concurring in part and dissenting in part:
I agree with the majority’s conclusion that “the computer modeling used by the EPA to calculate the safety of drinking water was neither contrary to law nor arbitrary and capricious.” Maj. Op. at 1050; see also The Lands Council v. McNair, 537 F.3d 981, 989-94 (9th Cir.2008) (en banc). I also agree with the majority’s holding that National Ass’n of Home Builders v. Defenders of Wildlife, — U.S. -, 127 S.Ct. 2518, 2529-30, 168 L.Ed.2d 467 (2007), forecloses petitioners’ argument “that the EPA acted arbitrarily and capriciously by establishing the pesticide tolerances for acetamiprid, mepiquat, and pymetrozine before receiving ... DNT studies.” Maj. Op. at 1051.
I disagree, however, with the majority’s conclusion that the tolerance regulations for acetamiprid, mepiquat, and pymetro-zine must be remanded to the EPA because “it is entirely unclear why the EPA chose safety factors of 3x for pymetrozine and acetamiprid, and lx for mepiquat,” and therefore “the Final Order [did] not provide enough information to demonstrate a rational connection between the factors the EPA examined and the eonclu-*1054sions it reached.” Maj. Op. at 1052. The petitioners failed to raise this argument in their objections to the EPA prior to the issuance of the Final Rule. See Order Denying Objections to Issuances of Tolerances, 70 Fed.Reg. 46,706 (Aug. 10, 2005) [hereinafter, Final Rule], and therefore the EPA was given no opportunity to explain its rationale for adopting the 3x and lx safety factors. Because we may review only the objections raised before the EPA, see 21 U.S.C. § 346a(g)-(h); 40 C.F.R. §§ 178.25(a)(2), 180.30(b), the issue is not properly before us. Even if the petitioners had raised this objection, there is no doubt the EPA could have provided information sufficient to demonstrate a rational basis for selecting these safety factors. Indeed, the record establishes that the EPA drew upon its expertise to select uncertainty factors based on principles well established in the scientific community. There is no basis for the majority’s conclusion that the agency’s path is not reasonably discernable in this case. See, e.g., Motor Vehicle Mfrs. Ass’n v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 43, 103 S.Ct. 2856, 77 L.Ed.2d 443 (1983). Accordingly, I dissent from Part IIB of the majority opinion.1
A
As noted by the majority, the tolerance rules at issue in this appeal have substantially similar histories.2 While the development of its safety factor policy was in process, the EPA received pesticide petitions requesting the issuance of tolerances under the FFDCA for the three pesticides at issue. As required in the FFDCA, the EPA reviewed the scientific data and other information, evaluated their completeness and reliability, and considered “the relationship of the results of such studies to human risk.” 21 U.S.C. § 346a(b)(2)(D)(iii); see, e.g., Pymetrozine; Pesticide Tolerance, 66 Fed.Reg. 66,786, 66,787 (Dec. 27, 2001). The EPA determined that the child safety factor was not necessary, or could be reduced to 3x, for each of the pesticides based on its analysis of potential exposure levels for the given pesticide, the extent to which the pesticide affected fetal development, and the effects, if any, on infants as observed in multi-generational studies of mammals. In each case, the EPA provided a reasoned explanation for eliminating or reducing the child safety factor.3
*1055In 2002, the NRDC filed objections to the final rules for each of the pesticides based on the EPA’s decision to reduce the lOx safety factor to 3x or lx. The NRDC’s objections were based on the theory that the EPA did not have “reliable data” which would allow it to reduce the child safety factor because the EPA had not yet received the results of a DNT study for each pesticide, and because the EPA relied on drinking water exposure models rather than on pesticide-specific drinking water monitoring data.4 Notably, the NRDC did not object on the ground that the EPA failed to justify its choice of a safety factor of 3x or lx relative to some other safety factor, such as “4x or 5x or 8x or 9x.” Maj. Op. at 1052.5 The EPA *1056rejected NRDC’s objections in the Final Rule, and provided a detailed explanation of its basis for concluding that it had reliable data notwithstanding the absence of certain DNT studies and the use of exposure models, and why such reliable data supported the EPA’s conclusion that it could safely reduce the child safety factor. See generally 70 Fed.Reg. 46,706. The Final Rule upheld the 3x child safety factor for acetamiprid and pymetrozine, and the lx child safety factor for mepiquat. Id. at 46,711-13, 46,736.
