This admiralty case against the government arises because plaintiffs' barges were damaged as a result of the allegedly negligent operation of a navigational dam along the Ohio River near Cincinnati by the United States Army Corps of Engineers during a prolonged period of cold weather and ice formation. Our decision turns on two interlocking questions: whether controversial decisions made in connection with the operation of Markland Dam by the Corps are covered by the "discretionary function" exception to governmental liability under the Federal Tort Claims Act, and whether any conduct by the Corps not covered by this exception was negligent and caused an ice "jam" or "gorge" to form which damaged petitioner's barges. The District Court concluded that the discretionary function exception was inapplicable and that the Corps' negligence caused an ice jam to form, in turn causing barges to sink or be swept downstream against Markland Dam. It awarded $2,711,649.50 to plaintiffs as damages, plus prejudgment interest.

Petitioners Walker Towing Corporation and B-R River Services sought indemnity from the United States under the Suits in Admiralty Act, 46 U.S.C.App. Secs. 741-52, for damages sustained by them and their shippers as a result of the January 1978 disaster. The case was tried to the District Court. On September 24, 1985, the District Court filed a 136-page opinion and order concluding that the Corps was negligent in several aspects of the operation of its facilities, that the negligence, and not an unavoidable "Act of God," was the actual and proximate cause of the plaintiffs' damages, that the Corps was not entitled to immunity under the discretionary function exception of 28 U.S.C. Sec. 2680(a), and that the plaintiffs thus were entitled to recover from the government. We conclude that the major decisions by the Corps which plaintiffs claim negligently caused their injury are protected by the discretionary function exception. Therefore, we reverse the judgment of the District Court.

The Corps has divided the Ohio River into three operating districts: Pittsburgh, Huntington, and Louisville. There are two dams involved in this case: Meldahl, which is in the Huntington District; and Markland, in the Louisville District. Meldahl is located at mile 436.2; Markland is located downstream at mile 531.5. Meldahl is 35 miles above Cincinnati; Markland is 61 miles below Cincinnati. In Corps parlance, the "Meldahl pool" is the portion of the river upstream from Meldahl Dam, the "Markland pool" is the portion upstream from Markland to Meldahl, and the "McAlpine pool" is the portion upstream from McAlpine Dam at Louisville to Markland.

Meldahl and Markland are both navigational dams. River traffic passes the dam by way of navigational locks. Markland, the site of the problem in this case, consists of two locks on the south (Kentucky) side of the river and twelve water control gates, called "tainter gates," that span the entire river. The locks and dam structure is approximately 1500 feet wide. Each of the twelve tainter gates is 100 feet wide and 42 feet high. The locks are capable of raising or lowering tows of up to 15 barges and a towboat a distance of 35 feet. In addition to the locks and dam structure, Markland includes a hydroelectric plant on the Indiana side of the river that services part of southern Indiana.

When Markland was built in the late 1950's and early 1960's, the Corps was concerned with potential problems in passing accumulations of ice and drift through the dam. As a solution, the Corps installed tainter gates at Markland that could be submerged as well as raised. When a gate is raised, there is a gap between the river bed and the bottom of the gate; water passes under the gate. Some ice and drift can be passed under a raised gate because the effect of raising a gate is to create suction which would pull ice and drift from the surface underneath the gate and through the dam. In theory, however, it is much easier to pass ice and drift over a submerged gate. When a gate is submerged, ice and drift simply flow over the gate; much less force-and thus less water volume-are required to pull ice and drift over a submerged gate than under a raised gate. Five of the twelve gates at Markland are of the submergible type. Submergible gates can also be operated in the raised position; they are thus two-way devices.

By the middle of 1964, the Corps encountered significant problems in operating submergible gates at some other dams on the Ohio. When the submergible gates were lowered, they had a tendency to vibrate. Although the Court found that the vibration problem at Markland was not significant, dam personnel were not permitted to operate the gates in the submerged position at any of the Corps' dams on the Ohio. Therefore, in 1978, the only way to pass ice at Markland using tainter gates was by raising them. The District Court found that gates must be raised at least 10-12 feet off their sills, which are at the bottom of the river, before enough suction could be generated to pass ice under the gates. Additionally, the Court found that at least two adjacent gates must be raised before the suction is great enough to pass ice. Thus, ice could be passed using the gates only when the flow of the river was great enough that two gates could be raised without causing the level of the pool to fall below navigable levels. Generally, according to the witnesses and the District Court's findings, one foot of gate opening on one 100-foot wide gate passes 3,000 to 5,000 cubic feet per second (cfs). (Gate opening is measured by the total number of feet all gates are raised. E.g., if all twelve gates are raised one foot each, the dam is "running" twelve feet of gate opening.)

