While James L. Wiggs was stopped for a traffic infraction, a drug-detection dog alerted on his vehicle. Deputies searched the vehicle and found cocaine inside. Wiggs entered a plea to the charge of possession of cocaine with intent to sell or deliver while reserving the right to appeal the denial of his dispositive motion to suppress. We reverse because the dog’s alert on Wiggs’ vehicle did not provide probable cause to search.
Just before midnight on August 14, 2007, a Sarasota County deputy stopped Wiggs’ vehicle for running a red light on U.S. 301. The deputy detained Wiggs in order to prepare a warning citation. During the detention, a drug-detection dog named Zuul alerted to Wiggs’ vehicle. A search of the vehicle revealed the cocaine that forms the basis for the charge in this case.
Wiggs filed a motion to suppress the cocaine in which he argued that Zuul’s alert did not provide probable cause to search his vehicle. Wiggs challenged Zuul’s reliability and cited evidence of numerous “false alerts” by the dog in the field. At the hearing on Wiggs’ motion to suppress, the State presented testimony and documents regarding Zuul’s training and field record.
The primary witness called by the State was Zuul’s handler, Deputy Indico. Deputy Indico testified that Zuul was acquired from a German Shepherd breeder in Hungary; the dog was selected based on his aggressiveness and motivation. Deputy Indico began training Zuul in February 2007. Deputy Indico and Zuul completed an eighty-hour narcotics training course offered by the Sarasota Sheriffs Office as well as a 400-hour patrol course. In addition, Deputy Indico and Zuul were certified by the Florida Department of Law Enforcement (FDLE) and National Police *156Canine Association (NPCA). Deputy Indi-co and Zuul’s training and certification process was completed on April 27, 2007.
Zuul lives with Deputy Indico in order to build a bond between them. The dog was trained to detect different scents of narcotics by placing the drug on top of a bean bag so the bag absorbed the scent. Then the bean bag was introduced to the dog in the form of a hide-and-seek game in which the dog used his nose to locate the bean bag. Eventually, the dog was introduced to the actual narcotic. Zuul was trained to detect marijuana, cocaine, methamphetamine, and heroin.
Zuul was trained as an aggressive alert dog, which means he scratches as his final response. When Zuul picks up a scent he snaps his head around and starts to work toward the scent. He begins sniffing harder, and his breathing then becomes louder, shallower, and quicker. Zuul’s body becomes more rigid, and he leans forward. There is an overall change in his demeanor until the final scratch response.
Zuul was trained on blank vehicles and rooms in a controlled environment to ensure he was not falsely alerting. The sheriffs trainers varied the amount of narcotic from .1 grams to over 100 grams. These trainers also used distracters like food, tennis balls, clothing, or anything commonly used or found in a vehicle. The NPCA trainers varied the amount of narcotic from eight to twenty-eight grams. Zuul did not falsely alert to any blank vehicles or rooms during any of his training.
To obtain NPCA certification Deputy Indico and Zuul had to meet specific NPCA training standards, which were admitted in evidence, and achieve 75 percent accuracy. The pair had to find narcotics in two out of four vehicles and from two out of three rooms within a building. The FDLE certification was for apprehension, tracking, and building searches but did not include any narcotics detection training.
After graduation Zuul and Deputy Indi-co continued to train on a weekly basis. They have been certified every year by the requirements set forth from the NPCA training standards. Deputy Indico kept monthly training and scent detection logs which the State introduced in evidence. The logs covered Zuul’s initial training in February 2007 as well as his weekly training up until Wiggs’ stop on August 14, 2007.
Deputy Indico also kept a monthly report of Zuul’s field activity from April 2007 until August 2007, which the State introduced in evidence. On the positive vehicle alerts that did not result in a drug find, or unverified alerts, Deputy Indico documented any history the vehicle or its passengers had with drugs. Deputy Indico acknowledged that Zuul had not been trained to refrain from alerting to residual odors. Thus, it was important for Deputy Indico to document the history of the driver and vehicle on unverified alerts. Deputy Indico explained that he obtained the histories by interviewing the driver or passengers. If, for example, the driver told the deputy that he had just picked up his brother and his brother was around people smoking marijuana, then Deputy Indico considered the unverified alert a positive alert. Deputy Indico did not document the details of the alleged drug histories.
Deputy Indico began using Zuul to search for narcotics during vehicle stops on May 14, 2007. During the first stop Zuul alerted and a marijuana pipe was discovered. The next day, May 15, Zuul alerted on another vehicle, but nothing was found. Deputy Indico documented that the passenger admitted using cocaine at some time before the stop. On May 17, 2007, Zuul alerted to a vehicle, but no *157narcotics were found. The deputy documented that the vehicle had a “narcotics history.” Similarly, Zuul alerted on May 18, 2007, and nothing was found, but the driver admitted to smoking marijuana. On June 2 and 9 Zuul sniffed vehicles for drugs but did not alert.
On June 17, 2007, Zuul alerted at the driver’s door of another vehicle but no drug that Zuul was trained to detect was found.1 This time Deputy Indico documented that the passengers admitted to smoking marijuana. On June 29 Zuul alerted on the driver’s side door of a vehicle but there was no find. According to Deputy Indico, the vehicle and owner had a narcotics history.
