Reservoir Level Rise under Extreme Driftwood Blockage at Ogee Crest

Le Lay and Moulin (2007) and STK (2017) studied blockage by driftwood of spillway inlets. Hartung and Knauss (1976) studied a spillway consisting of a 15-m-long ogee crest principal inlet and two lateral side bays as secondary inlets with higher crest elevations. The structure was organized to orient driftwood along the flow, break up long trunks, and force them into the downstream spillway tunnel. In addition, floating chains and interceptors were investigated to hold back driftwood in the reservoir and to keep the inlets free.

Godtland and Tesaker (1994) conducted model tests with driftwood on an ogee with and without piers. They indicated that without piers, the passing probability increases with increasing discharge. To avoid blockage, the free vertical opening between the crest and the bridge should be at least 15% of the maximum trunk length $L_M$, and the relative bay width should be $b/L_M\ge 0.8$, where, $b$ is the bay width and $L_M$ is the length of the largest trunk.

Johansson and Cederström (1995) investigated the passage of trees of various lengths in gated weir fields in a physical model. They found that a single tree approaching a single open gate has the highest probability of passing. If two trees arrive simultaneously or two gates are open, then the passing probability decreases. Furthermore, the probability of passing decreases as the tree length increases. A relative upstream head increase of up to 10% occurred due to blocking.

Hartlieb (2012a, b) conducted physical model tests in a 2.5-m-wide channel, including an ogee weir with piers forming three bays $b=0.5\mathrm{m}$ wide each. Driftwood of different lengths up to $2b$ was supplied either as a single trunk or in batches of five trunks. Hartlieb found that the approach flow velocity is of little significance for the blocking probability, whereas the relative trunk length is relevant. Furthermore, batches have a higher blocking probability than single trunks. For the head increase, the blocking probability is increased with large approach flow Froude numbers as a consequence of a denser driftwood packing at the weir.

Pfister et al. (2013) presented a physical model investigation of piano key weirs (PKWs) and driftwood, including individual trunk tests as well as batch tests. For the individual tests, the blockage probability was a function of the head $H$ and the trunk diameter $D$. For a reservoir-type inflow, trunks typically pass a PKW if the critical flow depth $h_c$ at the weir is larger than $D$ and block a PKW if $h_c<D$. The batch tests showed that the accumulation typically was loose, so the absolute head increase was small. Venetz (2014) continued the work of Pfister et al. (2013), including a study of channel approach flow upstream of the PKW.

Walker et al. (2018) performed model tests of a radial gated ogee crest spillway with two bays blocked by driftwood. For a full gate opening, they reported a reservoir level rise of 1.5 m and a discharge reduction of 35%. The approach to the bays corresponded to channel flow, proving a compact blocking similar to that from river applications.

Furlan (2019) used the experimental setup presented herein and conducted tests with artificial stems of well-defined characteristics. The outcomes were as follows: (1) driftwood experiments should be repeated at least 30 times for individual trunks to achieve errors smaller than 10%; for groups of 32 trunks, 10 test repetitions generated the same accuracy; (2) trunks with a high density (close to that of water) block more frequently than do light trunks; (3) pronounced weir heads $H$ (i.e., large discharges) tend to reduce the blocking probability; (4) numerous open bays tend to block less than one single open bay under the same head; (5) long trunks ($L/b>0.8$) block more frequently at bays of a given width than do short trunks; (6) equal test conditions can lead to a different blockage of a key trunk and thus to a different reservoir head increase; and (7) the horizontal extension of a driftwood carpet is not significant for the head increase under reservoir conditions.

Structures to reduce the impact of driftwood on the rating curve are numerous and usually are linked to a particular case. Perham (1986) discussed measures to keep racks free of driftwood at powerhouse inlets, including net-type racks and air bubbler systems. Perham reported that a particular floating debris boom had deficient performance and that conventional traveling rack cleaners have an insufficient cleaning capacity during intense debris occurrence.

Godtland and Tesaker (1994) described the case of Vinkelfalet Dam (Norway), in which a bridge along the weir was removed to let driftwood pass. USACE (1987, 1997) and FHWA (2005) include overviews of debris characterization, the estimation of debris quantities, debris impact on structures, trash racks, the removal of floating debris, debris passage at spillways, and the disposal of debris.

