- Email: [email protected]
Effects of proposed dam removal on ice jamming and bridge scour on the Clark Fork River, Montana
Cold Regions Science and Technology 55 (2009) 186–194
Contents lists available at ScienceDirect
Cold Regions Science and Technology j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / c o l d r e g i o n s
Effects of proposed dam removal on ice jamming and bridge scour on the Clark Fork River, Montana A.M. Tuthill a,⁎, K.D. White a, C.M. Vuyovich a, L.A. Daniels b a b
U.S. Army Cold Regions Research and Engineering Laboratory, 72 Lyme Road, Hanover, NH 03755-1290, United States Seattle District, Missoula Business Office, U.S. Army Corps of Engineers, 1600 North Avenue, Missoula, MT 59801-5512, United States
a r t i c l e
i n f o
Article history: Received 31 October 2007 Accepted 5 September 2008 Keywords: Ice jams Ice jam bed scour Ice jam modeling Dam removal Remediation of contaminated sediment Clark Fork River Superfund Project
a b s t r a c t The Milltown Dam, at the confluence of the Clark Fork and Blackfoot Rivers in Montana, lies at the downstream end of the nation's largest Superfund Project. Since its construction in 1906, several hundred thousand tonnes of metal-contaminated sediment from upstream mining activities, primarily copper, have accumulated in the dam impoundment. A large amount of this sediment was scoured during a 1996 ice jam event on Clark Fork and Blackfoot Rivers and transported downstream. The EPA remediation plan calls for phased removal and off-site disposal of much of the contaminated sediment, and removal of the Milltown Dam and a smaller mill dam upstream on the Blackfoot River. As much as possible, the river channels will be restored to their pre-project natural morphology. This study assessed ice impacts associated with the restoration plan, specifically where ice jams and related ice jam scour might occur with and without the dams in place. Also addressed was the effect of dam removal on potential ice-related scour around the piers of five bridges that cross the Blackfoot River just upstream of the Milltown Dam. Because this type of problem is relatively new, relevant engineering guidelines are lacking. Shortcomings of the tools and methods used are discussed. It is hoped that the study described herein will assist those addressing similar problems in the future, and also point to areas where analysis methods and tools could be improved. Published by Elsevier B.V.
1. Introduction The Milltown Dam, built in 1906 on the Clark Fork River, lies just downstream from the confluence of the Blackfoot River about 11 km east of Missoula, MT (Figs. 1 and 2). Since construction, some 5 million cubic meters of sediment have accumulated in the dam's mile-long impoundment. Much of this is material contaminated with metals (As, Cd, Cu, Pb, and Zn) as a result of the historic mining activities upstream at Butte and Anaconda. Since 1982, the EPA has listed the Milltown Reservoir–Clark Fork River on the National Priorities List, and it is now part of the nation's largest Superfund site. Extensive investigations have led to a remedial action plan consisting of phased removal of the Milltown Dam and the contaminated sediments in its impoundment. Ultimately, the dam decommissioning may result in a drop in bed elevation of up to 9 m. The small timber crib Stimson Dam, located 1.6 km upstream on the Blackfoot River, was removed in 2005 to allow fish passage and improve safety. The Clark Fork and Blackfoot Rivers experience infrequent ice events, the most severe in recent history occurring on February 9–10, ⁎ Corresponding author. E-mail address: [email protected] (A.M. Tuthill). 0165-232X/$ – see front matter. Published by Elsevier B.V. doi:10.1016/j.coldregions.2008.09.004
1996. On February 9, a massive ice run progressed down the Blackfoot River to jam about 500 m upstream of the Stimson Dam (Figs. 3 and 4). On February 10, a second breakup ice run on the Clark Fork jammed briefly before fracturing its way through the Milltown Dam impoundment. This ice event scoured large amounts of fine-grained metalcontaminated sediment and deposited them downstream (Moore and Landrigan, 1999). Our study evaluated potential effects of the Milltown Dam removal on the ice regime of the Clark Fork and Blackfoot Rivers in the vicinity of the project. Of particular interest were the potential consequences of displacing ice jams from their traditional locations upstream of the Milltown Dam impoundment to downstream sites, and the possibility of ice-related scour in the vicinity of the piers of the five bridges that cross the lower Blackfoot River. The study addressed this increasingly common problem using conventional ice engineering methods and models. It began with a review and analysis of past ice events and a program of field observations during the winter of 2005. The HEC-RAS hydraulic model (US Army, 2000) was then used to simulate freezeup covers and breakup ice jams for the pre- and post-dam removal cases. Using HECRAS and basic ice engineering theory, we evaluated the potential for post-dam removal ice jam scour around the bridge piers on the lower
A.M. Tuthill et al. / Cold Regions Science and Technology 55 (2009) 186–194
187
Fig. 1. Map of study area (from Moore and Landrigan, 1999).
Blackfoot. Finally, the study identified possible ways to mitigate adverse effects of the dam removal on the ice regime. Limitations of the approach and remaining uncertainties are discussed. 2. History of ice jams on the Clark Fork and Blackfoot Rivers near Missoula Although significant ice jams were reported on the lower Blackfoot in 1974 and on the Clark Fork below Missoula in 1984, the February 1996 event stands out as the most severe by a large margin. In the preceding three weeks, extreme cold caused thick ice covers on the Blackfoot and Clark Fork Rivers. A maritime weather system brought rapid thawing and rain to the region, releasing more than 65 km of ice on the Blackfoot as a dynamic breakup front that progressed downstream at 3.5–4.5 m/s. The Blackfoot ice run grounded and stopped about 2.4 km upstream of the Milltown Dam to form a 5-kmlong ice jam. After-the-event field inspections and photos revealed shear wall heights as great as 5 m, and top-of-ice elevations nearly level with the Highway 200 roadway along the S-curves above Bonner (Figs. 3 and 4). Likely reasons for the ice jam were the loss of momentum as the breakup front passed the sharp bend above Bonner and impacted the flat-lying sheet ice cover and gravel deposits of at the head of the Stimson Dam Pool. On the Clark Fork, a number of smaller ice jams occurred as a breakup front progressed downstream, ultimately fracturing its way through the 30-cm-thick sheet ice cover on the Milltown Dam pool on February 10. During the next several days, continued high flows and much more ice passed through the Clark Fork portion of the pool. Many thousands of tonnes of sediment were eroded from the impoundment through a combination of hydraulic scour and mechanical gouging of the bed and banks by moving ice floes. Moore and Landrigan (1999)
described the river as “running black with sediment amongst the ice blocks” and, after the event, they observed 20-cm-thick deposits of fine-grained sediment on ice blocks lining the river banks downstream of the reservoir. 3. Frequency and severity of ice events In order to approximate the probability of recurrence of a 1996-like event and identify possible unrecorded ice events, a hindcasting analysis was carried out using historic hydro-meteorological data from 1929 to 2005. For each winter, maximum ice thickness (ti) in cm was calculated from accumulated freezing degree (C)-days (AFDD) (US Army, 2002). pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ti ¼ C maxAFDD:
ð1Þ
C is a calibration coefficient calculated from observed ice thickness and AFDD data from the winters of 1996 and 2005. Following the February 1996 ice event, the observed thickness of the fractured sheet ice on the Milltown Dam Pool was about 30 cm. AFDD had peaked three days prior to the event at 350 °C, giving a calculated C value of 1.6. During the winter of 2005, AFDD peaked on January 17 at 222 °C when the average measured ice thickness on the Milltown Pool was 24 cm, again yielding a C value of 1.6. Note that these calculated ice thicknesses apply only to relatively quiescent sections of river characterized by thermal ice growth. Ice-hydraulic model calculations of thicker freezeup ice accumulations in faster-moving sections of river are described later. Based on the known historic events of 1996 and 1974, empirical criteria were developed to identify other likely ice jam events on the
188
A.M. Tuthill et al. / Cold Regions Science and Technology 55 (2009) 186–194
Fig. 2. Clark Fork–Blackfoot confluence and Milltown Dam in 2005.
lower Blackfoot River. These prediction criteria were based on a review of daily precipitation, air temperature and discharge data for all winters from 1922–2005. It is important to note that these criteria are site-specific and apply only to this case study and these particular discharges and ice thickness ranges. Fig. 5 plots base-to-peak discharge increase against calculated ice thickness, suggesting causal factors for historic ice jams on the lower Blackfoot River to include a pre-breakup ice thickness of 25 cm or greater and a flow increase equal to or greater than 57 cm. Ice jams were considered possible for ice thicknesses equal to or greater than 25 cm with discharge in the 40- to 57-cm range. In addition to the two known ice jams, one likely ice jam event (1971) and three possible events (1963, 1984, and 1986) were identified on the lower Blackfoot for the 77-year period of
record. The hydro-meteorological data were then reviewed as an independent check that the predictions were reasonable. Increases in discharge (ΔQ) and ice thicknesses (ti) were ranked and assigned probabilities (Fig. 6). Assuming ΔQ and ice thickness ti are independent variables, the exceedance probability of an historic or hindcast event can be estimated by multiplying the two probabilities. By this approach, the probability of a 1996-like event occurring on the lower Blackfoot was estimated at about 0.6%. Results of the hindcasting and probability analyses contain uncertainty as a result of the lack of historic event data (only two known events on the lower Blackfoot). In spite of this uncertainty, the analyses support the assertion that the 1996 event was quite rare, particularly in terms of discharge. Because of its low frequency, the 1996 ice jam was considered suitable for use as the design event in subsequent analyses. 4. Hydraulic modeling
Fig. 3. Blackfoot River ice jam near its downstream end, Feb. 1996.
