delivered a dissenting opinion, in which CHIEF JUSTICE HECHT, JUSTICE GREEN, and JUSTICE BOYD joined.
This inverse condemnation case concerns whether homeowners raised a fact *814question as to the elements of a taking. The homeowners blame their county and flood control district for approving new development without mitigating resulting runoff and drainage issues, causing their homes to flood. Of course, government entities do not have a duty to prevent all flooding. But, because the homeowners here presented evidence that the government entities knew unmitigated development would lead to flooding, that they approved development without appropriately mitigating it, and that this caused the flooding, I believe that the homeowners have raised a fact issue as to their takings claim. Because the Court concludes otherwise, I respectfully dissent.
I. Background and Procedural History
The plaintiffs in this case are more than 400 residents and homeowners in the upper White Oak Bayou watershed in Harris County. Their homes were built mostly in the mid-to-late 1970s and early 1980s. Despite a history of flooding in the area, they initially suffered little to no flood damage. This changed, however, when Tropical Storm Francis in 1998, Tropical Storm Allison in 2001, and an unnamed storm in 2002 flooded their homes one or more times. The homeowners blame Harris County and Harris County Flood Control District, asserting that they approved new upstream development without implementing appropriate flood-control measures, all the while being substantially certain flooding would result.
The government entities have long known that expanding development in the watershed could cause flooding. In 1976, the U.S, Army Corps of Engineers prepared a report on the upper White Oak Bayou. The report noted recurrent damaging floods, which it attributed primarily to “inadequate channel capacities of the streams.” This problem was “compounded by the continuing increases in suburban development which reduces the infiltration of rainfall and increases and accelerates runoff to the streams.” Inadequate street drainage and storm sewers also caused severe localized flooding. The report predicted: “Additional residential development is expected to occur with or without an adequate plan for controlling the floods. Although current local regulations require that new structures be built above the level of the 100-year flood, damages will increase substantially in the future with increased rainfall runoff rates.” Accordingly, the Corps proposed channel improvements and other changes to reduce flooding caused by existing and future development. The plan was to be funded primarily by the federal government. -
The government entities concurred with the Corps’ findings and agreed to facilitate the project. Their ostensible goal following the Corps’ report was to maintain or reduce the bayou’s 100-year flood plain. But federal funding was slow to materialize, even as thousands of acres were developed in the bayou’s watershed and the County continued approving more. Construction in the 100-year flood plain had been prohibited since the early 1970s, meaning that almost none of the new development or the plaintiffs’ homes were in the 100-year flood plain when built or approved.
The delay in federal funding for the Corps’ plan led County officials to develop their own flood-control plan. The District had already begun requiring new developments in the upper part of the watershed to provide on-site detention ponds, though the parties disagree on whether the District deviated from this policy. The government entities commissioned Pate Engineers to develop a plan. The “Pate Plan,” adopted by the County in 1984, proposed to eliminate flooding along the upper1 bay*815ou for 100-year flood events. Like the Corps’ plan, the Pate Plan called for channel improvements along the upper portion of the bayou (some near the plaintiffs’ homes). But it also called for more, such as for regional detention ponds to mitigate continuing development in the upper watershed. The plan was to be funded through local taxes and impact fees, but it was never fully implemented.
A 1989 flood led residents to • inquire about the flood-control measures. The District’s director assured thejn of plans to protect their property from. 100-year floods. Yet the 1989 flood revealed flaws in the Pate Plan’s engineering analysis, and the District commissioned another report on the upper White Oak Bayou watershed. Klotz Associates presented their findings in the early 1990s, concluding that the Pate Plan seriously underestimated the bayou’s flood flows and levels. Though the Klotz Plan’s features were more extensive in some regards than the Pate Plan, the Klotz Plan was modeled around containing 10-year (as opposed to 100-year) flood events. The government entities adopted these changes.
The new information led FEMA to update the bayou’s flood-plain maps in the 1990s. Revisions exposed an expanding flood plain, encompassing more and more of the plaintiffs’ homes. According to the homeowners, by 1999, all of their homes were within the 100-year flood plain—■ something the government entities do not dispute here. Indeed, from 1998 to 2002, most of their homes were inundated in three successive floods.
The homeowners filed an inverse condemnation suit. The government entities responded with a combined plea to the jurisdiction and motion for summary.judgment, contending that no • genuine issue of material fact- had been raised on the elements of .the takings claim. The trial court denied the motion, and the court of appeals affirmed the denial of the plea to the jurisdiction. 445 S.W.3d 242, 247 (Tex. App.-Houston [1st Dist.] 2013).