Built racks and nets were summarized by Hartlieb and Bezzola (2000) and Lange and Bezzola (2006). Möller et al. (2009) described the concept of a side-weir rack for Matteschwelle in Bern, Switzerland. Pfister (2010) applied the side-weir rack concept at a diversion tunnel inlet, and demonstrated that the upper part of the rack remains free for that setup (so that the rating curve is affected only slightly) if a lateral debris retention basin with a rotational flow is provided.

Schmocker and Hager (2013) presented a physical model study of driftwood accumulation at racks. The results indicated a significant effect of the approach-flow Froude number on the driftwood accumulation process, whereas the driftwood properties had only a minor effect on the resulting backwater rise. Schmocker and Weitbrecht (2013) presented an overview of driftwood risk analysis and retention measures for large alpine rivers.

Schmocker (2017) conducted model tests of a weir with two bays, without measures or with a rack installed slightly upstream of the crest. A wooden board with two openings representing a spillway with two weir fields was placed in the channel. Without a rack and for a natural emplacement of the trunks, a relative head rise of up to 30% was observed due to the blocked bays, and a relative head rise of less than 10% occurred if the rack was provided. When manually compacting the wood at the bay or the rack, a relative head rise of up to 40% was observed, versus 10% for the rack.

Schalko et al. (2018) provided a rack in a channel and supplied a predefined driftwood accumulation. The tests were performed partially with a mobile sediment bed. They predicted the channel flow depth increase due to blocking as a function of the initial flow depth and Froude number, as well as of the wood characteristics. Schalko et al. (2019a) presented a literature review of existing measures to reduce the driftwood accumulation probability at bridges and introduce bottom sills to reduce the accumulation effect.

Lassus et al. (2019) summarized model studies from the literature and conducted numerical simulations. They found overall agreement and specified the discharge reduction at an ogee crest equipped with a rack as 7%–22%, depending on the driftwood density, the rack type, and the reference head.

The present study focused on an extreme driftwood occurrence during a pronounced flood event. Only the reservoir head increase was considered, and a full blockage was assumed. As a reference, a gated ogee was chosen (with fully opened gates), because of its frequent application as a spillway inlet control structure. Physical model tests with quasi-natural driftwood were conducted, including different discharge scenarios and bay configurations. Tests were conducted without and with measures to counteract the hydraulic effect of the driftwood blockage.

Systematic experimental tests were performed on a straight channel at the Platform of Hydraulic Constructions (PL-LCH) of Ecole Polytechnique Fédérale de Lausanne (EPFL) (Bénet 2019) (Fig. 1). The channel was horizontal, 10 m long, $B=1.5\mathrm{m}$ wide, and 0.7 m high. At the channel end, an ogee crest with a design head of $H_D=0.15\mathrm{m}$ was transversely mounted. Its crest level was $W=0.42\mathrm{m}$ above the channel bottom to eliminate an effect of the approach flow on its rating curve, because $2H_D<W$ (Vischer and Hager 1999). The crest was equipped with 0.04-m-thick and round-nosed piers. They were fixed on a flexible frame so that $n=1–5$ open bays resulted. All the bays had the same width $b$ for a test, whereas the bay width was varied between 0.175 and 1.500 m between tests. The upstream pier front position either was aligned with the vertical weir front or was overhanging into the reservoir.

The discharge $Q$ was supplied by the in-house pump system and was measured by a magnetic inductive flowmeter (Krohne, Switzerland) up to $\pm 0.5\%$ full-scale. A point gauge fixed approximately 2 m upstream of the weir crest was used to measure flow depths up to $\pm 1\mathrm{mm}$. There, the maximum kinematic head (i.e., for $H_D$) in the channel was on the order of the measurement accuracy and thus negligible.

A flow tranquillizer was provided 6.0 m upstream of the weir to generate homogenous approach flow conditions, which were validated by Furlan (2019).

