullies, landslides, quarrying…) are counted — more than 1 billion m³ accumulating across ~5,000 reservoirs. This page extends that lens to nine other economies, estimating both the volume of annual sediment accumulation and the cost of removing it, from scholarly, government, and industry sources in seven languages.
Estimated annual sediment accumulation in reservoirs, by country/region. Bars show the plausible range assembled from national surveys, measured capacity-loss rates and modelling; the tick marks the central estimate. Note the logarithmic scale — the largest fleets (USA, Brazil, EU+UK, India) each trap around a thousand times more sediment than the smallest here.
Annual sediment accumulation in reservoirs
Million m³ per year · log scale · hover a bar for the basis of each estimate
EU+UK shown for all soil-loss processes (1,000–1,670 M m³/yr, per Panagos et al. 2024); water erosion alone contributes 135 M m³/yr of that flux. Volumes are bulk deposited sediment (in-situ density ≈1.2 t/m³ where converted from mass).
Notional annual cost of removing each year's sediment accumulation, obtained — exactly as in Panagos et al. — by multiplying the accumulation volume by the locally documented cost per cubic metre of sediment removal (low bound: cheapest applicable technique mix such as hydraulic flushing or dry excavation with aggregate resale; high bound: full mechanical/hydraulic dredging with transport and disposal at local prices). These are potential costs of keeping storage in balance, not current budgets: most countries today remove only a small fraction of what accumulates, and pay instead through permanently lost storage.
Estimated annual cost of removing the yearly sediment accumulation
€ billion per year · log scale · hover a bar for the unit costs behind it
EU+UK range spans the paper's published estimates: €5–8 B/yr with a conservative flushing-weighted mix up to €16–27 B/yr if all sediment were mechanically dredged at the EU mean of €16.8/m³; the €2.3 B/yr headline covers water-erosion sediment only.
Unit removal costs differ by an order of magnitude between economies — labour costs, disposal rules, haul distances, wet vs. dry removal, and whether dredged sand can be sold all matter. This spread, more than sediment volume itself, explains why Japan's small fleet costs nearly as much to maintain as Indonesia's rapidly silting one.
Documented sediment-removal unit costs
€ per m³ · linear scale · hover for the local prices and sources
Reference points: EU mean for mechanical dredging €16.8/m³ and flushing ≈€5/m³ [1]; dredging can fall to €2/m³ where fine sediment is simply discharged downstream (Tuyamuyun complex, Central Asia [37]).
Cumulative share of initial reservoir capacity already buried under sediment (2022) and projected by 2050, from the UNU-INWEH global assessment (Perera et al. 2023) — with Taiwan's directly measured national figure added. Once storage is gone, replacing it means new dams at far greater economic and environmental cost: this is the loss that removal budgets are (mostly failing) to prevent.
Reservoir storage lost to sedimentation, % of initial capacity
2022 estimate → 2050 projection · hover for details
UNU-INWEH model applies bathymetry-derived regional loss rates to 47,403 ICOLD-registered dams; where national surveys exist they can differ (e.g. Japan's MLIT survey of 794 dams shows ≈8% of gross storage silted, versus the model's 39% — see the Japan rationale below). Taiwan's 29.4% is a measured national average (Legislative Yuan / WRA).
Each card states the data used, the arithmetic, and the caveats. Bracketed numbers refer to the reference list at the bottom.
The anchor study [1]. Water erosion delivers 164 (±13) Mt/yr ≈ 135 (±10) M m³/yr of sediment to EU+UK river networks (WaTEM/SEDEM model); at the mean mechanical-dredging price of €16.8/m³, compiled from 33 studies in 20 countries (range €2.4–55/m³), removal costs €2.3 (±0.9) B/yr. Extrapolating measured capacity-loss rates (0.26–0.65 %/yr) across the 258–383 B m³ of European reservoir capacity captures all soil-loss processes and yields 1–1.67 B m³/yr — costing €5–8 B/yr with a conservative flushing-weighted mix (€5/m³), €10–18 B/yr with Italy's observed technique shares (€10.7/m³), or €16–27 B/yr if fully dredged. Italy (€1.05 B/yr), Spain (€0.39 B), and Germany (€0.32 B) dominate the regionalised water-erosion costs; the UK's is €33 M/yr.
