OPINION
SILVERMAN, Circuit Judge:This appeal arises after fourteen years of litigation concerning public transit in Los Angeles County. In 1994, the Labor/ Community Strategy Center and other Los Angeles County community organizations and local residents, known collectively as the “Bus Riders Union” or “BRU,” brought a civil rights class action against the County’s Metropolitan Transit Authority, charging the MTA with unlawfully discriminating against “inner-city and transit dependent bus riders” in its allocation of public transportation resources. The case did not go to trial; rather, in 1996, the parties agreed to, and the district court approved, a consent decree that committed MTA to implementing “a detailed plan to improve bus service.” See Labor/Community Strategy Ctr. v. L.A. County Metro. Trans. Auth., 263 F.3d 1041, 1043 (9th Cir.2001) ("Labor/ Community ”). The district court’s jurisdiction over the decree was explicitly set to expire in ten years.
Shortly before the tenth anniversary of the decree, BRU moved to extend the duration of the decree on the grounds that MTA had allegedly failed to comply with the decree’s overcrowding provisions. BRU *1117also sought civil contempt sanctions against MTA for MTA’s alleged failure to comply with a 2004 remedial order. Ruling that MTA had substantially complied with the decree, the district court denied BRU’s motion seeking these remedies and allowed the decree to expire.
We hold today that the district court did not abuse its discretion in denying BRU’s motion to extend the decree and for contempt sanctions. We therefore affirm the district court’s decision in all respects.
I. Background
A. The Underlying Lawsuit
In 1994, BRU filed a class action under several federal civil rights statutes,1 charging MTA with violating the Fourteenth Amendment by discriminating against low income and minority residents of Los Angeles County. The alleged discrimination included expending a disproportionately high share of its resources on commuter rail services, whose primary users were wealthy non-minorities, and a disproportionately low share on bus services, whose main patrons were low income and minority residents.2
In March 1995, the district court certified a plaintiff class of “[a]ll poor minority and other riders of MTA buses who are denied equal opportunity to receive transportation services because of the MTA’s operation of a discriminatory mass transportation system.” On the eve of trial the parties submitted a proposed consent decree, which the district court approved on October 29, 1996. See Labor/Community, 263 F.3d at 1043.
B. The Consent Decree
The decree committed MTA to a wide array of improvements in its bus services, including instituting new bus lines to and from centers of employment, education, and health care in the county; enhancing security on buses; improving bus shelters; and maintaining its fares at specific levels. BRU’s present appeal does not directly concern those aspects of the decree.
BRU does contest MTA’s compliance with Section II.A of the decree, which committed MTA to “[Reducing [o]vercrowding [b]y [ajdding [n]ew [sjervice.” The decree did not set out a specific number of buses or hours of service the MTA needed to add to achieve appropriate reductions in overcrowding. Instead, the decree set forth specific “load factor targets,” to be met by specific dates, and provided MTA with “discretion in determining how the targets w[ould] be met.” A “load factor” is “a numerical representation of the number of people standing [and sitting] on a bus in relation to the number of seats.” Labor/Community, 263 F.3d at 1044 n. 1. For example, a forty-seat bus which contained forty-eight passengers would have a load factor of 1.2. Under the terms of the decree, MTA was required to reduce the maximum load factor for each relevant bus line to 1.35 by December 31, 1997, 1.25 by June 30, 2000, and 1.2 by June 30, 2002.
In addition to these substantive provisions, the decree included a number of procedural mechanisms designed to assure implementation and enforcement. Four are relevant to this appeal.
First, the decree created a Joint Working Group, composed of representatives from BRU and MTA. The purpose of the Working Group was to “foster cooperation *1118in the implementation” of the decree and to resolve disputes between the parties.
Second, Section V of the decree established the position of “Special Master,” “to facilitate the resolution of disputes arising under any provision of this Consent Decree” that could not be settled by the Working Group. The parties agreed that Donald T. Bliss would serve as the initial Special Master.
