Opinion for the Court filed by Circuit Judge STEPHEN F. WILLIAMS.
STEPHEN F. WILLIAMS, Circuit Judge:The Assistant Secretary of Labor has exercised his authority under § 101(c) of the Federal Mine Safety and Health Act, 30 U.S.C. § 811(c) (1988), to exempt Cyprus Emerald Resources Corporation’s “Emerald No. 1” mine from a regulation governing the flow of air through underground coal mines. The order permits Emerald to ventilate the working face of the mine with air from the conveyor-belt area on the condition that it install carbon monoxide (“CO”) detectors throughout the mine. The United Mine Workers of America (“the Union”) challenges the order on a variety *168of grounds. On review, we leave the order in place but remand the case to the Assistant Secretary for more reasoned decision-making.
I
Underground coal mines generally have three or more “entries” leading from the surface. One of these is the “belt entry,” through which coal is transported by a conveyor belt from the mine to the surface. The belt is one of the chief sources of friction, dust, and flame in any mine. Thus the practice has long been to seal, the belt entry off from the working “face” of the mine, the place where the miners extract the coal.
When Congress enacted the Federal Mine Safety and Health Act of 1977, Pub.L. No. 95-164, codified in scattered sections of 30 U.S.C. §§ 801 et seq. (1988), it adopted interim standards based on the then prevailing learning, including one requiring separation of belt entry air:
In any coal mine opened after the operative date of this subchapter, the entries used as intake and return aircourses shall be separated from belt haulage entries, and each operator of such mine shall limit the velocity of the air coursed through belt haulage entries to the amount necessary to provide an adequate supply of oxygen in' such entries, and to insure that the air therein shall contain less than 1.0 volume per centum of methane, and such air shall not be used to ventilate active working places____
§ 303(y)(1), 30 U.S.C. § 863(y)(1). The requirement is restated in the regulations of the Mine Safety and Health Administration at 30 CFR § 75.326, and the parties generally refer to it by that designation.
At the same time, Congress authorized the Secretary of Labor both to replace any of the interim standards by rulemaking, § 101(a) of the Act, 30 U.S.C. § 811(a), and to exempt particular mines from specific standards, § 101(c), 30 U.S.C. § 811(c), as in this case. The Secretary exercises this power through the Mine Safety and Health Administration (“MSHA”), which is headed by the Assistant Secretary of Labor for Mine Safety and Health. See 29 U.S.C. § 557a.
Since its opening in 1977 Emerald’s No. 1 mine has satisfied § 75.326, in part by means of a canvas barrier between the belt entry and the face. Joint Appendix (“J.A.”) 5, 496-97. Claiming that the barrier had negative effects on ventilation and thus on safety, see, e.g., J.A. 44, Emerald concluded that it could improve the air flow by removing the barrier and ventilating the working face of the mine with belt air. To compensate for the associated loss in physical protection from the hazards of the belt area, Emerald proposed to install a sophisticated system of CO monitors that would enable miners to escape a developing fire more quickly than if the only warning came from conventional heat sensors. In 1982 it petitioned MSHA for a § 101(c) exemption of the mine from § 75.326.
In 1983 MSHA issued a favorable proposed decision on the petition, subject to a few stipulations. At the behest of the Union a Department of Labor administrative law judge reversed MSHA and denied Emerald’s petition. The Assistant Secretary in turn reversed the AD. The Union appealed to this court, and in International Union v. MSHA, 830 F.2d 289 (D.C.Cir.1987), we reversed the Assistant Secretary’s decision and remanded the case for him to explain, among other things, his interpretation of § 101(c)’s standard and his application of that standard in this case. On remand, the Assistant Secretary again granted Emerald’s petition.
II
Section 101(c) allows the Secretary to grant a petition to modify the application of a mandatory standard to a mine if she
determines that an alternative method of achieving the result of such standard exists which will at all times guarantee no less than the same measure of protection afforded the miners of such mine by such standard____
§ 101(c) of the Act, 30 U.S.C. § 811(c) (emphasis added).
There is quite a range of imaginable interpretations. At one end is a view es*169poused by the Union (if somewhat erratically) that the “result” is essentially a specific physical one, so that for § 75.326 the alternative method must achieve “a physical defense from the rapid propagation of fires.” 1 On this view few if any petitions could ever be granted, and changes with a high probability of increasing miners’ safety might be ruled out automatically. At the other end is a view that the “result” of a standard is equivalent to the total “measure of protection” that it affords, such that the success of a petition would depend entirely on a global calculation of its net effects on the mining environment.