On July 2, 2007, Zuul sniffed a vehicle for drugs but did not alert. On July 8 and July 9 Zuul alerted on vehicles, and narcotics were discovered. But on July 24, 25, and 30 Zuul alerted on vehicles in which no drugs were found. Similarly, on August 14 Zuul alerted on two different vehicles, but no drugs were discovered. In four of the five preceding instances, Deputy Indico documented that either an occupant of the vehicle had smoked marijuana or the vehicle had some type of narcotics history. In the fifth instance, Deputy Indico testified that he smelled the odor of burnt marijuana. Finally, on August 15, 2007,2 Zuul alerted to Wiggs’ vehicle and drugs were found inside.
In summary, Deputy Indico and Zuul had been summoned to seventeen vehicle stops between May and August of 2007. Ten of these encounters resulted in Zuul alerting on the vehicle with no discovery of drugs. Four post-alert vehicle searches, including Wiggs’, resulted in the discovery of drugs. And three encounters ended with no alert.
The specific question before this court is whether Zuul’s alert to the exterior of Wiggs’ vehicle provided probable cause to support a warrantless search of the vehicle’s interior. The framework for analyzing this issue has recently been set forth by the Florida Supreme Court in Harris v. State, 71 So.3d 756 (Fla.2011). In Harris, a drug-detection dog alerted to the driver’s door handle of the defendant’s vehicle during a traffic stop. Id. at 759-60. A search of the vehicle revealed over 200 pseu-doephedrine pills, eight boxes of matches, and muriatic acid; these items are commonly used to create methamphetamine. The State charged the defendant with possession of pseudoephedrine with the intent to use it to manufacture methamphetamine. The defendant entered a plea to the charge while reserving the right to appeal the denial of his dispositive motion to suppress. Id. at 762. The First District affirmed, and the supreme court accepted discretionary review of the case based on express conflict between the districts.
The question before the court “concern[ed] the evidence that the State must introduce to establish that probable cause existed for the warrantless search of a vehicle based on a drug-detection dog’s alert to the vehicle.” Id. In addressing this question, the court resolved a conflict among the district courts. The First, Fourth, and Fifth Districts had held that the State could establish a prima facie case of probable cause by providing evidence of the drug-detection dog’s proper training and certification. Id. at 762 (citing Harris v. State, 989 So.2d 1214, 1215 (Fla. 1st *158DCA 2008); State v. Laveroni, 910 So.2d 333, 336 (Fla. 4th DCA 2005); State v. Coleman, 911 So.2d 259, 261 (Fla. 5th DCA 2005)). The prima facie case of probable cause could be rebutted by the defendant’s presentation of evidence of the dog’s unreliability. Id. at 763. However, the Second District had held that the fact of a dog’s training and certification, in itself, was not sufficient to establish probable cause. Id. (citing Gibson v. State, 968 So.2d 631, 631 (Fla. 2d DCA 2007); Matheson v. State, 870 So.2d 8, 14 (Fla. 2d DCA 2003)).
The supreme court agreed with the Second District’s conclusion. Id. The court explained that “when a dog alerts, the fact that the dog has been trained and certified is simply not enough to establish probable cause to search the interior of the vehicle and the person.” Id. at 767. The court reasoned that the certification and training of drug-detection dogs was not subject to a uniform statewide or nationwide standard. Id. at 767-68. Additionally, the fact of the dog’s training and certification did not account for the possibility of false alerts, handler error, and alerts to residual odors. Finally, allowing the fact that a dog has been trained and certified to provide a prima facie case of probable cause would improperly place on the defendant the burden of production of evidence solely within the control of law enforcement. Id.
Thus, the supreme court adopted a “totality of the circumstances approach” that places the burden of producing evidence to establish the dog’s reliability on the State. Id. at 771.
The State’s presentation of evidence that the dog is properly trained and certified is the beginning of the analysis. Because there is no uniform standard for training and certification of drug-detection dogs, the State must explain the training and certification so that the trial court can evaluate how well the dog is trained and whether the dog falsely alerts in training (and, if so, the percentage of false alerts). Further, the State should keep and present records of the dog’s performance in the field, including the dog’s successes (alerts where contraband that the dog was trained to detect was found) and failures (“unverified” alerts where no contraband that the dog was trained to detect was found). The State then has the opportunity to present evidence explaining the significance of any unverified alerts, as well as the dog’s ability to detect or distinguish residual odors. Finally, the State must present evidence of the experience and training of the officer handling the dog. Under a totality of the circumstances analysis, the court can then consider all of the presented evidence and evaluate the dog’s reliability.
Id. The court placed special emphasis on the dog’s field records, explaining that a dog’s alert to residual odor, though different from a false alert, may not indicate that drugs are actually present at the time. Id. at 769. The court rejected the State’s argument that field records are meaningless because dogs do not distinguish between the odor of drugs that are present and residual odors. Instead, the court determined that evidence explaining unverified alerts would allow the trial court “to evaluate how any inability to distinguish between residual odors and drugs that are actually present bears on the reliability of the alert in establishing probable cause.” Id.
