KLC Farm v. Perdue

IN THE UNITED STATES DISTRICT COURT FOR THE DISTRICT OF KANSAS KLC FARM and GARETSON BROTHERS, Plaintiffs, v. Case No. 17-2484-DDC-KGG SONNY PERDUE, Secretary of the United States Department of Agriculture, et al., Defendants. _____________________________________ MEMORANDUM AND ORDER Plaintiffs KLC Farm, a Kansas partnership (“KLC Farm”), and Garetson Brothers, a Kansas partnership (“Garetson Brothers”) seek judicial review under the Administrative Procedure Act (“APA”), 5 U.S.C. § 706 . Plaintiffs filed a Complaint for Judicial Review and Declaratory Relief (Doc. 1) against defendants Sonny Perdue, the Secretary of the United States Department of Agriculture (“USDA”), Steven C. Silverman, the Director of the National Appeals Division of the USDA (“NAD”), and Denny Skiles, the Director of the Farm Service Agency (“FSA”). The parties have fully briefed the requested issues for review. See Docs. 20, 23, 24. Plaintiffs ask the court to enter a declaratory judgment finding that the NAD Director’s determinations upholding FSA’s calculations of plaintiffs’ payments under a financial assistance program “violated [p]laintiffs’ rights to due process of law, were not in accordance with governing statutes and regulations, were unsupported by substantial evidence, and were arbitrary and capricious.” Doc. 1 at 8. After reviewing the administrative record and considering the parties’ arguments, the court affirms the NAD Director’s decisions. The court explains why, below. I. Background Plaintiffs are corn producers and each run a farming operation in Haskell County, Kansas.1 Doc. 1 at 3 (Compl. ¶ 11); Doc. 14-6 at 1; Doc. 18-3 at 26. In March 2015, plaintiffs elected to enroll in the Agriculture Risk Coverage—County Option (“ARC-CO”) program established by the 2014 Farm Bill. That Bill provides financial assistance to producers during periods of market downturns by reimbursing a portion of farmers’ out-of-pocket losses. Doc. 1 at 3 (Compl. ¶¶ 10–11); Doc. 14-6 at 1; Doc. 18-3 at 26. Plaintiffs executed their ARC-CO contracts in September of 2015. Doc. 1 at 3 (Compl. ¶ 10); Doc. 14-6 at 1; Doc. 18-3 at 26. This request for judicial review centers on FSA’s calculation of plaintiffs’ ARC-CO payments for the 2014 crop year.2 A. ARC-CO Statutes and Regulations Participants in the ARC-CO program receive payments for a crop year when the actual crop revenue for the applicable county falls below the agriculture risk coverage guarantee. 7 U.S.C. § 9017 (a).3 “Actual crop revenue” is calculated by multiplying (A) “the actual average county yield per planted acre for the covered commodity, as determined by the Secretary” of the USDA times (B) the applicable price information determined under the statute. 7 U.S.C. § 1 Haskell County is located in southwest Kansas. The county seat is Sublette, which is located between Garden City and Liberal. 2 Producers could sign up retroactively for the 2014 crop year when they enrolled in 2015. See Doc. 14-6 at 1 n.4; Doc. 18-3 at 26 n. 4. 3 Citations to 7 U.S.C. § 9017 in this Order are to the § 9017 in effect from February 7, 2014 to December 19, 2018. 9017(b)(1) (bold text and emphasis added); see also 7 C.F.R. § 1412.3.4 The “agriculture risk coverage guarantee” is 86% of benchmark revenue. 7 U.S.C. § 9017 (c)(1) (bold text added); see also 7 C.F.R. § 1412.3 . “Benchmark revenue” is determined by multiplying (A) “the average historical county yield as determined by the Secretary [of the USDA] for the most recent 5 crop years, excluding each of the crop years with the highest and lowest yields” times (B) “the national average market price received by producers during the 12-month marketing year for the most recent 5 crop years, excluding each of the crop years with the highest and lowest prices.” 7 U.S.C. § 9017 (c)(2) (bold text and emphasis added). To calculate each participant’s ARC-CO payments, FSA multiples the producer’s payment acres by the lesser of (i) the difference between the actual crop revenue and the agriculture risk coverage guarantee or (ii) 10% of the benchmark revenue. See 7 U.S.C. § 9017 (d)–(e); 7 C.F.R. § 1412.3 . The dispute here centers around FSA’s calculation of the actual average county yield and the average historical county yield for the 2014 crop year in Haskell County. The average historical county yield is also referred to as the benchmark yield. Under the applicable regulations, the “[a]ctual average county yield is calculated as the crop year production of a covered commodity in the county divided by the commodity’s total planted acres for a crop year in the county, as determined by FSA.” 7 C.F.R. § 1412.3 (bold text and emphasis added). Under plaintiffs’ ARC-CO contracts, planted acres for corn include harvested acres plus unharvested acres. Doc. 14-2 at 1 and Doc. 18-1 at 26 (appendices to ARC-CO contracts).5 And, the 4 Citations to 7 C.F.R. § 1412.3 in this Order are to the § 1412.3 in effect from December 15, 2014 to August 15, 2018. 5 Though not in the regulations when plaintiffs signed their contracts, the current 7 C.F.R. § 1412.3 now specifies that planted acres for corn “are the harvested acres plus unharvested acres,” consistent with the 2015 ARCCO contracts. benchmark yield is “the 5-year Olympic average[6] of actual average county yields for the most recent 5 years.” 7 C.F.R. § 1412.3 (emphasis added). As explained in Part I.B. below, plaintiffs argue that FSA published a handbook further defining how it would calculate these yields. Plaintiffs assert FSA impermissibly failed to follow its handbook procedures and, instead, calculated the 2014 yields in an arbitrary and capricious fashion. Also, under 7 U.S.C. § 9017 (g)(1), the Secretary of the USDA must “to the maximum extent practicable, use all available information and analysis, including data mining, to check for anomalies in the determination of agriculture risk coverage payments.” And, under 7 U.S.C. § 9017 (g)(4), if the Secretary determines that the actual average county yield or benchmark yield is “an unrepresentative average yield for the county” then it must “assign an actual or benchmark county yield for each planted acre for the crop year . . . on the basis of yield history of representative farms in the State, region, or crop reporting district, as determined by the Secretary.” As explained in Part I.C. below, plaintiffs also argue FSA “failed to address and correct an obvious anomaly present in its data” when it calculated the 2014 actual average county yield for Haskell County, Kansas. Doc. 20 at 12. B. The FSA Handbook The FSA Handbook – Agriculture Risk Coverage and Price Loss Coverage Program, 1-ARCPLC (the “Handbook”) was published by FSA on January 8, 2015. Doc. 14-3 at 30. The Handbook was amended a number of times between January 2015 and September 2015—when plaintiffs signed their ARC-CO contracts—and FSA also has amended it since plaintiffs joined the program. See, e.g., Doc. 20 at 8, n.2; Doc. 14-3 at 29–39; Doc. 14-7 at 25–47; Doc. 14-8 at 1–2. When plaintiffs signed their ARC-CO contracts, Paragraph 116 of the Handbook described 6 The Olympic average is determined by excluding the crop years with the highest and lowest yields. See 7 U.S.C. § 9017 (c)(2). the actual average county yield as follows: “The actual average county yield is determined by [National Agricultural Statistics Service (“NASS”)] and is the county average yield for the crop year for the specific covered commodity.” Doc. 20 at 9; Doc. 14-7 at 31 and Doc. 18-4 at 33 (Handbook Paragraph 116 from Amendment 1 (January 8, 2015)). The Handbook also provided that the yields used to determine the benchmark yield would be “determined by NASS.” Doc. 14-7 at 27 and Doc. 18-4 at 29 (Handbook Paragraph 116 from Amendment 6 (September 1, 2015)). In their appeal to this court, plaintiffs describe this as the “Handbook I Method,” which they interpreted to mean FSA would calculate the actual average county yield by “dividing the county production by the county planted acres using NASS-determined data” only. Doc. 20 at 9. In October 2015, FSA amended the Handbook to include that the planted acres for corn means “the sum of harvested acres in a county as reported by NASS, plus the unharvested acres in a county as prepared by producers to [the Risk Management Agency (the “RMA”)].” Doc. 20 at 9; Doc. 14-6 at 2; Doc. 18-3 at 27; Doc. 14-7 at 38 and Doc. 18-4 at 40 (Handbook Definitions from Amendment 8 (October 1, 2015)). Plaintiffs characterize this as the “Handbook II Method,” which calculates the yields using both NASS data and RMA data. Doc. 20 at 9. Paragraph 116’s description was not amended, and still referred only to NASS data. Doc. 20 at 10; Doc. 14-4 at 34; Doc. 18-2 at 30. Plaintiffs contend that this amendment “retroactively changed the terms of the ARC-CO program and [p]laintiffs’ ARC-CO contracts” which—they assert—is impermissible. Id. To calculate plaintiffs’ 2014 crop year ARC-CO payments, plaintiffs assert that FSA used a combination of the Handbook II Method and a third, unofficial method that they characterize as the “Ad Hoc Method.” Doc. 20 at 10–12. This “Ad Hoc Method”—sometimes referred to as the “Cascade Method”—establishes a hierarchy of data sources and, if NASS data is unavailable, looks next to RMA data to calculate the actual average county yield. Doc. 20 at 10–11; Doc. 14-6 at 1–2; Doc. 18-3 at 26–27. In 2014, NASS did not publish any county-level statistics for production and planted acres for that crop year. See id. The Handbook formally was amended to include the Cascade Method on September 20, 2016. Doc. 14-3 at 36–39 (Handbook Paragraph 121); Doc. 14-4 at 35 n.1; Doc. 18-2 at 31 n.1. Plaintiffs also contend that applying this Cascade Method to the 2014 crop year is “an improperly retroactive rule.” Doc. 20 at 10. For plaintiffs’ 2014 ARC-CO payments, FSA calculated the benchmark yield using what plaintiffs describe as the Handbook II Method (a combination of NASS and RMA data) and the actual average county yield using what plaintiffs describe as the Cascade Method (using only RMA data). Doc. 20 at 10–12. NASS data was not used to calculate the 2014 actual average county yield because NASS “did not publish data sufficient to determine Haskell County’s actual county yield” for 2014. Doc. 20 at 11. And, as explained in more detail below, NASS data was combined with RMA data to calculate the benchmark yield because NASS data for 2009-2013 (the benchmark years) for total planted acres did not distinguish between corn grown for grain (a commodity under the program) and other types of corn. Plaintiffs argue that combining the two methods produced an ARC-CO payment that was much lower than what plaintiffs should have received. Doc. 20 at 12. Plaintiffs believe that using this method produced an arbitrary and capricious result. Id. And, they also assert, FSA should not be allowed to apply the Handbook II Method or Cascade Method retroactively after plaintiffs signed their contracts. Id. C. Haskell County Non-Irrigated Corn in 2014 Plaintiffs also emphasize a shift in the percentage of non-irrigated corn in Haskell County in 2014, arguing that this created an anomaly requiring FSA to assign a different yield under 7 U.S.C. § 9017 (g)(4). The applicable regulations provided that “separate yields will only be established where at least an average of 25% of a covered commodity[] . . . was irrigated . . . and at least an average of 25% of the same covered commodity[] was non-irrigated.” 