Other problems in passing ice also presented themselves. The Corps had determined that the concentration of all flows on two adjacent gates could cause riverbed erosion ("scour") immediately downstream of the dam unless the downstream pool was deep enough to alleviate the problem; otherwise, river flows had to be high enough that all gates could be raised. Additionally, the hydroelectric plant created ice-passing difficulties. The plant's full capacity was a flow of 35,000 cfs. The Markland lockmaster testified that the plant was routinely operated at full capacity. Flow used for the hydroelectric plant could not be used to allow the raising of gates; therefore, the 35,000 cfs used by the hydroelectric plant represented seven to twelve feet of gate opening that was unavailable.

The other ice-passing mechanism at Markland was the lock system. Markland has both a main and an auxiliary lock. The auxiliary lock, which is half as long as the main lock, is closest to the Kentucky shore; the main lock lies next to the auxiliary lock. The locks operate by means of "mitre gates," which swing out from the walls of the lock to allow filling or emptying, and swing back into the walls to allow vessels to enter and leave the lock. Additionally, each lock is equipped with one-piece "emergency leaves," which lift out of a recess built into the river bed. Ice can be passed through a lock by raising an emergency leaf to a level just below that of the upstream pool and allowing water and ice to flow over the leaf, thus "skimming ice." Due to a design problem, it was impractical to use the main chamber emergency leaf at Markland for passing ice. The auxiliary chamber could have been used; however, in 1976 a latch pin essential for use of the auxiliary chamber emergency leaf broke. The pin was not replaced until January 19, 1978. In any event, the Corps had permitted a "pusher boat" to become frozen in the chamber, thus eliminating the auxiliary lock from use as an ice-passing device.

The District Court's opinion gives a detailed, day-by-day account of weather forecasts and conditions, river flows and conditions, and Corps actions leading up to January 27, 1978. We need not recount all of that history here. Instead, we summarize only the facts relevant to our decision.

One of the sources relied upon by the District Court is the Corps' Ice Committee Report of June 1978 (Exhibit 179). An excerpt from that report provides a concise summary of the events that ultimately led to this litigation:

Tributary storage reservoirs had been effective in averting moderate flooding in several principal Ohio River tributaries during December. Much of the excess reservoir storage had been released and streamflows were approaching "normal" by the end of the month. However, below normal temperatures had begun to produce ice, particularly on tributaries. The continued development of ice both on tributaries and along the banks of the Ohio on into January would have to be considered normal for that time of year.

During 7-9 January a strong frontal system passed over the Ohio basin. Runoff producing rainfall was accompanied by rapidly falling temperatures which then produced snow. This began the accumulation of snow cover which was to ultimately set a record for the winter over much of the basin. It also signaled the advent of much colder temperatures which were to prevail for an extended period. More significantly, however, runoff from the rainfall set in motion much of the ice which was to later become such a problem. Runoff from uncontrolled areas below tributary storage reservoirs flushed substantial volumes of tributary ice out into the Ohio, where rising stages broke bank ice to join the overall mass....

In the middle reaches of the Ohio, streamflow increased threefold from the morning of 7 January until the morning of 12 January. Movement of ice resulted in progressively worsening conditions, and from evening of 16 January until morning of the 17th, no tows were able to get through the ice pack above Meldahl Dam. Thereafter, movement of boats and ice gradually reduced the congestion at Meldahl but the queue of waiting boats was not cleared until the 25th. All during this time, intermittent snowfall continued to add to overall coverage, but temperatures remained well below normal into the week of 22 January.

By 24 January, temperatures had risen enough that moisture moving into the basin that day fell in the form of light rain. Intensity increased only slightly in the next 24 hours but there was no evidence of runoff or increased streamflow. It appeared that the snow pack was absorbing and holding the rainfall. Weather forecasts on the morning of 25 January indicated continuation of relatively light rain for another 12-18 hours and then a return to very cold conditions. While rainfall intensity was expected to increase slightly, total amounts expected were not considered alarming. Rainfall intensity increased a great deal more during the day than had been anticipated and, by late afternoon, the rain was being forecast to turn to heavy snow by about midnight and to continue into the 26th.