II.< Analysis
Article I, Section 17 of the Texas Constitution provides:
No person’s property shall be taken, damaged or destroyed for or applied to public use without adequate compensation being made, unless by the consent of such person_.
Tex. Const, art. I, § 17(a).
Those seeking' recovery for a taking must prove the government “intentionally took or damaged their property for public use, or was substantially certain that would be the result.” City of Keller v. Wilson, 168 S.W.3d 802, 808 (Tex.2005). Sovereign immunity does not shield the government from liability for compensation under the takings clause. Gen. Servs. Comm’n v. Little-Tex Insulation Co., 39 S.W.3d 591, 598 (Tex.2001).1 To defeat the government entities’ plea to the jurisdiction, the homeowners need only raise a fact issue as to each element of their claim. See Tex. Dep’t of Parks & Wildlife v. Miranda, 133 S.W.3d 217, 228 (Tex.2004) (noting that plea to the jurisdiction procedure “mirrors” our summary judgment practice). That is, the homeowners must raise a fact issue as to-intent, causation, and public use. Little-Tex Insulation Co., 39 S.W.3d at 598. While determining whether they have- met this burden, we *816take as true all evidence favorable to the nonmovant, indulging every reasonable inference and resolving any doubts in the nonmovant’s favor. Miranda, 133 S.W.3d at 228.
A. Intent
In a takings case, “the requisite intent is present when a governmental entity knows that a specific act is causing identifiable harm or knows that the harm is substantially certain to result.” Tarrant Reg’l Water Dist. v. Gragg, 151 S.W.3d 546, 555 (Tex.2004). It is not enough that the act causing the harm be intentional—there must also be knowledge to a substantial certainty that the harm will occur. City of Dallas v. Jennings, 142 S.W.3d 310, 313-14 (Tex.2004). Intent, in takings cases as in other contexts, may be proven by circumstantial evidence. See Spoljaric v. Percival Tours, Inc., 708 S.W.2d 432, 435 (Tex.1986). But a taking cannot rest on the mere negligence of the government. City of Tyler v. Likes, 962 S.W.2d 489, 505 (Tex.1997). The issue here is whether the homeowners’ circumstantial evidence raises a fact question as to the government entities’ knowledge that “harm [was] substantially certain to result” to the homeowners’ homes because of the entities’ actions. See Gragg, 151 S.W.3d at 555.
The homeowners do not argue that the government entities have a general legal duty to prevent all flooding. Instead, they urge that the entities approved private development in the White Oak Bayou watershed without mitigating its consequences, being substantially certain the unmitigated development would bring flooding with it. The government entities respond that they did not intend the flooding; indeed, the District’s very purpose is to plan for storm-water runoff and control flooding. They assert that they cannot stop all flooding, and to the extent flooding does occur, their intent is to reduce or prevent it, not cause it.
The summary judgment evidence shows a fact question exists regarding whether the government entities were substantially certain their actions in approving development without appropriately mitigating it would cause the plaintiffs’ homes to flood. The Corps’ report, Pate Plan, and Klotz report confirm that the entities have known for several decades that development in the bayou’s watershed exacerbates flooding in the upper bayou. The homeowners point to evidence that the County approved development without the previously required on-site detention.2 The Klotz report led the entities to plan for 10-*817year flood events rather than 100-year floods as called for by the Pate Plan. And the homeowners’ expert witness, Dr. Larry Mays, opined that when the entities approved development, knowing that surface runoff would increase if the development was not mitigated, and then planned to contain only 10-year floods, the entities were substantially certain that flooding would result.
The government entities respond that their “knowledge must be determined as of the time [they] acted, not with benefit of hindsight.” City of San Antonio v. Pollock, 284 S.W.3d 809, 821 (Tex.2009). The entities urge that they changed from the Pate Plan to the Klotz Plan because the Pate Plan was flawed. There is no evidence they knew of the flaws at the time they adopted it or approved most of the development. In other words, no evidence exists that at the time they approved most of the development at issue, they were substantially certain the flood-control measures would fail and flooding would result. When the 1989 flood revealed the Pate Plan’s flaws, the entities followed their engineers’ advice and adopted the Klotz Plan. Since then, they have spent millions of dollars on drainage and infrastructure that, according to their experts, goes far beyond the Pate Plan, and no evidence exists that the Pate Plan would have stopped the floods the homeowners complain about. Michael David Talbott, an expert witness for the entities, explained that the District’s knowledge of the watershed “changed significantly over time” and continued to change “with the advent of new technologies and with observations of actual events.” The changing flood plains and changing plans were not premeditated; they resulted from new data and knowledge.