Third, Section VI, entitled “Modification of the Consent Decree,” set forth the conditions under which the decree may be modified. Under Section VI, a party seeking modification must show “that a significant change in circumstances warrants revision of the Consent Decree, and that the proposed revision or revisions are suitably tailored to the changed circumstances.”
Finally, Section VIII provided that the “District Court shall retain jurisdiction over this litigation for ten years from the date of approval of this Consent Decree in order to monitor compliance with this Consent Decree.” The court approved the decree on October 29, 1996, so the court’s obligatory retention of jurisdiction was set to terminate on October 29, 2006. Section VIII stated further that:
At the end of seven years, MTA may file a motion with the District Court to terminate the Consent Decree and the Court shall grant such motion if MTA shows to the Court’s satisfaction that it has substantially complied with the Consent Decree and that it has in place a service plan that will enable continued adherence to the principles and objective of the Consent Decree during the five years subsequent to the termination of this Consent Decree.
C. The 1999 and 2004 Orders
In September 1998, the members of the Working Group agreed that on seventy-five out of the seventy-nine bus lines monitored, MTA had not met the initial load factor target of 1.35 by the December 31, 1997 deadline. See Labor/Community, 263 F.3d at 1046. As the parties’ representatives could not agree on a remedial plan to address this failure, and also to ensure that MTA would meet the second target of 1.25 by June 30, 2000, the Special Master issued an order on March 6, 1999 mandating a plan which, as later revised, required, among other things, that MTA acquire 379 additional buses. Id. at 1047.
MTA sought review of the 1999 Order, questioning both the Special Master’s authority to impose it and the validity of the factual findings and interpretation of the decree on which it was premised. Id. The district court upheld the Special Master’s findings regarding the agency’s noncompliance with the 1997 1.35 load factor target as not clearly erroneous, and affirmed the 1999 Order as to the 248 buses that MTA was required to purchase to meet the 1997 target. At the same time, the district court declared it “too early to determine whether the MTA is incapable of meeting” the 2000 load factor target, so it vacated the Special Master’s Order with regard to the remaining additional buses and directed the Special Master to reevaluate the need for those buses once more up-to-date data were available. We affirmed the district court’s order in Labor/Community. 263 F.3d at 1051.
On January 12, 2004, the Special Master promulgated another order addressing measures necessary for MTA to achieve compliance with the load factor targets. The “Final Order on Remedial Service Plan to Meet 1.25 and 1.2 Load Factor Target Requirements” specified that “[i]n order to achieve compliance with the Consent Decree and to meet and maintain the 1.2 [load factor target],” MTA was required to provide the additional buses and service hours called for by the Working *1119Group. The Order permitted MTA to meet some of the additional requirements through changes to existing services and more efficient scheduling of its existing fleet. But it required MTA to purchase “the vehicular equivalent of 145 new 40-seat expansion buses” and to provide “an additional 290,145 annual in-service hours.” The Final Order stated that providing this expanded service, along with the other improvements specified, would constitute “substantial compliance with the load factor targets of the Consent Decree and create a presumption that [its] expansion bus procurement requirements ... have been met.”
D. The District Court’s Ruling
In February 2006, the Special Master resigned. Although BRU and MTA agreed on a replacement, the district court did not appoint a new Special Master, citing the limited “amount of time remaining before the Court is divested of jurisdiction over the ... Decree.” The court indicated, however, that if its “jurisdiction is extended, it will revisit this issue.”3
After the district court declined to appoint a new Special Master, BRU filed motions to extend the duration of the decree and for contempt sanctions.4 The district court denied both. It concluded that contempt sanctions were inappropriate because BRU had failed to satisfy its burden of “establishing], by clear and convincing evidence, that MTA either failed to substantially comply with the Final Order or failed to take all reasonable steps to insure compliance with the Final Order.” On the contrary, the court held, “it is clear that MTA has substantially complied, and taken all reasonable steps within its power to insure compliance, with the Final Order.”