B
In concluding that remand is required because the EPA failed to explain why it chose a 3x uncertainty factor for acetami-prid and pymetrozine and a lx uncertainty factor for mepiquat instead of “4x or 5x or 8x or 9x,” Maj. Op. at 1052, the majority is reaching an argument not properly before us. The regulations applicable to the FFDCA require exhaustion of administrative remedies. See 40 C.F.R. § 180.30(b) (providing that “judicial review is not available unless an adversely affected party exhausts the[] objection procedures, and any petition procedures preliminary thereto”). In order to exhaust the objection procedure provided by the FFDCA, petitioners must raise their objections to the agency with specificity. See 21 U.S.C. § 346a(g)(2)(A) (requiring objections to an EPA order establishing a tolerance to specify “with particularity the provisions of the regulation or order deemed objectionable”); 40 C.F.R. § 178.25(a)(2) (“To be considered by the Administrator, an objection must: Specify with particularity the provision(s) of the order, regulation, or denial objected to, the basis for the objeetion(s), and the relief sought.”). The purpose of this administrative exhaustion requirement is to give notice to an agency of specific errors or issues, so the agency can “bring the Agency’s experience to bear on a contested question,” Nader v. EPA, 859 F.2d 747, 754 (9th Cir.1988), and either correct the error or provide a reasoned explanation.
The NRDC objected to the EPA’s decision that a lOx safety factor was not required for each of the three pesticide tolerances, and based this objection on the ground that the EPA lacked reliable data because it had failed to obtain a DNT study and had relied on drinking water exposure models. The majority rests its decision that the tolerances for acetami-prid, mepiquat, and pymetrozine must be remanded to the EPA on a different ground, namely that the EPA failed to explain why it chose a 3x uncertainty factor for acetamiprid and pymetrozine and a lx uncertainty factor for mepiquat instead of “4x or 5x or 8x or 9x,” Maj. Op. at 1052. Because the NRDC did not object to the EPA’s decision on the ground that the EPA had failed to give a clear explanation of the 3x or lx safety factor in lieu of “4x or 5x or 8x or 9x,” id., the issue identified by the majority is not exhausted, and we are precluded from reviewing it.
The majority attempts to justify addressing issues that the NRDC did not raise to the EPA by turning the administrative exhaustion requirement on its head. As noted above, under the regulations requiring exhaustion of administrative remedies, parties objecting to proposed regulations bear the initial burden, and must raise their specific objections to the agency; any issues not raised are waived and need not be considered by the agency or by the court on appeal. 40 C.F.R. §§ 178.25(a)(2), 180.30. Here, the majority shifts the burden to the EPA to identi*1057fy, raise, and resolve additional issues that petitioners failed to raise at all, let alone frame with the requisite “particularity.” Id. § 178.25(a)(2).
The majority first notes that “the NRDC argued that, absent the DNT studies, the EPA had no reliable data to reduce the lOx child safety factor.” Maj. Op. at 1053. This is indeed the specific objection raised by the NRDC. The majority then recharacterizes this specific objection as being a more general “challenge to the lower child safety levels set by the EPA.” Maj. Op. at 1053.6 From this generalization, the majority leaps to the conclusion that the NRDC’s specific objection places a burden on the EPA to “explain why the data in its possession” supported its decision to set a 3x or lx child safety factor for the pesticides at issue “as opposed to 4x or 5x or 8x or 9x.” Maj. Op. at 1052, 1053. There is no basis in our case law for taking such a leap or placing such a burden on the EPA. We have never required an agency to interpret a specific objection as being, at some level of generality, a different, broader objection, and responding to that different objection. See, e.g., High Country Resources v. FERC, 255 F.3d 741, 745-46 (9th Cir.2001) (stating that a petitioner cannot “implicitly” raise an issue to a regulatory agency in order to exhaust administrative remedies, but must provide “much more specificity in the statement of objection” and present the issue “squarely” to the agency.). Nor would it be reasonable or consistent with the purpose of the administrative exhaustion requirement to do so. Here, the EPA responded to the NRDC’s specific objection by carefully explaining why the absence of DNT studies did not require the retention of the lOx safety factor. The EPA satisfied its burden, and had no additional obligation to make a further response.