The HEC-RAS model was used to analyze ice formation and breakup processes under existing and post-project conditions. HEC-RAS was chosen for the ice modeling due to its use in other phases of the project and the fact that the river geometry was provided in HEC-RAS format. The modeled reach included 32 km of the Clark Fork plus 11 km of the Blackfoot River from the Bonner Gage downstream to the mouth. The model was modified to simulate post-project channel geometry by removing the Milltown and Stimson Dams and adjusting the channel width and minimum bed elevation to conform to a preliminary restored channel design by Westwater Consultants et al. (2005). The restored channels had bottom widths of 60 m above the confluence and 75 m downstream. The restored Blackfoot channel followed the existing straight alignment while, for the Clark Fork restored channel, a sinuosity of 20% was assumed.
A.M. Tuthill et al. / Cold Regions Science and Technology 55 (2009) 186–194
189
Fig. 4. Map showing the 1996 ice jams (red) and possible post-project ice jam locations (blue and green). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
In this study, probable ice cover types were estimated initially from open water velocity calculated by HEC-RAS using the long-term average January discharge of 20 m3/s. Perham (1983) and others have used similar criteria to predict ice retention and freezeup ice cover
type in the design of ice control structures. The criteria predict thermally grown sheet and border ice for v ≤ 0.3 m/s, juxtaposed frazil ice accumulations for v between 0.3 and 0.7 m/s, and shoved frazil floes for v between 0.7 and 1.5 m/s. Where water velocity exceeds
Fig. 5. Blackfoot discharge increase vs. calculated ice thickness, with fields identifying likely and possible unrecorded ice jam events.
190
A.M. Tuthill et al. / Cold Regions Science and Technology 55 (2009) 186–194
Fig. 6. Probability of ice thickness and base-to-peak discharge increase for the lower Blackfoot River.
about 1.5 m/s, a section of river may remain open all winter. Though water velocity criteria provide a first estimate of ice cover type, it is important to note that many other factors influence river ice formation. These include channel geometry, water depth, turbulence, water and air temperature, wind velocity, sky cover and surface ice concentration (Matousik, 1984a,b). The HEC-RAS ice routine was used to calculate the thickness of the freezeup ice covers due to juxtaposition and shoving. The simulations used typical freezeup ice accumulation parameters, including an under-ice Mannings ni of 0.04, an ice accumulation porosity e of 0.4,
an angle of internal friction Φ for the granular ice material of 45°, and an under-ice erosion velocity veros of 1.2 m/s (White, 1999). Based on the historic AFDD analysis calibrated to the observations of 1996 and 2005, the sheet ice thickness on the pools and low velocity areas was assumed to be 30 cm. The freezeup simulations used the long-term January averages of 20 and 25 m3/s for the Blackfoot and Clark Fork Rivers, respectively. With the exception of the sheet ice on the dam impoundments, the HEC-RAS ice jam force balance predicted ice covers of juxtaposed and shoved frazil floes at most locations within the study reach. The
Fig. 7. HEC-RAS-simulated profile of the 1996 ice jam on the lower Blackfoot River under existing conditions. Blackfoot discharge = 113 cm.
A.M. Tuthill et al. / Cold Regions Science and Technology 55 (2009) 186–194
equilibrium frazil transport theories of Shen and Wang (1995) predicted some additional frazil deposition beneath the impoundment ice covers, but in terms of overall ice volume, this additional frazil volume was not significant. For the post-project case, HEC-RAS predicted frazil ice covers about 1.5 m thick on the restored river channels. Based on the simulations, the change from sheet ice covers on the existing impoundments to juxtaposed and shoved ice accumulations on the restored channels would increase the total ice volume within the study area by about 30%. The February 1996 event served as a worst-case scenario in terms of discharge and ice supply for the breakup ice jam modeling. Breakup jams for existing conditions were modeled at the upper end of the former Stimson Dam pool on the Blackfoot, and at the upstream end of the Clark Fork portion of the Milltown Dam reservoir. To model breakup on the Blackfoot, the February 9, 1996, daily average discharge of 113 m3/s was used. Clark Fork discharge was 198 m3/s above the Blackfoot confluence and 311 m3/s below. Fig. 4 shows locations of the observed ice jam on the Blackfoot River and the likely location of the jam at the upstream end of the Clark Fork portion of the Milltown Dam impoundment. Fig. 7 shows the simulated profile of the 1996 ice jam on the lower Blackfoot River using reasonable input parameters of erosion velocity veros = 1.2 m/s; angle of internal friction for the ice accumulation Φ = 45°; ice accumulation porosity e = 0.4; and Mannings resistance coefficient for the ice underside ni = 0.08 (White, 1999). The simulated profile matched observed top-of-ice heights fairly well along the Highway 120 S-curves, but the modeled top of jam was slightly higher than the observed top of ice near the upstream end of the jam (Fig. 7). The bulge in the ice jam profile between km 5 and 6 may result from the steeper water surface slope just upstream of this location. The HEC-RAS force balance under-predicted observed ice thickness and stage near the downstream end of the jam, possibly because of the inability of HEC-RAS to model ice thickening from momentum transfer as the ice run decelerates. Also, the erosion velocity of 1.2 m/s was reached at the toe, preventing further ice thickening. A limitation of the HEC-RAS ice routine is its inability to model porous flow or ice jam grounding. Use of a model with these capabilities such as RIVJAM (Beltaos, 1999) would likely produce a thicker and more realistic ice jam profile in the toe area. To compensate for these model deficiencies, the ice thickness at the most downstream cross section in the jam was artificially fixed at 2.1 m, the approximate average of the calculated ice thicknesses at the next few upstream cross sections. Another means of thickening the ice jam toe in HEC-RAS is to use an artificially high ice erosion velocity, though this runs the risk of making the modeled jam more stable than it would be in nature. Also, the use of unrealistically high under-ice erosion velocities could lead to overestimates of bed shear and underice scour. Because HEC-RAS does not simulate ice transport or ice jam initiation, several probable post-project jam sites were selected for breakup ice jam modeling on the Clark Fork (Fig. 4). The stability of these possible ice jams was evaluated in terms of HEC-RAS-calculated ice accumulation thickness and the under-ice water velocity. If the under-ice water velocity reached the ice erosion velocity, thinning the ice cover for most of the jam length, then an ice jam was considered unlikely at that location. The first possible jam was located in a bend 5.6 km downstream of the confluence and the second at an island 975 m downstream of the confluence. A third breakup jam site was selected 16 km below the confluence in downtown Missoula where multiple bridges and grade control structures cross the river. The same input parameters were used in these possible jam simulations as in the simulated 1996 ice jam on the lower Blackfoot River (Fig. 7). Ice erosion velocity veros, defined as the average under-ice water velocity at which ice pieces begin to erode from the underside of an ice jam, is a critical parameter in ice jam processes and modeling. In the HEC-RAS ice jam routine, further ice thickening is prevented once
191
under-ice water velocity reaches a user-specified veros. This analysis assumes that, once the flow velocity beneath a significant portion of the jam reaches the erosion velocity, further shove-thickening is countered by under-ice erosion, and the jam is probably not stable. In extreme cases, the entire ice accumulation may thin to the userspecified starting ice thickness (0.30 m in the case of these simulations). In addition to an indicator of jam stability, the concept of ice erosion velocity is important in terms of estimating bed shear since veros effectively limits the maximum water velocity beneath an ice accumulation and hence the calculated shear forces on the bed. Past numerical and physical modeling research suggest a possible range of 1 to 2 m/s for ice erosion velocity. Experiments and equations relating critical velocity to floe size by Tatinclaux and Gogus (1982) favor the lower end of this range, predicting an erosion velocity of about 1 m/s for assumed floe dimensions of 0.3-m-thick × 1.0-m-long. In a 1:25-scale physical model study of ice retention booms, Tuthill and Gooch (1998) found maximum full-scale equivalent erosion velocities of 1.0 and 1.1 m/s, respectively, for plastic and natural ice. Healy et al. (1997) cite an under-ice erosion velocity range of 1.0–2.0 m/s in their review of the ICEJAM ice jam model. Flato and Gerard (1986), whose jam model is the basis of the ice routine included in HEC-RAS, used a relatively high 1.5 m/s as the threshold for under-ice erosion. Also, Beltaos and Moody (1986) observed velocities under a jam of 1.3 m/s with no ice erosion. At the three plausible post-project ice jam sites modeled by HECRAS, calculated water velocities reached the 1.2-m/s erosion velocity and the ice accumulations thinned to (or never thickened beyond) their starting thickness of 0.30 m for much of their length. This suggests that, in the absence of a grounded toe condition, stable ice jams are unlikely at these locations. In spite of the previously discussed limitations of estimating ice jam stability by under-ice water velocity and ice thinning, the no-jam conclusion is supported by observation of the Clark Fork River between the Milltown Dam and Missoula, which has a continuously steep gradient and no known history of breakup ice jams. Ice jams have been observed downstream of Missoula where the channel flattens and becomes braided upstream of the Bitterroot confluence. As stated earlier, this method of predicting post-project ice jam occurrence and stability has limitations because HEC-RAS does not model porous flow or grounding, nor does it simulate ice transport or ice jam initiation. Given more time and resources, uncertainty could be reduced through the use of more sophisticated models such as DynaRICE with coupled ice dynamics and 2-dimensional unsteady hydraulics (Shen et al., 2000). The DynaRICE model has been used to simulate ice transport and jam initiation at the Mississippi–Missouri River confluence (Shen and Liu, 1998) and the Shokotsu River in Japan (Liu and Shen, 2003) with realistic results. In the current study, DynaRICE modeling would improve confidence in the predictions of post-dam-removal ice jam occurrence, location, and stability. 5. Ice-related bridge scour Although the previous analyses suggest that ice jams are unlikely in the restored Blackfoot channel and the Clark Fork downstream, the results contain enough uncertainty that the possibility cannot be completely ruled out. For this reason, the potential for ice-affected hydraulic scour in the vicinity of the Interstate 90 bridges was examined. Bed shear was calculated by a variety of methods, including the depth-slope product, for both a hypothetical ice jam case, and also the 100-year open water flood for comparison. Bed shear was also calculated using water velocity and drag equations with friction factors calculated from Mannings n, or estimated by friction factor ratios as described in Beltaos (2001). Calculated bed shear values were compared to published threshold velocities for bed material movement. As previously discussed, the models and methods used did not predict significant thickening at the
192
A.M. Tuthill et al. / Cold Regions Science and Technology 55 (2009) 186–194
Fig. 8. Simulated ice jam in restored Blackfoot channel (top) and average under-ice water velocity (bottom). Blackfoot discharge = 113 cm, Upper Clark Fork discharge = 200 cm.