Forecasts on the morning of 25 January had been based on the belief that a low pressure system in the Gulf states, associated with a warm front to the south and southeast of the Ohio basin, would pass to the southwest of the basin and up the Atlantic coast without significant impact on the Ohio basin. Coming into the western and northwestern portions of the basin was a cold front that was expected to move eastward over the basin, bringing much colder air and shutting off moisture. Later in the day, the low pressure system unexpectedly joined the cold front, intensified, and took a northeasterly path right across the Ohio basin. With record low barometric pressures, the storm passed through the Cincinnati area shortly after midnight with wind gusts reaching 50 mph. Gusts reached 80 mph by the time the storm exited the basin in the Cleveland area later in the day on the 26th.

Temperatures on the middle reaches of the Ohio plunged to around zero on the morning of 26 January and strong winds prevailed through much of the following day. Blizzard conditions paralyzed much of the area, but the most important factor relating to river ice had been the rainfall immediately preceding the enveloping cold air. In about 48 hours, rainfall intensity steadily increased from an insignificant drizzle to fairly heavy just before turning to snow. Rainfall averaged about 1 1/2 inches over most of the southern tributaries. With overflow of storage reservoirs cut to minimum, runoff from remaining uncontrolled portions of these watersheds brought rapid stream rises. The collective effects of this outpouring into the Ohio main stem were rapid rises on the main stem, and operators of the navigation structures had no alternative but to simply pass through whatever water or ice came to them from upstream.

The river stage at Cincinnati rose from 29.6 feet on the morning of 25 January to 35.3 feet by the next morning and then to 46.6 feet by the morning of the 27th. Thereafter, the rise slowed and a crest of 53.9 feet was subsequently reached on the 30th. Ice movement, with periodic stoppage and accumulation, had been a significant problem since the 16th. By the 25th, several miles of the river upstream from Markland Dam had become so jammed with ice that navigation had essentially ceased. The rapid rise beginning on the 25th and the forcing of more ice into that jam then produced a nightmare for the next several days. It is difficult to know with certainty just when the massive ice jam began to become a gorge jam causing a damming effect. Best indications are that it was late morning on the 27th, when localized flooding was reported at Aurora, Indiana. In any event, force of the rapidly increasing riverflow soon broke the gorge in the afternoon and moved much of the ice down against Markland Dam.

Exhibit 179 at App. 1, pp. 6-7.

When the ice jam described above broke, at 4:17 p.m. on January 27, a wall of ice and water swept downstream. Although the speed of the broken gorge was only four or five miles per hour, its force was sufficient to sweep away anything in its path. As the District Court described the scene:

What occurred was probably the greatest single disaster to hit the Ohio River in modern times. Barges, towboats, and docks were swept onto or through Markland Dam; barges and towboats were sunk; Big Bone Island [at mile 517] was simply shaved off, and does not now exist. Cleanup operations took weeks, and these events have earned a permanent place in Ohio River lore.

Dist.Ct. opinion at 90. Both Walker Towing and B-R River Services sustained property damage when the jam broke.

Several experts testified regarding the formation, characteristics, and dissipation of ice on the river. A general understanding of river ice is necessary before we discuss the specifics of the January 1978 ice.

Ice can form on a river only when the water reaches maximum density, which occurs at around four degrees Celsius or 39 degrees Farenheit. At this point, the formation of ice can occur in two different ways. Sheet ice will form if there is little turbulence. The water at the surface is cooled to zero degrees Celsius (the freezing point), while water below the surface is not quite as cold. The more quickly sheet ice forms, the thinner it will be, because the ice cover will insulate the river from colder air and thus trap the heat energy in the water. Sheet ice presents relatively few problems to navigation, primarily because of the structure of sheet ice crystals. The crystals are long and thin and arranged in a vertical plane. Thus, when sheet ice is broken (e.g., by a towboat), it simply fractures vertically. Once the ice is fractured, the thermal energy in the water below supplies enough heat to melt the ice.

