Rita Boucher v. AGRI , 934 F.3d 530 ( 2019 )


Menu:
  •                                In the
    United States Court of Appeals
    For the Seventh Circuit
    ____________________
    No. 16-1654
    RITA BOUCHER,
    Plaintiff-Appellant,
    v.
    UNITED STATES DEPARTMENT OF AGRICULTURE, et al.,
    Defendants-Appellees.
    ____________________
    Appeal from the United States District Court for the
    Southern District of Indiana, Indianapolis Division.
    No. 1:13-cv-01585 — Tanya Walton Pratt, Judge.
    ____________________
    ARGUED SEPTEMBER 21, 2018 — DECIDED AUGUST 8, 2019
    ____________________
    Before WOOD, Chief Judge, and FLAUM and HAMILTON, Cir-
    cuit Judges.
    HAMILTON, Circuit Judge. In the mid- to late-1990s, the late
    David Boucher cut down nine trees on his family farm in In-
    diana. For almost two decades, the United States Department
    of Agriculture (USDA) has disagreed, first with Mr. Boucher
    and now his widow, plaintiff Rita Boucher, about whether
    that modest tree removal converted several acres of wetlands
    2                                                     No. 16-1654
    into croplands, rendering the Bouchers’ entire farm ineligible
    for USDA benefits that would otherwise be available.
    Since at least 1985, federal law and regulatory policy have
    tried to remove financial incentives for destruction of environ-
    mentally important wetlands. In this case, however, the rec-
    ord shows arbitrary and capricious action by the agency. The
    USDA repeatedly failed to follow applicable law and agency
    standards. It disregarded compelling evidence showing that
    the acreage in question never qualified as wetlands that could
    have been converted illegally into croplands. And the agency
    has kept shifting its explanations for treating the acreage as
    converted wetlands. The USDA’s treatment of the Bouchers’
    acreage as converted wetlands easily qualifies as arbitrary, ca-
    pricious, and an abuse of discretion. See 
    5 U.S.C. § 706
    (2)(A).
    We reverse the district court’s affirmance of the USDA’s final
    determination and remand the case to the district court to en-
    ter judgment granting appropriate relief to plaintiff Rita Bou-
    cher.
    In Part I, we summarize the statutes, regulations, and
    agency guidance that govern the USDA’s wetland conserva-
    tion enforcement efforts. In Part II, we set forth the facts and
    history of this dispute. We explain in Part III the legal stand-
    ards for judicial review and explain in Part IV why this agency
    action was arbitrary, capricious, and an abuse of discretion.
    Along the way, we explain why the agency’s litigation posi-
    tion has strayed far from the applicable law and science.
    I. The USDA and Wetland Conservation
    A. The Statute, its Regulations, and the Agency
    Wetlands serve vital ecological and economic functions.
    They provide habitats for birds, fish, and unique species of
    No. 16-1654                                                           3
    wild plants; enhance drinking water supply and quality; pro-
    tect against loss of life and property from flooding; and offer
    significant recreational and commercial benefits from fishing,
    hunting, birdwatching, and other wetland-related activities
    that generate billions of dollars annually. 
    16 U.S.C. § 3901
    . Yet
    the continental United States has lost over half of its natural
    wetland habitats since the nation’s founding, with that loss
    having accelerated sharply from the 1950s through the 1970s.
    See 
    16 U.S.C. § 3901
    (7); Natural Resources Conservation Ser-
    vice, Introduction to Wetland Conservation Provisions. 1 Those
    losses have been felt acutely in the Midwest as large propor-
    tions of wetlands have been converted to agriculture and
    other uses. Regional Supplement to the Corps of Engineers Wet-
    land Delineation Manual: Midwest Region at 8 (Aug. 2010) (not-
    ing historic wetland loss in Indiana (87%), Illinois (85%), Iowa
    (89%), Minnesota (80%), Missouri (87%), and Ohio (90%)).
    Concerned about this precipitous loss of wetlands, Con-
    gress included wetland conservation provisions (known col-
    loquially as the “Swampbuster” provisions) in the Food Secu-
    rity Act of 1985. 
    16 U.S.C. §§ 3801
    , 3821–24. These laws condi-
    tion the availability of important USDA farm program bene-
    fits on farmers’ willingness to protect wetlands on their prop-
    erty. Farmers who convert (i.e., destroy) wetlands for agricul-
    tural purposes are denied those benefits. 
    16 U.S.C. § 3821
    (a);
    
    7 C.F.R. § 12.4
    ; see also Horn Farms, Inc. v. Johanns, 
    397 F.3d 472
    , 474 (7th Cir. 2005) (noting that initial “Swampbuster”
    provisions made loss of farm subsidies “proportional to the
    amount of wetland converted,” but 1990 amendment
    1  Available at https://www.nrcs.usda.gov/wps/portal/nrcs/detailfull/na-
    tional/water/wetlands/?cid=stelprdb1043554 (last visited Aug. 7, 2019).
    4                                                   No. 16-1654
    “provided that converting any wetland would cause the
    farmer to lose all agricultural payments”).
    Two USDA agencies implement this regulatory scheme.
    The Natural Resources Conservation Service (NRCS) is the
    USDA’s scientific arm charged with making technical deter-
    minations about whether wetlands exist or have been con-
    verted, as well as investigating failures to comply with the
    Swampbuster provisions. 
    7 C.F.R. §§ 12.2
    , 12.6(a)(2) & (c),
    12.30(a). And the USDA’s Farm Service Agency (FSA) relies
    on NRCS’s wetland determinations to make decisions regard-
    ing any violations and eligibility for benefits. 
    7 C.F.R. §§ 12.2
    ,
    12.6(a) & (b).
    B. Soil, Plants, and Water
    Farmers’ access to important financial benefits can thus
    turn on NRCS’s identification of “wetlands.” The statutory
    definition is somewhat technical, but it lies at the heart of our
    decision:
    The term ‘wetland’ … means land that —
    (A) has a predominance of hydric soils;
    (B) is inundated or saturated by surface or groundwa-
    ter at a frequency and duration sufficient to
    support a prevalence of hydrophytic vegetation
    typically adapted for life in saturated soil con-
    ditions; and
    (C) under normal circumstances does support a
    prevalence of such vegetation.
    
    16 U.S.C. § 3801
    (27) (emphasizing key terms discussed be-
    low); see also 
    7 C.F.R. § 12.2
    .
    No. 16-1654                                                   5
    Under this definition, in making wetland determinations,
    the NRCS must assess whether “the area of interest supports
    a prevalence of [1] hydrophytic vegetation, [2] a predomi-
    nance of hydric soils, and [3] wetland hydrology under nor-
    mal circumstances.” 
    7 C.F.R. § 12.30
    (c)(7). All three character-
    istics must be present for an area to be considered wetlands.
    B&D Land and Livestock Co. v. Schafer, 
    584 F. Supp. 2d 1182
    ,
    1194–95 (N.D. Iowa 2008) (“the statute plainly and unambig-
    uously defines these three requirements as separate, mandatory
    requirements”).
    The three terms are not self-explanatory to judges and
    other laypeople, so we look to the statute and its implement-
    ing regulations for further guidance:
    (1) Hydric Soil is soil that is conducive to potentially
    supporting the types of vegetation that might be
    found in wetlands, i.e., “soil that, in its un-
    drained condition, is saturated, flooded, or
    ponded long enough during a growing season
    to develop an anaerobic condition that supports
    the growth and regeneration of hydrophytic
    vegetation.” 
    16 U.S.C. § 3801
    (a)(12); 
    7 C.F.R. § 12.2
    .
    (2) Hydrophytic vegetation refers to plants known as
    “hydrophytes,” which can be found “growing
    in water or in soil too waterlogged for most
    plants to survive.” Webster’s Third New Int’l
    Dictionary 1109 (1993). The technical statutory
    definition describes this type of plant as those
    which grow “in water or in a substrate that is at
    least periodically deficient in oxygen during a
    6                                                   No. 16-1654
    growing season as a result of excessive water
    content.” 
    7 C.F.R. § 12.2
    ; 
    16 U.S.C. § 3801
    (a)(13).
    (3) Hydrology in this context is the fancy word for
    water, and lots of it. A wetland must be wet
    enough (at least for some period of the year)
    that typical wetland plants can grow or—in for-
    mal terms—the land must be observed to be “in-
    undat[ed] or saturat[ed] by surface or ground-
    water during a growing season at a frequency
    and duration sufficient to support a prevalence
    of hydrophytic vegetation.” 
