concurring in part, and dissenting in part:
My most fundamental objection to the majority opinion is that the facts set forth in the opinion bear little resemblance to the findings of fact made by the district court following a bench trial. Rather than adopting the findings of fact made by the district court, as we are mandated to do absent a showing of clear error, see Zivkovic v. Southern California Edison Co., 302 F.3d 1080, 1088 (9th Cir.2002), the majority completely rewrites the facts to such an extent that it decides a different case on different facts than that decided by the district court. Our precedent is to the contrary — the clear error standard of review “is significantly deferential, and we will accept the lower court’s findings of fact unless we are left with a definite and firm conviction that a mistake has been committed.” Lentini v. California Center for the Arts, 370 F.3d 837, 843 (9th Cir.2004) (citation and internal quotation marks omitted).
A few examples will illustrate my point:
• The district court found that at Site 1A, 154 soil samples were taken and only one sample exceeded the safe standard for thallium. Similarly, of the 99 soil samples taken from Site 2A, only two exceeded the safe standard for thallium. The district court found that two of the exceedences were slight and one appeared to be due to an unreliable test result. The district court specifically found that lead was of greater prevalence “and was considered to be the primary risk to health.” District Court Opinion, p. 4.
The majority opinion translates the district court’s finding into a histrionic description of the poisonous nature of thallium and a reference to the Navy’s Record of Decision (ROD) to disregard the district court’s finding. Majority Opinion, pp. 1024-25 and n. 2. This approach is the antithesis of our charge. Indeed, if there is “suppoH in inferences that may be drawn from facts in the record,” the district court’s finding cannot be clearly erroneous. USAA Federal Savings Bank v. Thacker (In Re Taylor), 599 F.3d 880, 888 (9th Cir.2010). The district court was not required to parrot the ROD or any other document or testimony that was part of the evidence presented. Rather, the district court, after considering all the evidence, made its ftndings. The district court specifically relied on evidence that “[sjampling data revealed few concentrations of thallium at very low levels.” District Court Opinion, p. 4. That finding was supported by the testimony of the industrial hygienist assigned to the project by the contractor.
• The district court found that the air monitors used by the contractor “were appropriate for the task.” District Court Opinion, p. 7.
The majority opinion makes what can only be characterized as an inappropriate contrary finding that “the air monitoring device specified did not monitor ‘total dust’ but only dust particles smaller than a certain size.” Majority Opinion, pp. 1025. However, the majority’s finding on appeal is contrary to the explicit testimony of the assigned industrial hygienist that total dust was measured. The district court acted completely within its dis *1039 cretion as the factfinder to credit this testimony rather than the testimony offered on behalf of the Plaintiffs.
• The district court found that the contractor’s industrial hygienist “had on-site responsibility and authority to stop work if working conditions presented a risk to health and safety.” District Court Opinion, p. 4.
The majority opinion concludes that because neither of the Navy’s industrial hygienists reviewed the health and safety plan, the Navy failed to meet its obligations under the governing agreement. See Majority Opinion, p. 1025-26. However, nothing in the governing agreement required the industrial hygienist to be a Navy employee.
• The district court found that:
[t]he Defendant never received any complaints in 1999 from residents of the adjacent housing complex or the elementary school about blowing dust from Box Canyon. Of the various health and safety officials that routinely visited the site, none observed conditions causing concerns. A teacher at the elementary school testified that her classroom faced the landfill. Daily, the teacher was in a position to notice whether dust was spreading towards either the school grounds or farther to the school buildings. The school teacher testified that over the life of the Box Canyon project, she observed dust coming over the fence onto the school grounds on only two or three occasions, and that each time the dust came over the perimeter fence, it did not approach the school
District Court Opinion, p. 7.
The district court credited the Navy’s proffered testimony regarding the measures taken to prevent dust on the site, and specifically credited the testimony of the assigned industrial hygienist in finding that “[a]ny visible dust was likely from uncontaminated soil.” Id. The district court acknowledged that “there were occasions where the action level for airborne dust was exceeded ...” Id., pp. 7-8. However, the district court attributed that to the fact that the “action level” set by the contractor was so low that there were exceedences where there was no visible dust.” Id., p. 8.