In short, we have no authority to reverse an agency decision based on an objection that was not raised. The Supreme Court has long warned us not to invent new issues as part of our review of objections raised to an administrative agency. “A reviewing court usurps the agency’s function when it sets aside the administrative determination upon a ground not theretofore presented and deprives the [administrative agency] of an opportunity to consider the matter, make its ruling, and state the reasons for its action.” Unemployment Comp. Com’n v. Aragon, 329 U.S. 143, 155, 67 S.Ct. 245, 91 L.Ed. 136 (1946); see also Woodford v. Ngo, 548 U.S. 81, 89, 126 S.Ct. 2378, 165 L.Ed.2d 368 (2006). The EPA addressed, with adequate thoroughness and detail, see, e.g., Balt. Gas and Elec. Co., 462 U.S. at 105-06, 103 S.Ct. 2246, each of the objections filed by the NRDC. These objections create the compass of our limited review. 21 U.S.C. § 346a(h)(l), (5); 40 C.F.R. § 178.25(a)(2).7
*1058C
Even were it proper for us to address the adequacy of the EPA’s consideration of an issue not raised, the record provides an adequate basis for us to conclude that the EPA’s selection of a 3x and lx uncertainty *1059factor was not arbitrary, as the majority erroneously concludes. Maj. Op. at 1052.
The record in this case contains the EPA’s published guidance for determination of an appropriate FQPA safety factor. See Office of Pesticide Programs, EPA, Determination of the Appropriate FQPA Safety Factor(s) in Tolerance Assessment 10 (Feb. 28, 2002). As explained in this guidance document, “when evaluating whether reliable data are available to support a FQPA safety factor different from the default lOx FQPA safety factor and what level of ‘different’ FQPA safety factor would be safe for infants and children,” id. at 19, the EPA may, in the exercise of its scientific judgment, consider the 3x uncertainty factor that has traditionally been used for evaluating risks posed by pesticide exposure in a related context. See id. at 26; see also id. at 9-10 (noting that in evaluating the risks posed by pesticide exposure, “often a value of 3x is used to address database deficiencies.... ”). However, in “the optimal case” in “which there is a high level of confidence that the hazard and exposure assessments are sufficiently conservative!,] and there are no residual uncertainties in the assessment,” the EPA will remove the safety factor, i.e., it will apply a “safety factor of IX,” deeming the existing threshold “sufficient to protect infants and children.” Id. at 51-52. See also Office of Pesticide Programs, EPA, Office of Pesticide Programs’ Policy on The Determination of The Appropriate FQPA Safety Factor(s) For Use in The Tolerance-setting Process: Response to Public Comments OPP Docket OPP-00610 97 (Feb. 28, 2002) (internal citation omitted) (further explaining the long-standing basis for “imposing a database uncertainty factor of 3X if one key study is missing from the database and a factor of 10X if more than one is missing.”).
Other EPA tolerance-setting regulations reflect the approach described in its guidance documents for selecting a FQPA safety factor different from the default lOx FQPA safety factor. See, e.g., Fenari-mol; Pesticide Tolerance, 71 Fed.Reg. 32,-841, 32,844-45 (June 7, 2006) (noting that the EPA originally reduced the FQPA Safety factor from lOx to 3x subject to receiving additional data, and subsequently recommended reducing the 3x FQPA Safety factor to lx when it obtained adequate data); Thiacloprid; Pesticide Tolerances, 68 Fed.Reg. 55,503, 55,509 (Sept. 26, 2003) (“Although the lack of morpho-metric assessments in the DNT raised some uncertainty, EPA determined that there were sufficient reliable data to select an additional safety factor of 3X instead of 10X.”). The EPA relied on this long-established approach in setting the safety factors for the three pesticides at issue here, see supra at nn. 2 & 3.
