ORDER AND AMENDED OPINION
THOMAS, Circuit Judge.ORDER
The opinion filed January 28, 2002, is amended as follows:
At slip opinion page 1170, the second sentence of the first full paragraph, which begins ‘While cases decided early ...” is deleted in its entirety and replaced with the following sentence: “These cases make clear that in order to perfect the legal title to swamp and overflowed lands, those lands had to be identified and patented.”
The footnote 5 at slip opinion page 1170 is deleted in its entirety.
*1059A footnote is added at slip opinion page 1170 in the first full paragraph, after the quote ending with “... when the land should be identified and the patent issued”, which shall read: “Appellees suggest that because perfected title to swamp lands is deemed effective as of the date of the Act, the appropriate starting date for the river analysis is 1850. We find this argument unpersuasive. The relation back doctrine may make the interest within the 1905 boundaries good as of 1850, but the doctrine cannot expand or contract the boundaries created by the 1905 patent. As the cases we cite in the text indicate, boundaries to swamp and overflow lands are determined exclusively by the process of identification and patent. Any other rule would create uncertainty among land owners, as all subsequent interests to land in swamp and overflow areas might fear that their property lines could be altered by earlier, unknown river movements. See Wright v. Roseberry, 121 U.S. 488, 505, 7 S.Ct. 985, 30 L.Ed. 1039 (1887) (stating that a patent is intended to be “an invaluable muniment of title, and a source of quite and peace to its possessor.”).”
At slip opinion page 1171 the following is deleted from the first partial paragraph: “and that the doctrine of relation back did not apply.”
With the foregoing amendments, the ap-pellee’s petition for rehearing is denied.
OPINION
In this appeal, we consider whether the district court had jurisdiction over a quiet title and ejectment action and, if so, whether it properly determined that the property in question was located in the State of California. We conclude that the district court erred in dismissing the case for lack of jurisdiction and in fixing the title to the lands on the basis of river movements that occurred prior to 1905, when the United States patented the disputed lands to the State of California. We reverse and remand for further proceedings.
I
Richard Bangs once observed that “Wild rivers are earth’s renegades, defying gravity, dancing to their own tunes, resisting the authority of humans, always chipping away, and eventually always winning.”2
In the early 1900’s, there were few better examples of this than the Colorado River, which flows for 1400 miles through the western United States before reaching the Gulf of California. It drains a 242,000 square mile area including parts of seven states: Colorado, Wyoming, Utah, New Mexico, Nevada, Arizona and California. There is a modest, triangular-shaped, 130-acre patch of land between Arizona and California, and the river runs through it. More accurately, it used to run through it. Therein lies the ownership problem.
Accretion and avulsion are, in a sense, the yin and yang of river course change. Accretion is “the gradual, imperceptible addition to land forming the banks of a stream by the deposit of waterborne solids or by the gradual recession of water which exposes previously submerged terrain.” State v. Jacobs, 93 Ariz. 336, 380 P.2d 998, 1000 (1963). When a river moves by accretion, the boundary line set by the river continues to run through the center of the river channel in its new location. Nebraska v. Iowa, 143 U.S. 359, 361, 12 S.Ct. 396, 36 L.Ed. 186 (1892), cited in Arizona v. Bonelli Cattle Co., 107 Ariz. 465, 489 P.2d 699, 701 (1971), approved as amended, 108 Ariz. 258, 495 P.2d 1312 (1972), rev’d on other grounds, 414 U.S. 313, 94 S.Ct. 517, 38 L.Ed.2d 526 (1973).
In contrast, avulsion occurs when a river abandons its old course and adopts a *1060new one “suddenly or in such a manner as not to destroy the identity of the land between the old and new channels.” Jacobs, 380 P.2d at 1001. When a river moves by avulsion, the boundary does not move with the river, but instead remains in the center of the old channel. Bonelli Cattle, 489 P.2d at 701; Jacobs, 380 P.2d at 1001.