ice jam toe. Greater toe thickness could produce a shear stress gradient sufficient to scour bed material in the ice jam toe region but this is difficult to quantify since ice-hydraulic models for moveable channels currently do not exist. Because of these uncertainties, the possibility of this type of local scour should be considered in bed protection design as well as the development of ice jam mitigation alternatives. Hydraulic and ice parameters such as water surface slope, average depth, and average channel velocity were taken from a HEC-RAS simulation with a hypothetical breakup ice jam located at the confluence and extending upstream beneath the bridges. Fig. 8 shows the jam profile with the restored Blackfoot River channel geometry and the existing channel thalweg for comparison. All input parameters were held the same as the previous simulations except the erosion velocity at the jam toe was increased to 2 m/s to prevent thinning and instability. As previously discussed, the use of an artificially high erosion velocity compensates for inability of HECRAS to model porous flow. Use of a model with porous flow might allow a thicker ice jam toe to form while maintaining water velocities below the ice erosion threshold. Note that, for most of the jam length, the under-ice water velocity is about 1 m/s, less than the assumed 1.2m/s upper bound for ice jam stability. All bed shear stress calculations assume a floating ice cover of uniform thickness across the river width and no ice grounding. Using the two-layer flow hypothesis for uniform flow beneath an ice jam, bed shear, τb was calculated using depth-slope product (e.g., Chow, 1959): τb ¼ γRib S
bed roughness ni or nb to the composite ice roughness nc, (Eqs. (3)–(5)) (US Army, 1998). Ri, the hydraulic radius with an ice cover, is approximated as half the under-ice flow depth yui. 2 3 323 3 n2i þ n2b 4 5 nc ¼ 2
Ric ¼
1:5 ni Ri nc
Rib ¼
ð4Þ
1:5 Ri :
ð5Þ
For the open water case at the 100-year discharge (680 m3/s), a 5.5-m hydraulic radius, and a water surface slope S of 0.00256, Eq. (2) gives an average bsed shear of 136 Pa. For the ice-hydraulic conditions of the HEC-RAS simulated jam at the confluence, S = 0.0025, a Mannings nb for the bed of 0.06, an under-ice flow depth yui of 2.3 m, and an ice Mannings ni of 0.08, Eqs. (2)–(4) give a bed shear τb of 22 Pa, an underice shear τi of 34 Pa, and a total shear τtot of 56 Pa. In the bed shear calculations using drag equations, the Mannings flow resistance coefficients ni and nb were converted to friction factors fi and fb using Eq. (6). For fi = 0.30 and fb = 0.20, and a HEC-RAScalculated flow velocity of 0.91 m/s, the Darcy–Weisbach Eq. (7) gave a bed shear value of 22 Pa, the same as the depth-slope product.
ð2Þ
where γ is the unit weight of water, Rib is the portion of the under-ice hydraulic radius affected by river bed resistance, and S is the water surface slope. The HEC-RAS ice option uses this approach to divide the under-ice hydraulic radius Ri into ice-affected and bed-affected components, Ric and Rib, respectively, based on ratios of the ice or
nb nc
ð3Þ
fði;bÞ ¼
τði;bÞ ¼
nð2i;bÞ 8g 1=3
Rðic;ibÞ 1 f ρU 2 : 8 ði;bÞ
ð6Þ
ð7Þ
A.M. Tuthill et al. / Cold Regions Science and Technology 55 (2009) 186–194
193
Table 1 Summary of bed shear calculations Depth-slope product τb = γRibS Q (m3/s)
S
ti (m)
y/yui (m)
ni
nb
Ric (m)
Rib (m)
τi (Pa)
τb (Pa)
τtot (Pa)
133 680
0.0025 0.00256
2.7 –
2.3 6.1
0.08 –
0.06 0.06
1.4 –
0.90 5.4
34
22 136
56 136
Darcy–Weisbach τ ði;bÞ ¼ 18 fði;bÞ ρU 2 with friction factors calculated from Mannings n Q (m3/s)
U (m/s)
ti(m)
yui(m)
ni
nb
Ric(m)
Rib(m)
fi
fb
τi(Pa)
τb(Pa)
τtot(Pa)
113 113
0.91 1.5
2.7 4.0
2.3 1.2
0.08 0.10
0.06 0.06
1.4 0.85
0.90 0.40
0.30 0.83
0.20 0.40
34 300
22 112
56 412
Darcy–Weisbach with empirical friction factor ratios Q (m3/s)
U (m/s)
ti(m)
yui (m)
fi/fo
fi/fb
τi (Pa)
τb (Pa)
τtot (Pa)
113
0.91
2.7
2.3
1.0–1.2
0.8–1.0
28–22
22–28
56
g is acceleration due to gravity and U is the average water velocity. The composite friction factor for flow under an ice cover:
ice run stops in its traditional location upstream of the restored Blackfoot channel.