This competing evidence demonstrates, in the words of the court of appeals below, “the factual complexity surrounding this issue and the difficulty of pinpointing precisely what [the government entities] knew about flooding in the upper watershed and when they knew it.” 445 S.W.3d at 258. Evidence exists that the entities approved development without appropriately mitigating it, and that at times they deviated from their early policy of requiring on-site detention ponds. There is also evidence that they abandoned plans protecting against 100-year floods for plans targeting only 10-year floods. Though none of the evidence on its own might raise a fact question, together it raises a question of fact as to the entities’ intent to take the homeowners’ property to facilitate new development without appropriate mitigation.
The homeowners also submit that the recurring nature of these floods is probative of the government entities’ intent and the extent of the taking. See Gragg, 151 S.W.3d at 555 (“In the case of flood-water impacts, recurrence is a probative factor in determining the extent of the taking and whether it is necessarily incident to authorized government activity, and therefore substantially certain to occur.”); see also Kopplow Dev., Inc. v. City of San Antonio, 399 S.W.3d 532, 537 (Tex.2013) (“With flood water impacts, recurrence is.a probative factor in assessing intent and the extent of the taking.”). Here, the entities knew that flooding occurred frequently in the bayou, and they originally planned for 100-year events. Yet there is evidence that they later planned only for 10-year events. The fact that storms sometimes surpass 10-year events is no surprise to the entities. The recurring nature of the flood events is evidence of intent and the extent of the taking.
The government entities invoke City of Keller v. Wilson, where the Court held that a city could rely on engineers’ conclusions that a drainage ditch was unneces*818sary, even- though, as it turned out, the ditch might have stopped flooding. See 168 S.W.3d at 829. Here, the entities did receive engineers’ advice. Yet the entities were also aware that the Klotz Plan would protect against 10-year events, whereas the prior plan had been to protect against 100-year floods. There is evidence the entities knew (based, in part, on engineers’ reports) that unmitigated development would lead to flooding. Unlike in, City of Keller, the engineers’ recommendations here informed the entities of the danger.
In sum, the homeowners have raised a fact question regarding intent. This case is unlike others where there was no “evidence that the act ... was substantially certain to lead to such damage.” Jennings, 142 S.W.3d at 315.
B. Causation
To prevail, the homeowners must also raise a fact issue as to whether the government entities’ actions “resulted in a ‘taking’ of property.” Little-Tex Insulation Co., 39 S.W.3d at 598. The entities assert that the homeowners failed to do so. -I disagree. • ,
The homeowners’ expert, Dr. Mays, performed a study that raises a fact question regarding whether the government entities’ approval of development and subsequent failure to properly mitigate it caused the homes to flood. Dr. Mays investigated the effects of unmitigated development on the watershed, performing his “Effect of Urbanization on Peak Discharges” analysis to demonstrate the difference between 1982 land use conditions and 1990 land use conditions. He compared peak discharges and runoff hydrographs from conditions in 1982 to peak discharges and runoff hydro-graphs from conditions in 1998, concluding that the primary cause of flooding along White Oak Bayou was development “without the implementation of proper storm-water management measures (such as detention and/or retention).” This caused “significant increases in flows and corresponding water surface elevations for an approximate 10-year flood event.” Dr. Mays determined that two of the three flood events between 1998 and 2002 were 10-year flood events. He concluded: “But for’-the actions by the County of approving this unmitigated development, and/or by not fully implementing the Pate .Plan (which would have eliminated the risk of flooding along the bayou for up to a 100-year event), the Plaintiffs’ properties would not have flooded during these three flood events.”
The- government entities point to purported “analytical gaps” in Dr. Mays’ study. For example, the entities point out that Dr. Mays freely admitted that he never modeled or tested the Pate Plan to show that it would have stopped the flooding. Without proving there was a way to stop the flooding, Dr. Mays could not prove the entities caused it. The entities present evidence that the expanding flood plain was caused (at least in part) by new knowledge about existing conditions as opposed to changes in conditions. The entities assert that intense, unexpected rainfall—not their own actions—caused the floods, observing that flooding occurred all over Harris County during these - storms, not just in the White Oak Bayou watershed. They also challenge Dr.- Mays’ method -of determining the size of the storms, arguing that they were larger.