7 C.F.R. §§ 1412.3 , 1412.53;7 see also Doc. 14-2 at 1 and Doc. 18-1 at 26 (appendices to ARC-CO contracts); Doc. 14-7 at 26 and Doc. 18-4 at 28 (Handbook Paragraph 116 from Amendment 6 (September 1, 2015)).8 In 2014, the percentage of non-irrigated corn in Haskell County was well below this 25% threshold. So, FSA was not required to calculate separate yields and, instead, it determined blended yields. While establishing separate yields was not required, plaintiffs contend that using blended data for irrigated and non-irrigated corn, “makes the actual county yields prone to anomalies when extreme fluctuations in the proportional makeup of the county occur.” Doc. 20 at 12. And, because the percentage of non-irrigated corn decreased for 2014 compared to the average percentage across the benchmark years, plaintiffs argue this created an anomaly for Haskell County and “skewed the applicable yields so much as to make the 2014 [yield] fundamentally unrepresentative.” Doc. 1 at 6–7. Plaintiffs assert that FSA was required to identify and correct this anomaly by assigning a new 2014 representative yield based on 7 Specifically, these regulations provided as follows: 7 C.F.R. § 1412.3 (“Separate irrigated and nonirrigated yields will be established in a county having a sufficient number of farms with P & CP acreage history of a covered commodity in 2009 through 2012, as determined by FSA. These separate yields will only be established where at least an average of 25 percent of a covered commodity’s P & CP acreage was irrigated in 2009 through 2012 and at least an average of 25 percent of the same covered commodity’s P & CP acreage in that county was non-irrigated in 2009 through 2012.”); 7 C.F.R. § 1412.53 (“In a county having a sufficient number of farms with P & CP acreage history of a covered commodity in 2009 through 2012, as determined by FSA, where at least an average of 25 percent of a covered commodity’s P&CP acreage was irrigated in 2009 through 2012 and at least an average of 25 percent of the same covered commodity’s P&CP acreage in that county was non-irrigated in 2009 through 2012, a separate irrigated and non-irrigated benchmark revenue, guarantee, and actual revenue will be maintained by FSA for the affected county.”) Citations to 7 C.F.R. § 1412.53 in this Order are to the § 1412.53 in effect from September 26, 2014 to August 15, 2018. — 8 — 7 U.S.C. § 9017 (g)(2) also required the Secretary “to the maximum extent possible” to “calculate a separate actual crop revenue and agriculture risk coverage guarantee for irrigated and non-irrigated covered commodities.” Plaintiffs never argue that the promulgated regulation setting forth the 25% thresholds was impermissible itself; instead they argue only that the change in this case made the yield unrepresentative under § 9017(g)(4). historical data, which would have increased their payments, but it failed to do so. Doc. 20 at 22– 27. Plaintiffs seek judicial review of (i) FSA’s methods used to calculate plaintiffs’ 2014 crop year ARC-CO payments, including whether FSA may apply those methods to plaintiffs retroactively, and (ii) the alleged failure of FSA to assign a different yield to correct for the change in irrigated acreage in the county. As summarized below, they already have appealed FSA’s calculations to the NAD and the NAD Director. D. Initial NAD Appeals In their first appeals to NAD, plaintiffs argued that NASS data was incomplete and unreliable, and RMA data—the data used to calculate the 2014 actual average county yield—also should have been used to calculate the benchmark yield. See Doc. 14-2 at 48; Doc. 18-1 at 70; Doc. 14-14 at 59–98 (hearing testimony of Jay Garetson). And, plaintiffs argued FSA should have assigned an actual average county yield rather than relying on the RMA calculated yield, due to an anomaly in the data caused by the decrease in acreage of non-irrigated corn. See Doc. 14-2 at 48–49; Doc. 18-1 at 70–71; Doc. 14-14 at 59–98 (hearing testimony of Jay Garetson). FSA responded, arguing that it had followed its established procedures and employed the Cascade Method appropriately to determine that “the actual crop revenue . . . was $17.99 per acre less than the ARC guarantee.” Doc. 14-2 at 49; Doc. 18-1 at 71. The NAD Administrative Judge found for FSA. Doc. 1 at 7 (Compl. ¶ 19); Doc. 14-3 at 8; Doc. 18-1 at 86. The Administrative Judge agreed that it would be reasonable to compare an actual average county yield based only on RMA county data to a benchmark yield also based on only RMA data. Doc. 14-3 at 2; Doc. 18-1 at 80. And, this would have produced a much higher payment for plaintiffs. Doc. 14-3 at 2; Doc. 18-1 at 80. But, the Administrative Judge concluded, the “statutes and regulations governing the ARC-CO program do not mandate a particular data set . . . and Congress delegated authority to the [FSA] to make such a determination.” Id. FSA’s policy used NASS county data if available, and if unavailable, to look to RMA county data next, then NASS district data, then state or committee determined yields—i.e., the Cascade Method. Doc. 14-3 at 1–3; Doc. 18-1 at 79–81. And, FSA applied this policy to calculate county/crop payments across the country. Doc. 14-3 at 3; Doc. 18-1 at 81. Because NASS county data for Haskell County was unavailable in 2014, FSA used RMA data exclusively to determine the 2014 actual average county yield. But, because NASS county data was available for the benchmark years, FSA used what plaintiffs call the Handbook II Method—i.e., a combination of NASS and RMA data to determine harvested and unharvested acres for 2009-2013, and ultimately benchmark yield. Doc. 14-2 at 51–52; Doc. 14-3 at 1–3; Doc. 18-1 at 73–74, 79–81. Both NASS data and RMA data were used because the actual average county yield is determined by dividing the “production of a covered commodity” by “the commodity’s total planted acres.” See 7 C.F.R. § 1412.3 . NASS provided production data for corn grown for grain (the covered commodity) and data on the total acres where corn—without distinguishing between corn grown for grain and other corn—was planted. Doc. 14-2 at 51–52; Doc. 14-3 at 1–2; Doc. 18-1 at 73–74, 79–80. To come up with the denominator of “total planted acres” that included corn grown only for grain, FSA combined the NASS data showing acres of harvested corn grown for grain plus RMA data showing acres of unharvested corn grown for grain. Id. The Administrative Judge found that this method combining NASS and RMA data was announced “as late as August 27, 2015.”9 Doc. 14-3 at 1; Doc. 18-1 at 79. But, 9 This date comes from teleconference notes of a call with the state committees discussing how they should establish average county yields in the absence of NASS or RMA data. If NASS or RMA data are available, FSA establishes the yields, and the notes reference the use of NASS data, when available, followed by RMA, as well as the Handbook II Method. If NASS and RMA data are unavailable, the state committees are tasked with determining he concluded, the way FSA calculated the actual average county yield and the benchmark yield was not unreasonable or improper. Doc. 14-3 at 3–5; Doc. 18-1 at 81–83. The Administrative Judge also determined that FSA “was not required to assign a yield because the calculated yield was not unrepresentative of the County.” Doc. 14-3 at 4–7; Doc. 18-1 at 83–85. He found that “[i]rrigated acres in the County yield[ed] greater than five times more corn on average.” Doc. 14-2 at 25; Doc. 14-3 at 4; Doc. 18-1 at 74, 82. But, though “the percentage of irrigated production could drastically impact the yield calculations,” the regulations did not require FSA to establish separate yields. Doc. 14-3 at 4–5; Doc. 18-1 at 82– 83. And, FSA was not required to assign a lower yield because the 2014 yield was not unrepresentative for the county. Doc. 14-3 at 5–7; Doc. 18-1 at 83–85. The Administrative Judge determined that despite FSA’s use of NASS and RMA data in the benchmark years and RMA data only in the 2014 year, and the differences in percentages of irrigated corn between the same periods, FSA didn’t commit clear error when it determined that the 2014 yield was still representative for the county. Id. Instead, the Administrative Judge found RMA and NASS data are comparable. And, while comparing non-irrigated acreage in 2014 to prior years shows an increase in irrigation, it does not necessarily inflate the yield to an extent that necessarily makes it unrepresentative. Id. So, FSA did not abuse its discretion by using an actual average county yield based on RMA data, rather than assigning a new yield. Id. Finally, the Administrative Judge found that FSA had fulfilled its obligation to check for anomalies in the data. Doc. 14-2 at 51; Doc. 14-3 at 4, 7; Doc. 18-1 at 73, 82, 85. He concluded that FSA’s Economic Policy and Analysis Staff (“EPAS”) had “calculated and reviewed about the yields. The teleconference notes also report that these methods of calculating were “stated at the September [2014] ARCPLC training meeting.” See Doc. 14-11 at 49–50, Doc. 14-12 at 3–4, and Doc. 18-7 at 9–10, 14–15 (Post Hearing Agency Exhibit 4). 