dissenting.
I agree that preemption initially occurred at the time the federal government significantly participated in the installation of refleetorized crossbucks.1 However, I diverge *1193from the majority’s conclusion reached in the last lengthy paragraph of the opinion that preemption continued past 1984, when the diagnostic team decided automatic gates were needed at the North Gabaldon crossing. Instead, I agree with Armijo that the decision to install automatic gates removed the preemptive effect of the crossbucks’ installation and, therefore, Armijo’s claims were not preempted at the time of her husband’s 1987 accident. I therefore respectfully dissent.
My disagreement with the majority concerns whether preemption was in place at the time of the accident in 1987. In CSX Transp., Inc. v. Easterwood, 507 U.S. 658, 113 S.Ct. 1732, 123 L.Ed.2d 387 (1993), the Supreme Court held that preemption applies when the preconditions for the application of either section (b)(3) or section (b)(4) have been met. 507 U.S. at 670-72, 113 S.Ct. at 1740-42. Looking at these regulations as applied to the North Gabaldon crossing at the time of the accident, the preconditions for section (b)(4) no longer were met in 1987. The regulation addressing section (b)(4) provides:
For crossings where the requirements of § 6It.6.21Jp(b)(3) are not applicable, the type of warning device to be installed, whether the determination is made by a State regulatory agency, State highway agency, and/or the railroad, is subject to the approval of FHWA.
23 C.F.R. § 646.214(b)(4) (emphasis added). Thus, a railroad may invoke preemption under section (b)(4) only so long as “the requirements of § 646.214(b)(3) are not applicable” to the particular crossing in question. One requirement of section (b)(3) is that a “diagnostic team recommends” automatic gates. See 23 C.F.R. § 646.214(b)(3)(I)(F). It is undisputed that in 1984 a diagnostic team recommended that automatic gates be installed at the North Gabaldon crossing. Therefore, although section (b)(4) at one time operated to preempt state claims because of the government’s significant participation in the installation of reflectorized crossbueks, this preemption ceased in 1984 once a diagnostic team recommended automatic gates.
The majority concludes that the diagnostic team’s 1984 recommendation of automatic gates did not remove the crossbucks’ preemptive effect because the Secretary of Transportation did not affirmatively abandon the crossbueks project and either withdraw federal funding or allow previously allocated funding to be used at another site. No case law requires such an affirmative withdrawal, and I believe that such a requirement is in conflict with the plain language of sections (b)(3) and (b)(4). Rather, I believe that the regulations recognize that as conditions surrounding a particular crossing change and make the crossing more dangerous to motorists, a determination that passive warning devices are sufficient to protect motorists in one year should not continue to preempt state claims when it is determined the passive devices provide insufficient warning in a later year. Also, considering the “presumption against preemption” applied by the Supreme Court in Easterwood, 507 U.S. at 667-68, 113 S.Ct. at 1739-40, I believe that we must conclude that neither section (b)(3) nor section (b)(4) operated to preempt Armijo’s claims at the time of the accident.2 Accord*1194ingly, I would reverse the decision of the district court and remand this action for further proceedings.
. Although I agree with the majority that preemption occurred as a result of the federal government’s participation in the installation of reflectorized crossbucks, I disagree with some of the language in the majority opinion concerning the time when such preemption first arose. The majority opinion concludes that preemption occurred, at the latest, on January 25, 1983, the date on which the Secretary of Transportation agreed to provide ninety percent of the funds required to install reflectorized crossbucks. However, we held in Hatfield v. Burlington N. R.R. Co., 64 F.3d 559, 562 (10th Cir.1995) ("Hatfield II ”), that preemption takes place when the federal government: (1) commits itself through a significant event or events, to a project to install active warning devices; and (2) expends significant federal resources on the project. Therefore, *1193it is insufficient under the second prong of the Hatfield II test merely to authorize that federal funds be spent on upgrading a crossing; instead Hatfield II requires “the actual expenditure of federal resources of more than a casual or de minimis nature” before preemption occurs. Id. (emphasis added.) See United States v. Zapata, 997 F.2d 751, 759 n. 6 (10th Cir.1993) (noting that a three-judge panel of the court cannot overturn an earlier opinion from the same court). Nevertheless, that date is not critical in this case because here it is clear that the federal government had reimbursed the state for the cost of engineering or installing the reflectorized cross-bucks. Therefore, prior to the time of this accident the federal government's participation was significant enough to trigger preemption under 23 C.F.R § 646.214(b)(4).
. I do not believe that preemption occurred under section (b)(3) at the time of the accident because in 1987, the federal government had not committed itself significantly to the installation of automatic gates at the North Gabaldon crossing. Instead, federal participation was limited to a single federal employee participating on a diagnostic team which selected 22 crossings requiring warning signal upgrades during the next five years. In contrast with Hatfield II, no engineering study had been performed at the crossing until after the accident occurred. Also, the second element of the Hatfield II test was not met at *1194the time of the accident. The federal resources expended here were limited to the sole federal employee participating on the diagnostic team. Although Santa Fe estimates that the federal government spent $1,250 on the engineer's salaiy while he served on the diagnostic team, this amount must be apportioned among the 22 crossings he considered in detail, as well as more than 100 crossings he considered on an original list of crossings.