The second way ice can form on a river, and the way the ice formed in January 1978, occurs when the water is turbulent and the air temperature is extremely cold. Turbulent water can reach a temperature slightly below the freezing point before solidifying. In this situation, most of the heat energy in the water has been dissipated. When the "supercooled" water (water in liquid form with a temperature below freezing) comes into contact with the subzero air, "frazil ice" can form. Frazil ice crystals are small and flat, and cling to one another and anything else with which they come into contact, creating a sort of slush. Once frazil ice has formed, individual particles of frazil ice come together to form larger pieces of ice, called "brash ice." Brash ice in turn may develop into even more solid pieces of ice called "black ice."

The formation of frazil, brash, and black ice creates more serious navigational problems than the formation of sheet ice. Ice cover on the river becomes an accumulation of floes. Because the thermal energy of the water must be largely dissipated before frazil ice will form, there is less potential for melting than with sheet ice. The longer-lasting nature of this type of ice, the creation of large moving floes, and the tendency of the ice to stick to anything it touches combine to cause a variety of navigational difficulties.

If the ice floes are unobstructed, the ice will simply move downriver. If, however, something obstructs the ice, a "jam" (also called a "gorge") can form. A variety of factors can independently or in unison cause ice to be arrested. Natural obstructions can include a narrowing in the river, bends in the river, islands or shoals. Man-made obstructions such as dams can also arrest the downstream progress of ice.

Once ice is obstructed, one other set of conditions must obtain before a jam forms. If the ice simply stays in place in a single layer, the river will continue to flow underneath the ice and relatively little harm will result. However, if the velocity of the ice moving downstream is great enough, and more and more ice comes downstream, the new ice will begin to pile up underneath the ice already present, creating a damming effect-an ice jam.

Engineers have developed a mode of analysis to aid prediction of when and where jams will form. The analysis was developed to predict when logs would jam on a river, but has been adapted as a predictive tool for ice. This analysis relies on a variable called the "Froude number." The formula for obtaining the Froude number for a given point in the river is:

where V is the velocity of the water, Y is the depth of the water, and g is the gravitational constant, 32.2 ft./sec². Thus, the faster and shallower the river is at a certain point, the higher the potential for jamming. According to the plaintiffs' ice expert, Samuel Lazier, a value greater than Fr = 0.08 indicates a high potential for jamming. The use of Froude numbers to predict jam formation is complicated by the further facts that upstream progression of an ice cover occurs more easily at lower Froude numbers, as does clogging of ice traveling downstream. Indeed, by the close of his testimony, Lazier acknowledged that Froude number analysis is merely an "indicator," and is not to be taken "literally." Tr. 826-27.

The development of an ice jam occurs over a period of time. New ice is constantly delivered to the jam, but the older ice is eroded somewhat by the flow of the river. Eventually, as more new ice arrives than old ice erodes, a jam reaches the point where it must either deteriorate or break. If the jam deteriorates, the ice simply continues to move downstream with the normal flow of the river. However, if the jam is large enough to cause a severe blockage of river flow, a differential head will form. A differential head is a large buildup of water on the upstream side of the jam combined with relatively little water on the downstream side. This creates a large force across the jam. The force of the differential head can be great enough to break the jam and sweep it quickly downstream, carrying anything in its path (e.g., barges and towboats) with it. The technical term, derived from the French, for this rush of water and ice is, appropriately, a "debacle."

After a lengthy bench trial, the District Court found that the government was liable for all of Walker Towing's damages and 40% of B-R River Services' damages. The District Court summarized its holding as follows:

For reasons which follow, we find and conclude that the United States was negligent in several aspects of its operation of facilities owned and operated by it, through the Army Corps of Engineers. The gist of the negligence findings are that the Government, under a duty to use due care in the operation of its locks and dams, breached that duty by, inter alia, failing to pass ice through Markland Locks and Dam as the ice came downstream during the middle weeks of January, 1978; failing to engage in routine surveillance of the reaches of the Ohio River upstream from Markland; failing to train its personnel in methods of handling and evaluating ice conditions; failing to ensure that the dam was in fact capable of dealing with reasonably foreseeable ice conditions; and failing to intermittently restrict diversion of river flows to a hydroelectric generating facility located on the Markland site so that ice could be passed through the dam.

We also conclude that the acts or omissions detailed herein constitute deviations from the appropriate standard of care and were a proximate cause of the damages sustained by petitioners Walker Towing Company, Inc. and B-R River Services, Inc. However, we conclude that B-R River Services, Inc., was itself negligent, and attribute to that petitioner sixty percent of the responsibility for its recoverable losses.