Sources: [1] Panagos et al. 2024; [2] ICOLD; [3] Basson 2009.
The US built ~970 km³ of reservoir capacity; net capacity has fallen from a peak of ~850 km³ in the late 1980s to ~810 km³ today — roughly 1.2 B m³ lost per year on average [4]. Gross sedimentation is somewhat higher: assumed loss rates are 0.1–0.5 %/yr for large and 0.4–2 %/yr for small reservoirs [4,8]; Reclamation's portfolio averages 0.19 %/yr [6], and a USACE sample of just 162 reservoirs measured 0.26 B m³/yr of deposition [5]. We adopt 1.2–2.0 B m³/yr. Unit prices: hydraulic dredging $8–24/yd³ and mechanical $13–48/yd³ in 2020 dollars (Illinois EPA, in the SedHyd national cost model [7]); flushing/sluicing can reach $2–5/m³ where feasible. Adopting $5–30/m³ (≈€4.5–26) gives €5.4–52 B/yr. Actual expenditure today is a small fraction of this — the dominant US "cost" is un-replaced storage, worth ~$38 B (2003 $) of lost capacity at Reclamation alone [6].
Sources: [4] Randle et al. 2021; [5] USACE 2025; [6] USBR; [7] Anchor QEA/SedHyd 2020; [8] Graf et al. 2010.
Canada has no national sedimentation survey; its boreal-shield catchments have among the world's lowest sediment yields, but its storage is enormous (Manicouagan, Williston, Caniapiscau…) and prairie reservoirs such as Lake Diefenbaker trap substantial loads. The UNU-INWEH model puts Canada's 491 large dams at ~20% storage lost by 2022 and 29% by 2050 (≈0.3 %/yr) [9] — almost certainly an overestimate for shield basins, since it transfers global regional rates. We conservatively adopt 30–100 M m³/yr (an implied fleet-average loss rate of only ~0.005–0.02 %/yr of Canada's ~500+ B m³ of storage), and North-American unit prices (€4.5–26/m³). Actual reservoir dredging in Canada is rare and mostly tied to navigation and intakes; hydropower operators primarily manage sediment passively [39].
Sources: [9] Perera et al. 2023; [39] Hydropower Sustainability Alliance guidelines; [8] North-American rates.
Japan surveys dam sedimentation annually. By FY2018, the 794 dams in MLIT's public survey had accumulated 1.76 B m³ (≈2.2 M m³ per dam) [11,12]; the largest single accumulator, Sakuma Dam, traps ~2.08 M m³/yr, and about 10% of dams gain ≥100,000 m³/yr [11]. Dividing the accumulated volume by the fleet's average service life implies a national accumulation of roughly 30–40 M m³/yr — small in volume (steep catchments, but limited fleet size: total storage ≈23 B m³, of which ≈8% is silted, far below the UNU model's 39% [9,11]). Costs, however, are the world's highest: excavation, haulage and disposal in Japan typically total ¥2,000–8,000/m³ (≈€12–46) once disposal fees of ¥1,000–5,000/m³ are included [13,14], and the national five-year emergency dredging programme (rivers + dams) is funded at ¥490 B for FY2020–24 — ≈€0.6 B/yr [13]. Sediment-bypass tunnels (Asahi, Miwa, Koshibu) are Japan's long-term answer precisely because dredging at these prices is unsustainable.
Sources: [11] 末次 2022 (水利科学); [12] MLIT堆砂状況; [13] serita.jp / 総務省 緊急浚渫推進事業; [14] disposal price lists (兵庫県 etc.).