    7 C.F.R. § 12.2
    ; see
    also Army Corps of Engineers Wetland Delineation
    Manual at A6 (1987) (“Hydrology” is “[t]he sci-
    ence of dealing with the properties, distribution,
    and circulation of water.”).
    The final key term is that the land must in fact be wetlands
    under “normal circumstances,” which means “the soil and
    hydrologic conditions that are normally present, without re-
    gard to whether the vegetation has been removed.” 
    7 C.F.R. § 12.31
    (b)(2)(i). As explained in agency guidance: “The prem-
    ise for the concept of normal circumstances is that for many
    wetlands where the vegetation has been removed, the soil and
    hydrological characteristics remain to the extent that hydro-
    phytic vegetation could return if vegetation management
    ceased.” NRCS National Food Security Act Manual § 514.2H
    (2010).
    The federal laws also address “converted” wetlands, de-
    fined by statute as:
    Wetland that has been drained, dredged, filled,
    leveled, or otherwise manipulated (including
    No. 16-1654                                                                 7
    any activity that results in impairing or reduc-
    ing the flow, circulation, or reach of water) for
    the purpose or to have the effect of making the
    production of an agricultural commodity possi-
    ble.
    
    16 U.S.C. § 3801
    (a)(7)(A). And relevant to this appeal, any
    “converted wetland” that was converted before December 23,
    1985 (i.e., before passage of the Swampbuster provisions), is
    simply deemed “non-wetland.” 
    7 C.F.R. § 12.2
    . 2
    NRCS agents are required to identify wetlands and imple-
    ment the Swampbuster provisions across the varying geogra-
    phy of the United States. Several sources of detailed guidance
    have been developed. Beyond the statutory provisions and
    regulations, NRCS agents rely on the Corps of Engineers Wet-
    lands Delineation Manual (1987) (the “Corps Manual”), the Re-
    gional Supplement to the Corps of Engineers Wetland Delineation
    Manual: Midwest Region (Version 2.0) (Aug. 2010) (“Corps Sup-
    plement”), the NRCS’s National Food Security Act Manual
    (2010) (“NRCS Manual”), and the NRCS’s Food Security Act
    2 The corresponding definition in USDA regulations is slightly altered
    from the statutory text, potentially pulling the definition away from the
    statute’s primary focus on hydrology: “Converted wetland is a wetland
    that has been drained, dredged, filled, leveled, or otherwise manipulated
    (including the removal of woody vegetation or any activity that results in im-
    pairing or reducing the flow and circulation of water) for the purpose of
    or to have the effect of making possible the production of an agricultural
    commodity without further application of the manipulations described
    herein.” 
    7 C.F.R. § 12.2
     (emphasis added).
    8                                                            No. 16-1654
    Wetland Identification Procedures (2010) (“NRCS Proce-
    dures”). 3
    Using a combination of off-site resources (e.g., aerial pic-
    tures, regional soil maps) and on-site investigations, NRCS
    agents or experts are charged with certifying a “wetland de-
    termination [that] is of sufficient quality to make a determina-
    tion of ineligibility for program benefits.” 
    7 C.F.R. §§ 12.30
    (c)(1); 12.6(c)(5)–(7); NRCS Manual at §§ 514.1A(1) &
    (3), B(1) (on-site visits are required in several circumstances
    that apply to the present case); NRCS Procedures at (1-3), (2-
    2), (2-3), (4-1), (5-2), (5-3).
    C. Making a Wetland Determination: Typical and Atypical Sit-
    uations
    To assess whether the USDA’s wetland determination for
    the Boucher farm is defensible, we set out in general terms the
    process and the rationales for that process in the agency’s
    governing statutes, regulations, and guidance manuals. As
    noted above, the overarching statutory command is that
    NRCS agents must determine whether all three of the re-
    quired characteristics—wetland soil, plants, and water—are
    present, or could be present, on disputed sites.
    There are scientific reasons why only one or two of the cri-
    teria are insufficient for a wetland determination. As one of
    NRCS’s primary sources of guidance cautions, NRCS agents
    3       The        Corps         Manual        is       available         at
    http://www.lrh.usace.army.mil/Potals/38/docs/USACE%2087%20Wet-
    land%20Delineation%20Manual.pdf; the Regional Supplement is availa-
    ble       at      https://usace.contentdm.oclc.org/utils/getfile/collection/
    p266001coll1/ id/7630; the NRCS Manual and Procedures are in the Ad-
    ministrative Record (AR) at 315–75.
    No. 16-1654                                                               9
    should be careful not to overweight the potentially “mislead-
    ing” presence of only hydric soils or wetland vegetation:
    “Many plant species can grow successfully in both wetlands
    and nonwetlands, and hydrophytic vegetation and hydric
    soils may persist for decades following alteration of hydrol-
    ogy that will render an area a nonwetland.” Corps Manual at
    6. Because “[i]ndicators of hydrophytic vegetation and hydric
    soil generally reflect a site’s medium- to long-term wetness
    history,” it is “[w]etland hydrology indicators [that] provide
    evidence that the site has a continuing wetland hydrologic re-
    gime and that hydric soils and hydrophytic vegetation are not
    relic[s] of a past hydrologic regime.” Corps Supplement at 68.
    Thus, the Corps Manual advises at page 6, a certification of
    “[t]he presence of hydric soils and wetland hydrology indica-
    tors in addition to vegetation indicators will provide a logical,
    easily defensible, and technical basis for the presence of wet-
    lands.” (Emphasis added.) 4
    4 Agency guidance repeatedly emphasizes this need to account for all
    three criteria to conform to the statute’s definition of wetlands, which was
    not followed here. See, e.g., NRCS Procedures (3-2) (definition of “wet-
    land” is “unique to the statute, and all decisions regarding the identifica-
    tion of wetlands must be based on this definition”); Corps Manual at 9
    (“The interaction of hydrology, vegetation, and soil results in the devel-
    opment of characteristics unique to wetlands [which] may be bordered by
    both wetter areas (aquatic habitats) and by drier areas (nonwetlands).”);
    NRCS Procedures (4-3) (“The ultimate decision if a site meets the [wet-
    land] criteria for any of the three diagnostic factors is made from a pre-
    ponderance of the evidence, best professional judgment, and the [agency]
    definitions, criteria, or both of hydrophytic vegetation, hydric soils, and
    wetland hydrology.”); NRCS Procedures (4-4) (a decision must be “based
    on the determination of a presence or absence of each of the three factors
    under [normal conditions]” and “wetlands” are only those “[a]reas
    10                                                        No. 16-1654
    As in this case, NRCS also may need to determine whether
    a specific area should be deemed a converted wetland after
    changes have been made to the land’s soil, vegetation, and/or
    hydrology. In these “atypical situations,” NRCS agents have
    guidance they can and should follow to determine whether
    such activity has indeed converted a wetland. Corps Manual
    at 73–74; Corps Supplement at 100 (“Atypical situations are
    wetlands in which vegetation, soil, or hydrology indicators
    are absent due to recent human activities or natural events.”);
    NRCS Procedures (2-4), (2-15) (again, alteration must have oc-
    curred after 1985).
    NRCS has some “flexibility” when “[s]tandard sampling
    methodology must be modified if it does not accurately assess
    the site,” but even in non-standard situations, “the basic ap-
    proach for making wetland determinations should not be al-
    tered (i.e., the determination should be based on the vegeta-
    tive, soil, and hydrologic characteristics of the area in ques-
    tion).” NRCS Procedures (5-5). “Any variation from the sam-
    pling methods should be fully documented.” Id. Agency guid-
    ance addresses typical and atypical situations when NRCS
    agents assess a parcel of land’s soil, vegetation, and water.
    1. Hydric Soil
    The presence of hydric soil alone is not controlling: “not
    all areas having hydric soils will qualify as wetlands.” Corps
    determined to support wetland hydrology, a prevalence of hydrophytic
    vegetation, and a predominance of hydric soils”); NRCS Manual § 514.3
    (“Wetlands are identified through the confirmation of three diagnostic
    factors: (i) A predominance of hydric soil; (ii) The prevalence of hydro-
    phytic vegetation; and (iii) Wetland hydrology”); NRCS Manual § 514.3(2)
    (“Wetlands must meet the definition of all three wetland factors (soils,
    plants, hydrology).”); NRCS Manual § 514.2H.