An implicit validation of the measured parameters is given in Fig. 2(a), comparing the measured rating curve in terms of discharge coefficient $C_d$versus the relative head $\chi$for the test setups with a calculated curve (Vischer and Hager 1999). Fig. 2(a) shows preliminary tests without driftwood and with the model geometries according to those of Tests 1–89. In Eq. (1), $b_e=nb-(n2K_{p}H)$ corresponds to the effective hydraulic width considering the pier contraction (Vischer and Hager 1999), where $K_P$ is the pier parameter. The model values of $C_d$ reach, on average, 98% of the calculated references with a standard deviation of $\sigma =0.02$. Scale effect affecting the rating curve are small because $H\ge 0.050\mathrm{m}$ for the main tests (Hager et al. 2020, Tables 1–4). Smaller heads of $H\ge 0.017\mathrm{m}$ had to be considered for Tests 70–73, 75–78, 82–86, and 99–100. For the driftwood, scale effects were present but not determinant. The trunk shape and surface were natural [Fig. 2(b)], the trunk stiffness was overestimated, and organic fine materials were absent.

Fig. 3 is a definition sketch of the model, showing the ogee crest with the piers. The figure includes some of the countermeasures that were installed during the second test phase, namely the pier overhang and the rack.

The following test procedure was applied:

The following parameters were varied systemically in the model: (1) model discharge $0.005{\mathrm{m}}^3/\mathrm{s}\le Q{\le 0.171\mathrm{m}}^3/\mathrm{s}$, corresponding to $0.098\le {\chi }_R\le 1.029$ (${\chi }_R=H_R/H_D$, based on reference discharges without driftwood); (2) supplied driftwood volume $1V$, $2V$, or $4V$ (with $V=0.038{\mathrm{m}}^3$); (3) model bay width $0.175\mathrm{m}\le b\le 1.500\mathrm{m}$ ($0.40\le b/L_M\le 3.46$), and the number of open bays $n=1–5$; (4) pier front overhang aligned or overhanging up to 0.080 m; and (5) rack type. Tables 1–4 list the test program and selected parameters.

Driftwood was supplied in batches of volumes $1V$, $2V$, and $4V$. The reference volume $V=0.038{\mathrm{m}}^3$ was chosen to focus on an extreme event. If applying a geometrical scale factor of $\lambda =35$ and the Froude similitude, then $1V=410{\mathrm{m}}^3$, $2V=820{\mathrm{m}}^3$ [e.g., Figs. 1(b and c)], and $4V=\mathrm{1,640}{\mathrm{m}}^3$ of solid wood (the loose batch volume is larger). Table 1 also gives the unit model driftwood volume $v$, indicating the solid wood volume which arrives per transverse meter of open bay [$0.025{\mathrm{m}}^2\le (v=\mathrm{\Sigma }V/nb)\le 0.175{\mathrm{m}}^2$ in model dimensions]. For $\lambda =35$, this corresponded to a range of $31–214{\mathrm{m}}^3$ solid wood /m open bay width in a prototype. A batch of $1V$ was composed of 690 trunks and 20 rootstocks, and the batches of $2V$ and $4V$ were composed of the related multiples. A volume of $4V$ thus included 2,760 trunks and 80 rootstocks.

The length $L$ mixture of the trunks within a batch was inspired by in situ observations of Bezzola and Hegg (2007) and Rickli and Hess (2009), who described driftwood accumulations at weirs after floods in Switzerland. The driftwood accumulations are shown in Fig. 4(a) and compared with the present mixture. A batch of $1V$ included model trunks of length $L_M=0.433\mathrm{m}$ (10 trunks, maximum length), $L=0.372\mathrm{m}$ (15 trunks), $L=0.367\mathrm{m}$ (20 trunks), $L=0.300\mathrm{m}$ (35 trunks), $L=0.233\mathrm{m}$ (50 trunks), $L=0.172\mathrm{m}$ (65 trunks), $L=0.167\mathrm{m}$ (85 trunks), $L=0.130\mathrm{m}$ (110 trunks), and $L=0.100\mathrm{m}$ (300 trunks) [Fig. 2(b)]. The supplied driftwood followed the coarse distribution of Rickli and Hess (2009) in accordance with the present approach of an extreme event. For the trunk diameter $D$, $L/D=20$ was assumed. At a scale factor of $\lambda =35$, prototype trunk lengths of $3.5\mathrm{m}\le L\le 15.2\mathrm{m}$ were reproduced.