The United States brought an action on behalf of the Fort Mojave Indian Tribe for quiet title, ejectment, and trespass damages, asserting that the lands at issue (“the disputed property”) attached by the natural process of accretion to land that the United States holds in trust for the Tribe. The defendants (“private landowners”) are successors to California’s 1905 patent, and they assert ownership interests in the disputed property on the basis of a chain of title stemming from this patent. The private landowners claim that the river changed course by avulsive movement prior to 1905, establishing the boundaries for the patent. In sum, as Norman MacLean might put it, the parties are haunted by waters.
The Arizona district court held that it lacked jurisdiction because the disputed property became part of the State of California due to avulsions in the Colorado River, but it nonetheless found that the private landowners held title to the disputed land.
II
The federal district courts’ jurisdiction over actions concerning real property is generally coterminous with the states’ political boundaries. Columbia River Packers’ Ass’n v. McGowan, 219 F. 365, 377 (9th Cir.1914). Because the remedies the United States seeks would act directly upon the land itself, jurisdiction is properly exercised in the state where the land is situated. The District Court of Arizona held that it lacked jurisdiction because the disputed property became part of the State of California due to an avul-sive movement of the Colorado River in 1857.
However, the Interstate Compact Defining the Boundary Between the States of Arizona and California, Pub.L. No. 89-531, 80 Stat. 340, art. II (1966) (“Boundary Compact”) fixed the boundary between Arizona and California for political purposes in the area encompassing the disputed property along the Colorado River, wherever it may run. The relevant section of the Boundary Compact provides:
Point No. 1. The intersection of the boundary line common to California and Nevada and the centerline of the channel of the Colorado River as constructed by the U.S. Bureau of Reclamation, said point being common to the boundaries of Arizona, California, and Nevada, where the 35th degree of north latitude intersects the centerline of said channel; thence downstream along and with the centerline of saidchannel ... [to a railway bridge at Topock, which is south of the disputed property] (emphasis added).
Because the disputed property now lies on the east side of the Colorado River, the Boundary Compact positions the disputed property within Arizona for political purposes.3 Thus, the District Court erred in *1061holding that it lacked jurisdiction.4
Ill
Although the Boundary Compact did establish the boundary between Arizona and California for political purposes, it did not affect private property titles. United States v. Aranson, 696 F.2d 654, 657 (9th Cir.1983). Thus, we must proceed to the merits of the issue.
The district court erred in basing its analysis on prepatent avulsive movements of the Colorado River, specifically an 1857 avulsion. The proper analytical starting point is 1905, when the United States patented the disputed property to the State of California. It was not until 1905 that California held legal title to the disputed property. Therefore, river movements before 1905 are not relevant to fixing title.
The district court, and the private landowners, reason that title to the disputed property vested in 1850, when California became a state. It was also in that year that California obtained at least an inchoate or equitable right to the disputed lands under the Swamp and Overflowed Lands Act, 9 Stat. 519 (1850), codified at 43 U.S.C. § 982 (“Swamp Act”). The Swamp Act required the Secretary of the Interior to list and survey all swamp and overflowed lands and, upon request, to patent to the states all the unsold swamp and overflowed lands within their territorial limits. Swamp Act § 2; 43 U.S.C. § 983. See Tubbs v. Wilhoit, 138 U.S. 134, 136, 11 S.Ct. 279, 34 L.Ed. 887 (1891).