    No. 16-1654                                                  11
    Manual at 21. “Only when a hydric soil supports hydrophytic
    vegetation and the area has indicators of wetland hydrology
    may the soil be referred to as a ‘wetland’ soil.” Id. If NRCS
    agents believe that ground or surface waters (i.e., hydrology)
    have been removed from otherwise hydric soils, they are in-
    structed to look for “Onsite evidence of drained soils,” such
    as the presence of ditches or canals, dikes, levees, or similar
    structures, or the “Presence of a tile system to promote sub-
    surface drainage.” Id. at 21–22. Whether such drainage will
    preclude a wetland designation depends on how successful it
    is; if such areas are not sufficiently drained, they “may con-
    tinue to have wetland hydrology.” Id. at 22.
    If NRCS agents face an atypical situation in which they be-
    lieve hydric soils themselves have been altered, they follow
    three investigative steps. First, they must describe the altera-
    tion (e.g., dredging, removal of surface layers). Corps Manual
    at 77. Then they should determine the date of the alteration
    and finally, characterize the soils that were present before the
    alteration. Id. at 77–79.
    2. Hydrophytic Vegetation
    The presence of hydric soil alone is not controlling, and
    neither is the growth of some potentially hydrophytic vegeta-
    tion—i.e., the type of plants found “growing in water or in
    soil too waterlogged for most plants to survive.” Webster’s
    Third New Int’l Dictionary 1109 (1993).
    The vegetation that NRCS agents encounter runs along a
    spectrum from species that will grow only in wetlands (called
    Obligate Wetland Plants, which occur more than 99% of the
    time in wetlands) to species that typically cannot survive in a
    wetland’s anaerobic soil conditions (called Obligate Upland
    12                                                           No. 16-1654
    Plants, which occur less than 1% of the time in wetlands).
    Corps Manual at 13–14 & Table 1. The categories of plants that
    lie between these extremes include Facultative Wetland Plants
    (occurring approximately 67% of the time in wetlands), Fac-
    ultative Plants (those “with a similar likelihood of occurring
    in both wetlands and nonwetlands”), and Facultative Upland
    Plants (occurring less than 33% of the time in wetlands). Id. 5
    Thus, the presence (or former presence) of some vegeta-
    tion should typically not control wetland status, except per-
    haps if it is Obligate Wetland or Obligate Upland vegetation.
    NRCS agents are provided with extensive guidance on the
    measurement and evaluation of vegetation in order to rely on
    it as a wetland indicator. See, e.g., Corps Supplement at 15–
    31. Key in such determinations are the “hydrologic factors
    [that] exert an overriding influence on [plant] species that oc-
    cur in wetlands.” Corps Manual at 13.
    NRCS agents may face atypical situations in which vege-
    tation has been removed or altered. For example, the predom-
    inant land use in the Midwest is agricultural. “Wetlands used
    for agriculture may be considered atypical because they gen-
    erally lack a natural plant community and may be planted in
    crops or pasture species or altered by mowing, grazing, or
    other management practices.” Corps Supplement at 101. Such
    parcels of land may indeed be “wetlands [that] exhibit
    5 See also Corps Manual at 11 n.1 (“Some species, due to their broad eco-
    logical tolerances, occur in both wetlands and nonwetlands.”); Corps Sup-
    plement at 16 (acknowledging “the broad tolerances of certain plant spe-
    cies that allow them to be widely distributed across the moisture gradi-
    ent”); Corps Supplement at 21–22 (noting that “the five basic levels of wet-
    land indicator status (i.e., OBL, FACW, FAC, FACU, and UPL) are used in
    hydrophytic vegetation indicators”).
    No. 16-1654                                                   13
    indicators of hydric soil and wetland hydrology but lack any
    of the hydrophytic vegetation indicators.” Id. at 104.
    In such circumstances, before proceeding with a further
    assessment, NRCS must “[v]erify that at least one indicator of
    hydric soil and one primary or two secondary indicators of
    wetland hydrology are present.” Corps Supplement at 105.
    “[U]nless soil and/or hydrology are also disturbed,” the ab-
    sence of “hydric soil or wetland hydrology” indicators means
    “the area is likely non-wetland.” Id. And if the NRCS agent
    believes that hydrology was also altered, then the agent
    should determine whether the disputed land could have wet-
    land hydrology and “[v]erify that the area is in a landscape
    position that is likely to collect or concentrate water,” such as
    a “[c]oncave surface (e.g., depression or swale).” Id. Thus, in
    such an atypical situation, an NRCS agent may proceed with
    a wetland determination only after the agent is convinced that
    the disputed area has: (1) hydric soil, (2) the requisite hydrol-
    ogy (or potential requisite hydrology, if altered) to support
    hydrophytic vegetation, and (3) hydrophytic vegetation that
    has been removed or altered.
    The agent’s first step is to “[e]xamine the area and describe
    the type of alteration that occurred” to the land’s vegetation.
    Corps Manual at 74–75. The agent can look for evidence of
    such alteration, including “clear cutting,” the use of drainage
    structures such as “dams [or] levees,” or the installation of
    drainage tiles. Id. Once the alteration has been identified, the
    agent’s second step is to “[d]etermine the approximate date
    when the alteration occurred.” Id. at 75.
    After reaching a conclusion on the type and date of alter-
    ation, the NRCS agent moves to the third step and describes
    the alteration’s effects on the vegetation and determines “the
    14                                                              No. 16-1654
    type of vegetation that previously occurred.” Corps Manual
    at 75. In some instances, direct evidence of the removed veg-
    etation may be available—perhaps through recent aerial pho-
    tography, in which species identification is “sometimes pos-
    sible,” or remnants of the plants are discovered through an
    onsite inspection (e.g., “stumps, roots”). Id.
    In a situation like the Boucher property, if no direct evi-
    dence is available, “[c]ircumstantial evidence … may some-
    times be obtained by examining the vegetation in adjacent ar-
    eas.” Corps Manual at 76. Critical to this case, in choosing a
    comparison site, the NRCS agent must select a site with “[1]
    the same topographic position, [2] soils, and [3] hydrology as
    the altered area.” Id. The instruction that a comparison site
    must have comparable hydrology is repeated throughout
    agency guidance, including a specific admonition directly on
    point here, because it was not followed with the Boucher
    property: NRCS agents may follow “Corps manual … STEP
    3” to select an “adjacent vegetation data source … ‘in the local
    area on the same hydric soil map unit,’” but that “comparison
    site should support hydrologic conditions that are similar to what
    existed on the altered site prior to the drainage.” NRCS Procedures
    (5-30) (emphasis added), quoting 
    7 C.F.R. § 12.31
    (b)(2)(ii). 6 In
    6 See also 
    7 C.F.R. § 12.31
    (b)(2)(i) (NRCS must determine if the land “un-
    der normal circumstances … supports a prevalence of hydrophytic vege-
    tation”; “normal circumstances” means “the soil and hydrologic conditions
    that are normally present, without regard to whether the vegetation has been re-
    moved”) (emphasis added); 
    7 C.F.R. § 12.31
    (b)(2)(ii) (if vegetation has been
    removed, NRCS can look to a site “on the same hydric soil map unit” that
    exists “under non-altered hydrologic conditions”—i.e., hydrologic conditions
    that are similar to the hydrologic conditions on the disputed site); NRCS
    Procedures (2-4) (NRCS agents can choose a “comparison site” on the
    No. 16-1654                                                                    15
    short, the existence or potential existence of sufficient hydrol-
    ogy—i.e., lots of water—to support hydrophytic vegetation is
    central to both the statutory definition of a wetland and all of
    the expert guidance for making an accurate wetland determi-
    nation.
    3. Hydrology
    “Numerous factors (e.g., precipitation, stratigraphy, to-
    pography, soil permeability, and plant cover) influence the
    wetness of an area,” the Corps Manual advises, but adds at
    page 29, in an eloquent example of technical and bureaucratic
    understatement, that it is unavoidable that “the characteristic
    common to all wetlands is the presence of an abundant sup-
    ply of water.” Thus “it is essential to establish that a wetland
    area is periodically inundated or has saturated soils during
    the growing season.” 
    Id.