The majority’s version reads:
“The district court found that there were ‘occasions’ in the course of the project when the air monitoring equipment registered dust levels in excess of the levels that were supposed to require work stoppages (called “exceedences” by the parties) but did not note that record evidence showed that such exceedences occurred more than two hundred times.... It is undisputed that the work was never stopped because of these exceedences.... [T]he Navy’s QAO for the project, Nars Ancog, never looked at the air monitoring data collected by [the contractor] nor did anyone else from the Navy. Residents in the nearby Wire Mountain Family Housing area testified that, at times during the remediation project in 1999, visible clouds of dust blew from the landfill ...” Majority Opinion, p. 1026.
This unmitigated and unauthorized appellate factfinding flies directly in the face of testimony from the assigned hygienist that the exceedences required work stoppage only if the dust could not be controlled. The district court’s acceptance of that testimony was completely consistent with its factfinding function.
• As to the Plaintiffs specific allegations, the district court found that tests conducted in 2000 at Box Canyon, and at the adjacent housing com*1040plex and school revealed “no elevated levels of thallium.” Soil samples from Plaintiffs residence and “swipe samples” from inside the residence contained no thallium. The district court noted that there was “mixed testimony” about the results of the urine samples, but found that the evidence failed to connect any exposure to “the Government’s negligence in employing and supervising its contractor.” District Court Opinion, p. 8.
The majority opinion contrasts the following “assertions” by the Navy and “contentions” by the Plaintiffs: 1) the samples were taken months after the project and following a windy, rainy winter; 2) the Navy’s report was inconsistent with the raw data; and 3) Plaintiffs’ expert’s opinion was that thallium was present in Plaintiffs’ residence and adjacent homes. Majority Opinion, p. 1026-27.
The existence of counter assertions and contentions between the parties invokes the quintessential factfinding function of a trial judge, which we review for clear error. See Zivkovic, 302 F.3d at 1088. Rather than doing so, the majority completely disregards the district court’s findings, conducts a retrial on appeal and finds for the Plaintiffs on the bases that the discretionary function exception did not apply and that the Navy acted unreasonably during the project. I disagree on both counts.
1. The Discretionary Function Exception
a. The Manual Provisions
The majority opinion’s analysis falters with its acknowledgment that the district court failed to make findings regarding whether the provision in the Manual was mandatory or whether it required review of the safety plans by Navy personnel. See Majority Opinion, p. 1029. Following a bench trial, if the findings are inadequate on a contested issue, it is our obligation to remand the issue to the district court to resolve the factual dispute. See Zivkovic, 302 F.3d at 1090-91. Rather than adhering to this basic precept of appellate review, the majority opinion completely usurps the function of the district court. See Fisher v. Roe, 263 F.3d 906, 912 (9th Cir.2001), abrogated on other grounds in Mancuso v. Olivarez, 292 F.3d 939, 944 n. 1 (9th Cir.2002) (“This case graphically illustrates one of the bread and butter principles of appellate review that governs the manner in which we measure the work of a trial court. Trial courts find facts. We do not.”)
The majority professes to make a legal conclusion that the Safety and Health Program Manual required review of the health and safety plan by a Navy hygienist. See Majority Opinion, p. 1030 n. 5. However, in truth no provision of the manual mandates use of a Navy hygienist. And the district court judge credited express testimony from the author of the Manual that there was never any intent to utilize a Navy hygienist. Once again, the majority completely usurps the factfinding function of the trial judge, disregards the evidence and retries the case.
Because I would remand for the district court to resolve the factual issue that must be resolved before a legal ruling can be made, I dissent from the majority’s conclusion that the Navy Manual provisions imposed a mandatory duty on the government.
b. The Federal Facility Agreement Provisions
The majority opinion concludes, and I agree, that the provisions in the agreement were not sufficiently specific to render them mandatory requirements. See Majority Opinion, p. 1030-31. The majority *1041then proceeds to consider whether the discretion conferred upon the Navy was based on considerations of public policy. See id., p. 1030, citing Terbush v.United States, 516 F.3d 1125, 1129 (9th Cir.2008).