109,000 different county/crop combinations,” then “conducted a secondary review of a sampling of counties,” and charged “state executive directors . . . with reviewing local data and requesting review if yields seemed suspect.” So, the Administrative Judge determined FSA had satisfied its obligation under 7 U.S.C. § 9017 (g)(1). Ultimately, FSA’s calculations of plaintiffs’ payments were upheld in these initial appeals. Based on FSA’s calculations, actual crop revenue was $17.99 below the benchmark guarantee, and under the statute and after sequestration, plaintiffs’ 2014 ARC-CO payments were $14.25 per acre, with payments totaling more than $50,000 for Garetson Brothers and more than $100,000 for KLC Farm. Doc. 14-2 at 54; Doc. 14-3 at 2; Doc. 18-1 at 76, 80; see also Doc. 1410 at 44; Doc. 18-6 at 59. E. NAD Director Appeals Plaintiffs then sought review of the NAD Appeal Determinations by the NAD Director. Doc. 1 (Compl. ¶ 20); Doc. 14-3 at 14–28; Doc. 14-4 at 1–29; Doc. 18-1 at 92–122; Doc. 18-2 at 1–25. There, plaintiffs argued that using RMA data to calculate their actual average county yield was erroneous as a matter of law because the Handbook in effect when they signed their ARCCO contracts required calculation based on NASS data. Doc. 14-3 at 19; Doc. 18-1 at 100. And, though the Handbook later was amended to include RMA data as an acceptable substitute, they argued retroactive application “is unquestionably improper.” Doc. 14-3 at 21; Doc. 18-1 at 103. Again, plaintiffs argued an anomaly due to the increase of irrigated corn required FSA to assign a representative yield for the 2014 crop year, and FSA failed to check for and correct such anomaly. Doc. 14-3 at 23–26; Doc. 18-1 at 104–107. In its response to the NAD Director, FSA argued that using different data in certain years is “a product of the best possible and available data” and is not an inconsistent application of different methodologies. Doc. 14-4 at 32; Doc. 18-2 at 28. And, while FSA since had updated the Handbook to clarify FSA’s procedures, it always has used the Cascade Method. Doc. 14-4 at 33–34; Doc. 18-2 at 29–30. FSA contends that requiring it to recalculate the yields used to determine the benchmark yield each year, based on what data source is available to determine the actual average county yield would be “unwieldly” because the data set available each year changes. Doc. 14-4 at 34; Doc. 18-2 at 30. And, no matter the source used to calculate the actual average county yield, that actual average county yield becomes the historical county yield used in future years to determine the benchmark yield (5-year Olympic average). Id. Finally, FSA maintained that it has procedures in place to check for anomalies in the data, and the increase in irrigated corn does not mean FSA failed to detect an anomaly. Doc. 14-4 at 35–36; Doc. 18-2 at 31–32. As discussed in more detail below, plaintiffs asked the NAD Director to overturn the Administrative Judge’s determinations for four different reasons—the same four reasons now argued in this judicial review action. But, the NAD Director determined that substantial record evidence existed to support the Administrative Judge’s conclusions that the calculations of plaintiffs’ 2014 crop year ARC-CO payments were correct. Doc. 16-6 at 6; Doc. 18-3 at 31. On July 7, 2017, Steven C. Silverman, the Director of the NAD, issued Director’s Determinations upholding the Administrative Judge’s determinations. Doc. 1 at 2, 7 (Compl. ¶¶ 7, 21); Doc. 145 at 37–38; Doc. 18-3 at 24–25. To calculate the benchmark yield, the NAD Director, like the Administrative Judge, found that FSA had used a combination of NASS data and RMA data—what plaintiffs call in this appeal the Handbook II Method. Doc. 14-6 at 2; Doc. 18-3 at 27. And, for the 2014 actual average county yield, no NASS county data was available. So, only RMA data was used—what plaintiffs here call the Cascade Method. Doc. 14-6 at 2–4; Doc. 18-3 at 27–29. But, in reality, both methods represent different applications of the Cascade Method—looking to RMA data to fill gaps where NASS data is unavailable or incomplete. See id. (explaining that RMA data was combined with NASS data to calculate the benchmark yield because NASS did not publish data on total acres on which corn for grain was planted). Because the statute provided FSA “discretion to calculate the actual average county yield and the historical average county yield,” the NAD Director reviewed the calculations under the abuse of discretion standard. Doc. 14-6 at 6; Doc. 18-3 at 31. The Director determined from the administrative hearing testimony and posthearing exhibits that FSA has used the Cascade Method to determine yields “since at least 2002 and continues to rely on this method.” Doc. 14-6 at 1–2; Doc. 18-3 at 26–27. And, as explained in more detail in Part III below, the Director concluded that the calculations of the benchmark yield and actual average county yield for 2014 using this method was not clear error or an abuse of this discretion, even when the Handbook referenced only NASS data and the end result was a comparison of combined NASS/RMA data against pure RMA data. Doc. 14-6 at 6–11; Doc. 183 at 31–36. On the question whether FSA had failed to check for and correct an anomaly in Haskell County, the NAD Director found that “FSA conducts a nationwide review” to check for “payment differences between counties.” Doc. 14-6 at 5; Doc. 18-3 at 30. “If the FSA determines that a county yield requires further review, the agency works with NASS and RMA to determine whether there are any statistical or actual production errors.” Id. And, FSA “directed state executive directors to review their respective county crop combinations . . . and identify any counties with yields that require further review.” Id. Also, the state executive directors could “request further review of county yields that are significantly different from the benchmark yields.” Id. But, FSA made no adjustments for the 2014 crop year for Haskell County. Id. The Director examined whether the change in irrigated acreage was certain to make the actual average county yield determined from pure RMA data an unrepresentative yield, but concluded it was not. Doc. 14-6 at 11–13; Doc. 18-3 at 36–38. The Director found that the evidence established the 2014 non-irrigated acreage was 2.8 percent compared to the “average of 8.6 percent . . . in the three years that constituted the Olympic average.” Doc. 14-6 at 12; Doc. 18-3 at 37. And, he found, “the average yield per irrigated acre was 7.5 times higher than the average yield per non-irrigated acre.” Id. But, the Director concluded FSA’s determination that the yield was representative was reasonable because a change in irrigated acreage like the one at issue here is just one of the factors affecting yield, and thus does not mandate a finding that the yield was inflated and unrepresentative. Doc. 14-6 at 12–13; Doc. 18-3 at 37–38. Looking exclusively at RMA data, the 2014 yield of 214 bushels was similar to the 2009 yield of 212 bushels (and combined NASS/RMA data produced a similar result in 2009 at 211 bushels), despite a higher percentage of irrigated acreage in 2014. Id. In short, the Director found that a decrease in non-irrigated acreage of this magnitude did not require FSA to find an anomalous actual average county yield and establish a new yield. So, the NAD Director affirmed the Administrative Judge’s findings, concluding “the 2014 actual average county yield was within the range of historical yields in Haskell County,” and thus “FSA did not err when it calculated [plaintiffs’] 2014 ARC-CO payment[s].” Doc. 14-6 at 12–13; Doc. 18-3 at 38. Plaintiffs now seek declaratory and injunctive relief from the court. Doc. 1 at 8. They ask the court to declare that the NAD Director’s Determinations upholding FSA’s actions calculating plaintiffs’ ARC-CO payments were arbitrary, capricious, and unlawful. Id. They also ask the court to order defendants to recalculate their ARC-CO payments in a fair manner. Id. II. Standard of Review Plaintiffs have filed their Opening Brief (Doc. 20) appealing the NAD Director’s Determinations. Under 7 U.S.C. § 6999 and 7 C.F.R. § 11.13 (a), final determinations of NAD are reviewable by this court under the APA. See also Payton v. USDA, 337 F.3d 1163 , 1167–69 (10th Cir. 2003). The APA grants federal courts authority to review agency decisions. See 5 U.S.C. § 702 . The reviewing court “shall decide all relevant questions of law, interpret constitutional and statutory provisions, and determine the meaning or applicability of the terms of any agency action.” 5 U.S.C. § 706 . The reviewing court must set aside an agency action that is “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law” or is “unsupported by substantial evidence.” 5 U.S.C. § 706 (2)(A), (E); see also Kobach v. Election Assistance Comm’n, 772 F.3d 1183, 1197 (10th Cir. 2014) (citations omitted). And, the reviewing court “will set aside an agency action if the agency has failed to follow required procedures.” Citizens’ Comm. to Save Our Canyons v. Krueger, 513 F.3d 1169 , 1176 (10th Cir. 2008). When a court applies the “arbitrary and capricious” standard of review under the APA, it “must ‘ascertain whether the agency examined the relevant data and articulated a rational connection between the facts found and the decision made.’” Kobach, 772 F.3d at 1197 (quoting Aviva Life & Annuity Co. v. FDIC, 654 F.3d 1129, 1131 (10th Cir. 2011)); see also Payton, 337 F.3d at 1168–69; Nw. Pipeline Corp. v. FERC, 61 F.3d 1479 , 1485 (10th Cir. 1995) (equating the substantial evidence test to the arbitrary and capricious standard of review). The Supreme Court describes the scope of review under this standard as a “narrow” one, and it cautions that the reviewing court must not “substitute its judgment for that of the agency.” Judulang v. Holder, 565 U.S. 42 , 52–53 (2011) (citations and internal quotation marks omitted); see also Kobach, 772 F.3d at 1197 (“This [arbitrary and capricious] standard of review is very deferential to the agency’s determination, and a presumption of validity attaches to the agency action such that the burden of proof rests with the party challenging it.” (citations and internal quotation marks omitted)); Nw. Pipeline, 61 F.3d at 1486 (“In reviewing the agency’s decision, a court is not free to substitute its own judgment for that of the agency, but must instead uphold the agency decision if there is a rational basis for that decision.”). Despite this deferential standard, a court’s review still plays an important role by “ensuring that agencies have engaged in reasoned decisionmaking.” Judulang, 565 U.S. at 53 . This standard requires the reviewing court to “assess, among other matters, whether the decision was based on a consideration of the relevant factors and whether there has been a clear error of judgment.” Id. (citations and internal quotation marks omitted); see also Lackey v. USDA, 384 F. App’x 741, 748 (10th Cir. 2010). The reviewing court should set aside agency action if the agency relied on factors which Congress has not intended for it to consider, entirely failed to consider any important aspect of the problem, offered an explanation for its decision that runs counter to the evidence before the agency, or is so implausible that it could not be ascribed to a difference in view or the product of agency expertise. Olenhouse v. Commodity Credit Corp., 42 F.3d 1560, 1574 (10th Cir. 1994) (quoting Motor Vehicle Mfrs. Ass’n of