     Ground may be considered “inun-
    dated” if it “is covered by water due to ponding, flowing, or
    flooded water.” NRCS Manual § 514.2E. And soils are consid-
    ered saturated with a “[v]isual observation” of water that “is
    “same hydric soil map unit as the subject site … to make a decision on the
    presence of hydrophytic vegetation when the subject site is altered and the
    plant community that occurred prior to the alteration cannot be deter-
    mined,” but such “comparison site should support hydrologic conditions that
    are similar to what existed on the subject site prior to the alteration”) (emphasis
    added); NRCS Procedures (2-11); Corps Supplement at 102 (“to determine
    the plant community that would occupy the site under normal circum-
    stances, if the vegetation were not cleared or manipulated,” NRCS can
    “[e]xamine the vegetation on an undisturbed reference area with soils, hydrol-
    ogy, landscape position, and other conditions similar to those on the site”) (em-
    phasis added); Corps Supplement at 109 (if there has been “clearing of
    woody species on rangeland or pasture land,” the NRCS may “[e]xamine
    the vegetation on a nearby, unmanaged reference site having similar soils
    and hydrologic conditions”) (emphasis added).
    16                                                            No. 16-1654
    12 in. (30 cm) or less from the soil surface,” which “must be
    associated with an existing water table located immediately
    below the saturated zone.” Corps Supplement at 76–77.
    Evidence of hydrology sufficient to support a wetland can
    be direct or circumstantial. For example, NRCS agents are ad-
    vised to look for “primary indicators” of hydrology, which
    “provide stand-alone evidence of a current or recent hydro-
    logic event,” with even “[o]ne primary indicator [being] suf-
    ficient to conclude wetland hydrology is present.” Corps Sup-
    plement at 72. There is no shortage of primary indicators for
    which NRCS agents can test—again, any one of which is suf-
    ficient to support a finding of wetland hydrology—including
    the presence of surface water, a high water table, or satura-
    tion. Corps Supplement at 72–98; see also Corps Manual at
    30–32. 7
    Some of these indicators can be gleaned from recorded
    data that may be available from federal, state, and local agen-
    cies, or NRCS can collect data onsite. For example, NRCS
    agents can make a “[v]isual observation of inundation” or
    “soil saturation” by digging a soil pit and “observ[ing] the
    level at which water stands in the hole.” Corps Manual at 31–
    32. And lots of water, on or very close to the land’s surface, is
    required to qualify as a wetland. For “soil saturation to impact
    7 Other primary indicators include water marks on vegetation or fixed ob-
    jects, sediment deposits, drift deposits, algal mats or crusts, iron deposits,
    inundation visible on aerial imagery, sparsely vegetated concave surfaces,
    water-stained leaves, aquatic fauna, true aquatic plants, hydrogen sulfide
    odor, oxidized rhizospheres along living roots, presence of reduced iron,
    recent iron reduction in tilled soils, thin muck surface, and gauge or well
    data. Corps Supplement at 72–98; Corps Manual at 30–32.
    No. 16-1654                                                                17
    vegetation, it must occur within a major portion of the root
    zone (usually within 12 inches of the surface).” Id. at 32. 8
    Therefore, digging a several-foot-deep hole and observing
    water pool only at the bottom—away from the root zone—
    would not show that the soil could support hydrophytic veg-
    etation.
    Absent at least one of the many “primary indicators,”
    NRCS agents can also look for “secondary indicators,” of
    which two or more “must be present to conclude that wetland
    hydrology is present.” Corps Manual at 34. Examples of sec-
    ondary indicators include an “FAC-neutral test of the vegeta-
    tion,” the “local soil survey hydrology data,” or “[g]eo-
    morphic position” of the area at issue. Id. and Corps Supple-
    ment at 98–99. 9 As relevant to this case, the geomorphic posi-
    tion indicator “is present if the immediate area in question is
    8 These hydrology criteria are related to the definitions of hydric soil and
    hydrophytic vegetation. Recall, to be considered hydric, the undrained
    soil must be “saturated, flooded, or ponded long enough … to develop an
    anaerobic condition” that supports wetland vegetation. 
    16 U.S.C. § 3801
    (a)(12); 
    7 C.F.R. § 12.2
    . And to be considered hydrophytic, the vege-
    tation must be capable of growing “in water or in a substrate that is at least
    periodically deficient in oxygen … as a result of excessive water content.”
    
    7 C.F.R. § 12.2
    ; 
    16 U.S.C. § 3801
    (a)(13). Therefore, if one digs a ditch and
    the observed water sits 12 inches or more below the surface, then such
    water will not be sufficient to reach or support the roots of hydrophytic
    plants.
    9 The “FAC-neutral test” looks to see whether more than 50% of dominant
    plant species in a given area are Facultative, Facultative Wetland, or Obli-
    gate Wetland Plants—even if the remaining vegetation falls into the Fac-
    ultative Upland and Obligate Upland Plant categories. If so, the vegetation
    can be used as a “secondary indicator” that the area may have wetland
    hydrology. Corps Manual at 17–18, 57–58 & n.1. See also Corps Supple-
    ment at 99.
    18                                                   No. 16-1654
    located in a depression.” Corps Supplement at 98. But absent
    evidence that hydrology has been drained, removed, or di-
    verted (as described below), here is the critical bottom line: “If
    no positive wetland hydrologic indicator is present, the area
    at the observation point is not a wetland.” Corps Manual at
    59.
    As with soil and vegetation, it is possible that hydrology
    has been altered or removed. In that case, the agent may fol-
    low the same investigative steps for atypical situations out-
    lined above regarding vegetation. It bears repeating that hy-
    drology must be determined independently because soil and
    vegetation indicators alone may not be reliable. Thus, if NRCS
    agents believe—as they did when assessing the Boucher’s
    farm—that a site’s “hydrology has been altered (increased or
    decreased) by manipulations,” then the “consideration of
    wetland hydrology is particularly important” because “some
    plant communities are slow to respond to changes in hydrol-
    ogy” and “soil features are very resistant to change.” NRCS
    Procedures (5-58) (“it becomes important that the agency ex-
    pert confirms that under [normal conditions] wetland hydrol-
    ogy is still present at the time of rendering [an agency] wet-
    land identification decision”).
    In such a situation, “STEP 1” is for the NRCS agent to
    “[e]xamine the area and describe the type of alteration that
    occurred.” Corps Manual at 80. Evidence of alteration such
    that “sufficient ground or surface water has been removed by
    artificial means” may include the construction of dams or lev-
    ees, ditching, land-leveling, diversion of water, ground-water
    extraction through pumping, or the “[p]resence of a tile sys-
    tem to promote subsurface drainage.” 
    Id. at 22, 80
    . As always,
    the agent must then “[d]etermine the approximate date when
    No. 16-1654                                                   19
    the alteration occurred.” 
    Id. at 80
    . Note the relevance of drain-
    age tile, for it figures prominently in this case.
    “STEP 2” requires the NRCS agent to record “a general de-
    scription of how the alteration … has affected the area,” such
    as whether there is more or less frequent inundation, or
    whether the duration of inundation and soil saturation has
    changed. Corps Manual at 80–81. In “STEP 3” the agent must
    “[o]btain all possible evidence that may be used to character-
    ize the hydrology that previously occurred.” 
    Id. at 81
    . The
    Corps Manual offers six potential sources of information that
    NRCS agents should consult if possible, including aerial pho-
    tography “taken during the growing season” that shows
    “whether the area was inundated,” stream data, historical rec-
    ords, floodplain management maps, and interviews of knowl-
    edgeable public officials. 
    Id.
     81–82. As relevant to this case,
    another potential source of information is “[f]ield hydrologic
    indicators” such as watermarks on structures or, “[i]f adjacent
    undisturbed areas are in the same topographic position and are
    similarly influenced by the same sources of inundation,” then the
    agent can “look for wetland indicators in these [comparison]
    areas.” 
    Id. at 81
     (emphasis added). If, however, after consult-
    ing all available sources outlined above, the NRCS agent con-
    cludes that “no indicators of wetland hydrology were found,”
    then “the original hydrology of the area was not wetland hy-
    drology.” 
    Id. at 82
    .
    Drawing from the above guidance for both typical and
    atypical situations, the NRCS agent must then certify a “wet-
    land determination [that] is of sufficient quality to make a de-
    termination of ineligibility for program benefits.” 