As stated previously, Furlan (2019) recommended repeating identical driftwood tests several times to achieve statistical relevance. The number of required repetitions decreased, however, with the size of the batches. The maximum number of trunks per batch that Furlan tested was 32, for which 10 repetitions were proposed. We tested batches of 690 ($1V$) to 2,760 ($4V$) trunks plus 20–80 rootstocks, suggesting that less than 10 test repetitions were necessary. Tests 1–3 thus were conducted four times [Tests 1–3, 1a–3a, 4–6, and 7–9 (Table 1) all under identical conditions] to compare the outcome in terms of discharge coefficient $C_d$ (derived from the measured $H$) as a function of the discharge (expressed as ${\chi }_R$). Fig. 4(b) shows that the data of Tests 1–9 were similar per ${\chi }_R$. A statistical analysis indicated that the maximum deviation from the mean head $H$ of all measurements per $Q$ was less than $\pm 2\%$. An absolute error below 4% seemed acceptable in the context of driftwood, so all further tests were conducted only once, without repetition. The rating curve without driftwood (Vischer and Hager 1999) is shown in Fig. 4(b).

Tests were performed with overhanging pier fronts protruding into the reservoir (Figs. 3 and 7). The suggested hydraulic effect of this countermeasure is that the driftwood blockage initiated at the pier front is distant from the critical flow section (located near the weir crest and dominating the rating curve).

The data included three different overhangs, $p$: (1) $p=0\mathrm{m}$ (zero overhang) for Tests 16–33, (2) $p=0.04\mathrm{m}$ (small overhang) for Tests 34–42 and 52–60, and (3) $p=0.08\mathrm{m}$ (large overhang) for Tests 43–51 and 61–69 (Table 2). A value of $p$ equivalent to one (0.04 m) or two (0.08 m) pier widths corresponds to classical values to decrease the pier effect on the discharge capacity of the weir (flow contraction). The pier overhang relative to the head without driftwood varied within $0\le p/H_R\le 1.64$. The following parameters were varied: (1) $n=4$ or 5 open adjacent bays of width $b=0.175$ or 0.335 m (corresponding to $b/L_M=0.40$ or 0.77); (2) discharges of approximately $\chi =H_R/H_D=0.33$, 0.67, or 1.00; and (3) driftwood volumes of $1V$, $2V$, and $4V$.

The data showed a clear positive effect of the pier overhang $p$. The larger the overhang, the less $C_d$ was affected by the blocked driftwood. Fig. 8(a) shows the rating curve for the three test overhangs. All data of the large overhangs (open symbols) nearly collapsed with the prediction without driftwood (Vischer and Hager 1999), whereas those tests without overhangs (solid symbols) had strongly reduced $C_d$ values. The positive effect of the pier overhang was related to the head $H_R$ expressed as $p/H_R$ [Fig. 8(b)].

Note thatis the discharge coefficient measured with wood divided by the computed reference value without wood (Vischer and Hager 1999). The effect of a full blockage under the herein tested extreme conditions was small (typically $\eta \ge 0.95$) if $p/H_R>0.35$.

The absence of piers refers to a physical model without piers (Fig. 9), so that $b=B=1.5\mathrm{m}$. In reality, this condition herein equated to $B/L_M=b/L_M=3.46$. Without piers, horizontally floating trunks face no vertical obstacle, so that a blockage is impossible. The driftwood only touches the transversely oriented weir crest, at least for small discharges.

The test procedure (Table 3, Tests 70–89) had to be modified because a full or partial blockage could be maintained only under the smallest discharges. The discharge thus was increased in small steps (roughly $+5\mathrm{L}/\mathrm{s}$ each), up to the values at which all the wood passed the crest. Three wood behaviors were identified: full blockage for minimal discharges, partial blockage under small discharges, and the passage of the driftwood under slightly higher discharges. Three wood volumes of $1V$, $2V$, and $4V$ were tested, as well as model discharges up to $0.041{\mathrm{m}}^3/\mathrm{s}$ (unit discharges of $q=0.027{\mathrm{m}}^2/\mathrm{s}$).