Moreover, the district court stated, extension of the decree was unnecessary because the “Decree did not require perfection,” and “[d]espite an increasing ridership, increasing traffic congestion and fiscal constraints, MTA has substantially complied with the Consent Decree while maintaining fares at reasonable levels.” The district court concluded that the decree “has served its purpose and will not be extended.” BRU timely appealed.
II. Standard of Review
We review a district court’s interpretation of a consent decree de novo, with “deference ... based on the court’s extensive oversight of the decree from the commencement of the litigation to the current appeal.” Nehmer v. U.S. Dep’t of Veterans Affairs, 494 F.3d 846, 855 (9th Cir.2007). Both a district court’s refusal to extend a consent decree and its denial of a motion for contempt are reviewed for abuse of discretion. See Thompson v. U.S. Dep’t of Hous. & Urban Dev., 404 F.3d 821, 827 (4th Cir.2005); Hallett v. Morgan, 296 F.3d 732, 749 (9th Cir.2002); David C. v. Leavitt, 242 F.3d 1206, 1210 (10th Cir.2001); Holland v. N.J. Dep’t of Corr., 246 F.3d 267, 281 (3d Cir.2001); Vanguards of Cleveland v. City of Cleveland, 23 F.3d 1013, 1017 (6th Cir.1994). A district court abuses its discretion when “its equitable decision is based on an error of law or a clearly erroneous factual finding.” Ken*1120ney v. United States, 458 F.3d 1025, 1032 (9th Cir.2006).
III. Discussion
A. Extension of the Decree
By its express terms, the decree provided for the district court’s retention of jurisdiction over compliance with the decree only until October 29, 2006, ten years after its execution.5 Because the decree contains an express expiration date for the court’s retention of jurisdiction, any change to that date entails a modification of the decree. See Thompson, 404 F.3d at 824 (holding that extension of a consent decree’s termination date required modification of the decree). The first issue, then, is whether, and under what conditions, a modification of the decree is permitted.
To answer this question, we turn first to the decree itself. See United States v. Asarco Inc., 430 F.3d 972, 980 (9th Cir.2005) (noting that the meaning of “[a] consent decree, like a contract, must be discerned within its four corners”). According to Section VI of the decree, modification of the decree is permitted only if the party seeking modification establishes that (1) “a significant change in circumstances warrants revision” and (2) “the proposed revision or revisions are suitably tailored to the changed circumstances.” The decree further provides that modification “may be warranted when changed factual conditions make compliance with the Consent Decree unworkable or substantially more onerous, and when the changed factual conditions were unforeseen at the time of the entry into this Consent Decree.” Perhaps not surprisingly, these requirements are essentially identical to those articulated by the Supreme Court in Rufo v. Inmates of Suffolk County Jail, 502 U.S. 367, 393, 112 S.Ct. 748, 116 L.Ed.2d 867 (1992), and applied by this Court in Asarco and other cases involving modifications to consent decrees. See Asarco, 430 F.3d at 979; Hook v. Arizona, 120 F.3d 921, 924 (9th Cir.1997). So, as provided both by the decree and by the Rufo case law, BRU’s requested modification — i.e., extension of the district court’s jurisdiction over the decree — is warranted only if the following four conditions are met. First, BRU must establish that a “significant change either in factual conditions or in the law” occurred after execution of the decree. See Asarco, 430 F.3d at 979. Second, it must demonstrate that the change was not “anticipated at the time it entered into [the] decree.” Id. Third, it must show that the changed factual circumstance makes “compliance with the consent decree more onerous, unworkable, or detrimental to the public interest.” Id. (internal citations omitted). Finally, the proposed extension of the decree’s termination date must be “suitably tailored to resolve the problems created by the changed ... conditions.” Id.