Finally, we conclude that Walker Towing Company is entitled to recover for its losses in the amount of $1,500,000.00, and B-R River Services, Inc. is entitled to recover for its losses in the amount of $711,649.50, both together with prejudgment interest at the statutory rate from the dates on which petitioners settled with their claimants.

Dist.Ct. opin. at 4-5.

The United States has appealed the decision of the District Court on several grounds. First, the government asserts that all of its actions are shielded from liability because all of the actions upon which liability was predicated were discretionary functions. Second, the government asserts that the Corps' handling of the ice conditions in January 1978 did not cause the breaking of the ice jam and that the District Court was clearly erroneous in so holding. Third, the government argues that the District Court held the Corps to an unreasonable standard of care. Additionally, the government asserts that even if the disaster can in some way be attributed to the actions of the Corps, the disaster was too remote a consequence of those actions to be regarded as the proximate cause of the disaster and was otherwise unforeseeable. Finally, Walker Towing cross-appeals from the District Court's denial of its request for attorneys' fees under the Equal Access to Justice Act, 28 U.S.C. Sec. 2412. Because we hold that the District Court erred in finding that no Corps actions causally related to the disaster were shielded by the discretionary function exception, we need not reach the government's arguments on proximate cause. Additionally, we do not decide the cross-appeal because the plaintiffs are no longer prevailing parties.

When Congress enacted the Federal Tort Claims Act, it preserved the government's sovereign immunity for

any claim based upon an act or omission of an employee of the Government, exercising due care, in the execution of a statute or regulation, whether or not such statute or regulation be valid, or based upon the exercise or performance or the failure to exercise or perform a discretionary function or duty on the part of a federal agency or an employee of the Government, whether or not the discretion involved be abused.

28 U.S.C. Sec. 2680(a). We have held that the discretionary function exception applies to the Suits in Admiralty Act as well as to the FTCA. Chotin Transp., Inc. v. United States, 819 F.2d 1342, 1347 (6th Cir.1987) (en banc). The question whether the discretionary function exception shields the government from liability is one of subject matter jurisdiction; therefore, we examine that issue first. Feyers v. United States, 749 F.2d 1222,1225 (6th Cir.1984); Carlyle v. United States Department of the Army, 674 F.2d 554, 556 (6th Cir.1982).

In Dalehite v. United States, 346 U.S. 15, 73 S. Ct. 956, 97 L. Ed. 1427 (1953), the Supreme Court first addressed and elaborated on the scope of the discretionary function exception. It enumerated three types of administrative activities which are covered by discretionary function: the initiation of programs, planning the operation of programs and carrying out programs in accordance with directions. In holding the government immune from liability for damages resulting from negligent drafting of fertilizer export regulations, negligence in the manufacture of fertilizer, and the negligent failure to police the shipboard loading of fertilizer, the Court held that all three types of alleged negligence were protected by the exception.

It is unnecessary to define, apart from this case, precisely where discretion ends. It is enough to hold, as we do, that the "discretionary function or duty" that cannot form a basis for suit under the Tort Claims Act includes more than the initiation of programs or activities. It also includes determinations made by executives or administrators in establishing plans, specifications, or schedules of operations. Where there is room for policy judgment and decision there is discretion. It necessarily follows that acts of subordinates in carrying out the operations of government in accordance with official directions cannot be actionable. If it were not so, the protection of Sec. 2680 would fail at the time it would be needed, that is, when a subordinate performs or fails to perform a causal step, each action or nonaction being directed by the superior, exercising, perhaps abusing, discretion.

Id. at 35-36, 73 S.Ct. at 967-968 (footnotes omitted). The Court found that government decisions on which the plaintiffs sought to base liability "were all responsibly made at a planning rather than operational level and involved considerations more or less important to the practicability of the Government's fertilizer program." Id. at 42, 73 S.Ct. at 971.