fo ¼ ðfi þ fb Þ=2
6. Possible ice mitigation measures
ð8Þ
For purposes of comparison, methods described in Beltaos (2001) were used to estimate bed shear stress based on empirically derived ratios of fi/fo and the concept of apparent roughening of the bed due to the ice cover. Bed shear τb is expressed in terms of τtot and the ratio fb/fo in Eq. (9): τb ¼ ðfb =fo Þðτ tot =2Þ
ð9Þ
Eqs. (3)–(6) suggest that for fi NN fb, the point of maximum velocity occurs well below mid-depth and Ric NN Rib. As a result, the calculated τi will be much greater than τb. Vertical velocity profiles measured beneath freezeup ice jams by Gerard and Andres (1982) found the point of maximum velocity to lie close to the mid-depth even where the ice underside was much rougher than the bed. Similarly, Beltaos (2001) reported a fi / fo range of 1–1.2, in applications of the RIVJAM model. By Eq. (8), fb/fo falls in the 0.8–1.0 range and, assuming a total shear τtot of 56 Pa, Eq. (9) gives τb values in the 22 to 28 Pa range. This suggests two-layer flow hypothesis, without adjustment for apparent roughening of the bed, may in some cases be unconservative. It is possible that the post-project ice jams, if they occurred in the vicinity of the I-90 bridges, could be thicker in the toe area than those predicted by HEC-RAS, for reasons previously discussed. This could result in local water velocities well in excess of 0.91 m/s. A final shear stress calculation considers this thickened toe condition by assuming a 4-m-thick ice accumulation at the I-90 piers and an ice erosion velocity near the upper end of the accepted range of 1.5 m/s. Using Mannings resistance coefficients ni and nb of 0.1 and 0.06, respectively, the calculated bed shear of 112 Pa is close to the 136 Pa estimate from the 100-year open water event. Bed shear calculations by various methods are summarized in Table 1. Assuming clear water scour conditions, recommended values of limiting shear stress for coarse gravel (D50 = 50 mm) and sandy silt (D50 = 1 mm) found at the site are about 38 and 2 Pa, respectively (Zipparo and Hasen, 1992). Based on this information, the sandy silt would be expected to erode under the predicted 22–28 Pa bed shear beneath the simulated 1.8-m-thick breakup ice accumulation, while the gravel bed would resist erosion under these conditions. The silt and some of the gravel would be expected to erode during open water season floods or in the event of a 4-m-thick ice jam at the I-90 Bridge piers with an under-ice water velocity of 1.5 m/s, as discussed in the previous paragraph. In addition to hydraulic scour, the possibility for mechanical scour exists. Though unlikely, and difficult to predict with any certainty, the formation and abrupt release of an ice jam in the vicinity of the piers could mechanically displace bed and bank material. Because this possibility cannot be completely ruled out, ice jam mitigation measures are proposed, namely ensuring that the Blackfoot
Although it is not expected that the removal of the Milltown and Stimson Dams will significantly increase ice jams and ice-related problems on the lower Clark Fork, it is possible that ice run on the Blackfoot River could continue downstream and jam in the vicinity of the five bridges instead of stopping in its 1996 location. This possibility increases the potential for ice-related scour (both hydraulic and mechanical) in this section of river. One strategy would be to design the bed protection to withstand the expected under-ice hydraulic scour in the vicinity of the bridges. Based on the above bed shear analysis, it is probable that the armor required for the 100-year open water event would be adequate for protecting the bed during a severe breakup ice jam event. A more conservative approach would be to construct grade control and possibly ice retention piers to ensure that the Blackfoot breakup ice run stops in its traditional location about 2.1 km upstream of the confluence with the Clark Fork. Monitoring post-project conditions would help gage whether ice retention piers are needed in addition to other mitigation measures. It may be that the Blackfoot ice run will continue to stop in its traditional location with no additional structures, since the causal factors such as the sharp bend along Highway 200 and some of the gravel deposits at the head of the former Stimson Dam impoundment will remain after the dams are removed and the channel restored. Some head cutting and downstream transport of these gravel deposits are expected following dam removal. 7. Conclusions The results of this study suggest that the removal of the Milltown and Stimson Dams will probably not increase the potential for breakup ice jam problems in the restored or existing downstream river channels. The dam removals will change the ice cover type from the sheet ice on the existing impoundments to accumulations of shoved frazil ice on the restored channels, causing a 30-percent increase in total ice volume within the study area. In the case of a high-magnitude, low-frequency breakup event, the Blackfoot ice run would likely jam in its 1996 location upstream of the former impoundment area because the causal factors (sharp bend, slope reduction and gravel flats) remain relatively unchanged. Based on ice jam stability analysis using HEC-RAS, and field observation of existing ice conditions, the relatively steep restored and downstream channels on the Clark Fork will convey breakup ice downstream past Missoula without jamming. The level of confidence in the no-jamming prediction could be improved through the use of more sophisticated models that simulate ice transport and porous flow through the ice accumulation.
194
A.M. Tuthill et al. / Cold Regions Science and Technology 55 (2009) 186–194
Removal of the Milltown Dam and the change from an impoundment to free-flowing restored channels will increase the potential for under-ice scour in the vicinity of five bridges that cross the lower Blackfoot. Based on bed shear calculations by a range of methods, bed and bank protection designed for the 100-year open water flood would likely resist hydraulic scour for the expected range of ice and hydraulic conditions. Though unlikely, the possibility exists for an ice jam and its sudden release in the vicinity of the bridges. This direct contact of moving ice could result in local bed and bank shear stresses in excess of those resulting from hydraulic scour beneath a stationary floating jam. These conclusions, which are based on simple models and theory, limited field observations, and engineering judgment, contain some uncertainty. A more in-depth analysis with additional field data and the use of more sophisticated models would reduce, but not eliminate, this uncertainty. Structural ice control would provide a margin of safety against ice jams and related scour in the restored channel. Possible measures include piers to retain the breakup ice run above the restored Blackfoot channel, and grade control to prevent head cutting of the gravel flats upstream of the former Stimson Dam. Acknowledgments This study was supported by the Missoula, MT, Business Office of the US Army Corps of Engineers, Seattle District. The ice evaluation is a component of the larger EPA effort to remediate contaminated sediment in the Clark Fork River. References Beltaos, S., 1999. Flow through voids of breakup ice jams. Canadian Journal of Civil Engineering 26, 117–185. Beltaos, S., 2001. Hydraulic roughness of breakup ice jams. Journal of Hydraulic Engineering 127 (8) (August 2001). Beltaos, S., Moody, W.J., 1986. Measurements of the configuration of a breakup ice jam. NWRI Contribution, pp. 86–123. March 1986. Chow, V.T., 1959. Open-channel Hydraulics. McGraw-Hill Book Co., New York. Flato, G., Gerard, R., 1986. Calculation of ice jam thickness profiles. Proceedings, Fourth Workshop on Hydraulics of River Ice, Montreal, Canada, June 19–20, 1986.
Gerard, L., Andres, D., 1982. Hydraulic roughness and freezeup accumulations: North Saskatchewan River through Edmonton. Canadian Committee on River Ice. CGU-HS, Edmonton, Canada, pp. 62–87. Healy, D., Hicks, F., Beltaos, S., 1997. A comparison of the ICEJAM and RIVJAM ice jam profile models. Proceedings, 9th Workshop on River Ice, Fredericton, NB, Canada, September 24–26, 1997. Liu, L., and Shen, H.T. (1998) “A numerical model for river ice jam evolution.” Proceedings of the 14th International Symposium on Ice, Potsdam, NY, USA, 27–31 July 1998. H.T. Shen (ed.), A.A. Balkema, Rotterdam, 1999. Matousik, V., 1984a. Regularity of the freezing-up of the water surface and heat exchange between the water body and the water surface. Proceedings IAHR Ice Symposium, vol. 1. Hamburg, Germany, p. 187. August 27–31, 1984. Matousik, V., 1984b. Types of ice run and conditions for their formation. Proceedings IAHR Ice Symposium, vol. 1, p. 315. August 27–31, 1984. Moore, J.N., Landrigan, E.M., 1999. Mobilization of metal-contaminated sediment by ice jam floods. Environmental Geology 37 (1–2) (January 1999). Perham, R.E., 1983. Ice sheet retention structures. US Army Cold Regions Research and Engineering Laboratory, Hanover, New Hampshire, CRREL Report 83–30. Shen, H.T., Liu, L., 2003. Shokotsu River ice jam. Journal of Cold Regions Science and Technology 37 (1), 35–49 (July 2003). Shen, H.T., Wang, D.H., 1995. Under cover transport and accumulation of frazil granules. Journal of Hydraulic Engineering 121 (2). Shen, H.T., Su, J., Liu, L., 2000. SPH simulation of river ice dynamics. Journal of Computational Physics 165 (2), 752–777. Tatinclaux, J.C. and Gogus, M. (1982) “Stability of floes beneath a floating ice cover”, IAHR, 1981, vol. 2, pp 298–311. Tuthill, A.M., Gooch, G., 1998. A physical model study of ice retention Booms. Proceedings, IAHR 14th International Symposium on Ice, pp. 61–66, Potsdam, NY, 27–31 July 1998. US Army, 1998. HEC-RAS River Analysis System” Hydraulic Reference Manual Version 2.2. US Army Corps of Engineers, Hydrologic Engineering Center. September 1998. US Army, 2000. HEC-RAS River Analysis System” Users Manual Version 3.1.2. US Army Corps of Engineers, Hydrologic Engineering Center. October 2000. US Army, 2002. Ice Engineering” Engineer Manual 1110-2-1612. US Army Corps of Engineers, Washington DC. 30 October 2002 http://www.usace.army.mil/publications/eng-manuals/em1110-2-1612/toc.htm. Westwater Consultants et al. (2005) “Draft Restoration Plan for Restoring the Clark Fork River and Blackfoot River following the Removal of the Milltown Dam.” Westwater Consultants, Corvallis, MT, River Design Group, Whitefish, MT, and Geum Environmental Consulting, Hamilton, MT, April 13, 2005. White, K.D., 1999. Hydraulic and physical properties affecting ice jams. CRREL Report 99-11. US Army Cold Regions Research and Engineering Laboratory, Hanover, NH. Zipparo, V.J., Hasen, H., 1992. Davis' Handbook of Applied Hydraulics, 4th edition. McGraw-Hill. November 1992.