Relative to its size, Taiwan has the world's most severe reservoir sedimentation: the national average siltation ratio is 29.4% of capacity [20], driven by typhoons on young mountain geology (Morakot 2009 added ~90 M m³ to Tsengwen Reservoir alone). Since 2016 the WRA's capacity-maintenance programme has removed an average of 14.6 M m³/yr (2017–2025), a record 19.3 M m³ in 2025, and 10.2 M m³ in the first half of 2026 — roughly balancing average annual inflow of sediment [15,16]. Removal blends dry excavation, cross-lake pumping barges, and hydraulic desilting during typhoon floods. Costs: full dry excavation with screening at Shihmen runs ≈NT$500/m³, partly offset by selling screened sand at ≈NT$198/m³ [17]; hydraulic desilting tunnels are far cheaper per m³ once built (the Amuping tunnel's first seven operations avoided ≈NT$562 M of excavation [18]), and Taipower is investing >NT$5 B in hydraulic desilting at Wushe [19]. We adopt NT$100–500/m³ (≈€3–14.5).
Sources: [15] 水利署; [16] CNA/經濟日報 2026; [17] CNA 2026-08; [18] 自由財經 2025; [19] CNA 2020; [20] 立法院.
India's Central Water Commission compendia (243 surveyed reservoirs, 2015; updated 2020) find average annual capacity loss of ~0.4 %/yr fleet-wide, and ~0.9 %/yr of live storage in re-surveyed reservoirs [21,22]; a quarter of reservoirs older than 50 years have lost ≥30% of capacity, and UNU projects a 26% national loss by 2050 [9,10]. Applied to India's ~257 B m³ of live storage in large dams, 0.3–0.5 %/yr gives 750–1,300 M m³/yr. Unit prices are low: the Dredging Corporation of India charges ₹180–240/m³, and commercial reservoir-dredging services list ₹150–200/m³ [23]; disposal and deep-reservoir work push toward ₹300/m³ (≈€1.5–2.9/m³ adopted). Large-dam desilting is nonetheless rarely executed — Bhakra's first desilting came after 62 years — so nearly all of this cost is currently "paid" as lost storage.
Sources: [21] CWC Compendium 2015; [22] CBIP/CWC 2019; [23] DCI & market rates; [9,10] UNU-INWEH.
No national compendium exists, but measured per-reservoir rates are among the world's highest: Mrica (PB Soedirman) receives 4–6.6 M m³/yr and has shrunk from 150 to ~15 M m³ of capacity [24]; Wonogiri/Gajah Mungkur ~3 M m³/yr [28]; Sengguruh is 95% full at ~2 M m³/yr [26]; Kedungombo ~1.95 M m³/yr [27]. With 200+ large dams (~20–25 B m³) on volcanic, deforestation-prone catchments and the UNU rate of ~0.35 %/yr as a floor, we adopt 70–200 M m³/yr nationally. Documented dredging contracts are cheap by world standards: Sutami ≈Rp4.8 B for ~410,000 m³ (≈Rp12,000/m³) and Sengguruh ≈Rp15,000/m³ [25]; confined disposal or long pumping raises this several-fold, so we adopt Rp12,000–50,000/m³ (≈€0.6–2.7). Budgets, not prices, are the binding constraint — dredging at Mrica removes under 1 M m³/yr against 4+ M m³/yr of inflow [24].
Sources: [24] Radar Banyumas / Banyumas Ekspres; [25] ANTARA Jatim; [26] Malang Posco Media; [27] Siklus J. Teknik Sipil; [28] Wikipedia ID (Gajah Mungkur); [9] UNU.
Malaysia's sedimentation is concentrated: the Cameron Highlands scheme (Ringlet reservoir) is the infamous case — land-clearing for agriculture has forced continuous dredging, and TNB reports spending more than RM100 million a year maintaining its dams against sedimentation [29]. Large lowland and Borneo storages (Kenyir, Bakun, Murum) have low specific yields [30,31], keeping the national volume modest; with ~70 large dams and >50 B m³ of (mostly young) storage, UNU estimates only ~8% lost by 2022 [9]. We adopt 10–30 M m³/yr and regional dredging prices of RM10–30/m³ (≈€2–6).