Because the petitioners did not challenge the absence of an explanation as to why the EPA picked a 3x or lx factor rather than a “4x or 5x or 8x or 9x” factor, we are deprived of the EPA’s more detailed explanation as to why the scientific community relies on the 3x uncertainty factor when database deficiencies are identified. Nevertheless, the record indicates that use of the 3x uncertainty factor is generally accepted by the scientific community. See Michael L. Dourson et al., Evolution of Science-Based Uncertainty Factors in Noncancer Risk Assessment, 24 Regulatory Toxicology and Pharmacology 108 (1996) (generally discussing the science underlying the use of lOx and 3x default uncertainty factors). Indeed, the lack of any controversy regarding the rationale underlying the selection of a 3x safety factor rather than a “4x or 5x or 8x or 9x” factor may explain why the NRDC did not argue in its objections to the EPA that the agency needed a special justification for selecting a 3x safety factor, once the EPA determined that the additional *1060lOx factor was not required. In light of the EPA’s reliance on a long-established and widely accepted protocol, the majority’s statement that “the EPA chose these lower safety levels arbitrarily” is not supported by the record, and is contrary to our obligation to defer to the scientific analysis and judgments made by an agency operating within its area of special expertise. Balt. Gas and Elec. Co., 462 U.S. at 103, 103 S.Ct. 2246.
As we have recently emphasized, “[w]e are to be most deferential when the agency is making predictions, within its area of special expertise, at the frontiers of science.” Lands Council, 537 F.3d at 993 (internal quotation marks and alterations omitted). Lands Council prohibits us from indulging the temptation to “act as a panel of scientists.” Id. at 988. Nor can we require the agency “to always demonstrate the reliability of its scientific methodology or the hypotheses underlying the [agency’s] methodology.” Id. at 990 (internal quotation marks omitted). Lands Council teaches that our proper role is simply to ensure that the agency, in its expertise, made no clear error of judgment rendering its action arbitrary and capricious. See id. at 992-94. The EPA’s choices in this case were neither arbitrary nor capricious and we must defer to its judgment. See 5 U.S.C. § 706(2)(A); Balt. Gas and Elec. Co., 462 U.S. at 103-05, 103 S.Ct. 2246. Accordingly, I would deny the petition for review in its entirety.
. I also dissent from footnote 6 of the majority opinion. The majority’s suggestion that the "EPA may wish to revisit[its] methodology” regarding its "model-based water exposure testfs]” conflicts directly with the teaching of our en banc panel in Lands Council. Lands Council emphasized that it "is not a proper role for a federal appellate court” to "act as a panel of scientists that instructs the [agency] how to validate its hypotheses,” "chooses among scientific studies,” and requires the agency to "explain every possible scientific uncertainty.” 537 F.3d at 988. The majority errs in this case by doing that which Lands Council expressly forbids.
. The three pesticides at issue are acetami-prid, mepiquat, and pymetrozine. For the individual tolerance rulings, see Acetamiprid; Pesticide Tolerance, 67 Fed.Reg. 14,649 (Mar. 27, 2002); Mepiquat; Pesticide Tolerance, 67 Fed.Reg. 3,113 (Jan. 23, 2002); Pymetrozine; Pesticide Tolerance, 66 Fed.Reg. 66,786 (Dec. 27, 2001).
.In its order establishing the tolerance for acetamiprid, the EPA explained:
[T]he safety factor could be reduced to 3x for acetamiprid because the toxicology database is complete; there is no quantitative or qualitative evidence of increased susceptibility following in útero exposure of rat and rabbit fetuses; the dietary (food and water) and residential exposure assessments will not underestimate the potential exposures for infants, children, and/or women of childbearing age; and the requirement of a developmental neurotoxicity study is not based on criteria reflecting special concern for the developing fetuses *1055or young which are generally used for requiring a DNT study and a safety factor.