    7 C.F.R. § 12.30
    (c)(1). That is at least what is supposed to happen. We
    20                                                No. 16-1654
    now move to the facts of this case, which strayed far from
    these norms.
    II. The USDA and the Boucher Farm
    A. Round One (1980s–2003)
    The Boucher farm in Hancock County, Indiana, has been
    used to produce livestock and grain for more than 150 years.
    The Bouchers purchased the farm in the late 1980s, just after
    the Swampbuster provisions went into effect. The Bouchers
    received a USDA notice in November 1987 that the farm con-
    tained hydric soils and thus had potential for a wetland clas-
    sification. But a National Wetland Inventory taken in 1989 did
    not identify any wetlands on the Boucher property.
    Around 1994, Mr. Boucher began to clean up two areas of
    the farmland, formally designated in the record with the Dr.
    Seuss-like names Un1 and Un2. These two fields were being
    used for illegal dumping by persons unknown. In 1994, Mr.
    Boucher removed five trees to reduce cover for covert dump-
    ing. A few years later, he removed four more trees. All told,
    the nine trees had occupied approximately 12/10,000ths of an
    acre. As described in the later agency hearing, one could fill
    an area the size of a pick-up truck bed by laying flat a slice
    from each of the nine trees’ trunks.
    In June 2002, a USDA representative visited the Boucher
    farm to consider a request to establish a conservation filter
    strip around the perimeter of the farm. The USDA representa-
    tive reported a potential wetland violation based on Mr. Bou-
    cher’s tree removal. FSA directed NRCS to investigate. An
    NRCS resource conservationist, Karen Hauer, completed a
    routine wetland determination for Un1 and Un2 in late 2002.
    No. 16-1654                                                             21
    Only brief notes are provided in the record, but Ms.
    Hauer’s form indicates that the site was not “significantly dis-
    turbed” and did not present an “Atypical Situation.” She be-
    gan her investigation by following the above-outlined agency
    guidance and assessing the hydrological properties of Un1
    and Un2. She found no surface water. Any soil saturation was
    at a depth of greater than 12 inches (i.e., not close enough to
    the root zones to support hydrophytic vegetation). Despite
    the lack of hydrology or a significant disturbance (making for
    an “atypical” situation), Ms. Hauer apparently assumed the
    hydrology had been drained through the installation of tile.
    When agency experts were later asked why they assumed the
    fields had been tiled—as we will see, they had not been—the
    reason given was “because typically these fields are tile
    drained” and “they would not be farmable normally without
    drainage.”
    Ms. Hauer thus selected a comparison area in an “off-site
    adjacent” area. She chose “Field 7,” which was entirely un-
    suitable for comparison. It was an unfarmed area in a depres-
    sion and was indisputably a wetland. 10 If Ms. Hauer believed
    that she needed a comparison site due to alteration of hydrol-
    ogy in Un1 and Un2, the process for testing that hypothesis is
    outlined above: verify that such an alteration was made and
    the date of that alteration, assess the effect of that alteration,
    and then obtain all possible evidence to characterize the pre-
    vious hydrology (including the use of primary and secondary
    indicators). Her assessment did not make it past step 1. Not
    10 There is no dispute that Field 7 is a wetland of some type. The Bouchers
    originally challenged NRCS’s determination that Field 7 was not an “arti-
    ficial wetland,” but were unsuccessful on that claim in the underlying
    agency proceedings and have not raised the issue on appeal.
    22                                                No. 16-1654
    surprisingly, Ms. Hauer had no trouble finding the requisite
    hydrology or hydrophytic vegetation on Field 7’s wetland, in-
    cluding plant species that were almost all categorized as
    FACW (very likely to occur in wetlands), and just two FAC
    species (just as likely to occur in wetlands as not).
    NRCS sent Mr. Boucher the preliminary determination on
    February 7, 2003, which concluded that Un1 and Un2 con-
    tained 2.8 acres of converted wetland. The agency was under
    the impression that Mr. Boucher planned to accept a remedi-
    ation plan, and on February 27, 2003 mailed him a plan that
    would have required Mr. Boucher to plant 300 trees per acre
    to compensate for the removal of the nine trees. Mr. Boucher,
    however, decided to challenge the decision. He hired an attor-
    ney and requested reconsideration and a site visit. Mr. Bou-
    cher did not feel that the follow-up visit was “adequate,” and
    on April 17, 2003, he appealed to the Farm Service Agency
    (FSA) County Committee to request a review by the NRCS
    State Conservationist. The FSA Committee docketed his re-
    quest, but noted that the committee “must feel that there is
    merit to your request for the wetland review” to be ordered.
    Mr. Boucher pointed out that the only trees he removed
    had been Facultative Upland Plants—i.e., those unlikely to be
    found in wetlands—and that no leveling or drainage work
    had been performed on this site. The FSA Committee found
    merit in the appeal and the case was referred to the NRCS
    State Conservationist for a further site visit and assessment.
    The Bouchers recalled meeting with State Conservationist
    Phil Bousman on September 9, 2003 and communicated about
    the planned meeting with their attorney. The USDA also has
    records that Mr. Bousman was in communication with Ms.
    Hauer for the Boucher aerial slides in anticipation of the site
    No. 16-1654                                                    23
    visit. The USDA, however, “has no documentation that the
    site visit ever took place,” with a USDA representative ex-
    plaining that those records seem to have “disappeared.” Hrg.
    Tr. 45.
    Mr. Boucher left his meeting with the State Conservation-
    ist “feeling that he had proved his position.” No further com-
    munication was received from the USDA for nearly ten years,
    until late 2012. In February 2004, Mr. Boucher passed away.
    B. Round Two (2012–2013)
    Mrs. Boucher took over the farm and leased it to a new
    tenant. In 2012, her tenant asked if he could remove an old
    house and barn from a portion of the property called Field 8.
    Mrs. Boucher agreed and on July 2, 2012, sought permission
    from the USDA for that removal. Unfortunately, as we will
    see, this request prompted the USDA to discover in Novem-
    ber 2012 that it had never completed a Final Technical Deter-
    mination for Un1 and Un2 after Mr. Boucher’s 2003 FSA ap-
    peal for a State Conservationist’s visit.
    A new site visit was scheduled for early 2013 to appraise
    Field 8, per the recent request, at which time the 2003 prelim-
    inary determination for Un1 and Un2 could be finalized. Con-
    ditions for the site visit were unusual. Over three inches of
    rain fell on January 13 and January 14, the day of the visit.
    That rain melted eleven inches of snow on the ground. NRCS
    sent a soil scientist, an area easement specialist, and a district
    conservationist, but no hydrologist, to assess the Boucher
    farm.
    On this visit, the NRCS experts observed standing water
    and puddles in several fields, which can of course serve as a
    “primary indicator” of wetland hydrology. Corps
    24                                                No. 16-1654
    Supplement at 73. According to agency guidance applicable
    to this situation, “Care must be used” when basing a determi-
    nation on such observations “because surface water may be
    present in non-wetland areas immediately after a rainfall
    event or during periods of unusually high precipitation,” and
    absent “other hydrology indicators … a follow-up visit dur-
    ing the growing season may be needed.” 
    Id.
     The agency ex-
    perts did not schedule a follow-up visit but did snap some
    pictures of puddled fields they believed to be Un1 and Un2
    and noted in their assessment form the (unsurprising) “evi-
    dence that water collected at the surface after heavy rains.”
    A critical point, as we will see, was that the experts also
    noted that “Drainage tile has been added to the site,” and thus
    “Hydrology data from comparison site Field 7 … was used
    for hydrology as per policy.” App. 52. In a separate place on
    the assessment forms, the experts again recorded their belief
    that “[t]he site has been cleared, drained, and is currently
    farmed.” App. 51 (emphasis added).
    Based on these assumptions, the experts in 2013 concurred
    with Ms. Hauer’s 2002 decision to use “[a] comparison site
    (reference site) to the south (Field 7) … to determine Vegeta-
    tion and Hydrology as per policy” outlined in 
    7 C.F.R. § 12.31
    (b)(2)(ii) and the Corps Manual guidance on atypical
    situations. (Emphasis added.) The experts agreed with Ms.