I believe the homeowners have raised a fact issue regarding whether the entities’ actions caused the flooding. The homeowners presented evidence'that the Klotz Plan targeted 10-year events, whereas the Pate Plan had aspired to contain 100-year floods. Dr. Mays presented data supporting his conclusion that the flooding was caused by water exceeding the bounds of *819the bayou.rather than by the pooling of water in the homeowners’ yards. Regarding the size of the flood events, the entities’ own expert, Andy Yung, created ,a fact issue. In an affidavit, Yung stated that the rainfall for the 2001 storm exceeded a 100-year event. But in a report, he concluded that the 1998 and 2001 floods were about 50-year events and the “magnitude of the 2002 flood event was considerably less.” He then concluded that “in all three of these events, areas were inundated which have been historically identified in the 100-year flood plain.” In other words, the entities’ own expert provides evidence that the floods were less than 100-year events, but that the flooding extended into areas not anticipated even for 100-year events. A fact question exists regarding the cause of the floods.
In sum, several unresolved fact questions exist. For example, the parties dispute the nature and degree of the three flood events. The government entities have contended that two of the storms produced unprecedented rainfall, yet evidence exists that these storms were within the parameters planned for by the Pate Plan and the Corps’ study. The parties also' dispute the degree to which unmitigated development contributed to the 100-year flood plain’s expansion, the relative benefits and shortcomings of the Pate and Klotz Plans, and the entities’ ability to manage and control development in the bayou’s watershed, as well as the entities’ efforts to mitigate the effects of new development. Experts have weighed in on both sides of the debate, but the questions remain unsettled, and the merits of these homeowners’ claims depend on their resolution.
C. Public Use
The government entities assert that no evidence- exists of a taking for public use. They argue that property is only “damaged for public use” if the “governmental entity is aware that its action will necessarily cause physical damage to certain private property, and yet determines that the benefit to the public outweighs the harm caused to that property.” Jennings, 142 S.W.3d at 314. The entities’ arguments on this point, however, merely repeat their arguments concerning intent and causation: the entities made no conscious decision to sacrifice -the plaintiffs’ property for public use, and their conduct did not cause the flooding, so it was not for public use. In response, the homeowners contend the' entities approved development and made drainage-plan decisions for the sake of the public.-'
At least some - evidence exists that in approving new development and drainage plans causing flooding, the entities’ were acting for a public use. To the extent the government entities were substantially certain the homeowners’ homes would flood because of unmitigated development, but sacrificed their homes for the sake of new development, this'was for a public use. If the homeowners bore the burden of the growing community; ■ then, “in all fairness and justice, [the burden] should be borne by the' public as a whole.” Gragg, 151 S.W.3d at 554.
III. Conclusion
A fact question exists as to each element of the homeowners’ takings claim. To conclude that there is a question of fact is not to hold that government entities have a duty to prevent all flooding. Rather, it means the plaintiffs have, presented at least some evidence of their takings claim, and a jury should evaluate the evidence to reach a conclusion. In other words, the government entities’ plea to the jurisdic*820tion should be denied. Because the Court holds otherwise, I respectfully dissent.
. For convenience, I use "sovereign immunity” throughout this opinion to refer to the related doctrines of sovereign immunity and governmental immunity. See City of Houston v. Williams, 353 S.W.3d 128, 134 n.5 (Tex. 2011) (explaining that governmental immunity protects political subdivisions of the State).
. The government entities note that the homeowners’ pleadings specifically blamed flooding on development "[sjince 1984.” The pleadings stated that prior to 1984, the District “required all new development projects to incorporate on-site storm water detention into their development plans.” The entities assert that the homeowners judicially admitted that on-site detention accompanied all pre-1984 development. A judicial admission, however, must be clear and unequivocal. See Horizon/CMS Healthcare Corp. v. Auld, 34 S.W.3d 887, 905 (Tex.2000). A possible interpretation of the homeowners’ pleadings is that the District had a requirement of on-site detention, but this says nothing about the District’s adherence to that requirement. Indeed, the homeowners’ response to the entities’ plea to the jurisdiction and summary judgment motion specifically claimed that "Defendants continued to approve new development without on-site mitigation even after this 1976 study had clearly shown them that the bayou did not have the capacity to handle any additional storm water runoff.” As evidence that some pre-1984 development was not mitigated, the homeowners rely on a table supplied by the entities' expert, Melvin Spinks. The parties interpret the table differently, and the table itself is not clear on its face, but the homeowners’ pleadings do not contain an unequivocal admission on the subject.