In Miller v. United States, 583 F.2d 857, 867 (6th Cir.1978), we applied Dalehite to hold the discretionary function exception applicable to a case arising from the release of large quantities of water from Lake Superior to Lake Huron through the lock and dam system at Sault Ste. Marie. We pointed out that Dalehite itself relied on several lower federal court cases immunizing the government from liability in connection with the operation of dams and locks:

In this regard, the Court cited with approval a number of District Court decisions involving claims of water damage. In Lauterbach v. United States [95 F. Supp. 479 (W.D.Wash.1951) ], the court rejected the plaintiffs' claim of damage due to the release of flood waters from a dam, based in part on the discretionary function exception as applied to the design and construction of the dam, and in part on a finding that there was in fact no negligence shown in its operation. In Olson v. United States [93 F. Supp. 150 (D.N.D.1950) ], the court granted a motion to dismiss plaintiff's claim that employees of the United States negligently opened the floodgates of a dam, destroying plaintiff's livestock; the court concluded that "When flood waters are to be released and how much water is to be released certainly calls for the exercise of judgment.... The Government's agents did not open the gate in the dam in a negligent manner. They merely abused their discretion as to when to open it." [93 F.Supp. at 152-53]

583 F.2d at 867 (footnotes omitted) (emphasis original). See also Coates v. United States, 181 F.2d 816 (8th Cir.1950).

The viability of Dalehite was reaffirmed in United States v. S.A. Empresa de Viacao Aerea Rio Grandense (Varig Airlines), 467 U.S. 797, 104 S. Ct. 2755, 81 L. Ed. 2d 660 (1984). In that case, the Court held that the negligent failure of the Federal Aviation Administration to inspect certain aspects of aircraft design during the certification process was protected under the discretionary function exception. Most recently, a unanimous Supreme Court emphasized that the discretionary function exception does not preclude liability for any and all acts by a federal agency, but "applies only to conduct that involves the permissible exercise of policy judgment." Berkovitz v. United States, --- U.S. ----, 108 S. Ct. 1954, 1960, 100 L. Ed. 2d 531 (1988)(delegation to private sector of vaccine safety testing in violation of FDA regulations not covered by exception).

The District Court rejected the government's contention that all of the actions for which it was held negligent were, under the principles of Dalehite, Varig Airlines, and Miller, protected by the discretionary function exception. The actions which the District Court deemed negligent and unprotected by immunity were: 1) failure of the Corps to conduct periodic inspections of the Markland pool; 2) failure to replace the broken latch pin in a timely manner and allowing the pusher boat to become frozen in the auxiliary lock chamber; 3) failure to sufficiently train lock personnel; 4) failure to restrict the flow to the hydroelectric plant in order to have more flow through the gates, thereby increasing Markland's ice-passing capacity; 5) failure to coordinate ice-passing activities with Meldahl and to follow the same policy that Meldahl followed in maximizing the passage of ice; and 6) failure to compensate for the immobilization of the submergible tainter gates.

Applying the principles of these cases to this case, we conclude that the District Court correctly viewed the first three of the Corps' actions as involving operational, non-policy decisions. The District Court erred, however, in its analysis of the Corps' decision not to restrict the flow to the hydroelectric plant and as a consequence not to follow the same policy of maximizing ice passage that the Corps followed at Meldahl. The Court also erred in its analysis of the Corps' decision not to compensate for the immobilization of the submergible gates.

The District Court decided that the decision not to cut back flow to the hydroelectric plant was made entirely by the Markland lockmaster:

Senior Corps personnel testified that, during the period here in question, the Corps was under pressure from the Governor of Indiana to maintain maximum output from the Markland generating plant. Power was needed because the United Mine Workers were on strike, and coal was at a premium. Thus, there are overtones of intergovernment relations, matters likely discretionary, which affect consideration of this point.

However, Markland Lockmaster Sheldon testified that he had the authority to cut the amount of water diverted to the power plant down to 5,000 cfs. He did not testify that he had been instructed that the plant must be kept on line, nor did he testify that he was aware that pressure had been brought to bear upon his superiors to maintain the plant at full capacity.

Had the Corps, through e.g., Mr. Whitlock [one of Sheldon's superiors], either formally or informally ordered that the power plant be fed the maximum amount of water under all circumstances, it is likely that we would see that as a policy determination isolated from challenge by the discretionary function exception. However, because the point was not translated into an operation guideline for personnel at the dam, and because it is unquestioned that as the matter stood in January 1978, workers at the dam had the authority to cut back flow to the plant, we find that decisions, or omissions, regarding the extent to which water would be diverted to the hydroelectric plant fall outside the purview of the discretionary function exception.

Dist.Ct. opin. at 110-11.