    Hauer’s 2002 assessment that Field 7 has a “similar landscape
    position and soil type” as Un1 and Un2. The vegetation ob-
    served on Field 7 was indeed hydrophytic vegetation, with al-
    most all species categorized as FACW—i.e., the types of spe-
    cies likely to occur in wetlands—and one species of FAC
    (equally likely to be in wetlands or not) and one FACU (less
    likely to be in wetlands).
    No. 16-1654                                                            25
    The agency issued on January 30, 2013 a new preliminary
    technical determination that Un1 and Un2 were converted
    wetlands. And on March 1, 2013, the NRCS issued its final
    technical determination, affirming its 2003 preliminary deter-
    mination, but slightly reducing the acreage it had designated
    in 2003 as converted wetland to 2.6 acres total—0.7 acres on
    Un1 and 1.9 acres on Un2. The NRCS explained that during
    its site visit to “confirm the original findings … in 2003,” it
    used Field 7 as a “reference site” to evaluate Un1 and Un2 be-
    cause “the vegetation [was] no longer visible for identification
    and drainage tile has been added to these locations.” App. 45
    (emphasis added). 11
    After some prompting, the USDA notified Mrs. Boucher of
    its final determination on March 27, 2013. Mrs. Boucher ap-
    pealed the determination to the USDA’s National Appeals Di-
    vision. On May 3, 2013, NRCS completed an “Atypical Situa-
    tion Data Sheet” to document its use of Field 7 as a compari-
    son site for Un1 and Un2. In case there was any doubt about
    the basis for the NRCS experts’ use of a comparison site, and
    even though the agency has been trying to run away from this
    evidence for years now, the Atypical Situation Data Sheet
    showed clearly that the experts believed Un1 and Un2 were
    “drained allowing crop production,” and thus “Hydrology
    11 This language implies that the USDA personnel in 2013 were not aware
    of the 2003 appeal and the evidence Mr. Boucher had provided to contra-
    dict Ms. Hauer’s preliminary findings. In the agency hearings and the dis-
    trict court, the USDA blamed the Bouchers for the passage of time and the
    disappearance of USDA records, claiming the Bouchers should have real-
    ized the USDA had never closed its file after Mr. Boucher provided the
    evidence showing that the Un1 and Un2 parcels were not converted wet-
    lands. Dkt. 49 at 24–25. On appeal, at least, the government has dropped
    this unusual line of argument.
    26                                                   No. 16-1654
    was removed.” They also noted that “Soils were minimally
    disturbed from land clearing and the site was not filled.” App.
    57. The experts had thus used a “Comparison Site (reference
    site)” in order “to determine Vegetation and Hydrology,” re-
    lying upon the above-detailed agency guidance manuals.
    C. Agency Hearings
    1. National Appeals Division Hearing and Decision
    On appeal to the National Appeals Division, Mrs. Boucher
    needed to show that the “agency’s adverse decision [was] er-
    roneous by a preponderance of the evidence,” which she
    could do by submitting additional evidence. 
    7 C.F.R. § 11.8
    (e).
    Mrs. Boucher submitted evidence that Un1 and Un2: (1) did
    not appear to have any drainage tiling; (2) to the extent they
    did have any tiling, the tiles were installed before 1985; (3) did
    not demonstrate inundation or saturation with water;
    (4) were not in a depression; and (5) previously had trees that
    were not hydrophytic.
    The agency continued to assert that Un1 and Un2 had been
    drained and that drainage required the use of Field 7 (the un-
    farmed wetland in a depression) as a comparison site. In its
    pre-hearing filing, the agency explained that at the time of the
    site visit, “Hydrology was noted as being altered,” and
    “therefore, reference site hydrology data and secondary indi-
    cators noted provided the support to meet the wetland hy-
    drology criteria.” An evidentiary hearing was conducted by
    telephone on June 20, 2013.
    Mrs. Boucher hired a company, B. Thompson Associates
    (Thompson), to make an independent assessment of Un1 and
    Un2. Relying on the Purdue University Indiana Drainage
    Guide for the best method to expose drainage tile, Thompson
    No. 16-1654                                                 27
    used a trenching machine to dig trenches five and a half feet
    deep (a foot or two deeper than the Guide’s recommended
    minimum level). Thompson dug those trenches in large cross
    patterns that reached the full distance across both Un1 and
    Un2.
    The trenches discovered no drainage tile. What’s more, in
    the wetter-than-normal spring of 2013, the open trenches did
    not pond or demonstrate any saturation. The water table
    never rose above the four-foot mark—far below the maxi-
    mum twelve inches from the surface required for wetland hy-
    drology. This evidence was consistent with NRCS agent Ka-
    ren Hauer’s 2002 assessment, before she assumed—incor-
    rectly—that drainage tile had been installed.
    In response, the Agency’s Appeals Coordinator contended
    that the cross-shaped trenches were dug into the middle of
    Un1 and Un2 and that—because hydrology still appeared ab-
    sent—she believed there must be tiling around the periphery
    of Un1 and Un2 responsible for the change in hydrology. Such
    an arrangement of tiling, however, would not be effective in
    draining water from a saturated field. Thompson and the Pur-
    due Drainage Guide advised that effective drainage tiling
    needs to be in a grid pattern with tiles (i.e., pipes) that are
    spaced no more than 70 to 120 feet apart.
    To be sure, a few pieces of tile had been recovered, but
    they were made of concrete. That type of tile was not used
    after the early 1980s when, if wetlands had already been con-
    verted, the lands were no longer considered wetlands by fed-
    eral law. Mrs. Boucher also secured a note from the prior own-
    ers of the farm, who explained that any drainage tile on the
    property was installed around 1981 or 1982.
    28                                                 No. 16-1654
    Mrs. Boucher also pointed out that, to the extent the
    agency experts believed they had taken photographs of stand-
    ing water in fields Un1 and Un2 during the wet January 14,
    2013 visit, they had been disoriented as to several locations on
    the property—including for those areas in the pictures. The
    pictures believed to be of Un1 and Un2 actually showed
    standing water primarily on a different portion of the farm—
    Field 1—that had been determined to be prior converted non-
    wetland. Hrg. Tr. 25–26 & 48.
    Thompson (the company that trenched Un1 and Un2) also
    conducted a GPS survey of the property. The survey con-
    firmed that Un1 and Un2 are not in a depression and thus can-
    not meet the depressional geomorphic position criterion. Hrg.
    Tr. 25. In fact, the only depressional field was the wetland in
    Field 7, the comparison site selected by the NRCS experts. At
    the risk of stating the obvious, this was strong evidence that
    Field 7 did not provide a relevant comparison to Un1 and
    Un2.
    Mrs. Boucher also contested the agency’s assumption that
    the nine removed trees were hydrophytic. She provided the
    agency with a letter from Mr. Boucher to their attorney that
    explained: “The species of tree that I removed are consistent
    with Facultative Upland and Soils developed under Prairie
    vegetation soils, ‘trees can be removed’ without an Act of
    Conversion.” AR 404. In fact, she had contested the hydro-
    phytic nature of the nine trees throughout the agency process.
    See, e.g., App. 87 (“My husband documented in his notes the
    trees removed were non-hydrophytic”); App. 88 (“According
    to my husband the trees removed were non-hydrophytic” and
    “were not wetlands to begin with”); App. 94 (“hydrophytic
    [v]egetation was not found [on a comparable parcel] as
    No. 16-1654                                                  29
    always contended by my husband”); App. 100 (“My husband
    cleared 9 non-hydrophytic trees over 6 years in the farmstead
    area referred to as Fields Un1 and Un2. The brush being re-
    ferred to is actually herbaceous plants not woody plants.”);
    App. 105 (“I proved the only woody vegetation removed were
    nine non-hydrophytic trees and herbaceous vegetation.”).
    The only response from the agency was an assertion that some
    trees on the Field 7 comparison site, although classified as
    FACW and not FACU, were in the same “family” as the re-
    moved FACU trees. Hrg. Tr. 40.
    The agency thus faced substantial evidence that one of two
    situations was true. The first was that there was insufficient
    (or perhaps no) drainage tile on Un1 or Un2, so that hydrol-
    ogy had never existed on these plots to make them wetlands.
    That finding would be consistent with the trenching, topog-
    raphy measurements, and the older evidence that the re-
    moved trees were not hydrophytic. The second, less likely sit-
    uation, was that the agency experts were correct that drainage
    tile had removed the hydrology that would have supported
    hydrophytic vegetation, but that the tile was installed before
    1985, meaning that Un1 and Un2 could not legally be deemed
    wetlands. In either situation, Un1 and Un2 were not wetlands.