The Assistant Secretary has adopted a “two-step” process that appears to fall somewhere between these poles. The first step addresses “the hazards associated with belt entry fires,” J.A. 855, and the second involves consideration of all safety consequences, id. at 856. At one point he speaks of the first step as one the mine must satisfy to meet the statutory burden and the second step as “discretionary.” Id. But at the same time the Assistant Secretary states firmly that determining whether a proposal provides “the same measure of protection” as the interim standard, as the statute insists, “requires an evaluation of all safety benefits resulting from the standard and all the safety benefits resulting from the alternate method.” Id.; see also id. at 857-58. He supported this approach by stressing the interrelationship of the various standards with one another. Thus, he wrote, any alternative to the global view would be contrary to “the reality inherent in mining,” namely, that
the mandatory standards which regulate the industry are interrelated and together provide protection from various types .of mine hazards. Mandatory standards therefore cannot be examined individually through a microscope____ The true comparison in this case is between the work environment created by compliance with § 75.326, versus the environment resulting from the use of belt air and the installation of a CO monitoring system.
J.A. 857.
The first step appears aimed at meeting § 101(c)’s requirement that the alternative “achievfe] the result” of the original standard in the sense of addressing the hazards at which it was aimed, the second at assuring “the same measure of protection.” The “result” clause is less stringent than the requirement of the “same measure of protection,” and thus is reasonably read as requiring that the modification promote the specific safety goals of the original standard (here, protection from the effects of belt entry fires) with roughly comparable success. Id. In step two, the Assistant Secretary then determines whether the modification achieves a net gain in mine safety (or at least equivalence), taking all effects into account.
As we must defer to an agency’s reasonable interpretation of an ambiguous statute that it must administer, see Chevron U.S.A. v. Natural Resources Defense Council, 467 U.S. 837, 842-44, 104 S.Ct. 2778, 2781-83, 81 L.Ed.2d 694 (1984), we need not pass on whether there are valid alternative readings of § 101(c). The section is ambiguous at least in the sense of not clearly precluding the Assistant Secretary’s approach. Thus the only question before us is whether his interpretation of § 101(c) is reasonable. We find that it is.
The Union insists that the interim standards collectively sought to create a kind of “double shield”: physical separation, which shields the miners from flame and smoke (§ 75.326), and fire detection, which provides them timé to escape (§ 75.1103-4(b)). (As Judge Bork put it in oral argument of the 1987 case, it gives the equivalent of a belt and suspenders.) As the Union sees it, Emerald’s petition merely replaces one form of fire detection for another (CO detectors for the prescribed heat sensors) and *170rescinds the other method, physical separation.
The obvious trouble with the Union's position is that under it any petition that proposes ventilating the face with belt air must be rejected even if it demonstrably improves net mine safety — even if the enhancement in fire detection clearly offsets the losses from removal of the shield. Granted, Congress was concerned about specific mine safety problems, and the parties agree that particular risks — those of belt fires — were the focus of § 75.326. But it also recognized that its expertise was limited and subject to obsolescence; thus the delegation of power to exempt mines from the interim standards where appropriate. Congress could not have intended for its baseline standards to become methodological fetishes that frustrate MSHA’s efforts to improve mine safety. We cannot find the Assistant Secretary’s reading unreasonable.
The Union further argues that even if MSHA could rescind § 75.326’s physical barrier requirement, it must do so through rulemaking under § 101(a) rather than case-by-case adjudication under § 101(c). But the courts have always accorded agencies broad discretion in choosing between rulemaking and adjudication as a means of addressing issues, see NLRB v. Bell Aerospace Co., 416 U.S. 267, 94 S.Ct. 1757, 40 L.Ed.2d 134 (1974); SEC v. Chenery Corp., 332 U.S. 194, 201-03, 67 S.Ct. 1575, 1579-81, 91 L.Ed. 1995 (1947); Wisconsin Gas Co. v. FERC, 770 F.2d 1144, 1166 (D.C.Cir.1985), and we do so here. The Union suggests that this mine is indistinguishable from any other that could be the subject of a similar § 101(c) petition. Even if this were true, it would read quite a gloss on the statute to require the Assistant Secretary to deny a § 101(c) petition whenever petitioner failed to prove the unique character of the mine. As a change might be suitable for various proportions of the country’s mines (e.g., one mine, 10%, 90%), the Union’s position would pose a whole set of issues whose litigation would little advance the cause of mine safety. In any event, here the Assistant Secretary found that the proposed change presented the most effective solution to particular ventilation problems present at this specific mine, though not necessarily unique to it. See J.A. 871; see also id. at 823-24 (noting that diesel equipment, which Emerald No. 1 apparently does not employ, arguably undercuts the effectiveness of CO monitors). We detect no basis for seeing the use of § 101(c) as an attempted end run around the statutory provision for rulemakings.