The failure of substantial compliance with the terms of a consent decree can qualify as a significant change in circum*1121stances that would justify the decree’s temporal extension. See Thompson, 404 F.3d at 828-29; David C, 242 F.3d at 1212; Vanguards of Cleveland, 23 F.3d at 1019-20; Holland, 246 F.3d at 283-84. Here, however, after overseeing this case for more than twelve years, the district court concluded that MTA had substantially complied with the decree. The court ruled as follows:
For the past ten years — -the entire term of the Consent Decree — the parties have disagreed as to how to implement the Consent Decree, how to reach its objectives, and how to measure its success. In hindsight, the Consent Decree was a less than perfect document. As a result, it is impossible to achieve absolute compliance. Indeed, the Special Master informed the parties that the Consent Decree did not require perfection. [Citation] However, it was possible for MTA to substantially comply with the Consent Decree. Despite an increasing ridership, increasing traffic congestion and fiscal constraints, MTA has substantially complied with the Consent Decree while maintaining fares at reasonable levels. The Consent Decree has served its purpose and will be not extended.... As a result of the Consent Decree and the efforts of all of the parties, the quality of life has improved for Los Angeles’s public transit dependant poor population.
The court’s finding that the decree “had served its purpose” reflected a conclusion that it was no longer necessary to involve the federal courts in the day-to-day operation of the Los Angeles County bus system. The district court, armed with a decade of knowledge about the case, was uniquely positioned to determine whether there had been substantial compliance. We accord that court’s decision deference on this issue. See Thompson, 404 F.3d at 827 (noting deferential standard of review in institutional reform cases and stating that “[o]ver time, the district court gains an intimate understanding of the workings of an institution and learns what specific changes are needed within that institution in order to achieve the goals of the consent decree” (quoting Navarro-Ayala v. Hernandez-Colon, 951 F.2d 1325, 1338 (1st Cir.1991))); see also Ruiz v. Lynaugh, 811 F.2d 856, 861 (5th Cir.1987) (discussing deference due district court in institutional reform cases “because it is intimately involved in the often complex process of institutional reformation” and “has the personal knowledge, experience, and insight necessary to evaluate the parties’ intentions, performances, and capabilities”).
Moreover, it is BRU’s burden to show otherwise. BRU must demonstrate that MTA failed to substantially comply with the decree in order to justify its extension. Asarco, 430 F.3d at 980 (stating that burden is on the moving party). In arguing that MTA’s level of compliance was insubstantial, BRU focuses exclusively on just one of the decree’s several requirements, and uses an imperfect and misleading metric to evaluate compliance. Further, BRU ignores the many ways in which MTA met or exceeded its obligations.
Although the underlying data measuring bus overcrowding are not in dispute, the parties interpret that data in vastly different ways. BRU argues that the data demonstrate compliance rates as low as 9% with the 1.20 load factor target during 2005 and 2006. Meanwhile, MTA argues the same data supports the conclusion that it achieved a 94-99% compliance rate with the load factors during the same time period. In some sense, both figures are correct.
BRU’s compliance figures are based on the standard set forth in the decree for *1122identifying instances of violations of the load factor targets. According to the decree, peak load factors are to be determined “by computing the highest ratio of total number of passengers to total number of seats achieved during any 20 minute weekday peak period in the peak direction of travel on each bus line.” Under this metric, if a bus line exceeds the load factor just once during a given quarter, that line is deemed noncompliant for the entire quarter, even if that line met the target at every other measurement that quarter. MTA argues that this standard greatly overstates its level of noncompliance, and it offers its own figures to compensate for this limitation. MTA states that in 2005, it monitored 123,154 20-minute time periods and found that 96.89% of those periods experienced no load factors above the 1.20 target.
BRU correctly notes that the figures it cites are based on the compliance standard written into the decree and affirmed by this court in Labor/Community. 263 F.3d at 1048-^49. But that standard measures only strict compliance with the load factor targets — not compliance with the decree overall — and does so in an imprecise manner. That coarse-grained metric is useful for certain types of analyses, such as determining whether there has been full and absolute compliance (MTA concedes there has not been), but it is not particularly helpful in measuring levels of compliance below 100%, and it fails to accurately capture the extent to which MTA did meet the targets during the relevant time periods.