The United States Supreme Court addressed the issue of title to lands conveyed under the Swamp Act in a series of cases from the late nineteenth and early twentieth centuries. These cases make clear that in order to perfect the legal title to swamp and overflowed lands, those lands had to be identified and patented. In United States v. O’Donnell, 303 U.S. 501, 58 S.Ct. 708, 82 L.Ed. 980 (1938), for instance, the Court noted that the Swamp Act “invested] the state in praesenti with an inchoate title to those lands falling within the description of the act, to be perfected as of the date of the act when the land should be identified and the patent issued.” 5 Id. at 509, 58 S.Ct. 708. Similarly, in Joanna Little v. J.J. Williams, 231 *1062U.S. 335, 34 S.Ct. 68, 58 L.Ed. 256 (1913), the Court noted that it was only when the Secretary of the Interior identified and patented land granted under the Swamp Act that the fee-simple title vested in the state and the state’s title became perfect. Id. at 340, 34 S.Ct. 68. The Court had applied the same reasoning to several earlier cases. In Niles v. Cedar Point Club, 175 U.S. 300, 20 S.Ct. 124, 44 L.Ed. 171 (1899), the Court held that legal title of certain swamp land never passed to the state of Ohio, because the land was never patented to the state. Id. at 309, 20 S.Ct. 124. Brown v. Hitchcock, 173 U.S. 473, 19 S.Ct. 485, 43 L.Ed. 772 (1899), clearly announced that “[u]nder the swamp-land act, the legal title passes only on delivery of the patent.” Id. at 476, 19 S.Ct. 485. Michigan Land & Lumber Co. v. Rust, 168 U.S. 589, 18 S.Ct. 208, 42 L.Ed. 591 (1897), explained that “wherever the granting act specifically provides for the issue of a patent [as does the Swamp Act], then the rule is that the legal title remains in the government until the issue of the patent.” Id. at 593, 18 S.Ct. 208. Following this logic, we have held in a quiet title action that the patent date is “the controlling date” for determining a river’s position. United States v. 62.57 Acres of Land in Yuma County, Arizona, 449 F.2d 5, 8 (9th Cir.1971).6
In short, the district court’s analysis should have commenced with the patent date, not with pre 1905 avulsive river movements. Given the undisputed testimony that both accretion and avulsion occurred in the area after that date, a remand is required to permit the district court to conduct whatever proceedings it deems necessary for a re-analysis using the correct premise.
REVERSED AND REMANDED.
. Richard Bangs, Whitewater Adventure: Running America’s Great Rivers, 1 (1990).
. Sherrill v. McShan, 356 F.2d 607 (9th Cir.1966), is inapposite, because it was decided in February 1966, about six months before the Boundary Compact was enacted. In that case, the district court concluded, as in the present case, that a parcel of land on the east side of the Colorado River remained part of California, and dismissed the case for lack of jurisdiction. Id. at 608. On appeal, this Court affirmed, accepting the district court's determination that avulsive movements had placed the disputed lands in California, as well as the district court’s analysis of both states’ constitutions that led to the same re-*1061suit. Id. at 610. However, the Boundary Compact, once adopted, definitively established the political boundary between Arizona and California.
. Although the district court held that it lacked jurisdiction because the property is located within the political boundary of California, it proceeded to decide the merits. Arguably, the district court should not have attempted to resolve the title question, given that resolution of the jurisdictional issue did not necessarily depend on resolution of title. See Steel Co. v. Citizens for a Better Environment, 523 U.S. 83, 94-95, 118 S.Ct. 1003, 140 L.Ed.2d 210 (1998). However, given that the district court actually had jurisdiction under the Boundary Compact, any error in this respect was harmless.
. Appellees suggest that because perfected title to swamp lands is deemed effective as of the date of the Act, the appropriate starting date for the river analysis is 1850. We find this argument unpersuasive. The relation back doctrine may make the interest within the 1905 boundaries good as of 1850, but the doctrine cannot expand or contract the boundaries created by the 1905 patent. As the cases we cite in the text indicate, boundaries to swamp and overflow lands are determined exclusively by the process of identification and patent. Any other rule would create uncertainty among land owners, as all subsequent interests to land in swamp and overflow areas might fear that their property lines could be altered by earlier, unknown river movements. See Wright v. Roseberry, 121 U.S. 488, 505, 7 S.Ct. 985, 30 L.Ed. 1039 (1887) (stating that a patent is intended to be “an invaluable muniment of title, and a source of quiet and peace to its possessor.”).
. The private landowners rely on Arkansas v. Tennessee, 246 U.S. 158, 38 S.Ct. 301, 62 L.Ed. 638 (1918), a case involving the boundary line between the two states, for the proposition that river movement analysis begins on the date on which a boundary line between two states is established (in the present case, the boundary was created in 1850, when California became a state). However, Arkansas v. Tennessee is not relevant to the present case, as it concerned the location of a state boundary rather than private title claims.