The District Court's view that the power plant decision was nondiscretionary thus hinges on its finding that the Corps did not order the lockmaster to feed the maximum amount of water to the plant under all circumstances. But the record is clear that the decision was made at a higher level.

It is true that under normal circumstances, the lockmaster could vary the flow to the power plant; however, the record is clear that once the ice emergency and severe weather conditions began to create problems on the river in late January 1978, Corps policymakers became involved. General Heiberg, who was at the time the Division Engineer for the Ohio River Division, testified that the competing needs for power and navigability were weighed and that he decided that the need to generate power should prevail in that situation. See Tr. 4--372-75, 4--431 ("[I]t was essentially my decision."). District Engineer Whitlock also testified that the decision to continue maximum diversion to the power plant was made by balancing the competing needs for electricity and maintenance of navigation. Tr. 3--262-64, 3--337. Larry Dickson, who was directly under Whitlock and was the immediate supervisor of Lockmaster Sheldon, testified that while the plant was operating at near capacity, he discussed the plant with both Sheldon and Whitlock, and the decision was made to tell Sheldon to "keep on doing what you're doing." Tr. 3--321-22. Sheldon testified that he was in communication with Whitlock and Dickson several times a day, and that Whitlock was making the decisions about passing ice. Tr. 763-64, 715. Thus, Sheldon was not exercising judgment in January 1978; he was merely carrying out the discretionary decisions of his superiors.

We emphasize that our inquiry into the factual question whether it was Sheldon or his superiors who were responsible for the power plant decisions is relevant only to determine whether a concededly discretionary decision was implemented. In undertaking this inquiry, we do not deviate from the principle that the character of the decision, and not the identity of the decisionmaker, determines whether the discretionary function exception applies. "It is the nature of the conduct, rather than the status of the actor, that governs." Varig Airlines, 467 U.S. at 813, 104 S.Ct. at 2764. The District Court correctly recognized that the character of the power plant decision by the Corps policymakers was discretionary, but it clearly erred in concluding that this decision was never effectuated. We are convinced that a definite mistake has been committed, and we reverse on this point. United States v. United States Gypsum Co., 333 U.S. 364, 395, 68 S. Ct. 525, 541, 92 L. Ed. 746 (1948).

The District Court's ultimate conclusion was that "[f]ailure to pass ice at Markland proximately caused the ice gorge at Big Bone Island." Dist.Ct. opin. at 134. This decision was heavily influenced by the Court's view of action taken by personnel upstream at Meldahl. The District Court noted that "large amounts of ice were continually passed through Meldahl into the Markland pool," id. at 133, and concluded that "[a]ll this ice--ice coming through Meldahl, ice coming from tributaries, and ice forming in the Markland pool--had to go somewhere." Id. at 131. Thus, according to the District Court, the passage of ice at Meldahl combined with the failure to pass ice at Markland to cause the disaster.

This finding is rooted in two separate negligence findings. First, the Corps was found negligent for failing to coordinate ice passage between Meldahl and Markland. Second, the Corps was found negligent for failing to pass ice at Markland.

The record is clear that ice-passing efforts at Meldahl were not coordinated with efforts to manage the situation at Markland. The District Court found that

[T]he lockmaster at Meldahl did not know what was going on at Markland, that the lockmaster at Markland did not know how ice was being passed at Meldahl, and that the Division personnel were unwilling to override the decision of District personnel in Huntington to maintain operations at Meldahl, although the Division personnel had better knowledge regarding the problems those operations were causing downstream.

Id. at 63.

While the District Court's conclusion that failure to coordinate activities between Meldahl and Markland contributed to the disaster may be correct, that failure is protected by the discretionary function exception. The District Court found that a "lack of oversight control" existed, but the Division personnel were exercising oversight control when they were apprised of the situation and decided not to override decisions made by lower-ranking personnel. That action by Division personnel may or may not have been made by considering all of the pros and cons of allowing Meldahl to keep passing ice, but "[e]ven the negligent failure of a discretionary government policymaker to consider all relevant aspects of a subject matter under consideration does not vitiate the discretionary character of the decision that is made." Myslakowski v. United States, 806 F.2d 94, 97 (6th Cir.1986), cert. denied, 480 U.S. 948, 107 S. Ct. 1608, 94 L. Ed. 2d 793 (1987). Division personnel decided that they would not instruct Meldahl to stop passing ice in order to "help out" Markland. Whether this was the correct decision is not for us to second-guess. The purpose of the exception is "to prevent the courts from intruding, through the vehicle of tort suits, upon the decisionmaking authority of the other branches of government." Estate of Callas v. United States, 682 F.2d 613, 620 (7th Cir.1982).