Acetamiprid; Pesticide Tolerance, 67 Fed. Reg. at 14,655.
The EPA removed the additional child safety factor for mepiquat (reducing it to lx) because it had already satisfied itself of the safety of another "salt” of mepiquat that it determined was identical in terms of its environmental distribution and toxicology. Mepi-quat; Pesticide Tolerance, 67 Fed.Reg. at 3,115-16. It explained:
The Agency published a risk assessment for mepiquat chloride on January 12, 2000, that discusses use on cotton as well as all other registered uses of mepiquat chloride. In that analysis risk estimates for exposure to mepiquat chloride were below the Agency’s level of concern. The Agency has reviewed a dissociation study for mepiquat pentaborate that demonstrates that mepi-quat pentaborate dissociates in an identical physical manner to mepiquat chloride in water. Therefore, the analysis performed for mepiquat chloride or the "chloride salt,” also pertains to this mepiquat "pentaborate salt” use because the use rate, maximum seasonal use rate and other pertinent use factors remain the same for mepiquat chloride or the "chloride salt.” ...
Margins of safety are incorporated into EPA risk assessments either directly through use of a margin of exposure (MOE) analysis or through using uncertainty (safety) factors in calculating a dose level that poses no appreciable risk to humans.
The Agency has determined that the FQPA safety factor for mepiquat is IX. See the Agency’s risk assessment for mepiquat chloride dated January 12, 2000, for details. The facts are that mepiquat pentaborate is another “salt” of mepiquat and that mepi-quat pentaborate disassociates to mepiquat and therefore the basic toxicology data base for mepiquat chloride pertains to mepiquat pentaborate.
Id.
In establishing the tolerance for pymetro-zine, the EPA stated:
There is a complete toxicity database for pymetrozine and exposure data are complete or are estimated based on data that reasonably accounts for potential exposures. EPA determined that the 10X safety factor to protect infants and children could be reduced to 3. The FQPA factor is reduced after assessing the potential for additional sensitivity of infants and children to residues of pymetrozine in the following studies: developmental toxicity studies in rabbit and rat and two-generation reproduction study in the rat. There was no evidence of increased susceptibility in these studies. The FQPA safety factor was not reduced to one due to the need for a developmental neurotoxicity study.
Pymetrozine; Pesticide Tolerance, 66 Fed. Reg. at 66,791.
. Natural Resources Defense Council, Objections to the Establishment of Tolerances for Pesticide Chemical Residues: Halosulfuron-methyl and Pymetrozine Tolerances (filed Feb. 25, 2002); Natural Resources Defense Council, Objections to the Establishment of Tolerances for Pesticide Chemical Residues: Imidacloprid, Mepiquat, Bifenazate, Zetacy-permetkrin, and Diflubenzuron Tolerances (filed May 19, 2002); Natural Resources Defense Council, Objections to the Establishment of Tolerances for Pesticide Chemical Residues: Isoxadifen-ethyl, Acetamiprid, Pro-piconazole, Furilazole, Fenhexamid, and Fluazinam Tolerances (filed May 20, 2002).
. The NRDC also objected to the three tolerances on a number of other grounds, including the failure to consider farm children adequately, a failure to consider worker risk adequately, an inadequate aggregate risk assessment, an improper reliance on a "lowest observed adverse effect level” (LOAEL) rather than a "no observed adverse effect level” (NOAEL), a residue "percentile” assumption *1056that inadequately protected infants and children, and an improper assessment of chronic dietary exposure, among others. The NRDC did not raise these issues to us on appeal. See supra Note 4; Final Rule, 70 Fed.Reg. at 46,713-14, 46,717-36.
. As noted supra, however, a valid objection cannot be this general; it must "[sjpecify with particularity the provision(s) of the order, regulation or denial objected to, the basis for the objection(s), and the relief sought.” 40 C.F.R. § 178.25.