In applying the totality of the circumstances test to the facts of that case, the supreme court held that the trial court erred in determining that the State established probable cause based on the dog’s alert. Id. at 770. The court noted that, *159while the dog had a “ ‘really good’ ” success rate during training, the State had not established whether the dog had any false alerts. Id. Additionally, the court noted the absence of evidence regarding the details of the dog’s training, such as whether the handler knew the location of the drugs and whether a variety of environments and distractions were used during training. The State also failed to present evidence setting forth the criteria for the dog’s certifications.
Moreover, the State failed to present any quantification of the dog’s success in the field because it did not introduce the dog’s field performance records. Id. at n. 12. Without such evidence, the State could not explain the significance of the dog’s unverified field alerts. Id. Similarly, the State failed to present evidence regarding the dog’s ability to detect residual odors and how the dog’s alert to a residual odor on the door handle provided a fair probability that drugs would be found inside the defendant’s vehicle.
The court concluded that, because the State failed to establish a reliable alert, it failed to establish probable cause. Id. at 774-75. Therefore, the trial court erred in denying the defendant’s motion to suppress. Id. at 775. The supreme court thus reversed and remanded for further proceedings.
The totality of the circumstances analysis leads to the same conclusion in this case. Here, the State presented evidence of Deputy Indico and Zuul’s national training and certification which included the NPCA training standards that .were met. The State established that Zuul had to be at least 75 percent accurate in his certification training and never alerted on any blank vehicles during training. Through Deputy Indico, the State provided details regarding Zuul’s police training, including discussion of the various environments and distractions.
The State also produced Zuul’s field performance records. However, the dog’s field performance records were problematic. Zuul had conducted seventeen vehicle sniffs in the field and alerted fourteen times. Drugs were only found after four of those fourteen alerts. Based solely on the number of sniffs in which Zuul’s alerts uncovered narcotics, Zuul’s field accuracy rate is four out of fourteen, or approximately 29 percent. This accuracy rate is clearly insufficient to establish reliability, that is, a fair probability that drugs would be found in a vehicle following an alert.
The State asserts that Zuul’s field accuracy rate is actually 100 percent because Deputy Indico had documented some type of narcotics history associated with each vehicle on which Zuul alerted but in which no drugs were found. While Deputy Indi-co’s explanation of these unverified alerts is relevant to the issue of whether an alert is a false alert or an alert to a residual odor, Harris, 71 So.3d at 769-70, the explanations were not specific enough to establish the existence of residual odors on which Zuul should have alerted.
In five of the unverified alerts, the deputy merely testified that the vehicle had a “drug history.” However, the deputy did not explain what this history entailed or why it ensured that drugs had once been present in those vehicles. In four of the unverified alerts, Deputy Indico testified that someone in each vehicle had used narcotics recently. However, there was no testimony regarding how long before the stop the drugs had been used or how residual an odor Zuul could be expected to detect. The absence of this information leaves us unable to evaluate the significance of these unverified alerts. The evidence simply does not explain why the alerts to those residual odors would give *160rise to probable cause to search Wiggs’ vehicle. Thus, we are not inclined to consider these nine unverified alerts in Zuul’s field accuracy rate.
For the last unverified alert, the deputy testified that he actually smelled the odor of burnt marijuana in the vehicle. Adding this alert to Zuul’s positive alert calculation makes Zuul’s accuracy rate five out of fourteen, or approximately 36 percent. We also do not find this rate sufficient to establish reliability.
Another omission that is problematic in this case is the State’s failure to set forth the details of Zuul’s alert on Wiggs’ vehicle. No evidence was presented about the nature of the alert, the search of Wiggs’ vehicle, or the location of cocaine therein. Thus, it is impossible to tell if Zuul alerted on a residual odor, as did the dog in Harris, or whether he alerted on the actual cocaine itself. The State has failed to make a connection between Zuul’s alert and the discovery of the drugs in this case.
Under the totality of these circumstances, the trial court erred in determining that the State established probable cause based on Zuul’s alert on Wiggs’ vehicle. We acknowledge that the State presented sufficient evidence regarding Deputy Indico and Zuul’s training and certification and the meaning of the training and certification. However, we are unconvinced of the sufficiency of the State’s evidence of Zuul’s field performance records as it relates to his alerts on residual odors. And the State’s lack of evidence regarding Zuul’s alert made it impossible for the trial court to complete a totality of the circumstances analysis of the reliability of Zuul’s alert as a basis for probable cause to search Wiggs’ vehicle. The evidence did not establish probable cause to believe that contraband actually would be found in the vehicle after Zuul’s alert. Accordingly, the trial court erred in denying Wiggs’ motion to suppress, and we reverse.
Reversed.
LaROSE, J., Concurs. ALTENBERND, J., Concurs specially.. The police did find methadone during the search of the vehicle.
. While the vehicle stop occurred shortly before midnight on August 14, 2007, Deputy Indico and Zuul conducted the sniff on August 15, 2007.