U.S., Inc. v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 43 (1983)). But, an agency’s decision will be upheld if it has a rational basis. See Nw. Pipeline, 61 F.3d at 1486. Factual determinations may be set aside only “if they are unsupported by substantial evidence. The substantial-evidence standard does not allow a court to displace the agency’s choice between two fairly conflicting views, even though the court would justifiably have made a different choice had the matter been before it de novo.” Trimmer v. U.S. DOL, 174 F.3d 1098, 1102 (10th Cir. 1999) (internal quotations and citations omitted). Questions of law, on the other hand, are reviewed de novo. Id. The court “give[s] strict effect to the unambiguous intent of Congress, if Congress has clearly spoken to the issue before [the court].” Wyo. Farm Bureau Fed’n v. Babbitt, 199 F.3d 1224 , 1231 (10th Cir. 2000). But, where “Congress is silent on the issue and has delegated authority over the subject matter to the [agency], [the court] defer[s] to the [agency’s] construction, unless . . . it is unreasonable or impermissible.” Id. The court “give[s] great deference to [an agency’s] interpretation of its own regulations, and [ ] will only reject those interpretations when they are unreasonable, plainly erroneous, or inconsistent with the regulations’ plain meaning.” Lackey, 384 F. App’x at 748 (internal quotation and citation omitted). Ultimately, the reviewing court “may affirm agency action, if at all, only on the grounds articulated by the agency itself.” Olenhouse, 42 F.3d at 1565 ; see also Motor Vehicle Mfrs. Ass’n of U.S., Inc., 463 U.S. at 43, 50 (explaining that the court “may not supply a reasoned basis for the agency’s action that the agency itself has not given” (internal quotation and citation omitted)). III. Analysis Plaintiffs assert four arguments supporting their request that the court set aside the NAD Director’s Determinations. First, plaintiffs contend that FSA’s use of a new method to determine the actual average county yield, rather than following the Handbook I Method violated the Accardi doctrine10 because it failed to follow FSA’s own established procedures. Second, plaintiffs argue that applying the Cascade Method (and Handbook II Method) to determine their 10 Explained in Part III.A. below, the Accardi doctrine is derived from caselaw that provides agencies must follow their own procedures and regulations, and if an agency fails to do so that action is generally invalid. ARC-CO payments is an impermissible retroactive application of FSA’s rules and procedures. Third, plaintiffs argue that using the Cascade Method (with only RMA data) to determine the actual crop revenue, but using the Handbook II Method (combining NASS and RMA data) to calculate the benchmark revenue was arbitrary and capricious. Last, plaintiffs argue that FSA erred by failing to assign a different representative yield due to an anomaly created from the decrease in non-irrigated corn planted in Haskell County in 2014. These are the same arguments plaintiffs presented to the NAD Director. See Doc. 14-6 at 6; Doc. 18-3 at 31. There is no dispute about how FSA calculated plaintiffs’ payments. Instead, the question is whether the methods used were arbitrary, capricious, or unlawful. The court addresses each of plaintiffs’ arguments in subsections A through D, below. The court concludes that none of the four arguments permit the court to set aside the agency’s decision. A. FSA’s Departure from the Handbook I Method Plaintiffs first argue that Paragraph 116 of the Handbook required FSA to use only NASS data to determine their 2014 ARC-CO payments. Doc. 20 at 15 (reciting the Handbook language which provided that “the actual average county yield is determined by NASS and is the county average yield for the crop year”). Plaintiffs assert that the departure from the Handbook I Method by using RMA data to determine the 2014 actual average county yield, “automatically renders FSA’s determination arbitrary and capricious and not in accordance with law.” Id. Their position contends that the Accardi doctrine requires FSA to follow the procedure it established in the Handbook. Id. The NAD Director concluded that using RMA data was not an impermissible departure from FSA’s established procedures. Doc. 14-6 at 7; Doc. 18-3 at 32 (“I am not convinced that the plain language of [Paragraph 116] prohibits FSA from using any data other than NASS data to calculate the actual average county yield. At most, the Handbook provision suggests that FSA, in the ordinary course, will use NASS data to calculate the average yield.”). He noted that, FSA, under the Cascade Method, relies on NASS data when available, but, to fill in gaps in NASS data, FSA turns to RMA data. Id. And, the NAD Director determined that the Handbook I Method was not a “mandatory rule” but a “non-binding policy statement that provides guidance to agency personnel.” Id. Plaintiffs disagree. They contend the Handbook language referencing NASS data only is not a “policy statement” but a “procedure” affecting their rights that FSA must follow. Doc. 20 at 16. The Accardi doctrine “‘require[s] an agency to adhere to the policy and regulations it [ ] promulgates.”’ Jagers v. Fed. Crop Ins. Corp., 758 F.3d 1179, 1186 (10th Cir. 2014) (quoting Unites States v. Thompson, 579 F.2d 1184, 1191 (10th Cir. 1978)). For example, in Accardi, the court considered regulations granting the Board of Immigration Appeals discretionary authority over applications seeking to suspend deportations. United States ex rel. Accardi v. Shaughnessy, 347 U.S. 260 , 264–68 (1954). The Accardi petitioner alleged the Board had not followed its established procedures and exercised that authority because the Attorney General had issued a public statement and list dictating the petitioner be deported before the Board considered his application. Id. The Supreme Court held that, because the regulations required the Board to exercise its own discretion, the petitioner should have the opportunity to try to prove his allegation that the Board failed to do so. Id. at 268 . But not every agency policy confers a right that requires the agency to follow the policy. See Motor Vehicle Mfrs. Ass’n of U.S., Inc., 463 U.S. at 42 (explaining that “an agency must be given ample latitude to adapt their rules and policies to the demands of changing circumstances” (internal quotation and citation omitted)). The Tenth Circuit has considered this issue, albeit in a different setting. In United States v. Thompson, our Circuit decided whether a Department of Justice policy manual—issued for the guidance of United States Attorneys and discussing when a defendant may be federally prosecuted after a state prosecution—conferred a right on a defendant such that the court should overturn his conviction. 579 F.2d 1184, 1185, 1189 (10th Cir. 1978). The Circuit concluded that the “departmental policy” was “at most a guide for the use of the Attorney General and the United States Attorneys in the field” and thus was not “capable of conferring a right upon” defendant. Id. at 1189 ; see also Vietnam Veterans of Am. v. Sec’y of the Navy, 843 F.2d 528 , 536–39 (D.C. Cir. 1988) (discussing whether a Secretary of Defense memorandum created a “binding rule”/“procedure” or a “policy statement,” which is not binding on the agency, and concluding that the memorandum created no “binding right” requiring upgrades for the discharge status of certain members of the armed forces). But cf. Morton v. Ruiz, 415 U.S. 199 , 232–36 (1974) (explaining where Bureau of Indian Affairs (“BIA”) manual required all eligibility requirements to be published publicly, but the eligibility requirement to live on a reservation was included only in the internal manual, the BIA must “comply . . . with its own internal procedures” before invoking that eligibility requirement to deny assistance benefits because, “[w]here the rights of individuals are affected, it is incumbent upon agencies to follow their own procedures”); Thompson, 579 F.2d at 1191 (Seth, J. dissenting) (explaining that courts “have held that agencies must follow their statements of policy and directives” in a “variety of situations . . . if the statements are applicable, are made known, and are of guidance for the execution of the agency policies” regardless of “what they are called or how they are adopted” because the public “has a right to rely on them as against some inconsistent case by case subjective determination by a public official”); United States v. Heffner, 420 F.2d 809 , 811–12 (4th Cir. 1969) (holding that IRS agents did not follow IRS instructions to special agents describing the “procedure for protecting the Constitutional rights of persons suspected of criminal tax fraud” when they interrogated the defendant because the agents did not warn defendant they investigate tax fraud or advise defendant that he could retain counsel, which violated the Accardi doctrine that requires an agency to observe the “rules, regulations, [and] procedures which it has established”). Here, the court is not persuaded that Handbook language and the Accardi doctrine require FSA to use only NASS data. First, plaintiffs have not shown the Handbook reference to calculating average county yield using NASS data was intended to confer a binding right upon plaintiffs, thus implicating the Accardi doctrine. See Vietnam Veterans, 843 F.2d at 536–39; see also Jagers, 758 F.3d at 1186 (explaining that the Accardi doctrine may not apply to RMA manuals and handbooks that were not promulgated regulations, but “appear to be departmental housekeeping provisions that do not give rise to an enforceable right” (internal quotations omitted)); Barrie v. Fed. Aviation Admin., 16 F. App’x 930, 934 (10th Cir. 2001) (“Accardi stands for the proposition that an agency must adhere to its own rules and regulations when an individual’s due process interests are implicated. . . . [T]he Accardi holding applies to regulations that exist to protect the rights of those regulated by the agency.”); Wilkinson v. Legal Servs. Corp., 27 F. Supp. 2d 32 , 48–64 (D.D.C. 1998) (concluding that personnel handbook “did not create binding obligations” on the agency or “substantive rights for its employees” under the facts of the case, and explaining the Accardi doctrine is implicated when an agency has conferred a procedural right on a person by regulation or otherwise, including self-imposed rules, and has allegedly failed to provide the due process required by the right created). To determine whether a procedural right has been conferred that is binding on the agency, the court should “inquire[] into the purpose” of the rule and determine “for whose benefit” the rule was