. The majority also relies in part upon the dissent of Judge Wright in Center for Auto Safety v. Peck, 751 F.2d 1336, 1373 (D.C.Cir.1985) (Wright, J., dissenting). Maj. Op. at 1052-53 n. 7. However, Judge (now Justice) Scalia’s opinion for the majority in that case captures the dissent’s error, and, concomitantly, the majority’s mistake in this case when it warns that “we must be implacably skeptical of belated recognition at the appellate stage that elements of scientific analysis unchallenged during a contested proceeding are incomprehensible without further explanation. To credit such post-appeal pleas of inadequate information is to threaten the integrity of all rulemaking in fields beyond our own limited scientific ken.” Id. at 1361. The entirety of the passage bears repeating:
The dissent’s response is equally unsatisfactory, consisting essentially of the assertion that late presentation of the issue must be *1058disregarded because validity of the [data] curves is essential to validity of the agency’s conclusion. It is simply not the case, however, that all of the essential postulates for an agency rule must be contained in the record. Every judgment of any consequence is constructed upon an infinitude of other judgments, of greater or lesser certitude, in a progression of logical dependency terminating in a first principle the equivalent of 1 + 1=2. They cannot all possibly be included in the statement of basis and purpose for a rulemaking. We do not have authority to require that all elements underlying a rule be set forth in a fashion “understandable to a layman,” see Vermont Yankee Nuclear Power Corp. v. Natural Resources Defense Council, Inc., 435 U.S. 519, 557, 98 S.Ct. 1197, 1218, 55 L.Ed.2d 460 (1978). And it may as well be disclosed that in scientific fields we judges ourselves are laymen, ill equipped to determine where the line falls between requisite explanation of problematic analysis and useless replication of what for the cognoscenti amounts to a textbook on basic physics. Thus, our "feel” for the adequacy of an analysis in such cases is necessarily governed by the reaction that it elicits from knowledgeable commenters. From the point at which it reaches common ground we can reasonably assume that no further explication is required. We will hear on appeal assertions that needful elaborations fairly requested were not provided; but we must be implacably skeptical of belated recognition at the appellate stage that elements of scientific analysis unchallenged during a contested proceeding are incomprehensible without further explanation. To credit such post-appeal pleas of inadequate information is to threaten the integrity of all rulemaking in fields beyond our own limited scientific ken. The present challenge to the effectiveness curves presents the threat in a particularly flagrant form. NHTSA specifically asked the petitioners (and other rulemaking participants) in 1979: "Do the existing analyses represent the most appropriate methods of approaching a study of bumper standards at different impact speeds and levels of damage resistance? If not, what method should be used?” Though State Farm and IIHS, among others, responded with substantial comments there was not even a suggestion that the effectiveness curves represented a fundamentally invalid methodology. The dissent provides a list of reasons why petitioners might not have raised these objections during the rulemaking proceeding. Dissent at 1387-88 all of which boil down to petitioners' probable reluctance to upset favorable determinations. Even if that were the only consideration, it would be of questionable wisdom to reward their tactical decision to leave the agency in the dark — and thus encourage benighted agency action in the future. But in any event, the dissent's speculations do not explain why the proponents of the 2.5 mph standard did not raise these arguments. In fact, some commenters on the June 1979 Final Assessment (which supported the 5.0 mph standard) did object that the "effectiveness ratios for 5.0 mph systems are overstated to a considerable extent,” but that objection was based upon the lack of correspondence with repair cost data rather than theoretical inadequacy of the effectiveness curves.
We conclude that the curves were considered part of the common ground of expert analysis that required no further explanation. Petitioners' objections on this score put us in mind of the Supreme Court’s injunction in Vermont Yankee that administrative proceedings should not be a game or a forum to engage in unjustified obstructionism by making cryptic and obscure reference to matters that "ought to be” considered and then, after failing to do more to bring the matter to the agency’s attention, seeking to have that agency determination vacated on the ground that the agency failed to consider matters "forcefully presented.”
435 U.S. at 553-54, 98 S.Ct. at 1217. These instructions are even more appropriate when cryptic and obscure reference to theoretical inadequacy — or in fact, even less than that, mere assertion of lack of explanation of theoretical adequacy — is first made on appeal.
Id. at 1361-62 (emphases in original, internal citations to the record omitted).