I. OVERVIEW

This appeal concerns regulations promulgated by the National Oceanic & Atmospheric Administration (NOAA) regarding the incidental taking of porpoise by purse seine tuna fishermen. The regulations derived from a formal rule-making proceeding conducted under § 4 of the Administrative Procedure Act (APA). 5 U.S.C. § 553 . The American Tunaboat Association (ATA) filed this action in district court to set aside or modify the final regulations adopted by NOAA. The district court, on cross-motions for summary judgment, ruled for the ATA, finding NOAA’s decision unsupported by substantial evidence. The district court ordered NOAA to revise its regulations in accordance with a prior decision entered by an administrative law judge. NOAA appeals. We affirm.

II. FACTS

The members of the ATA fish commercially for yellowfin tuna in the eastern tropical Pacific Ocean. Over the years, the fishermen have observed that yellowfin tuna tend to associate with various species of porpoise. Where a school of porpoise is observed at the surface, a school of tuna will generally be found below. The fishermen take advantage of this phenomenon by guiding the porpoise to an area where nets have been set, knowing the tuna will follow.

The nets used are the “purse seine” variety, which can be closed with a sort of purse string, encircling the tuna below. Unfortunately, porpoise are often incidentally caught in the nets and, as they are air-breathers, they drown. It was an unacceptably high mortality level of porpoise incidentally captured in purse seine nets that was primarily responsible for passage of the Marine Mammal Protection Act of 1972 (MMPA). 16 U.S.C. §§ 1361 et seq. Section 1373(a) of the MMPA provides in relevant part:

The MMPA generally imposes a moratorium on intentional takings of porpoise species most often associated with yellowfin tuna. Recognizing, however, that certain of the species exist in abundance, the MMPA delegates to the Secretary of Commerce the duty to issue permits for the taking of these porpoise incidental to commercial fishing. The Secretary of Commerce, in turn, delegates to NOAA the duty of determining which species of por poise are “disadvantaged” or “depleted” within the meaning of the MMPA. The basic approach of the statute and regulations has been to impose yearly quotas on the taking of abundant porpoise species by tuna fishermen, and to prohibit the taking of any species deemed depleted. All parties agree that, through the cooperation of the tuna industry, the statute and regulations have vastly reduced the mortality levels of porpoise.

On March 20, 1980, the ATA applied for a general permit for the period January 1, 1981 through December 31, 1985. Supp. ER 4-33. The application and NOAA’s related regulatory proposal, Supp. ER 34-45, were assigned to an administrative law judge (AU) for hearing. Most of the testimony adduced at hearing concerned estimates of current and historic porpoise populations. The parties agreed that porpoise population is calculated primarily from three measurements: (1) mean school size, (2) density of schools, and (3) range of ocean within which the schools are found. In spite of this, the ATA and NOAA’s population estimates differed widely due to their disagreement on how the three measurements should be conducted.

Both the ATA and NOAA called numerous lay and expert witnesses attempting to present to the AU the “best scientific evidence available” to determine porpoise populations within the meaning of the MMPA. The NOAA staff’s population estimates were premised on evidence that (1) aerial observations plus limited observations from research vessels provided an adequate basis for determining mean school size; (2) 100% of the schools consisting of 15 or more porpoise are observed along the trackline of an airplane flying 900 feet above the surface at 130 mph in all weather conditions; (3) only data collected prior to 1977 need be used to establish the range inhabited. Relying on these data bases, the NOAA staff proposed that the northern offshore spotted porpoise be deemed depleted and incidental takings thereof banned. The staff also proposed to revise the aggregate quota for other porpoise stocks downward from 31,150 in 1980 to 18,640. Supp. ER 35-45.

The AU disagreed with NOAA’s data collection methods. He found that the “best scientific evidence available,” 16 U.S.C. § 1373 (a), (1) included observations by federal observers placed aboard tuna-boats, along with aerial and research vessel observation; (2) discredited the 100% sighting of porpoise schools theory; and (3) required utilization of available range data more current than 1977. The AU estimated that if these data collection methods were employed, the northern offshore spotted porpoise would not be found depleted. Therefore, the AU included that species in his recommended total allowable quota of 20,500.

Although NOAA adopted the AU’s quota recommendation and his suggestion that the northern offshore spotted porpoise was not depleted, Supp. ER 154, it disagreed with the AU’s recommendations concerning population data collection techniques. NOAA’s final decision rejected these recommendations and instead approved the NOAA staff’s data collection techniques.

The ATA filed suit in district court seeking a declaration that the AU’s findings of mean school size, density of schools, and range of schools were the “best scientific evidence available.” ER 13. On cross-motions for summary judgment, the district judge ruled that NOAA should have accepted the following AU recommendations: (1) a mean school size of 328 should be used; (2) in calculating porpoise school density, not all schools are seen on the trackline of the aircraft; and (3) the range of schools is larger than that adopted in NOAA’s final decision. The district court granted ATA’s declaratory relief and ordered NOAA to submit recalculated population estimates based on the AU’s recommendations. NOAA complied on April 26, 1982. ER 146-51. This appeal followed.

III. DISCUSSION

A formal rule-making proceeding is reviewed under the APA’s substantial evidence standard. 5 U.S.C. § 706 (2)(E); Citizens to Preserve Overton Park, Inc. v. Volpe, 401 U.S. 402, 414 , 91 S.Ct. 814, 822 , 28 L.Ed.2d 136 (1971). As with all agency action under the APA, rule-making must be set aside if arbitrary, capricious, or an abuse of discretion. 5 U.S.C. § 706 (2)(A); Overton Park, 401 U.S. at 413-14 , 91 S.Ct. at 822 . While we should be influenced by the district court’s decision, we are not bound by it. See Washington State Farm Bureau v. Marshall, 625 F.2d 296, 302 (9th Cir.1980); Asarco, Inc. v. EPA, 616 F.2d 1153, 1161 (9th Cir.1980).