Contents lists available at ScienceDirect
Cold Regions Science and Technology j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / c o l d r e g i o n s
Effects of proposed dam removal on ice jamming and bridge scour on the Clark Fork River, Montana A.M. Tuthill a,⁎, K.D. White a, C.M. Vuyovich a, L.A. Daniels b a b
U.S. Army Cold Regions Research and Engineering Laboratory, 72 Lyme Road, Hanover, NH 03755-1290, United States Seattle District, Missoula Business Office, U.S. Army Corps of Engineers, 1600 North Avenue, Missoula, MT 59801-5512, United States
a r t i c l e
i n f o
Article history: Received 31 October 2007 Accepted 5 September 2008 Keywords: Ice jams Ice jam bed scour Ice jam modeling Dam removal Remediation of contaminated sediment Clark Fork River Superfund Project
a b s t r a c t The Milltown Dam, at the confluence of the Clark Fork and Blackfoot Rivers in Montana, lies at the downstream end of the nation's largest Superfund Project. Since its construction in 1906, several hundred thousand tonnes of metal-contaminated sediment from upstream mining activities, primarily copper, have accumulated in the dam impoundment. A large amount of this sediment was scoured during a 1996 ice jam event on Clark Fork and Blackfoot Rivers and transported downstream. The EPA remediation plan calls for phased removal and off-site disposal of much of the contaminated sediment, and removal of the Milltown Dam and a smaller mill dam upstream on the Blackfoot River. As much as possible, the river channels will be restored to their pre-project natural morphology. This study assessed ice impacts associated with the restoration plan, specifically where ice jams and related ice jam scour might occur with and without the dams in place. Also addressed was the effect of dam removal on potential ice-related scour around the piers of five bridges that cross the Blackfoot River just upstream of the Milltown Dam. Because this type of problem is relatively new, relevant engineering guidelines are lacking. Shortcomings of the tools and methods used are discussed. It is hoped that the study described herein will assist those addressing similar problems in the future, and also point to areas where analysis methods and tools could be improved. Published by Elsevier B.V.
1. Introduction The Milltown Dam, built in 1906 on the Clark Fork River, lies just downstream from the confluence of the Blackfoot River about 11 km east of Missoula, MT (Figs. 1 and 2). Since construction, some 5 million cubic meters of sediment have accumulated in the dam's mile-long impoundment. Much of this is material contaminated with metals (As, Cd, Cu, Pb, and Zn) as a result of the historic mining activities upstream at Butte and Anaconda. Since 1982, the EPA has listed the Milltown Reservoir–Clark Fork River on the National Priorities List, and it is now part of the nation's largest Superfund site. Extensive investigations have led to a remedial action plan consisting of phased removal of the Milltown Dam and the contaminated sediments in its impoundment. Ultimately, the dam decommissioning may result in a drop in bed elevation of up to 9 m. The small timber crib Stimson Dam, located 1.6 km upstream on the Blackfoot River, was removed in 2005 to allow fish passage and improve safety. The Clark Fork and Blackfoot Rivers experience infrequent ice events, the most severe in recent history occurring on February 9–10, ⁎ Corresponding author. E-mail address: [email protected] (A.M. Tuthill). 0165-232X/$ – see front matter. Published by Elsevier B.V. doi:10.1016/j.coldregions.2008.09.004
1996. On February 9, a massive ice run progressed down the Blackfoot River to jam about 500 m upstream of the Stimson Dam (Figs. 3 and 4). On February 10, a second breakup ice run on the Clark Fork jammed briefly before fracturing its way through the Milltown Dam impoundment. This ice event scoured large amounts of fine-grained metalcontaminated sediment and deposited them downstream (Moore and Landrigan, 1999). Our study evaluated potential effects of the Milltown Dam removal on the ice regime of the Clark Fork and Blackfoot Rivers in the vicinity of the project. Of particular interest were the potential consequences of displacing ice jams from their traditional locations upstream of the Milltown Dam impoundment to downstream sites, and the possibility of ice-related scour in the vicinity of the piers of the five bridges that cross the lower Blackfoot River. The study addressed this increasingly common problem using conventional ice engineering methods and models. It began with a review and analysis of past ice events and a program of field observations during the winter of 2005. The HEC-RAS hydraulic model (US Army, 2000) was then used to simulate freezeup covers and breakup ice jams for the pre- and post-dam removal cases. Using HECRAS and basic ice engineering theory, we evaluated the potential for post-dam removal ice jam scour around the bridge piers on the lower
A.M. Tuthill et al. / Cold Regions Science and Technology 55 (2009) 186–194
187
Fig. 1. Map of study area (from Moore and Landrigan, 1999).
Blackfoot. Finally, the study identified possible ways to mitigate adverse effects of the dam removal on the ice regime. Limitations of the approach and remaining uncertainties are discussed. 2. History of ice jams on the Clark Fork and Blackfoot Rivers near Missoula Although significant ice jams were reported on the lower Blackfoot in 1974 and on the Clark Fork below Missoula in 1984, the February 1996 event stands out as the most severe by a large margin. In the preceding three weeks, extreme cold caused thick ice covers on the Blackfoot and Clark Fork Rivers. A maritime weather system brought rapid thawing and rain to the region, releasing more than 65 km of ice on the Blackfoot as a dynamic breakup front that progressed downstream at 3.5–4.5 m/s. The Blackfoot ice run grounded and stopped about 2.4 km upstream of the Milltown Dam to form a 5-kmlong ice jam. After-the-event field inspections and photos revealed shear wall heights as great as 5 m, and top-of-ice elevations nearly level with the Highway 200 roadway along the S-curves above Bonner (Figs. 3 and 4). Likely reasons for the ice jam were the loss of momentum as the breakup front passed the sharp bend above Bonner and impacted the flat-lying sheet ice cover and gravel deposits of at the head of the Stimson Dam Pool. On the Clark Fork, a number of smaller ice jams occurred as a breakup front progressed downstream, ultimately fracturing its way through the 30-cm-thick sheet ice cover on the Milltown Dam pool on February 10. During the next several days, continued high flows and much more ice passed through the Clark Fork portion of the pool. Many thousands of tonnes of sediment were eroded from the impoundment through a combination of hydraulic scour and mechanical gouging of the bed and banks by moving ice floes. Moore and Landrigan (1999)
described the river as “running black with sediment amongst the ice blocks” and, after the event, they observed 20-cm-thick deposits of fine-grained sediment on ice blocks lining the river banks downstream of the reservoir. 3. Frequency and severity of ice events In order to approximate the probability of recurrence of a 1996-like event and identify possible unrecorded ice events, a hindcasting analysis was carried out using historic hydro-meteorological data from 1929 to 2005. For each winter, maximum ice thickness (ti) in cm was calculated from accumulated freezing degree (C)-days (AFDD) (US Army, 2002). pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ti ¼ C maxAFDD:
ð1Þ
C is a calibration coefficient calculated from observed ice thickness and AFDD data from the winters of 1996 and 2005. Following the February 1996 ice event, the observed thickness of the fractured sheet ice on the Milltown Dam Pool was about 30 cm. AFDD had peaked three days prior to the event at 350 °C, giving a calculated C value of 1.6. During the winter of 2005, AFDD peaked on January 17 at 222 °C when the average measured ice thickness on the Milltown Pool was 24 cm, again yielding a C value of 1.6. Note that these calculated ice thicknesses apply only to relatively quiescent sections of river characterized by thermal ice growth. Ice-hydraulic model calculations of thicker freezeup ice accumulations in faster-moving sections of river are described later. Based on the known historic events of 1996 and 1974, empirical criteria were developed to identify other likely ice jam events on the
188
A.M. Tuthill et al. / Cold Regions Science and Technology 55 (2009) 186–194
Fig. 2. Clark Fork–Blackfoot confluence and Milltown Dam in 2005.
lower Blackfoot River. These prediction criteria were based on a review of daily precipitation, air temperature and discharge data for all winters from 1922–2005. It is important to note that these criteria are site-specific and apply only to this case study and these particular discharges and ice thickness ranges. Fig. 5 plots base-to-peak discharge increase against calculated ice thickness, suggesting causal factors for historic ice jams on the lower Blackfoot River to include a pre-breakup ice thickness of 25 cm or greater and a flow increase equal to or greater than 57 cm. Ice jams were considered possible for ice thicknesses equal to or greater than 25 cm with discharge in the 40- to 57-cm range. In addition to the two known ice jams, one likely ice jam event (1971) and three possible events (1963, 1984, and 1986) were identified on the lower Blackfoot for the 77-year period of
record. The hydro-meteorological data were then reviewed as an independent check that the predictions were reasonable. Increases in discharge (ΔQ) and ice thicknesses (ti) were ranked and assigned probabilities (Fig. 6). Assuming ΔQ and ice thickness ti are independent variables, the exceedance probability of an historic or hindcast event can be estimated by multiplying the two probabilities. By this approach, the probability of a 1996-like event occurring on the lower Blackfoot was estimated at about 0.6%. Results of the hindcasting and probability analyses contain uncertainty as a result of the lack of historic event data (only two known events on the lower Blackfoot). In spite of this uncertainty, the analyses support the assertion that the 1996 event was quite rare, particularly in terms of discharge. Because of its low frequency, the 1996 ice jam was considered suitable for use as the design event in subsequent analyses. 4. Hydraulic modeling
Fig. 3. Blackfoot River ice jam near its downstream end, Feb. 1996.