    Rather than grappling with this evidence, the hearing of-
    ficer used transparently circular logic, asserting that the
    agency experts had appropriately found hydric soils, hydro-
    phytic vegetation, and wetland hydrology, “using a similar
    adjacent property [i.e., Field 7, the wetland in the depression]
    because [the] Property was converted and no longer had any
    natural fauna.” App. 42. Without mentioning Mrs. Boucher’s
    evidence that the removed trees were FACU—i.e., a category
    of trees unlikely to be found in wetlands—the hearing officer
    30                                                No. 16-1654
    found that Mrs. Boucher “failed to prove what type of natural
    vegetation was native to the Subject Property and did not
    show that [NRCS] inaccurately chose the species of natural
    vegetation that would have grown on” Un1 and Un2.
    Oddly, the decision did not mention the agency experts’
    erroneous assumption from 2002 forward that Un1 and Un2
    had been drained. Nor was there an effort to account for the
    role hydrology should have played, per all the agency guide-
    lines, in the agency experts’ assessment. The hearing officer
    did not mention, let alone reconcile, Mrs. Boucher’s evidence
    that Un1 and Un2 had not been drained and still did not
    demonstrate sufficient inundation or saturation. Nor did the
    hearing officer note the site topography study, which showed
    that Un1 and Un2 were not in a depression, unlike Field 7—
    which meant, under all applicable expert guidance for the
    agency, that the comparison site could not serve as a “similar
    adjacent property.” 
    Id. 2
    . Agency Director Review
    Mrs. Boucher then requested a Director review of the hear-
    ing officer’s decision. She again protested that Field 7 was a
    wetland and not comparable to Un1 and Un2. She also ob-
    jected that the “Hearing Officer did not adequately con-
    sider … the following of the program regulations that apply
    in this situation,” nor did the officer “cite any of the evidence
    and arguments I submitted or [gi]ve any reason he considered
    them not to be valid.”
    In response, the agency continued to “disagree[] that these
    areas have not been impacted by drainage” and insisted that
    “[d]rainage did exist on the tract as documented by the clay
    tile.” (The source of this assertion remains a mystery. Again,
    No. 16-1654                                                   31
    the tile fragments found on the farm were cement, not clay,
    and to the extent there was any tiling, it would have been in-
    stalled before 1985.) Any evidence offered by Mrs. Boucher to
    the contrary did “not prove the hydrology of the site has not
    been drained/altered.” AR 165. The agency now also asserted,
    apparently for the first time, but important on judicial review,
    that “Removal of woody vegetation … is considered a hydro-
    logical alteration.” AR 163. The agency further disputed Mrs.
    Boucher’s evidence that the removed trees were non-hydro-
    phytic because: (1) different species of trees, but possibly
    those in the same family as the removed trees, were located
    on Field 7 and classified as FACW; and (2) it could have been
    possible for “trained Agency staff” to identify the removed
    trees from aerial photos (although there was no suggestion
    they had done so).
    The following month, an agency deputy director found
    that the decision was “supported by substantial evidence”
    and affirmed. App. 29–30, 36. In an overdue admission, the
    deputy director abandoned any underlying NRCS expert de-
    termination that Un1 and Un2 had tiling or any drainage,
    finding: “No drainage has been installed since the early
    1980s.” App. 30. In fact, the deputy director’s factual “Back-
    ground” statement recounting NRCS’s wetland assessment
    omitted any reference to the agency experts’ conclusions re-
    garding, and reliance upon, the existence of drainage tile,
    quoting only the portions of the record and documentation
    that referred to tree removal. App. 30–32.
    The deputy director acknowledged that Mrs. Boucher
    “specifically argues that Un1 and Un2 do not exhibit the re-
    quired hydrology or hydrophytic vegetation necessary to
    identify land as wetland.” Relying on 
    7 C.F.R. § 12.31
    (b)(2)(ii),
    32                                                   No. 16-1654
    the deputy director concluded that use of a comparison site
    was warranted once the agency reviewed slides of aerial pho-
    tography that showed some trees had been removed. Aston-
    ishingly, the deputy director wrote that all that was required
    of the comparison site was that it be on the same hydric soil map.
    App. 34. Further, Un1 and Un2 were appropriately consid-
    ered “converted wetland” because, per 
    7 C.F.R. § 12.2
    (a), all
    that is required for the agency to deem land “converted wet-
    land” is proof that “woody hydrophytic vegetation has been
    removed from hydric soils.” App. 35. It made no difference
    that Mrs. Boucher “successfully argued below, and Agency
    conceded, that drainage did not exist on Un1 and Un2 as
    Agency originally claimed.” Because aerial photography
    showed some type of woody vegetation had been removed,
    Un1 and Un2 “were correctly labeled” converted wetland.
    App. 35.
    The deputy director did not mention, let alone reconcile,
    Mrs. Boucher’s evidence that: (1) even without any drainage
    on Un1 and Un2, the fields did not demonstrate sufficient in-
    undation or saturation; (2) the site topography study revealed
    that Un1 and Un2 were not in a depression, unlike Field 7;
    and (3) the removed trees were FACU—i.e., trees unlikely to
    be found in wetlands.
    D. District Court Proceedings
    Mrs. Boucher sought judicial review in the Southern Dis-
    trict of Indiana under the Administrative Procedure Act, 
    5 U.S.C. § 702
    . The parties filed cross-motions for summary
    judgment based on the administrative record. Mrs. Boucher
    repeated her arguments from the agency proceedings. By this
    time, the agency stopped mentioning drainage or drainage
    tile at all. The agency argued that any removal of vegetation
    No. 16-1654                                                      33
    permits use of a comparison site that is merely on the same
    hydric soil map, and a finding that woody vegetation has
    been removed is sufficient to deem an area “converted wet-
    land.” Dkt. 49 at 6, 12, 17, citing 
    7 C.F.R. §§ 12.31
    (b)(2)(ii) and
    12.2(a).
    The district court rejected Mrs. Boucher’s argument that
    any tiling was present prior to 1985 because that argument
    “ignore[d] the other evidence that she herself presented dur-
    ing the administrative proceedings”—namely, that there was
    no tiling in Un1 and Un2. App. 16. In some logical tension
    with that holding, though, the district court rejected Mrs. Bou-
    cher’s objection to Field 7 as a comparison site because the
    agency’s evidence indicated that Un1 and Un2 “had been
    cleared, drained, and cropped, which necessitated the use of a
    comparison or reference site.” App. 18 (emphasis added).
    Concluding that “NRCS followed the required regulatory
    procedures [and] appropriate technical manuals,” as well as
    “completed the necessary wetland determination data forms
    and atypical situation data sheet to document and support its
    work and findings,” the district court held that sufficient evi-
    dence supported the agency’s wetland determination. The
    district court granted summary judgment for the USDA.
    III. Legal Standards
    We review de novo an appeal from a summary judgment
    disposition of an administrative action. Habitat Education Cen-
    ter, Inc. v. U.S. Forest Service, 
    673 F.3d 518
    , 525 (7th Cir. 2012).
    And we may overturn the underlying final decision of the
    USDA only if it is “arbitrary, capricious, an abuse of discre-
    tion, or otherwise not in accordance with law.” 
    5 U.S.C. § 706
    (2)(A). This deferential standard of review stems from
    Congress’s directive to defer to an agency’s expertise because
    34                                                    No. 16-1654
    “an agency must have discretion to rely on the reasonable
    opinions of its own qualified experts even if, as an original
    matter, a court might find contrary views more persuasive.”
    Marsh v. Oregon Natural Resources Council, 
    490 U.S. 360
    , 378
    (1989).
    But even deferential review should “be searching and
    careful” when considering “whether the decision was based
    on a consideration of the relevant factors and whether there
    has been a clear error of judgment.” 