Sources: [29] TNB Genco 2022; [30] MDPI Water 2023 (Kenyir); [31] WIT Trans. (Murum); [9] UNU.
Brazil's ~900 large dams hold ~600 B m³ — the largest storage in this comparison. Measured national loss rates span 0.2 %/yr (de Araújo et al.) to the UNU projection of ~0.35 %/yr (59 B m³ lost 2022–2050) [9]; Brazilian literature commonly cites ~0.5 %/yr for smaller reservoirs, and the semi-arid Northeast's açudes are already heavily silted [32]. That yields 1.2–2.1 B m³/yr. Unit prices: dredging/desassoreamento contracts typically run R$20–60/m³ (≈€3.2–9.6), partly offset where dredged sand is sold to construction [32,33]; the São Paulo metropolitan region alone spends ≈R$160 M/yr on de-silting its water-supply reservoirs [32]. As elsewhere, most of the potential cost is unpaid: hydropower operators dredge only around intakes, while dead storage quietly fills.
Sources: [9] Perera et al. 2023; [32] ABRH / Ceará water-loss studies; [33] market dredging rates (LF Ambiental etc.).
Mexico's dams store ~150 B m³ (210 major presas plus thousands of smaller ones). Bathymetric re-surveys show pervasive azolvamiento: of 25 major reservoirs assessed with a standard diagnostic, 17 had already exhausted their design dead-storage allowance [34,35], and 35 re-surveys in the centre-northeast confirm high, variable rates [35]. UNU projects 28% lost by 2050 (≈0.3 %/yr) [9]; we adopt 0.2–0.33 %/yr → 300–500 M m³/yr. Desazolve (de-silting) market prices run ~MX$40–120/m³ (≈€1.9–5.6) depending on wet/dry method and haul [36]. Actual removal is negligible against inflow: CONAGUA's clean-up campaigns extract only ~0.5 M m³ of azolve a year nationally [38].
Sources: [34] Ingeniería del Agua; [35] SciELO (35 batimetrías); [36] análisis de precios unitarios; [38] CONAGUA/El Universal; [9] UNU.
Mirroring the EU study so the countries are comparable.
Step 1 — Annual sedimentation volume. Preference order: (a) national measured surveys (Japan's MLIT annual survey, Taiwan's WRA programme data, India's CWC compendia); (b) measured national capacity-loss rates × installed storage (USA, Brazil, Mexico); (c) modelled regional rates from the UNU-INWEH global assessment, sanity-checked against local bathymetry (Canada, Indonesia, Malaysia). Where mass is reported, a bulk density of 1.2 t/m³ converts tonnes to cubic metres, as in the EU study.
Step 2 — Unit removal price. Local documented prices only: contract values, utility budgets divided by removed volumes, dredging-market rate cards, and disposal-fee schedules — in the local language wherever possible. The low bound reflects the cheapest technique that plausibly scales (flushing, sluicing, dry excavation with aggregate resale); the high bound reflects mechanical/hydraulic dredging with transport and confined disposal.
Step 3 — Cost. Cost = volume × price (Panagos et al., Eq. 2), reported as a low–high range with a central estimate at the geometric mean. We deliberately do not net out aggregate-resale revenue or add environmental externalities; both can shift results materially in either direction.
Known limitations. (i) Volumes and prices are correlated with technique choice — multiplying extremes overstates the true spread; (ii) "potential removal cost" is a what-if, not a budget: the realistic alternative mix includes bypassing, sluicing, watershed conservation and, sometimes, planned storage retirement; (iii) exchange-rate and inflation noise of ±10–15% is smaller than the data uncertainty; (iv) model-based storage-loss figures (UNU) can diverge sharply from national surveys — Japan is the clearest example — and we flag every such case in the rationale.
Language of each source is tagged. Links verified August 2026.