The HEC-RAS model was used to analyze ice formation and breakup processes under existing and post-project conditions. HEC-RAS was chosen for the ice modeling due to its use in other phases of the project and the fact that the river geometry was provided in HEC-RAS format. The modeled reach included 32 km of the Clark Fork plus 11 km of the Blackfoot River from the Bonner Gage downstream to the mouth. The model was modified to simulate post-project channel geometry by removing the Milltown and Stimson Dams and adjusting the channel width and minimum bed elevation to conform to a preliminary restored channel design by Westwater Consultants et al. (2005). The restored channels had bottom widths of 60 m above the confluence and 75 m downstream. The restored Blackfoot channel followed the existing straight alignment while, for the Clark Fork restored channel, a sinuosity of 20% was assumed.
A.M. Tuthill et al. / Cold Regions Science and Technology 55 (2009) 186–194
189
Fig. 4. Map showing the 1996 ice jams (red) and possible post-project ice jam locations (blue and green). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
In this study, probable ice cover types were estimated initially from open water velocity calculated by HEC-RAS using the long-term average January discharge of 20 m3/s. Perham (1983) and others have used similar criteria to predict ice retention and freezeup ice cover
type in the design of ice control structures. The criteria predict thermally grown sheet and border ice for v ≤ 0.3 m/s, juxtaposed frazil ice accumulations for v between 0.3 and 0.7 m/s, and shoved frazil floes for v between 0.7 and 1.5 m/s. Where water velocity exceeds
Fig. 5. Blackfoot discharge increase vs. calculated ice thickness, with fields identifying likely and possible unrecorded ice jam events.
190
A.M. Tuthill et al. / Cold Regions Science and Technology 55 (2009) 186–194
Fig. 6. Probability of ice thickness and base-to-peak discharge increase for the lower Blackfoot River.
about 1.5 m/s, a section of river may remain open all winter. Though water velocity criteria provide a first estimate of ice cover type, it is important to note that many other factors influence river ice formation. These include channel geometry, water depth, turbulence, water and air temperature, wind velocity, sky cover and surface ice concentration (Matousik, 1984a,b). The HEC-RAS ice routine was used to calculate the thickness of the freezeup ice covers due to juxtaposition and shoving. The simulations used typical freezeup ice accumulation parameters, including an under-ice Mannings ni of 0.04, an ice accumulation porosity e of 0.4,
an angle of internal friction Φ for the granular ice material of 45°, and an under-ice erosion velocity veros of 1.2 m/s (White, 1999). Based on the historic AFDD analysis calibrated to the observations of 1996 and 2005, the sheet ice thickness on the pools and low velocity areas was assumed to be 30 cm. The freezeup simulations used the long-term January averages of 20 and 25 m3/s for the Blackfoot and Clark Fork Rivers, respectively. With the exception of the sheet ice on the dam impoundments, the HEC-RAS ice jam force balance predicted ice covers of juxtaposed and shoved frazil floes at most locations within the study reach. The
Fig. 7. HEC-RAS-simulated profile of the 1996 ice jam on the lower Blackfoot River under existing conditions. Blackfoot discharge = 113 cm.
A.M. Tuthill et al. / Cold Regions Science and Technology 55 (2009) 186–194
equilibrium frazil transport theories of Shen and Wang (1995) predicted some additional frazil deposition beneath the impoundment ice covers, but in terms of overall ice volume, this additional frazil volume was not significant. For the post-project case, HEC-RAS predicted frazil ice covers about 1.5 m thick on the restored river channels. Based on the simulations, the change from sheet ice covers on the existing impoundments to juxtaposed and shoved ice accumulations on the restored channels would increase the total ice volume within the study area by about 30%. The February 1996 event served as a worst-case scenario in terms of discharge and ice supply for the breakup ice jam modeling. Breakup jams for existing conditions were modeled at the upper end of the former Stimson Dam pool on the Blackfoot, and at the upstream end of the Clark Fork portion of the Milltown Dam reservoir. To model breakup on the Blackfoot, the February 9, 1996, daily average discharge of 113 m3/s was used. Clark Fork discharge was 198 m3/s above the Blackfoot confluence and 311 m3/s below. Fig. 4 shows locations of the observed ice jam on the Blackfoot River and the likely location of the jam at the upstream end of the Clark Fork portion of the Milltown Dam impoundment. Fig. 7 shows the simulated profile of the 1996 ice jam on the lower Blackfoot River using reasonable input parameters of erosion velocity veros = 1.2 m/s; angle of internal friction for the ice accumulation Φ = 45°; ice accumulation porosity e = 0.4; and Mannings resistance coefficient for the ice underside ni = 0.08 (White, 1999). The simulated profile matched observed top-of-ice heights fairly well along the Highway 120 S-curves, but the modeled top of jam was slightly higher than the observed top of ice near the upstream end of the jam (Fig. 7). The bulge in the ice jam profile between km 5 and 6 may result from the steeper water surface slope just upstream of this location. The HEC-RAS force balance under-predicted observed ice thickness and stage near the downstream end of the jam, possibly because of the inability of HEC-RAS to model ice thickening from momentum transfer as the ice run decelerates. Also, the erosion velocity of 1.2 m/s was reached at the toe, preventing further ice thickening. A limitation of the HEC-RAS ice routine is its inability to model porous flow or ice jam grounding. Use of a model with these capabilities such as RIVJAM (Beltaos, 1999) would likely produce a thicker and more realistic ice jam profile in the toe area. To compensate for these model deficiencies, the ice thickness at the most downstream cross section in the jam was artificially fixed at 2.1 m, the approximate average of the calculated ice thicknesses at the next few upstream cross sections. Another means of thickening the ice jam toe in HEC-RAS is to use an artificially high ice erosion velocity, though this runs the risk of making the modeled jam more stable than it would be in nature. Also, the use of unrealistically high under-ice erosion velocities could lead to overestimates of bed shear and underice scour. Because HEC-RAS does not simulate ice transport or ice jam initiation, several probable post-project jam sites were selected for breakup ice jam modeling on the Clark Fork (Fig. 4). The stability of these possible ice jams was evaluated in terms of HEC-RAS-calculated ice accumulation thickness and the under-ice water velocity. If the under-ice water velocity reached the ice erosion velocity, thinning the ice cover for most of the jam length, then an ice jam was considered unlikely at that location. The first possible jam was located in a bend 5.6 km downstream of the confluence and the second at an island 975 m downstream of the confluence. A third breakup jam site was selected 16 km below the confluence in downtown Missoula where multiple bridges and grade control structures cross the river. The same input parameters were used in these possible jam simulations as in the simulated 1996 ice jam on the lower Blackfoot River (Fig. 7). Ice erosion velocity veros, defined as the average under-ice water velocity at which ice pieces begin to erode from the underside of an ice jam, is a critical parameter in ice jam processes and modeling. In the HEC-RAS ice jam routine, further ice thickening is prevented once
191
under-ice water velocity reaches a user-specified veros. This analysis assumes that, once the flow velocity beneath a significant portion of the jam reaches the erosion velocity, further shove-thickening is countered by under-ice erosion, and the jam is probably not stable. In extreme cases, the entire ice accumulation may thin to the userspecified starting ice thickness (0.30 m in the case of these simulations). In addition to an indicator of jam stability, the concept of ice erosion velocity is important in terms of estimating bed shear since veros effectively limits the maximum water velocity beneath an ice accumulation and hence the calculated shear forces on the bed. Past numerical and physical modeling research suggest a possible range of 1 to 2 m/s for ice erosion velocity. Experiments and equations relating critical velocity to floe size by Tatinclaux and Gogus (1982) favor the lower end of this range, predicting an erosion velocity of about 1 m/s for assumed floe dimensions of 0.3-m-thick × 1.0-m-long. In a 1:25-scale physical model study of ice retention booms, Tuthill and Gooch (1998) found maximum full-scale equivalent erosion velocities of 1.0 and 1.1 m/s, respectively, for plastic and natural ice. Healy et al. (1997) cite an under-ice erosion velocity range of 1.0–2.0 m/s in their review of the ICEJAM ice jam model. Flato and Gerard (1986), whose jam model is the basis of the ice routine included in HEC-RAS, used a relatively high 1.5 m/s as the threshold for under-ice erosion. Also, Beltaos and Moody (1986) observed velocities under a jam of 1.3 m/s with no ice erosion. At the three plausible post-project ice jam sites modeled by HECRAS, calculated water velocities reached the 1.2-m/s erosion velocity and the ice accumulations thinned to (or never thickened beyond) their starting thickness of 0.30 m for much of their length. This suggests that, in the absence of a grounded toe condition, stable ice jams are unlikely at these locations. In spite of the previously discussed limitations of estimating ice jam stability by under-ice water velocity and ice thinning, the no-jam conclusion is supported by observation of the Clark Fork River between the Milltown Dam and Missoula, which has a continuously steep gradient and no known history of breakup ice jams. Ice jams have been observed downstream of Missoula where the channel flattens and becomes braided upstream of the Bitterroot confluence. As stated earlier, this method of predicting post-project ice jam occurrence and stability has limitations because HEC-RAS does not model porous flow or grounding, nor does it simulate ice transport or ice jam initiation. Given more time and resources, uncertainty could be reduced through the use of more sophisticated models such as DynaRICE with coupled ice dynamics and 2-dimensional unsteady hydraulics (Shen et al., 2000). The DynaRICE model has been used to simulate ice transport and jam initiation at the Mississippi–Missouri River confluence (Shen and Liu, 1998) and the Shokotsu River in Japan (Liu and Shen, 2003) with realistic results. In the current study, DynaRICE modeling would improve confidence in the predictions of post-dam-removal ice jam occurrence, location, and stability. 5. Ice-related bridge scour Although the previous analyses suggest that ice jams are unlikely in the restored Blackfoot channel and the Clark Fork downstream, the results contain enough uncertainty that the possibility cannot be completely ruled out. For this reason, the potential for ice-affected hydraulic scour in the vicinity of the Interstate 90 bridges was examined. Bed shear was calculated by a variety of methods, including the depth-slope product, for both a hypothetical ice jam case, and also the 100-year open water flood for comparison. Bed shear was also calculated using water velocity and drag equations with friction factors calculated from Mannings n, or estimated by friction factor ratios as described in Beltaos (2001). Calculated bed shear values were compared to published threshold velocities for bed material movement. As previously discussed, the models and methods used did not predict significant thickening at the
192
A.M. Tuthill et al. / Cold Regions Science and Technology 55 (2009) 186–194
Fig. 8. Simulated ice jam in restored Blackfoot channel (top) and average under-ice water velocity (bottom). Blackfoot discharge = 113 cm, Upper Clark Fork discharge = 200 cm.