Derivative inferences drawn by an AU are entitled to no special weight by the reviewing court, although testimonial inferences are. Penasquitos Village, Inc. v. NLRB, 565 F.2d 1074, 1076-79 (9th Cir.1977). It is equally clear that our standard of review does not change simply because the agency has disagreed with the AU. Brooks v. NLRB, 538 F.2d 260, 261 (9th Cir.1976). In essence, “[t]he significance of [the AU’s] report [will] depend largely on the importance of credibility in the particular case.” Universal Camera Corp. v. NLRB, 340 U.S. 474, 496 , 71 S.Ct. 456, 468 , 95 L.Ed. 456 (1951).

Despite the narrow scope of review under this standard and the broad discretion afforded NOAA in administering the MMPA, we affirm the decision of the district court. In doing so, we have in mind the rule that, even though an agency decision may have been supported by substantial evidence, where other evidence in the record detracts from that relied upon by the agency we may properly find that the agency rule was arbitrary and capricious. Bowman Transportation, Inc. v. Arkansas-Best Freight System, Inc., 419 U.S. 281, 284 , 95 S.Ct. 438, 441 , 42 L.Ed.2d 447 (1974) (agency decision supported by substantial evidence may still be arbitrary and capricious). See also B. Mezines, J. Stein & J. Gruff, Administrative Law § 51.03, at 51-59 (1982) (footnotes omitted); Packer Transportation Co. v. United States, 596 F.2d 891 , 894 n. 4 (9th Cir.1979).

A. Mean School Size

The AU relied on a broad factual basis in determining that the data collected by federal observers on tuna vessels are trustworthy and that NOAA’s reasons for rejecting that data were speculative and not supported by the record. The AU did not rule that observer data was the only data to be utilized; rather he determined that “[t]he best scientific evidence available consists of a combination of data from aerial survey, research vessel, and federal observers on board tuna vessels.” ALJ’s Recommended Decision, Supp. ER at 88. We agree.

Although receiving some support from the record, NOAA’s decision ignores a comprehensive data base that is the product of many years’ effort by trained research personnel. Pursuant to MMPA regulations, these federal employees have been aboard tuna vessels for several years in their research and observation capacity. 50 C.F.R. 216.24(f). This collection of data is authorized by the regulations and its use in enforcement proceedings has recently been upheld by this circuit. Balelo v. Baldrige, 724 F.2d 753 (9th Cir.1984) (en banc), cert. denied, — U.S. —, 104 S.Ct. 3536 , 82 L.Ed.2d 841 (1984). No evidence was presented to indicate that schools of target species encountered by tuna vessels were not representative of the overall size of schools in the population. Evidence was presented that porpoise did not provide the only sighting cue: birds, logs and other floating debris were also used. In brief, no evidence presented substantiated NOAA’s charges that an upward bias tainted the observer data.

We are aware, as the AU noted, that a combination of the three data sources (aerial survey, research ship, and federal observer) with each bearing equal weight in the calculation probably underestimates the true mean school size. This results from the fact that a greater percentage of weight is then accorded to the small aerial and research vessel data collections as opposed to the large and geographically diverse collection of observer samples. We do not find, however, that this method of ascertaining mean school size should be rejected. On the contrary, we find that the AU’s population figure deals with the potential deficiencies on all the data sources while NOAA’s does not. We cannot agree with NOAA’s decision because our review of the whole record reveals evidence (and some speculation) that so detracts from the decision as to make it arbitrary and capricious.

B. Density of Schools

NOAA’s density calculations were based on the assumption that all schools of 15 or more porpoise will be observed in aerial surveys under all conditions. After reviewing the evidence, the AU concluded that this was not a proper assumption upon which to base the density calculations. The district court agreed and held that the NOAA formulation was not supported by substantial evidence, reversing the agency decision. We affirm.

We have reviewed the record and conclude that the basic assumption underlying the NOAA calculation was not supported by substantial evidence. In arriving at its formulation, NOAA relied on testimony that 100 percent of all schools of 15 or more porpoise will be observed from the track line of an aircraft flying at an altitude of 900 feet and a speed of 130 miles per hour. This testimony was never substantiated: photographic verifications proved inconclusive and other evidence revealed that meteorological and other conditions do in fact generally affect sighting potential. See Supp. ER 58, 70-74. Questions have also arisen concerning the validity of the sampling sizes NOAA utilized. Supp. ER at 74-75.

C. Range Inhabited

Although noting that an extension of the stock range would have little effect on the overall population estimate, the AU determined that the “best scientific evidence” of those ranges would include data collected by federal observers after 1977. The district court agreed, holding that NOAA should have accepted the AU’s findings in this regard. We affirm.

In light of the comprehensive and reliable nature of the data collected by the federal observers, it was arbitrary for the agency to have simply disregarded it. Without offering any reason, the agency refused to consider the post-1977 data, data collected on over'300 expeditions and data which revealed that the relevant range of the porpoise was broader than originally imagined. Even though the net effect of this data on the overall population estimates would not have been dramatic, due to the low density of the populations in the western ranges, we conclude that the agency acted arbitrarily in refusing to consider this information in its decision-making process.

IY. CONCLUSION

We have reviewed the “whole record” in this matter and conclude that the agency acted arbitrarily in failing to utilize the best scientific evidence available in arriving at population and range estimates and also arrived at a conclusion not supported by substantial evidence in determining school density. Therefore, the decision of the district court is

AFFIRMED.