If the question here were simply whether MTA had achieved full compliance with the decree, we would use BRU’s proposed standard. But the question is whether there was substantial compliance, a less precise standard that cannot be satisfied by reference to one particular figure, while ignoring alternative information. Our analysis requires we do more than simply count the number of technical deviations from the decree. Instead, we must determine, using a holistic view of all the available information, whether MTA’s compliance with the Decree overall was substantial, notwithstanding some minimal level of noncompliance.
In addition to using an imperfect metric to evaluate compliance with the load factor targets, BRU also focuses narrowly on that one requirement, as does the dissent, at the expense of giving due weight to the various other requirements under the decree which MTA met, and in some cases, exceeded. There is no question that the reduction of bus overcrowding was an important part of the decree. But so were other requirements. As BRU’s counsel described, the decree imposes three “essential” and “core” requirements on MTA. These are the reduction in bus overcrowding through the load factor targets, new service through an expedited pilot project followed by a five-year new service plan, and a roll-back and lowering of bus fares for at least five years. MTA exceeded its obligations as to the bus fare and pass requirements, maintaining fares at specified levels for years longer than it was required to do so. And MTA has now met its obligations relating to new service requirements. To give some idea of the extent of MTA’s efforts, since the start of the decree, it has added 1.2 million in-service hours annually, and added over 545 buses to its peak fleet in order to expand its bus service. BRU argues that MTA’s achievements in these areas are outweighed by its imperfect compliance with the load factor targets. Of course, there is no precise formula describing how best to weigh the various obligations under the decree, and there is no indication that the district court abused its discretion in weighing them as it did.
*1123We note that the de minimis level of noncompliance here is nowhere close to the near total noncompliance in cases in which courts concluded that extensions of the consent decrees were warranted. In Thompson, 404 F.3d at 834, there was a “near total failure” of some defendants to comply with their obligations. Those defendants failed to do “almost [anything] that they were required to do under the Decree[.]” Id. at 828; see also David C., 242 F.3d at 1212-13 (noting that defendant was “20 percent in compliance and 80 percent in noncompliance”). In contrast, aside from the imperfect compliance with the load factor targets, MTA complied fully with its numerous obligations under the decree.
We hold that BRU has failed to demonstrate that the district court abused its discretion in finding that MTA had substantially complied with the consent decree. The evidence presented supported the district court’s finding that the imperfections with respect to load factor targets were de minimis in relation to the overall scheme of things. Because the first prong of the Rufo test fails, we hold that the district court did not abuse its discretion in refusing to extend the decree.
Our decision is consistent with the principle that federal court intervention in state institutions is a temporary measure and may extend no longer than necessary to cure constitutional violations. See Bd. of Ed. of Okla. City Pub. Sch. v. Dowell, 498 U.S. 237, 248, 111 S.Ct. 630, 112 L.Ed.2d 715 (1991); Toussaint v. McCarthy, 801 F.2d 1080, 1087 (9th Cir.1986). In this case, as the district court found, perhaps every last wish and hope of the decree was not achieved, but the decree accomplished its essential purposes and the situation improved greatly. These improvements strongly inform our assessment that the district court was within its discretion in holding that it no longer needed to oversee the running of the Los Angeles County bus system.
B. BRU’s Motion for Contempt Sanctions
BRU also contends that the district court abused its discretion in denying its motion for civil contempt sanctions against MTA for MTA’s alleged failure to comply with the 2004 Final Order. For issuance of a contempt order against MTA to be proper, BRU must establish “(1) that [MTA] violated the court order, (2) beyond substantial compliance, (3) not based on a good faith and reasonable interpretation of the order, (4) by clear and convincing evidence.” In re Dual-Deck Video Cassette Recorder Antitrust Litig., 10 F.3d 693, 695 (9th Cir.1993). BRU argues that MTA violated the Final Order in two ways: (1) by failing to allocate all of the newly purchased buses and additional service hours to peak time periods, and (2) by failing to hire any new mechanics.