ice jam toe. Greater toe thickness could produce a shear stress gradient sufficient to scour bed material in the ice jam toe region but this is difficult to quantify since ice-hydraulic models for moveable channels currently do not exist. Because of these uncertainties, the possibility of this type of local scour should be considered in bed protection design as well as the development of ice jam mitigation alternatives. Hydraulic and ice parameters such as water surface slope, average depth, and average channel velocity were taken from a HEC-RAS simulation with a hypothetical breakup ice jam located at the confluence and extending upstream beneath the bridges. Fig. 8 shows the jam profile with the restored Blackfoot River channel geometry and the existing channel thalweg for comparison. All input parameters were held the same as the previous simulations except the erosion velocity at the jam toe was increased to 2 m/s to prevent thinning and instability. As previously discussed, the use of an artificially high erosion velocity compensates for inability of HECRAS to model porous flow. Use of a model with porous flow might allow a thicker ice jam toe to form while maintaining water velocities below the ice erosion threshold. Note that, for most of the jam length, the under-ice water velocity is about 1 m/s, less than the assumed 1.2m/s upper bound for ice jam stability. All bed shear stress calculations assume a floating ice cover of uniform thickness across the river width and no ice grounding. Using the two-layer flow hypothesis for uniform flow beneath an ice jam, bed shear, τb was calculated using depth-slope product (e.g., Chow, 1959): τb ¼ γRib S
bed roughness ni or nb to the composite ice roughness nc, (Eqs. (3)–(5)) (US Army, 1998). Ri, the hydraulic radius with an ice cover, is approximated as half the under-ice flow depth yui. 2 3 323 3 n2i þ n2b 4 5 nc ¼ 2
Ric ¼
1:5 ni Ri nc
Rib ¼
ð4Þ
1:5 Ri :
ð5Þ
For the open water case at the 100-year discharge (680 m3/s), a 5.5-m hydraulic radius, and a water surface slope S of 0.00256, Eq. (2) gives an average bsed shear of 136 Pa. For the ice-hydraulic conditions of the HEC-RAS simulated jam at the confluence, S = 0.0025, a Mannings nb for the bed of 0.06, an under-ice flow depth yui of 2.3 m, and an ice Mannings ni of 0.08, Eqs. (2)–(4) give a bed shear τb of 22 Pa, an underice shear τi of 34 Pa, and a total shear τtot of 56 Pa. In the bed shear calculations using drag equations, the Mannings flow resistance coefficients ni and nb were converted to friction factors fi and fb using Eq. (6). For fi = 0.30 and fb = 0.20, and a HEC-RAScalculated flow velocity of 0.91 m/s, the Darcy–Weisbach Eq. (7) gave a bed shear value of 22 Pa, the same as the depth-slope product.
ð2Þ
where γ is the unit weight of water, Rib is the portion of the under-ice hydraulic radius affected by river bed resistance, and S is the water surface slope. The HEC-RAS ice option uses this approach to divide the under-ice hydraulic radius Ri into ice-affected and bed-affected components, Ric and Rib, respectively, based on ratios of the ice or
nb nc
ð3Þ
fði;bÞ ¼
τði;bÞ ¼
nð2i;bÞ 8g 1=3
Rðic;ibÞ 1 f ρU 2 : 8 ði;bÞ
ð6Þ
ð7Þ
A.M. Tuthill et al. / Cold Regions Science and Technology 55 (2009) 186–194
193
Table 1 Summary of bed shear calculations Depth-slope product τb = γRibS Q (m3/s)
S
ti (m)
y/yui (m)
ni
nb
Ric (m)
Rib (m)
τi (Pa)
τb (Pa)
τtot (Pa)
133 680
0.0025 0.00256
2.7 –
2.3 6.1
0.08 –
0.06 0.06
1.4 –
0.90 5.4
34
22 136
56 136
Darcy–Weisbach τ ði;bÞ ¼ 18 fði;bÞ ρU 2 with friction factors calculated from Mannings n Q (m3/s)
U (m/s)
ti(m)
yui(m)
ni
nb
Ric(m)
Rib(m)
fi
fb
τi(Pa)
τb(Pa)
τtot(Pa)
113 113
0.91 1.5
2.7 4.0
2.3 1.2
0.08 0.10
0.06 0.06
1.4 0.85
0.90 0.40
0.30 0.83
0.20 0.40
34 300
22 112
56 412
Darcy–Weisbach with empirical friction factor ratios Q (m3/s)
U (m/s)
ti(m)
yui (m)
fi/fo
fi/fb
τi (Pa)
τb (Pa)
τtot (Pa)
113
0.91
2.7
2.3
1.0–1.2
0.8–1.0
28–22
22–28
56
g is acceleration due to gravity and U is the average water velocity. The composite friction factor for flow under an ice cover:
ice run stops in its traditional location upstream of the restored Blackfoot channel.