Ill
We must now consider whether the Assistant Secretary correctly applied his own standard. Here the key issues are whether he has identified a reasonable basis in the record for concluding that Emerald’s proposal achieves the general results of § 75.326 and produces a net gain in overall mine safety, and whether he has responded reasonably to serious attacks on the proposal. See generally Greater Boston Television Corp. v. FCC, 444 F.2d 841, 850-52 (D.C.Cir.1970). Because we find no explanation at all for his removal of any limit on permissible belt entry air velocities, and no adequate response to the Union’s evidence as to false alarms and non-CO smoke, we remand the case to him for additional consideration of those issues.
First, we note that the Assistant Secretary did articulate a reasonable basis for several of his principal findings. There was, for instance, substantial evidence in the record for the conclusion that, leaving the question of false alarms to one side, CO detectors are generally reliable and demonstrably more effective than heat point sensors. See J.A. 170ff., 482, 499-500, 522, 661-62. Moreover, there was sufficient evidence to sustain the conclusions that CO monitors could effectively cope with air velocities of at least 300 feet per minute (fpm), see, e.g., J.A. 490, 505, 607, 640-44, that there was no viable alternative to solving the ventilation problem at Emerald No. 1, see, e.g., J.A. 585-94, and that granting the petition would increase the dust levels at the mine face by no more than a small percentage of the maximum allowable concentration, J.A. 204-06, 491-92.
*171Further, we reject the Union’s claim that the entire case should be reconsidered in light of Emerald’s recent decision to move from four to three active entries. Not only does the Union not suggest a single specific way in which this change would affect the relevant conditions at the mine, but the conjectures that it offers on the subject are mutually contradictory, with a claim at one moment that the entry reduction “may actually aggravate the alleged track air problem,” Union Brief at 45, and at another that it will completely cure it, Union Reply Brief at 11-12 (“there is no evidence that the air pressure imbalances ... persist”).
We find, however, that the Assistant Secretary did not adequately resolve three issues necessary for fulfillment of his duty of reasoned decisionmaking.
1. Unlimited Velocities.
Apparently on his own initiative, the Assistant Secretary removed all limits on air velocity within the mine. J.A. 868 n. 15. For his twin claims that unlimited velocities would neither impair the effectiveness of the CO monitors nor increase flame propagation, he pointed to three bits of evidence. As to impaired effectiveness, he “rel[ied] upon the testimony of [MSHA witnesses] Turcic and Miller establishing the increased effectiveness of carbon monoxide systems at higher air velocities.” Id. But Turcic and Miller testified about the effectiveness of CO monitors only up to 300 fpm. See J.A. 487, 640, 643-44. In fact, the study that they conducted together, which formed the basis of their expertise on the matter, addressed velocities only up to 300 fpm. J.A. 170. The questions that they answered about the effect of granting the petition presupposed that any approval would be subject to a 300 fpm limit, as MSHA itself had proposed. See, e.g., J.A. 594.
The Assistant Secretary also cited as support “Assistant Secretary Exhibit No. 2,” a report by C.D. Litton of the Interior Department’s Bureau of Mines. This was a document that the Assistant Secretary on his own hook added to the record after the hearing closed, along with Assistant Secretary No. 1. See J.A. 765-67 (order receiving the exhibits); 876-77 (invoking No. 2 in his handling of the effect on detection); 787-807 (the exhibits themselves). It is not self-evident on the face of the Litton report that it shows equal detection capability at velocities in excess of 300 fpm. The Union demanded that if the Assistant Secretary was to rely on the exhibits, he reopen the hearing to allow them an opportunity for cross-examination and rebuttal testimony. He denied this, on the grounds that the exhibits were not critical to the grant of the petition and that the Union had not shown that witness credibility was at stake. But the first response was not substantively true, for the exhibits are the sole record basis for concluding that CO monitors function as well above 300 fpm as below, and the second point hardly exhausts the need for cross-examination or rebuttal.
If on remand the Assistant Secretary should rely on the Litton report, he must explain how it represents substantial evidence for the proposition that CO monitors provide the same detection capacity at speeds above 300 fpm as between 50 and 300 fpm, the speeds used in the Turcic & Miller study. Similarly, he must explain why he has not permitted the Union to confront this rather complex study in a supplemental oral hearing.
For the related finding that unlimited velocities would not cause greater flame propagation or lessen escape time, the Assistant Secretary relied not only on the Litton report but also on another post-hearing exhibit, Assistant Secretary No. 1, which is a letter to MSHA from an official at Interior’s Bureau of Mines. See J.A. 876 (the Assistant Secretary’s analysis); 761-66 (text of the letter). Again, the reasons the Assistant Secretary offered for denying the Union a chance either to cross-examine the author or to offer rebuttal were unresponsive. Accordingly we remand for him either to address the procedural and substantive issues raised, or to rescind his removal of the 300 fpm limit.2
*1722. Non-CO Smoke.