In Terbush, we recognized a clear distinction between decisions involving routine maintenance, that do not involve policy considerations and more extensive undertakings that do implicate policy considerations. See id. at 1133-34.
We cited as examples of routine maintenance snow removal and removing mold from a commissary meat department. See id. In contrast, we observed that “repairing [a] roadside wall involved balancing several policy considerations.” Id. at 1134.
The majority opinion concludes that the outcome of this case is dictated by Marlys Bear Medicine v. United States, 241 F.3d 1208 (9th Cir.2001). See Majority Opinion, pp. 1032-33. However, the majority’s conclusion is irretrievably tainted by the impermissible factfinding reflected in its statement of facts. To support its conclusion, the majority opinion relies on “the Navy’s failure to have [the contractor’s health and safety plan] reviewed by the Navy’s own [industrial hygienist] or other competent person and the failure of the Navy’s [Quality Assurance Officer] to inspect any ah’ monitoring ...” Majority Opinion, p. 1033. However, as the majority previously noted, the district court made no findings regarding whether the Navy Manual required use of a Navy hygienist or whether use of the contractor’s hygienist would satisfy the requirement. Rather than acknowledging that the lack of findings precludes appellate review, and remanding for appropriate findings, the majority engages in appellate factfinding to conclude that the Navy failed to meet its obligations to review the contractor’s health and safety plan. See Majority Opinion, pp. 1029-30. I cannot co-sign this blatant departure from the confines of appropriate appellate review.
2. Reasonableness of the Navy’s Conduct1
As an initial matter, the majority concludes that failure by the district court to make adequate findings on this issue would support a finding of clear error. See Majority Opinion, p. 1033 n. 6. Although the majority cites United States v. Hinkson, 585 F.3d 1247, 1262 (9th Cir.2009) to support this conclusion, Hinkson says no such thing. Rather, Hinkson reiterates that on appeal the factual findings of the district court are reviewed. See id. at 1263 (“[W]e look to whether the trial court’s resolution ... resulted from a factual finding that was illogical, implausible, or without support in inferences that may be drawn from the facts in the record.”). This test presupposes the existence of reviewable factual findings. In the absence of adequate factual findings, Zivkovic instructs us to remand to the district court.
Having cleared the way by misplaced reliance on Hinkson, the majority then unabashedly retries the case. For example, the majority opinion accuses the district court of “mischaracteriz[ing] the test results regarding the presence of thallium.” Majority Opinion, pp. 1036-34. However, the record reflects that the district court credited the testimony of the Navy’s experts over that of the Plaintiffs’ experts. See District Court Opinion, p. 8. In fact, the majority concedes that there *1042was a dispute between the parties on this point. See Majority Opinion, p. 1035 n. 9. When a trier of fact resolves such a dispute, no clear error occurs. See Anderson v. City of Bessemer City, 470 U.S. 564, 574, 105 S.Ct. 1504, 84 L.Ed.2d 518 (1985) (‘Where there are two permissible views of the evidence, the factfinder’s choice between them cannot be clearly erroneous.”) (citations omitted).
The other two bases supporting the majority’s conclusion are similarly deficient. See Majority Opinion, pp. 1036-37. As previously discussed, because the majority recognizes that there were not adequate findings made regarding the industrial hygienist requirement, the matter should be remanded. The discussion regarding the air monitoring samples is simply a reiteration of the impermissible reweighing of the facts described above. In sum, the majority opinion’s analysis and final conclusions are irreversibly tainted by its failure to adhere to the fundamental precepts of appellate review. Although I agree that additional factfinding is warranted, it should be done by the district court and not by this panel.
Accordingly, I concur in that portion of the majority opinion concluding that additional factfinding is warranted. However, I would remand the case to the district court for that additional factfinding rather than direct a verdict for Plaintiffs as the majority has done. I also agree that there is no need to assign the case to a different judge on remand. For the reasons discussed, I respectfully dissent from the balance of the majority opinion.
. I also disagree profoundly with the majority’s discussion of foreseeability. However, I will not unnecessarily lengthen my dissent by detailing my different view on this point because the majority’s wholesale disregard of the district court's factual findings (and lack of findings where applicable) sufficiently call into question the analysis employed by the majority.