fo ¼ ðfi þ fb Þ=2
6. Possible ice mitigation measures
ð8Þ
For purposes of comparison, methods described in Beltaos (2001) were used to estimate bed shear stress based on empirically derived ratios of fi/fo and the concept of apparent roughening of the bed due to the ice cover. Bed shear τb is expressed in terms of τtot and the ratio fb/fo in Eq. (9): τb ¼ ðfb =fo Þðτ tot =2Þ
ð9Þ
Eqs. (3)–(6) suggest that for fi NN fb, the point of maximum velocity occurs well below mid-depth and Ric NN Rib. As a result, the calculated τi will be much greater than τb. Vertical velocity profiles measured beneath freezeup ice jams by Gerard and Andres (1982) found the point of maximum velocity to lie close to the mid-depth even where the ice underside was much rougher than the bed. Similarly, Beltaos (2001) reported a fi / fo range of 1–1.2, in applications of the RIVJAM model. By Eq. (8), fb/fo falls in the 0.8–1.0 range and, assuming a total shear τtot of 56 Pa, Eq. (9) gives τb values in the 22 to 28 Pa range. This suggests two-layer flow hypothesis, without adjustment for apparent roughening of the bed, may in some cases be unconservative. It is possible that the post-project ice jams, if they occurred in the vicinity of the I-90 bridges, could be thicker in the toe area than those predicted by HEC-RAS, for reasons previously discussed. This could result in local water velocities well in excess of 0.91 m/s. A final shear stress calculation considers this thickened toe condition by assuming a 4-m-thick ice accumulation at the I-90 piers and an ice erosion velocity near the upper end of the accepted range of 1.5 m/s. Using Mannings resistance coefficients ni and nb of 0.1 and 0.06, respectively, the calculated bed shear of 112 Pa is close to the 136 Pa estimate from the 100-year open water event. Bed shear calculations by various methods are summarized in Table 1. Assuming clear water scour conditions, recommended values of limiting shear stress for coarse gravel (D50 = 50 mm) and sandy silt (D50 = 1 mm) found at the site are about 38 and 2 Pa, respectively (Zipparo and Hasen, 1992). Based on this information, the sandy silt would be expected to erode under the predicted 22–28 Pa bed shear beneath the simulated 1.8-m-thick breakup ice accumulation, while the gravel bed would resist erosion under these conditions. The silt and some of the gravel would be expected to erode during open water season floods or in the event of a 4-m-thick ice jam at the I-90 Bridge piers with an under-ice water velocity of 1.5 m/s, as discussed in the previous paragraph. In addition to hydraulic scour, the possibility for mechanical scour exists. Though unlikely, and difficult to predict with any certainty, the formation and abrupt release of an ice jam in the vicinity of the piers could mechanically displace bed and bank material. Because this possibility cannot be completely ruled out, ice jam mitigation measures are proposed, namely ensuring that the Blackfoot
Although it is not expected that the removal of the Milltown and Stimson Dams will significantly increase ice jams and ice-related problems on the lower Clark Fork, it is possible that ice run on the Blackfoot River could continue downstream and jam in the vicinity of the five bridges instead of stopping in its 1996 location. This possibility increases the potential for ice-related scour (both hydraulic and mechanical) in this section of river. One strategy would be to design the bed protection to withstand the expected under-ice hydraulic scour in the vicinity of the bridges. Based on the above bed shear analysis, it is probable that the armor required for the 100-year open water event would be adequate for protecting the bed during a severe breakup ice jam event. A more conservative approach would be to construct grade control and possibly ice retention piers to ensure that the Blackfoot breakup ice run stops in its traditional location about 2.1 km upstream of the confluence with the Clark Fork. Monitoring post-project conditions would help gage whether ice retention piers are needed in addition to other mitigation measures. It may be that the Blackfoot ice run will continue to stop in its traditional location with no additional structures, since the causal factors such as the sharp bend along Highway 200 and some of the gravel deposits at the head of the former Stimson Dam impoundment will remain after the dams are removed and the channel restored. Some head cutting and downstream transport of these gravel deposits are expected following dam removal. 7. Conclusions The results of this study suggest that the removal of the Milltown and Stimson Dams will probably not increase the potential for breakup ice jam problems in the restored or existing downstream river channels. The dam removals will change the ice cover type from the sheet ice on the existing impoundments to accumulations of shoved frazil ice on the restored channels, causing a 30-percent increase in total ice volume within the study area. In the case of a high-magnitude, low-frequency breakup event, the Blackfoot ice run would likely jam in its 1996 location upstream of the former impoundment area because the causal factors (sharp bend, slope reduction and gravel flats) remain relatively unchanged. Based on ice jam stability analysis using HEC-RAS, and field observation of existing ice conditions, the relatively steep restored and downstream channels on the Clark Fork will convey breakup ice downstream past Missoula without jamming. The level of confidence in the no-jamming prediction could be improved through the use of more sophisticated models that simulate ice transport and porous flow through the ice accumulation.
194
A.M. Tuthill et al. / Cold Regions Science and Technology 55 (2009) 186–194
Removal of the Milltown Dam and the change from an impoundment to free-flowing restored channels will increase the potential for under-ice scour in the vicinity of five bridges that cross the lower Blackfoot. Based on bed shear calculations by a range of methods, bed and bank protection designed for the 100-year open water flood would likely resist hydraulic scour for the expected range of ice and hydraulic conditions. Though unlikely, the possibility exists for an ice jam and its sudden release in the vicinity of the bridges. This direct contact of moving ice could result in local bed and bank shear stresses in excess of those resulting from hydraulic scour beneath a stationary floating jam. These conclusions, which are based on simple models and theory, limited field observations, and engineering judgment, contain some uncertainty. A more in-depth analysis with additional field data and the use of more sophisticated models would reduce, but not eliminate, this uncertainty. Structural ice control would provide a margin of safety against ice jams and related scour in the restored channel. Possible measures include piers to retain the breakup ice run above the restored Blackfoot channel, and grade control to prevent head cutting of the gravel flats upstream of the former Stimson Dam. Acknowledgments This study was supported by the Missoula, MT, Business Office of the US Army Corps of Engineers, Seattle District. The ice evaluation is a component of the larger EPA effort to remediate contaminated sediment in the Clark Fork River. References Beltaos, S., 1999. Flow through voids of breakup ice jams. Canadian Journal of Civil Engineering 26, 117–185. Beltaos, S., 2001. Hydraulic roughness of breakup ice jams. Journal of Hydraulic Engineering 127 (8) (August 2001). Beltaos, S., Moody, W.J., 1986. Measurements of the configuration of a breakup ice jam. NWRI Contribution, pp. 86–123. March 1986. Chow, V.T., 1959. Open-channel Hydraulics. McGraw-Hill Book Co., New York. Flato, G., Gerard, R., 1986. Calculation of ice jam thickness profiles. Proceedings, Fourth Workshop on Hydraulics of River Ice, Montreal, Canada, June 19–20, 1986.
Gerard, L., Andres, D., 1982. Hydraulic roughness and freezeup accumulations: North Saskatchewan River through Edmonton. Canadian Committee on River Ice. CGU-HS, Edmonton, Canada, pp. 62–87. Healy, D., Hicks, F., Beltaos, S., 1997. A comparison of the ICEJAM and RIVJAM ice jam profile models. Proceedings, 9th Workshop on River Ice, Fredericton, NB, Canada, September 24–26, 1997. Liu, L., and Shen, H.T. (1998) “A numerical model for river ice jam evolution.” Proceedings of the 14th International Symposium on Ice, Potsdam, NY, USA, 27–31 July 1998. H.T. Shen (ed.), A.A. Balkema, Rotterdam, 1999. Matousik, V., 1984a. Regularity of the freezing-up of the water surface and heat exchange between the water body and the water surface. Proceedings IAHR Ice Symposium, vol. 1. Hamburg, Germany, p. 187. August 27–31, 1984. Matousik, V., 1984b. Types of ice run and conditions for their formation. Proceedings IAHR Ice Symposium, vol. 1, p. 315. August 27–31, 1984. Moore, J.N., Landrigan, E.M., 1999. Mobilization of metal-contaminated sediment by ice jam floods. Environmental Geology 37 (1–2) (January 1999). Perham, R.E., 1983. Ice sheet retention structures. US Army Cold Regions Research and Engineering Laboratory, Hanover, New Hampshire, CRREL Report 83–30. Shen, H.T., Liu, L., 2003. Shokotsu River ice jam. Journal of Cold Regions Science and Technology 37 (1), 35–49 (July 2003). Shen, H.T., Wang, D.H., 1995. Under cover transport and accumulation of frazil granules. Journal of Hydraulic Engineering 121 (2). Shen, H.T., Su, J., Liu, L., 2000. SPH simulation of river ice dynamics. Journal of Computational Physics 165 (2), 752–777. Tatinclaux, J.C. and Gogus, M. (1982) “Stability of floes beneath a floating ice cover”, IAHR, 1981, vol. 2, pp 298–311. Tuthill, A.M., Gooch, G., 1998. A physical model study of ice retention Booms. Proceedings, IAHR 14th International Symposium on Ice, pp. 61–66, Potsdam, NY, 27–31 July 1998. US Army, 1998. HEC-RAS River Analysis System” Hydraulic Reference Manual Version 2.2. US Army Corps of Engineers, Hydrologic Engineering Center. September 1998. US Army, 2000. HEC-RAS River Analysis System” Users Manual Version 3.1.2. US Army Corps of Engineers, Hydrologic Engineering Center. October 2000. US Army, 2002. Ice Engineering” Engineer Manual 1110-2-1612. US Army Corps of Engineers, Washington DC. 30 October 2002 http://www.usace.army.mil/publications/eng-manuals/em1110-2-1612/toc.htm. Westwater Consultants et al. (2005) “Draft Restoration Plan for Restoring the Clark Fork River and Blackfoot River following the Removal of the Milltown Dam.” Westwater Consultants, Corvallis, MT, River Design Group, Whitefish, MT, and Geum Environmental Consulting, Hamilton, MT, April 13, 2005. White, K.D., 1999. Hydraulic and physical properties affecting ice jams. CRREL Report 99-11. US Army Cold Regions Research and Engineering Laboratory, Hanover, NH. Zipparo, V.J., Hasen, H., 1992. Davis' Handbook of Applied Hydraulics, 4th edition. McGraw-Hill. November 1992.