created. Wilkinson, 27 F. Supp. 2d at 48–64. The Handbook recites that it was “issued to provide the 2014 Farm Bill policy and procedure” to state and county offices. Doc. 14-3 at 30 (Handbook page explaining the “Reason for Issuance”); see also Doc. 14-14 at 32–33, 112–113 (hearing testimony explaining the Handbook was issued to provide the 2014 Farm Bill policy and procedure to state and county offices to help with “enrolling producers, approving contracts, and ensuring that the payments go out”). And, Paragraph 116, as part of an illustration explaining how the ARC-CO payments are calculated, provided that the “actual average county yield is determined by NASS.” Doc. 14-7 at 31. But, this provision does not grant producers any “procedural protections or substantive rights” about the data FSA must use. Barrie, 16 F. App’x at 934 (concluding that the Accardi doctrine did not apply to FAA Handbooks issued to aid inspectors in carrying out FAA policies). Instead, as the Director found, “the statute provides FSA with discretion to calculate the actual average county yield and the historical average county yield.” Doc. 14-6 at 6; Doc. 18-3 at 31 (citing 7 U.S.C. § 9017 (b)); 7 C.F.R. § 1412.3 (delegating authority to FSA to determine the yields). So, FSA is tasked with determining the actual average county yield. Naturally, this calculation affected plaintiffs’ payments under their ARC-CO contracts and they thus possess some interest in how FSA determines the actual average county yield and benchmark yield. But, plaintiffs have not shown that Paragraph 116 or 7 C.F.R. § 1412.3 were intended to confer any right on them about the data FSA must use, so long as FSA’s choice is not irrational. And, as discussed below, plaintiffs also have not shown that the agency intended for this Handbook provision to create a binding rule. Second, even if Paragraph 116 of the Handbook had conferred a “right” on plaintiffs requiring FSA to consider NASS data, no NASS county data was available in 2014. The Director determined that Paragraph 116 did not mandate FSA to use solely NASS data to calculate plaintiffs’ ARC-CO payments, despite the plain language of the provision. Doc. 14-6 at 7; Doc. 18-3 at 32. And, under the facts of this case, this interpretation is not unreasonable or plainly erroneous. Under the applicable regulations, FSA had discretion to determine what data to use. And, Paragraph 116 issued by FSA referenced NASS alone. But, FSA cannot reasonably be required to use NASS data when, sometimes, it simply is not available. And, the Director found, FSA did use NASS county data when available and it looked to RMA data only to fill gaps in the data. See Doc. 14-6 at 7; Doc. 18-3 at 32. So, FSA followed the Handbook provision to the extent possible. Citing Vietnam Veterans, the Director determined Paragraph 116 was a “non-binding policy statement,” and thus FSA’s failure to follow it strictly was not a violation of the Accardi doctrine. This interpretation is not erroneous given that the regulations gave FSA discretion to determine what data to use. And, exercising this authority, FSA determined it should use NASS county data when available, but, where NASS data is not available or is incomplete, existing FSA policy called for it to apply the Cascade Method. See Vietnam Veterans, 843 F.2d at 536– 38 (discussing the difficulty in determining whether an agency statement outside of promulgated regulations is binding on the agency and explaining the “binding quality” depends “on whether the agency intended to establish a substantive rule . . . which creates or modifies rights that can be enforced against the agency” and determining the interpretation of the Department of Defense memorandum as a “flexible policy as opposed to a binding rule” was consistent with the discretion granted to the review boards under the regulations (internal quotations and citations omitted)); United States v. Lockyer, 448 F.2d 417, 421 (10th Cir. 1971) (holding that an IRS Handbook did not create due process rights favoring the taxpayer and implicating the Accardi doctrine, where the purpose of the handbook was “to guide the revenue agent in the execution of his investigation” and did not exist “for the protection of the taxpayer’s rights and interests”); Rosenburg v. Comm’r of Internal Revenue, 450 F.2d 529 , 531–33 (10th Cir. 1971) (explaining that the IRS is not always bound to its own procedures because some procedures are “directory rather than mandatory”); see also Wilkinson, 27 F. Supp. 2d at 64–65 (when determining if an agency has created a binding rule under the Accardi doctrine, the court should consider whether the agency “intended [itself] to be bound” by the rule, and “[a]n agency . . . has significant discretion to determine whether a provision is binding ‘law’ or precatory policy,” i.e., not “every general guideline is a binding rule”). Cf. United States v. Leahey, 434 F.2d 7, 11 (1st Cir. 1970) (holding that agencies do not “always violate due process when they fail to adhere to their procedures” but in this case the IRS “had a duty to conform to its procedure” that it had announced in a news release because the purpose of the news release was “designed to protect taxpayers by setting a clear and uniform standard” about what warnings must be given in a tax fraud investigation; however, the result may “have been different” if the procedure that was violated was “designed to promote some other agency goal”). The evidence in the record established FSA employed the Cascade Method consistently to determine yields—and did so in setting both the 2014 benchmark yields and, later, the actual average county yields across the country. See, e.g., Doc. 14-11 at 49–50, Doc. 14-12 at 3–4, and Doc. 18-7 at 9–10, 14–15 (email with teleconference notes from August 2015 discussing the Cascade Method and noting this formula was “stated at the September [2014] ARCPLC training meeting”); Doc. 14-14 at 24–28, 30–35 (FSA employee testimony that FSA has used the Cascade Method since 2002 and that the benchmark yields were published before the enrollment period); Doc. 14-12 at 6–8 and Doc. 18-7 at 17–19 (April 2015 memorandum to State Committees discussing Cascade Method and hierarchy of data sources when NASS county yield data is unavailable); Doc. 14-12 at 9–14 and Doc. 18-7 at 20–25 (USDA Notice to State and County offices from December 2014 discussing calculations of yields using county NASS and RMA data and process for State Committees to establish yields when no such data is available). No evidence was presented showing that FSA intended Paragraph 116 to create a mandatory, inflexible rule. See Vietnam Veterans, 843 F.2d at 539 (explaining that the language of a provision, its context, and agency treatment may help determine whether it creates a binding right); see also Wilkinson, 27 F. Supp. 2d at 64–65. Instead, the evidence established the Cascade Method was used consistently to calculate actual and benchmark yields. And, FSA separately and consistently communicated this method to state and county offices for which the Handbook was issued. Cf. Heffner, 420 F.2d at 812 (explaining that the purpose of the Accardi doctrine is to “prevent the arbitrariness which is inherently characteristic of an agency’s violation of its own procedures”). In sum, the Director’s determination that the Handbook provision was a non-binding policy statement and not a mandatory rule enforceable against the agency was neither arbitrary, capricious, an abuse of discretion, unlawful, nor unsupported by substantial evidence. The court thus upholds the NAD Director’s Determinations that FSA was not required to rely exclusively on NASS data. B. Retroactive Application of the Cascade Methods Second, plaintiffs contend FSA cannot apply the Cascade Method to determine their ARC-CO payments because doing so would represent an impermissible retroactive application of a later published FSA rule and procedure. This argument stems from the undisputed fact that FSA amended the Handbook to include explicitly the Cascade Method in September 2016—i.e., a year after plaintiffs signed their ARC-CO contracts (and a year after they received their payments). Plaintiffs assert that “[j]udicial rules against retroactive rulemaking are well established.” Doc. 20 at 19. They assert it is “fundamentally unfair” for FSA to change the official methodology set forth in its Handbook without notice to plaintiffs as they were relying on and had “settled expectations” that only NASS data would be used. Doc. 24 at 5–6. Plaintiffs’ argument that “FSA fundamentally changed the official methodology” and the retroactive application of this change is “improper, erroneous, and should be overturned” is intertwined with their argument, above, that Paragraph 116 created a binding rule the agency must follow. Doc. 20. at 19–20. Plaintiffs contend that not only was this improper retroactive rulemaking, but it also was arbitrary and capricious agency action. Doc. 24 at 6. The NAD Director concluded that using the Cascade Method “to fill a data gap that arose when the agency did not have NASS data for Haskell County for the 2014 crop year” did not impermissibly apply a new methodology retroactively. Doc. 14-6 at 7–8; Doc. 18-3 at 32–33. Rather, the NAD Director found the record established that FSA updated the Handbook “to clarify its policies and procedures.” Doc. 14-6 at 8; Doc. 18-3 at 33. He concluded FSA did not fundamentally change its rules or policies, as FSA did not have rules “constrain[ing] the agency from using data other than NASS data to fill a data gap.” Id. And, the Director found FSA had been using the Cascade Method “since at least 2002.” Id. So, the Director concluded, “FSA simply clarified its longstanding policy of using alternative sources of data when the agency’s preferred source of data, i.e., NASS data, is either insufficient or unavailable.” Doc. 14-6 at 8–9; Doc. 18.3 at 34. Plaintiffs are correct that retroactive rulemaking is generally disfavored. See JuradoGutierrez v. Greene, 190 F.3d 1135, 1148 (10th Cir. 1999). “A provision has retroactive effect if it, for example, ‘would impair rights a party possessed when he acted, increase a party’s liability for past conduct, or impose new duties with respect to transactions already completed.”’ Id. (quoting Landgraf v. U.S.I. Film Prods., 511 U.S. 244, 280 (1994)). The question is ‘“whether the new provision attaches new legal consequences to events completed before its enactment.”’ Id. (quoting Landgraf, 511 U.S. at 269–70). And “[t]he general standard is that [a] rule changing the law is retroactively applied only if Congress expressly authorized retroactive rulemaking and the agency clearly intended the rule to have retroactive effect.” Cherry v. Barnhart, 125 F. App’x 913, 915 (10th Cir. 2005) (internal quotations and citation omitted); see also NetworkIP, LLC v. FCC, 548 F.3d 116 , 122–23 (D.C. Cir. 2008) (explaining that “[t]hough agencies are entitled to deference, they may not retroactively change the rules at will” because “individuals should have an opportunity to know what the law is and conform their conduct accordingly” and the “[t]raditional concepts of due process” prohibit “an agency from penalizing a private party for violating a rule without first providing adequate notice of the substance of the rule”). But, the problem with plaintiffs’ argument is the Cascade Method was not a new rule or policy that FSA applied retroactively. As determined in Part III.A. above, there was no binding rule requiring FSA to use only NASS data, nor did plaintiffs have a right that NASS data be used in all circumstances. Instead, under the applicable regulations, FSA was tasked with determining the actual average county yield and benchmark yield, and FSA policy looked first to NASS data. Then, when NASS data was unavailable or incomplete, FSA decided where to go to fill the gaps. The use of the Cascade Method to calculate the yields under the facts presented here was not a retroactive application of a new policy changing an existing binding policy, nor was it arbitrary and capricious. See Pope v. Shalala, 998 F.2d 473 , 482–86 (7th Cir. 1993) (applying regulations that clarified policy on evaluation of pain in social security cases even though they were promulgated after the administrative hearing because a “rule simply clarifying an unsettled or confusing area of the law . . . does not [retroactively] change the law, but restates the law” according to what the agency policy has been), overruled on other grounds by Johnson v. Apfel, 189 F.3d 561 (7th Cir. 1999). Though the Cascade Method was not formally included in the Handbook until September 2016, the evidence in the record established that FSA consistently has applied the Cascade Method since well before plaintiffs joined the ARC-CO program. See, e.g., Doc. 14-11 at 49–50, Doc. 14-12 at 3–4, and Doc. 18-7 at 9–10, 14–15 (email with teleconference notes from August 2015 discussing the Cascade Method and noting this formula was “stated at the September [2014] ARCPLC training meeting”); Doc. 14-14 at 24–28, 30–35 (FSA employee testimony that FSA has used the Cascade Method since 2002 and that the benchmark yields using this method were published before the enrollment period); Doc. 14-12 at 6–8 and Doc. 18-7 at 17–19 (April 2015 memorandum to State Committees discussing Cascade Method and the hierarchy of data sources when NASS county yield data is unavailable); Doc. 14-12 at 9–14 and Doc. 18-7 at 20– 25 (USDA Notice to State and County offices from December 2014 discussing calculations of yields using county NASS and RMA data and the process for State Committees to establish yields when no such data is available); Doc. 14-4 at 22–25 (addition of Cascade Method to the Handbook), 46 (Handbook “Reasons for Amendment” stating that “Paragraph 121 has been added to clarify policy on ARC-CO yields”); see also Pope, 998 F.2d at 482–86 (explaining that to determine “whether a rule is a clarification or a change in the law” courts should look at “the intent and interpretation of the promulgating agency” as well as the agency’s earlier positions to verify that it is indeed a clarification). Thus, while the Handbook later was clarified in policy, the use of the Cascade Method was not a retroactive application of a new policy, but an application of existing FSA policy. The evidence in the record supports the Director’s conclusion that the Handbook amendment clarified the agency’s policy of what data is used in the event NASS data is unavailable or incomplete. The court agrees. No impermissible retroactive rulemaking occurred here. The court again upholds the NAD Director’s Determinations as it does not find the Director’s conclusions to be arbitrary, capricious, unlawful, or unsupported by the evidence.11 C. Calculation of the Benchmark Yield and Actual Average County Yield Using Different Data Sources Third, plaintiffs argue the way FSA calculated the benchmark yield and the actual average county yield was arbitrary and capricious because it combined two different methods or data sources—using the Cascade Method (with only RMA data) to determine the actual crop revenue, but using the Handbook II Method (combining RMA and NASS data) to calculate the benchmark revenue. Plaintiffs argue that if NASS data was unavailable in 2014 necessitating the use of RMA data to calculate the actual average county yield, then FSA also should have used RMA data exclusively to calculate the benchmark yield. Doc. 20 at 20. Plaintiffs’ brief demonstrates the difference in calculating the benchmark using RMA data—they believe it would have increased their payments from $17.99/acre to $54.39/acre. Doc. 20 at 20–21. Plaintiffs’ position is that FSA should not have been able to deviate from the Handbook in the 11 Defendants spend a portion of their Response Brief detailing the sections of plaintiffs’ ARC-CO contracts where, plaintiffs agreed, they are subject to changes in the law. These included statutory and regulatory changes occurring after they signed their contracts. See Doc. 23 at 15–17 (citing ARC-CO contracts and appendices in the record like provisions stating that “all payments are subject to changes in statutory and regulatory provisions (including any and all new statutory or regulatory provisions) irrespective of whether those amendments or provisions or changes occurred after the signing of this contract”). Plaintiffs conspicuously do not address these provisions in their briefing. The court does not consider these arguments because it upholds the NAD Director’s Determinations for the same reasons articulated by the Director. first place, but if it does, it must do so in a manner that is not arbitrary and capricious. Doc. 20 at 21. Plaintiffs argue that RMA data trends higher, “artificially deflating the current year’s payment” so by not matching the actual yield and benchmark yield data it “is unquestionably erroneous, arbitrary and capricious, and not in accordance with law.” Doc. 20 at 21. Defendants contend FSA consistently has applied the Cascade Method to establish both benchmark yields and actual average county yields, and this methodology is not arbitrary and capricious. Doc. 23 at 17–19; Doc. 14-14 at 24–25 (hearing testimony explaining that “[t]o propose some sort of methodology policy amendment that would require [FSA] to have the same source data for all years . . . could cause a real disruption and could be somewhat unwieldly for [FSA] to even determine county average yields” because FSA has “no control over what data is going to be available” each year and FSA must calculate more than 100,000 benchmark and actual yields each year). FSA published the benchmark yields it intended to use for each county across the country before producers chose to enroll in the program. See Doc. 14-14 at 30–31, 43, 51–52. And while the benchmark years for Haskell County had NASS data available each year, this data was not available for all counties and other benchmark yields across the country were determined using other data. Id.; see also Doc. 23 at 17–19. Whatever source is used to determine the actual average county yield for the year, that yield then is used in future years to calculate the benchmark yield. See Doc. 14-14 at 25; see also Doc. 23 at 17. Finally, defendants argue that though RMA data “may trend higher than NASS data,” this isn’t always the case. Doc. 23 at 18. For example, in 2012, the combined NASS/RMA yield in Haskell County was higher than the pure RMA yield. Id.; Doc. 14-6 at 3–4; Doc. 18-3 at 28–29. So, defendants argue, their calculations of the actual average county yield and the benchmark yield—though not based on an identical data source for Haskell County in 2014—were not arbitrary, capricious, or unlawful calculations. The NAD Director again upheld the initial NAD appeals. Doc. 14-6 at 9–11; Doc. 18-3 at 34–36. Because Congress had delegated authority to FSA to determine the benchmark and actual yields and the statutes and regulations did not mandate FSA to use particular data sources, the NAD Director reviewed FSA’s actions under an abuse of discretion standard. Doc. 14-6 at 9–10; Doc. 18-3 at 34–35. The Director determined that FSA calculated the actual average county yield in accordance with the regulation’s definition. Doc. 14-6 at 10; Doc. 18-3 at 35. And, using of the “well established” Cascade Method was appropriate, given NASS data was unavailable or incomplete and FSA was mandated under the statute to make payments “‘beginning October 1, or as soon as practicable thereafter”’ if it determined that a payment was required for the crop year. Doc. 14-6 at 10; Doc. 18-3 at 35 (quoting 7 U.S.C. § 9017 (f)). Finally, the Director determined that “nothing in the record substantiates [plaintiffs’] argument that FSA’s use of the [C]ascade [M]ethod artificially deflated the ARC-CO payment calculation[s].” Doc. 14-6 at 10; Doc. 18-3 at 35. The Director compared the NASS data and RMA data available for each of the benchmark years and found no discrepancies showing that RMA data and NASS data are so different that the use of the two sources must have “artificially deflated” the payment. Doc. 14-6 at 10–11; Doc. 18-3 at 35–36. Instead, the differences ranged from 0.8% to less than 5%, and, in one of the years used to calculate the benchmark yield, NASS data was higher than RMA data. Id. So, the Director upheld the agency’s decision, finding no failure to consider relevant factors or clear error of judgment. Like the Administrative Judge, this court agrees with plaintiffs that it would be logical to calculate the benchmark and actual yields using the same data source. See Doc. 14-3 at 2; Doc. 18-1 at 80. But, using different data sets depending on what is available each year in each county across the county also is logical. To the calculate actual average county yield, FSA had discretion to determine what data to use. See 7 C.F.R. § 1412.3 (“Actual average county yield is calculated as the crop year production of a covered commodity in the county divided by the commodity’s total planted acres for a crop year in the county, as determined by FSA.”). It established a hierarchy of data sources to use with the Cascade Method. It then employed the Cascade Method across the country when what it had determined was the best source—i.e., the NASS county data referenced in the Handbook—was unavailable or incomplete to calculate the actual and benchmark yields as defined in the regulation. The applicable regulation provided the benchmark yield means the “the 5-year Olympic average of actual average county yields for the most recent 5 years.” 