The Final Order required that, among other things, MTA purchase “the vehicular equivalent of 145 new 40-seat expansion buses” and add an additional 290,145 annual in-service hours. MTA did add 145 buses and 298,985 in-service hours to the bus system. But according to BRU, MTA violated the Final Order by failing to allocate all of these resources to peak time periods. MTA does not dispute that some of the new hours and buses went to non-peak times, but contends that the Final Order gave MTA discretion regarding how to perform the allocation.
Upon review of the Final Order, it is not immediately obvious whether the hours and buses were required to be allocated to peak time periods. There is no express language directing MTA to do so and the Final Order provides MTA with “discretion” to deploy the buses “throughout the *1124bus system.” Even assuming that MTA violated the Final Order by allocating some resources to non-peak time periods, the district court was within its discretion in ruling that contempt sanctions were not warranted. There is no evidence that MTA’s conduct was “not based on a good faith and reasonable interpretation of the .order.” See id.
While MTA did not add all of the additional in-service hours to peak periods, it did add an estimated 162,947 annual in-service hours during peak periods and did increase its total annual in-service hours by more than the amount required by the Final Order. MTA also added 145 buses to its fleet as required (though it is not clear how many of those buses were added to peak time periods). These efforts rebut any inference that MTA was acting in bad faith. In addition, even if MTA’s interpretation of the Final Order were ultimately deemed incorrect, it was not unreasonable. MTA is right that the Final Order does not literally state that the additional hours and buses must be added during peak times.
BRU also argues that MTA’s failure to hire new mechanics violated the Final Order, which mandated that MTA “[h]ire additional mechanics as needed to meet the expansion service requirements!!]” The language “as needed” renders the directive conditional, and we have no basis to second-guess MTA’s conclusion that it had no need to hire more mechanics. It is, for example, undisputed that MTA added nearly 300,000 annual in-service hours after Special Master Bliss issued the Final Order. The fact that MTA increased its in-service hours so significantly without adding mechanics lends substantial credence to its argument that no new mechanics were needed to meet its obligations under the Final Order.
We hold that BRU has not demonstrated that the district court abused its discretion in declining to sanction MTA for its alleged violations of the Final Order.
IV. Conclusion
For the foregoing reasons, the District Court’s denial of Plaintiffs’ Motion to Extend and its denial of Plaintiffs’ Motion for Civil Contempt Sanctions are AFFIRMED.
. The complaint relied upon 42 U.S.C. §§ 1981, 1983, and 2000d.
. The allegations made in the underlying complaint are discussed in more detail at Labor/Community, 263 F.3d at 1043.
. BRU asserts that the district court’s failure to appoint a new Special Master violates the decree. Neither BRU nor MTA timely appealed that order, so we lack jurisdiction to review it. See 28 U.S.C. § 2107.
. BRU originally brought motions to extend the duration of the decree and for contempt before the Special Master in 2004. The Special Master denied the former, without prejudice, as premature; the latter remained pending at the time of his resignation.
. BRU argues that "[t]he Decree contains no termination date” and therefore that the district court lacked the discretion to end its jurisdiction until MTA complied fully with the terms of the decree. BRU did not raise this issue before the district court and we generally will not consider an issue raised for the first time on appeal. See Bolker v. Comm’r, 760 F.2d 1039, 1042 (9th Cir.1985). BRU argues that the issue is one of law and can be resolved on the record as it presently exists. We decline to exercise our discretion to consider the issue because it may turn upon facts that MTA did not have an opportunity to develop before the district court, namely, extrinsic evidence of the parties’ intent. See id. (declining to address issue not raised below because of possibility of relevant facts not developed in the record).