The second way the District Court's decision was influenced by actions taken at Meldahl is seen in the finding that the failure to pass ice at Markland proximately caused the gorge. Essentially, the District Court reasoned as follows: Meldahl passed ice, and no jam formed above Meldahl. Markland did not pass ice, and a jam formed above Markland. Therefore, if Markland had emulated Meldahl and passed ice, no jam would have formed above Markland.

The problem with the District Court's analysis is that the situations at Markland and at Meldahl should not be treated as equivalent. As our discussion of the hydroelectric plant illustrates, the problems at Markland were unique. Meldahl personnel never had to balance the competing interests of navigation and generation of electricity, because the Meldahl facility did not have a power plant. Thus, the decision to pass ice and methods of passing ice at Meldahl are not comparable to the decisions made at Markland. Once the discretionary decision to continue supplying the hydroelectric plant with water was made at Markland, the courses of action open to Markland personnel were far more limited than the courses of action open to Meldahl personnel.

The low flows to the dam that resulted from the discretionary decision not to divert water from the hydroelectric plant left the Markland personnel with only three possible ways to pass ice: through the auxiliary lock chamber, through the main lock chamber, or by concentrating the flow to two gates and attempting to pass ice under those gates.

The District Court found that the auxiliary chamber could not be used because it was inoperable due to Corps negligence. We do not disturb this finding. The significance of this negligence is small, however. The record indicates that use of the auxiliary chamber to pass ice was "of limited utility" and "interfered with locking procedure." Testimony of Patrick Carigan, Tr. 45. Carigan also testified that "you shouldn't lock vessels through [the main chamber] at the same time you're passing ice [through the auxiliary chamber] because it's too dangerous." Id. at 11.

The main chamber was not devoted to passing ice because the decision was made to continue locking boats through the chamber in an effort to maintain navigation. Lockmaster Sheldon was instructed to "lock boats and ice." Tr. 3-321. Carigan testified that the view of the Louisville District was that "everyone's interest would be better served by attempting to get as many of the boats out of that pool as possible...." Tr. 13. He stated that the choice was "to lock rather than pass ice." Id. at 14. Although some ice was passed along with boats that were locked, the amount was small. The decision to maintain navigation at the expense of passing ice, however, is a decision the Corps is uniquely qualified to make, and is protected by the discretionary function exception. See Estate of Callas, 682 F.2d at 620.

The District Court also held that the Corps' failure to compensate for the immobilization of the submergible tainter gates by maintaining the structure in a condition that would permit underflow passage of ice at relatively low flows was negligent. Dist.Ct. opin. at 125. This holding is also in error. The record was clear that Corps personnel could not concentrate flow to two gates in an attempt to pass ice because of concern about "scour." The Corps' weighing of the need to pass ice against concern for the structural integrity of the dam is protected by the discretionary function exception. The District Court correctly held that the creation of the condition that precipitated the decision--the immobilization of the submergible gates--was protected by the exception. Dist.Ct. opin. at 104-06. It was error for the Court to then hold that the Corps' failure to compensate for the immobilization was negligent. Again, the Court slipped into the error of comparing Meldahl with Markland. At Meldahl, the lockmaster was able to "walk the gates"--concentrate flows for short periods at different gates--and thus compensate somewhat for the immobilization of the submergible gates. At Markland, though, the flows were simply not high enough to pursue that tactic because much of the flow was diverted to the hydroelectric plant.

We have held that only the negligence concerning surveillance, training, and the auxiliary lock are unprotected by the discretionary function exception. It is clear from the District Court's findings that these negligent acts, standing alone, were not substantial factors in causing the disaster. The finding of causation is very close even when the Corps' conduct which we have found protected is included as a permissible cause. Once the actions that are protected by the discretionary function exception are removed from the causation equation, the District Court's finding can no longer stand.

This was a long and difficult case, and the District Court striggled valiantly with many complex issues. We hold, however, that the District Court erred in its discretionary function decision, and that error led to an erroneous finding of liability. Therefore, the judgment of the District Court is reversed.