Belt friction can produce smoke that does not contain carbon monoxide and would therefore not set off CO monitors. Such smoke is demonstrably hazardous to miners. See J.A. 507, 680, 878. The Assistant Secretary dismissed the threat of non-CO. smoke on several grounds, none of which appears persuasive. The Litton report, on which the Assistant Secretary relied for other purposes, appears to suggest that the risk from such smoke is great enough to justify insistence on use of additional sensors not relying on CO concentrations. See J.A. 797-98.
First he asserted that non-CO smoke was generally not associated with flames. J.A. 878-79. This observation appears irrelevant under the Assistant Secretary’s basic decision, under the “same measure of protection” clause, to consider safety effects not related to the precise hazard at which a standard is aimed. Second, he argued that heat point sensors would not have picked up non-CO smoke any better than CO detectors. Id. This too is unresponsive. In the absence of a grant of Emerald’s petition, heat sensors would be used in conjunction with a physical barrier that would probably keep at least some of the dangerous fumes from reaching the face. Granting the petition removes that protection without adding a substitute. Third, the Assistant Secretary contended that unless the petition were granted, ventilation problems created “the possibility ... that rather than being coursed directly into the return, much of the smoke would end up at the face or in the intake escapeway.” J.A. 879. The fact that such exposure is a possibility without the petition hardly justifies turning it into a certainty by granting the petition. (It does, however, suggest that the difference between the two solutions is not as great as might otherwise appear.)
Finally, the Assistant Secretary noted that various technologies in use at the belt can “reduce the likelihood” of non-CO smoke. Id. Here, at least, the Assistant Secretary is on the right track, focussing on the net safety effect. It is possible that under current conditions the risk that non-CO smoke will harm miners is very small— and worth taking in light of the safety benefits of granting the petition, especially if additional controls may reduce the non-CO smoke risk. Yet the Assistant Secretary never actually made this argument. We therefore remand to him so that he may evaluate this risk, explain how it affects the safety analysis involved in granting the petition, and address any claims that he should impose further conditions.
3. False Alarms.
Finally, the Union submitted evidence that CO monitors suffer from frequent false alarms and, in effect, cry wolf to the miners. See J.A. 556ff., 625, 808ff., 823ff. For this reason, the Union argues, CO monitors alone cannot provide enough protection for the workers at the face.
The Assistant Secretary ignored this problem altogether. He confined his treatment of the reliability of CO detectors to an analysis of false negatives, without discussion of false positives. On remand, therefore, he should assess the significance of the issue as part of his analysis.
IV
The Assistant Secretary has granted Emerald’s petition and Emerald has implemented its terms. We have found that his disposition fails to give an adequate treatment of certain issues, i.e., shows a want of reasoned decisionmaking, with some possibility that substantial evidence may be missing on some points.
We have commonly remanded without vacating an agency’s rule or order where the failure lay in lack of reasoned decision-making, see generally United Mine Workers of America v. Dole, 870 F.2d 662, 673-74 (D.C.Cir.1989), but also where the order was otherwise arbitrary and capricious, *173see, e.g., Maryland People’s Counsel v. FERC, 768 F.2d 450, 455 (D.C.Cir.1985). Relevant to the choice are the seriousness of the order’s deficiencies (and thus the extent of doubt whether the agency chose correctly) and the disruptive consequences of an interim change that may itself be changed. Cf. Washington Metropolitan Area Transit Comm’n v. Holiday Tours, Inc., 559 F.2d 841, 844 (D.C.Cir.1977) (describing analogous factors to be considered in deciding whether to grant preliminary injunction); American Hospital Supply Corp. v. Hospital Products Ltd., 780 F.2d 589, 593-94 (7th Cir.1986).
The record before us does not appear to speak to the effects of an interim change— except in the sense that those effects include the safety effects of the order itself (and thus its substantive validity). As the record affords us no basis for concluding that the deficiencies of the order will prove substantively fatal, we remand the case but do not vacate. The court would reconsider vacating the order, however, upon a showing that the Assistant Secretary was not proceeding with reasonable diligence on remand.
So ordered.
. Asserted in support of the ALJ’s decision on review to the Assistant Secretary. See J.A. 128, 130-31; see also Union Brief at 35. Before the AU the Union espoused another theory, under which satisfaction of § 101(c) would depend entirely, or almost entirely, on how well the proposal would address the risks for which the standard was designed, though not necessarily requiring that the proposal mimic the standard’s physical means. See J.A. 75; see also Reply Brief at 6-7.
. Emerald has drawn our attention to a recent • administrative decision, Clinchfield Coal Co., Case Nos. 89-MSA-2, 89-MSA-3 (ALJ Von Brand, August 24, 1990), appeal to the Assistant *172Secretary pending, that on remand may, together with its underlying record, support adherence to the limit removal.