7 C.F.R. § 1412.3 (bold and emphasis added). So, for example, if NASS data was available in 2013 to calculate the actual average county yield, it makes sense that the actual average county yield from 2013 based on such data is used to determine the 2014 benchmark yield (if it isn’t excluded from the Olympic average). If only RMA data was available in 2014, and thus the 2014 actual average county yield was based on RMA data alone, under the regulation’s definition FSA should use this actual average county yield when the 2015 benchmark is calculated. The 2013 actual average county yield based on NASS data and the 2014 actual average county yield based on RMA data, though drawn from different data sources, both would be used to determine the 2015 benchmark yield. This comports with the regulation’s definition for “average historical county yield” (i.e., benchmark yield). And, it supports the court’s finding that FSA’s calculation of the 2014 payments was a rational implementation of its directive to calculate the yields. Plaintiffs argue FSA should go back and recalculate a new historical actual average county yield for each of the benchmark years to match the data source used under the Cascade Method for the current year actual average county yield. But, this does not mean FSA’s different approach is arbitrary and capricious. For example, applying plaintiffs’ rationale to every county across the country could burden FSA and make it difficult to comply with the statute’s mandate to make payments as soon as practicable after the end of the applicable crop year. What if the particular data source from the current year is not available in all five of the benchmark years? Should FSA calculate yields for each county across the country differently depending on the data available over a six year period? Logical arguments exist to support either position, but choosing between rational alternatives is not a choice the court should make. The statutes and regulations conferred this discretion to FSA, not the courts. And, the court does not find calculation of the 2014 Haskell County actual average county yield and benchmark yield, though based in part on different data sets, to be unlawful, arbitrary, or capricious. The NAD Director compared the results of the two data sources and concluded that comparing the two does not produce clear error. The court agrees the record supports a finding that comparing the two sources is not arbitrary, capricious, or unlawful. Sometimes RMA is higher, sometimes NASS is higher, sometimes they are the same. But, overall the differences between the two sources do not lead the court to conclude that using the best data source available under FSA’s Cascade Method, depending on availability in each year, even if it results in a benchmark based on one source but an actual average county yield based on another, is irrational. The agency acted within the scope of its authority here. The Director did not “rely on factors . . . Congress” did not intend for him to consider, fail “to consider any important aspect of the problem,” offer an explanation “that runs counter to the evidence,” or reach an implausible conclusion. Olenhouse, 42 F.3d at 1574 . Instead, the evidence supports the Director’s conclusion that the data sources are comparable and FSA’s calculations were acceptable. The court thus upholds the NAD Director’s Determinations against plaintiffs’ third argument. D. FSA’s Alleged Failure to Assign a Representative Yield For their final argument, plaintiffs assert FSA was required by statute to assign a different representative yield due to an anomaly created from a decrease in non-irrigated corn planted in Haskell County in 2014. Plaintiffs assert that FSA failed to honor this statutory duty. Plaintiffs argue FSA “failed to address and correct an obvious anomaly present in its data” when it calculated the 2014 actual average county yield because non-irrigated corn “fell by 66% from its average proportion.” Doc. 20 at 12–13. Plaintiffs assert the average percentage of non-irrigated corn for the benchmark years (2010, 2012, 2013) was approximately 9%,12 but in 2014, the percentage of non-irrigated corn in the county dropped to about 3%. Id. at 23. Because the acreage planted of non-irrigated corn in Haskell County fell well below the benchmark average, plaintiffs argue this produced an inflated blended yield “that was not representative of the historical data.” Id. at 13, 24. Plaintiffs believe that FSA should have adjusted the actual average county yield based on an average 9% non-irrigated yield which, they contend, would have increased their payments from $17.99/acre to $57.95/acre. Id. at 24–25. Plaintiffs agree FSA’s use of a blended yield comports with the regulations, but, they assert, where a blended yield is used, FSA’s “duty to monitor for anomalies is that much more 12 Plaintiffs’ calculation uses rounded percentages and excludes the crop years with the highest and lowest yields (i.e., looks only at the years used to determine the Olympic average). Using the percentages provided in defendants’ brief, the result is roughly the same. The average for 2010, 2012, and 2013 years is 8.32%. And, the non-irrigated percentage for 2014 was 2.78%, producing the 66% decrease plaintiffs describe. The average across all five benchmark years using defendants’ figures gives a non-irrigated average of 7.18%; producing a 61.2% decrease from the average. See Doc. 23 at 21. critical.”13 Doc. 24 at 8. Plaintiffs argue no evidence in the record shows that FSA satisfied its duty to check for anomalies under the statute. Doc. 20 at 26. And, by failing to check for the alleged anomaly and assign a different yield based on historical data, they argue FSA violated 7 U.S.C. § 9017 (g), making FSA’s calculation of plaintiffs’ ARC-CO payments arbitrary and capricious. Doc. 20 at 26. Defendants, on the other hand, assert that FSA did check for anomalies as required by the statute. And, they contend, finding no anomaly or unrepresentative average yield, FSA was not required to assign a representative yield. Doc. 23 at 19–20. FSA’s position is that it was not required to establish separate irrigated and non-irrigated yields under the regulations. Doc. 23 at 20–21. And, the record evidence establishes that FSA checked for anomalies in the data after calculating the blended yield. Defendants assert that FSA’s EPAS calculated and reviewed the yields for each county/crop combination, then “[w]hen price data became available conducted a secondary review by sampling counties.” Doc. 23 at 21. Then, “FSA’s state executive directors were each charged with reviewing local data and requesting review if the yields appeared suspect.” Id. But, no anomalies were discovered for Haskell County. Id. Plaintiffs believe these actions amount to no more than an “eye test” and are insufficient to comply with 7 U.S.C. § 9017 (g)(1)’s explicit directive to “‘use all available information and analysis, including data mining, to check for anomalies.”’ Doc. 24 at 8–9 (quoting 7 U.S.C. § 9017 (g)(1)). Defendants also argue the 2014 actual average county yield was not unrepresentative for the county, as it was similar to the 2009 actual average county yield. Doc. 23 at 22. Defendants note that the non-irrigated acreage across the five benchmark years ranged from 4.60% to 10.67% in Haskell County. Doc. 23 at 21. The non-irrigated percentage in 2014 was 2.78%. Id. 13 As explained in Part I.C. above, the regulations did not require FSA to establish separate yields for irrigated and non-irrigated corn here. And, defendants argue, a change in percentage of non-irrigated corn is not an anomaly and nothing “in the statute or regulation . . . specifies that adjustments to the data be performed . . . due to percentage change in acres irrigated each year.” Doc. 23 at 22. Plaintiffs respond, asserting that 2009 was “eliminated from consideration by the ARC program’s Olympic-averaging method” and “casually looking at the number” compared to earlier years is “insufficient to show compliance with FSA’s obligations to monitor for anomalies like a significant fluctuation in allocation of irrigated/non-irrigated acres.” Doc. 24 at 9–10. Plaintiffs argue that the statute obligated defendants to dig deeper and had defendants done so, simply comparing the irrigation percentages from 2009 with 2014 would have identified an anomaly and that it created an unrepresentative yield. Id. The Director summarized how FSA conducted a review for anomalies. Doc. 14-6 at 5; Doc. 18-3 at 30. And, he determined FSA was not required to assign a representative yield. Doc. 14-6 at 11–13; Doc. 18-3 at 36–38. While a change in the percentage of irrigated versus non-irrigated yield affects the blended average yield, the Director found that many other factors can influence the average county yield as well. So, a change in percentage of non-irrigated acres does not necessarily mean that the actual average county yield is unrepresentative for the county. Doc. 14-6 at 12–13; Doc. 18-3 at 37–38. The Director concluded that comparing the 2014 actual average county yield to the 2009 yield was appropriate, despite the differences in the nonirrigated acreage percentage between the two years. Id. And, he determined, “the record shows that the 2014 actual average county yield was within the range of historical yields in Haskell County,” so “FSA was not required to assign a representative yield for the 2014 corn crop in Haskell County.” Doc. 14-6 at 12–13; Doc. 18-3 at 38. As explained below, the court finds the Director’s determinations were not arbitrary, capricious, or an abuse of discretion. Section 9017(g)(1) of Title 7 of the United States Code obligated the Secretary of the USDA “to the maximum extent practicable, [to] use all available information and analysis, including data mining, to check for anomalies in the determination of agriculture risk coverage payments.” This statute also provided that, if the Secretary determines that the actual average county yield or benchmark yield is “an unrepresentative average yield for the county,” then he shall “assign an actual or benchmark county yield” based on representative yield history for the area. 7 U.S.C. § 9017 (g)(4). These obligations are distinct, though, as plaintiffs argue, the existence of an anomaly could mean an unrepresentative yield exists.14 As summarized in Part I.D., the Administrative Judge made separate findings on whether FSA had satisfied its obligations under § 9017(g)(1) and (g)(4). See Doc. 14-3 at 4–7; Doc. 18-1 at 82–85. But, unlike the Administrative Judge, the NAD Director did not explicitly analyze whether FSA satisfied its obligation under § 9017(g)(1). He did make findings about FSA’s efforts to review and adjust yields as summarized in Part I.E, above. See Doc. 14-6 at 5; Doc. 18-3 at 30. And, in his analysis, the Director determined FSA need not assign a new yield. See Doc. 14-6 at 11–13; Doc. 18-3 at 36–38. But, he did not analyze § 9017(g)(1) and § 9017(g)(4) separately. See id. Similar to plaintiffs’ arguments in their Request for Director Review, 15 the 14 The identification of an anomaly such as an error in the data used also could result in a correction and recalculation under § 9017(b)(1), without needing to assign a yield. And, the plain reading of the statute does not require an assigned yield anytime an anomaly exists. Instead, § 9017(g)(4) directs the Secretary to assign a yield only if the Secretary determines the actual or benchmark yield is unrepresentative. 