The discharge coefficient $C_d$ of the fully or partially blocked crest behaved as shown in Fig. 10(a). The $C_d$ value clearly decreased when blockage occurred, as long as ${\chi }_R<0.30–0.35$. For larger discharges (higher ${\chi }_R$), the wood passed, and the $C_d$ value equaled that of the free crest, $C_{dR}$ [Fig. 10(a), Vischer and Hager (1999)]. Pfister et al. (2013) showed that the blocking probability of driftwood on piano key weirs is related to the ratio of the maximum trunk diameter $D_M$ relative to weir head $H_R$. Fig. 10(b) shows the normalized discharge coefficient $\eta$ versus $D_M/H_R$. As in Pfister et al. (2013), three regimes developed: (1) full blockage for $D_M/H_R>0.60$ with $\eta \approx 0.75–0.85$, (2) individual trunk passage for $0.35\le D_M/H_R\le 0.60$ with $\eta \approx 0.75–0.90$, and (3) a free ogee without blockage for $D_M/H_R<0.35$ with $\eta \to 1$. Pfister et al. (2013) indicated almost similar limits of $D_M/H_R<0.35$ for a blockage probability of zero and of $D_M/H_R>1$ for a blockage probability of 1.

Racks were inserted $0.5b$ upstream of the weir (Figs. 3 and 11) for bay widths of $b/L_M=0.77$ and 0.40. The pier overhang was $p=0\mathrm{m}$, and relative heads of ${\chi }_R=0.17$ to 1.02 were provided.

First, a full rack with one bar per pier exactly positioned upstream of the pier axes was studied [Fig. 11(a), Tests 90–98 and 118–120; Table 4]. This rack acted similarly to overhanging piers because its transverse spacing was $b$. Second, a reduced rack including only a bar upstream of every second pier was tested [Fig. 11(b), Tests 99–117; Table 4], with a transverse spacing of $2b$. It was suggested that this second arrangement motivates the rotation of the driftwood parallel to the flow so that the trunks would potentially pass the weir without generating a significant reservoir level rise.

All bars were circular in plan view, with a diameter equal to the pier width (e.g., 0.04 m). The bars thus were massive, but this setup allowed a direct comparison of the present results with those of the overhanging piers. Fig. 12 shows $C_d$ versus ${\chi }_R$ for both rack types.

The full rack included tests with driftwood volumes of $1V–4V$. The tests indicated that the wood had only a small effect on the discharge capacity, so $C_d$ values larger than 95% of the reference without driftwood were achieved. This is in line with the results of the tests considering overhanging piers. The rack was positioned at $0.57\le (0.5b/H_R)\le 3.35$, which was much farther upstream (in the reservoir) than the minimum distance of $p/H_R>0.35$ required for an uninfluenced operation in the context of the pier overhang [Fig. 8(b)].

The reduced-rack tests included only the maximum wood volume of $4V$, because the wood volume had no significant effect in previous tests. The rack was positioned at $0.58\le (0.5b/H_R)\le 6.76$, upstream of the minimum distance $p/H_R>0.35$ required for free operation in the context of the pier overhang [Fig. 8(b)]. The outcome of the reduced-rack tests was inhomogeneous. The presence of the wood partially reduced the discharge capacity, and minima values of 90% of the reference $C_d$ without driftwood were achieved. Some trunks leaned against the pier nose where the rack bar was absent, and thus reached into the critical flow zone on the weir crest, affecting the weir capacity. A systematic rotation of the trunks in the streamwise direction, combined with their passage of the weir, was not observed (Schalko et al. 2019a). This certainly was due to the wood supply mechanism chosen herein, in which the full volume was provided in one single batch. We suppose that a reduced rack could be more efficient and (partially) could orient the trunks if the wood were supplied continuously and in small batches of, e.g., five trunks.

Full racks maintained the discharge capacity expressed by $C_d$ by trend slightly better than did reduced racks. However, fewer data were available for full than for reduced racks. Reduced racks clearly improved the situation up to large discharges (${\chi }_R<0.90$), compared with the setup without countermeasures [Fig. 5(b)], but were less efficient for the design case (${\chi }_R\approx 1$).