15 In the introductions to plaintiffs’ Requests for Director Review, plaintiffs asserted that “FSA failed to employ statutorily mandated techniques to identify and correct for anomalies” as required under 7 U.S.C. § 9017 (g)(1). Doc. 14-3 at 15; Doc. 18-1 at 96. In plaintiffs’ argument section to the Director they intertwined this obligation with the obligation under 7 U.S.C. § 9017 (g)(4) to assign a new yield if the determined yield is unrepresentative for the county. Doc. 14-3 at 23–26; Doc. 18-1 at 104–107. They argued that the change in irrigated acreage was an anomaly that “skewed the yield so much as to make it fundamentally unrepresentative,” so FSA needed to assign a yield “that adjusts for the deviation.” Id. Still, plaintiffs recognize, as they must, that the obligations under each subsection in § 9017(g)(1)-(4) are independent obligations. See Doc. 14-3 at 26; Doc. 18-1 at 107. Director instead considered whether the change in irrigated acreage was an anomaly that made the yield unrepresentative. Affirming the initial NAD Appeal Determinations, the Director articulated how the yield may be found representative for the county even with a difference in current year’s irrigated acreage compared to a benchmark irrigated acreage like the one here. The court could infer from the Director’s explicit reasoning that he concluded FSA had satisfied its statutory obligation to check for anomalies under § 9017(g)(1). Alternatively, the court could infer that he did not believe the change in irrigation of this magnitude is an anomaly at all.16 Regardless, the court does not find the gap between the Administrative Judge’s conclusions and those of the Director sufficient to warrant remand for a specific finding on this issue. At minimum, the NAD Director’s Determinations show he concluded that a change in irrigation of this magnitude is not an anomaly large enough to merit a finding of unrepresentative Similarly, in their briefing to the court, plaintiffs’ arguments about § 9017(g)(1) are tied to their belief that a new yield must be established under § 9017(g)(4). See Doc. 1 at 6 (Compl. ¶ 18); Doc. 20 at 12–13, 22–27; Doc. 24 at 8–10. And, plaintiffs contend, if FSA had done a more in-depth review for anomalies, it would have identified one and corrected for it by finding the determined yield unrepresentative and assigning a new yield. Id. at 25–27. 16 Plaintiffs contend the change in 2014 (66% decrease from the Olympic average) was an “extreme fluctuation.” Doc. 24 at 8. The percentage of non-irrigated acreage in 2014 (2.8%) was lower than the average of all five benchmark years (7.12%) (and the Olympic average (8.27%)). But, it appears that the percentage of nonirrigated corn in Haskell County changes from year to year. And, changes of a similar magnitude are not uncommon. The chart below demonstrates the court’s calculation of the change, year to year, for each of the benchmark years and 2014. Non-irrigated percentages are taken from NAD Director’s findings, Doc. 18-3 at 29 and Doc. 14-6 at 4. The parties each used similar percentages rounded different ways in their briefs. See Doc. 20 at 23; Doc. 23 at 21. Under the statute’s Olympic averaging method, the highest and lowest yields, which may at times correlate with the irrigation data below, will be eliminated when calculating the benchmark yield. Regardless, as the Director demonstrated and as explained below, FSA could rationally conclude the 2014 yield was representative for the county despite the 2014 decrease. Year Non-Irrigated Percentage Percentage Fluctuation from Prior Year 2009 6.3% 2010 10.6% 68.3% increase from 2009 2011 4.5% 57.5% decrease from 2010 2012 6.5% 44.4% increase from 2011 2013 7.7% 18.5% increase from 2012 2014 2.8% 63.6% decrease from 2013 yield. And, the record establishes the Director considered FSA’s efforts to review the data for anomalies. See Olenhouse, 42 F.3d at 1574 (explaining the reviewing court should set aside agency action if the agency “entirely failed to consider any important aspect of the problem”). As explained in more detail below, the Director’s determination that a new yield need not be determined under § 9017(g)(4)—i.e., that the 2014 determined yield was representative—is a rational determination. Thus, the court concludes, FSA is not required to recalculate plaintiffs’ ARC-CO payments. And, the court need not consider whether it would be appropriate or practicable for FSA to establish additional procedures to monitor annual fluctuations in irrigated acreage percentage vis-à-vis the benchmark yield’s average irrigated percentage. The Secretary has discretion to determine if the yield is unrepresentative, and the court finds he did not abuse that discretion in this case. See 7 U.S.C. § 9017 (g)(4). The Director determined no anomaly existed that caused an unrepresentative average yield for the county, i.e., even though a decrease in the percentage of non-irrigated acreage occurred, the Director found the 2014 Haskell County corn actual average county yield determined with RMA data still was a representative yield for the county. Doc. 14-6 at 12–13; Doc. 18-3 at 37–38. So, the change in non-irrigated acreage present here was not an anomaly that made the actual yield unrepresentative for the county. And, FSA was not required to determine a new yield for 2014. The evidence in the record is sufficient to support this determination. The record includes evidence that FSA checked for anomalies. See, e.g., Doc. 14-14 at 41–42, 55–56, 102– 103 (hearing testimony that the EPAS checks the NASS and RMA data to see if anything appears amiss by comparing those data with other data for the commodity in the relevant state and county, then contacts NASS or RMA if the data appears to have an error to determine if it is credible, and the state committees also are asked to check the yields); Doc. 14-11 at 48 and Doc. 18-7 at 8 (post-hearing exhibit explaining review procedures which involve (1) EPAS first checking for errors in the NASS and RMA data compiled, then, when price information becomes available, sampling counties to determine why there are differences in payments between counties, (2) directing the state executive directors to review their counties to determine if any require further review by the Deputy Administrator for Farm Programs, and (3) if further review is requested, additional work with NASS and RMA to check for errors and make adjustments for mathematical errors or use of an incorrect data source or inaccurate data). And, plaintiffs have not carried their burden to show that these efforts were insufficient such that, if FSA had done more, it would have discovered an anomaly requiring FSA to establish a new representative yield changing their payments. Instead, the agency has discretion to determine whether the yield is unrepresentative requiring an assigned yield and the court finds no abuse of this discretion or clear error of judgment. See 7 U.S.C. § 9017 (g)(4). While the non-irrigated percentage fell 66% from the average percentage across years 2010, 2012, and 2013, this fluctuation does not mean that FSA must assign a different yield. The NAD Director compared the 2014 actual average county yield to the actual average county yields in the benchmark years. He concluded that the 2014 yield was similar to the 2009 yield, despite the difference in non-irrigated acreage between the two years (6.3% in 2009; 2.8% in 2014). Though 2014 had a larger proportion of irrigated acreage, the average yields were similar (211 bushels in 2009 (212.2 with RMA data), 214 bushels in 2014). This means a difference in irrigated acreage does not always produce a substantially higher average yield. The Director also examined the yields across two years with similar percentages of nonirrigated acreages, which again produced substantially different actual average yields. He compared 2009 and 2012, which had non-irrigated acreages of 6.3% and 6.5% respectively, but much different actual average yields, 212.2 bushels and 177.2 bushels respectively (using RMA data). So, while the difference in percentage of irrigated acres can affect the yields, this difference does not necessarily mean the 2014 yield was “an unrepresentative average yield for the county” just because the non-irrigated acreage differed from the benchmark Olympic average irrigated acreage. The Director determined rain and a variety of other factors also influence the yields. And, because the 2014 yield was similar to the 2009 yield, he upheld FSA’s determination that it was a representative yield for Haskell County. Essentially, this difference in irrigated and non-irrigated acreage compared to the benchmark data does not mean that FSA must establish a new yield under the facts here because the 2014 yield remained a representative yield. This conclusion is supported by evidence in the record. The Director’s analysis was rational, and not arbitrary or capricious. In sum, after reviewing the administrative record and relevant evidence within, the court agrees with defendants. The NAD Director’s Determinations were not arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law. The court thus affirms. IV. Conclusion For the reasons explained above, the court upholds the NAD Director’s Determinations about FSA’s calculations of plaintiffs’ 2014 ARC-CO payments. The Director’s findings were not arbitrary and capricious or unlawful. The court thus affirms the agency decision and dismisses the case. IT IS THEREFORE ORDERED BY THE COURT THAT the NAD Director’s Determinations are affirmed. IT IS SO ORDERED. Dated this 25th day of November, 2019, at Kansas City, Kansas. s/ Daniel D. Crabtree Daniel D. Crabtree United States District Judge