    Id.,
     quoting Citizens to
    Preserve Overton Park, Inc. v. Volpe, 
    401 U.S. 402
    , 416 (1971); see
    also Motor Vehicle Manufacturers Ass’n v. State Farm Mutual Au-
    tomobile Ins. Co., 
    463 U.S. 29
    , 43 (1983) (standard requires that
    “the agency must examine the relevant data and articulate a
    satisfactory explanation for its action including a ‘rational
    connection between the facts found and the choice made’”),
    quoting Burlington Truck Lines, Inc. v. United States, 
    371 U.S. 156
    , 168 (1962). To fulfill this responsibility, we must “review
    the entire record” and, “even if we disagree with an agency’s
    action,” we must “uphold the action if the agency considered
    all of the relevant factors and we can discern a rational basis
    for the agency’s choice.” Israel v. U.S. Dep’t of Agriculture, 
    282 F.3d 521
    , 526 (7th Cir. 2002); see also Marsh, 
    490 U.S. at 378
    (courts must “carefully review[] the record and satisfy[] them-
    selves that the agency has made a reasoned decision,” includ-
    ing an “evaluation of the significance—or lack of signifi-
    cance—of [any] new information”); see also 
    5 U.S.C. § 706
    (2)
    (in reviewing agency decision, “the court shall review the
    whole record or those parts of it cited by a party”).
    An agency decision will be found “arbitrary and capri-
    cious” if it “relied on factors which Congress has not intended
    it to consider, entirely failed to consider an important aspect
    No. 16-1654                                                  35
    of the problem, offered an explanation for its decision that
    runs counter to the evidence before the agency, or is so im-
    plausible that it could not be ascribed to a difference in view
    or the product of agency expertise.” State Farm, 
    463 U.S. at 43
    .
    If a reviewing court is faced with “such deficiencies,” it
    “should not attempt itself to make up” for those gaps by “sup-
    ply[ing] a reasoned basis for the agency’s action that the
    agency itself has not given.” 
    Id.
     “A contrary approach
    would … render judicial review generally meaningless.”
    Marsh, 
    490 U.S. at 378
    ; see also SEC v. Chenery Corp., 
    318 U.S. 80
    , 92 (1943) (declining to uphold administrative decision
    when “the considerations urged [on appeal] in support of the
    [agency’s] order were not those upon which its action was
    based”).
    IV. Analysis
    With that legal framework in mind, we ask: Did the USDA
    examine relevant factors and relevant data, or articulate a sat-
    isfactory explanation for its action including a rational con-
    nection between the facts found and the choice made? Were
    the opinions of its own qualified experts reasonable, such that
    we should defer to their expertise? Did the agency evaluate
    the significance—or lack of significance—of any new infor-
    mation presented during the agency appeals process? The an-
    swers to all of these questions are no.
    The agency guidance addresses the situation the NRCS
    agents faced at the Boucher farm: fields on a Midwestern farm
    appeared to “lack a natural plant community and [were]
    planted in crops or pasture species or altered by mowing,
    grazing, or other management practices.” Corps Supplement
    at 101. Before proceeding further, NRCS agents were advised
    to “[v]erify that at least one indicator of hydric soil and one
    36                                                 No. 16-1654
    primary or two secondary indicators of wetland hydrology
    [we]re present.” Corps Supplement at 105. That step was cru-
    cial: “Unless soil and/or hydrology are also disturbed,” the
    absence of “hydric soil or wetland hydrology” indicators
    means “the area is likely non-wetland.” 
    Id.
    If the NRCS agents suspected that hydrology had been al-
    tered, then their obligation before proceeding to a comparison
    site was to determine whether the fields could have had wet-
    land hydrology—for example by “Verify[ing] that the area is
    in a landscape position that is likely to collect or concentrate
    water,” such as a “[c]oncave surface (e.g., depression or
    swale).” 
    Id.
     The NRCS agents could have opted to look for one
    of almost two dozen primary or secondary indicators of wet-
    land hydrology, such as making a “[v]isual observation of in-
    undation” or “soil saturation” by digging a soil pit and “ob-
    serving the level at which water stands in the hole” to see if
    the soil saturation occurred “within a major portion of the
    root zone (usually within 12 inches of the surface).” Corps
    Manual at 32.
    That is not what happened, in 2002 or 2013. Instead the
    NRCS agents assumed that Un1 and Un2 had been drained of
    hydrology by the installation of field drainage tile and that
    Un1 and Un2 were in a depression. Both assumptions are, on
    this record, demonstrably wrong. The agents then walked
    down-gradient to Field 7, which is in a depression, into the
    midst of hydrophytic vegetation in what is obviously a wet-
    land. They used those observations of Field 7 to declare that
    Un1 and Un2 were converted wetlands. That is not agency ex-
    pertise that deserves deference. It is arbitrary and capricious
    disregard for both the facts and the law. The only evidence in
    the administrative record regarding the actual hydrology of
    No. 16-1654                                                   37
    Un1 and Un2 is Ms. Hauer’s initial 2002 measurements
    (which did not find wetland hydrology) and the work of the
    experts hired by Mrs. Boucher. They discovered neither
    drainage tiles nor sufficient wetland hydrology, and they
    measured the topography showing that Un1 and Un2 are not
    concave or in a depression.
    Instead of defending these shortcomings, the government
    on appeal has tried to change the subject. The USDA contends
    that it is Mrs. Boucher who is making the case “all about
    drainage tiles, but USDA’s decision had little if anything to do
    with tiles,” and that “throughout the administrative ap-
    peal … the agency relied not just primarily but exclusively on
    tree removal to justify [its] decision.” It was only “[t]he woody
    vegetation consideration, not the drainage assumption, [that]
    was the reason NRCS prepared an Atypical Situation Data
    Sheet and was the driving factor in the use of Field 7 to assess
    hydrology.” We are not sure what 600-page administrative
    record the government is looking at, but it does not appear to
    be the same one we have.
    We should be done—or rather the USDA should have
    been done when Mrs. Boucher provided evidence that the
    USDA experts should have found or recognized a decade ear-
    lier. But the USDA has continued to press two points on ap-
    peal—justifying its choice of a comparison site and its claim
    that it may forgo any assessment of wetland hydrology on a
    disputed site. We must address these points in a little more
    detail.
    A. Choice of Comparison Site
    The USDA argues that 
    7 C.F.R. § 12.31
    (b)(2)(ii) authorizes
    it to select a comparison site solely on the basis of the
    38                                                       No. 16-1654
    comparison site’s location “on the same hydric soil map
    unit”—without any reference as to whether the sites have
    similar hydrologic features. “In fact,” the agency argues, “that
    regulation does not govern the hydrology-comparison-site
    question.” USDA Br. at 30. The USDA contends that the ap-
    propriate “rules field experts follow for when and why to use
    a comparison site to assess hydrology come from the applica-
    ble policy manuals,” including—it says—the Corps Manual,
    the Corps Supplement, and the NRCS Procedures. 
    Id.
     at 30–
    31. Furthermore, the USDA asserts, “The Atypical Situation
    Data Sheet for Fields Un1 and Un2 specifically explained that
    NRCS used such guides to analyze Boucher’s property.” 
    Id. at 31
    . Mrs. Boucher’s positions are “no match for the straightfor-
    ward expertise and guidance of the Army Corps Manuals,”
    says the government. 
    Id.
    Let’s go to the tape, as they say. Not only would such an
    interpretation cut against a regulatory scheme designed to
    conserve wetlands, but it is directly contrary to the very guid-
    ance the USDA cites. The NRCS Procedures do indeed say
    that the NRCS may rely on an “adjacent vegetation data
    source … ‘in the local area on the same hydric soil map’” as
    authorized by the “Corps manual … Step 3.” (5-30), quoting 7
    C.F.R. 12.31(b)(2)(ii). But we cannot stop reading yet: The
    “comparison site should support hydrologic conditions that are
    similar to what existed on the altered site prior to the drainage.” 
    Id.
    (emphasis added); see also footnote 6, supra (citing some of
    the extensive agency guidance which directs that comparison
    sites have hydrology comparable to disputed sites).
    The USDA also points to a second passage in a guidance
    manual, both in its appellate brief and in oral argument, that
    it contends supports this position. Quoting from the Corps
    No. 16-1654                                                     39
    Supplement: “The evaluation of wetland hydrology requires
    special care on any site where indicators of hydrophytic vegetation
    and hydric soil are present but hydrology indicators appear to
    be absent,” which is particularly important in the Midwest.
    USDA Br. at 31, citing Corps Supplement at 118 (emphasis
    added here). In such a situation, the government wrote on ap-
    peal: “The applicable manuals prescribe ‘a number of ap-
    proaches that can be used to determine whether wetland hy-
    drology is present on sites where indicators of hydrophytic vege-
    tation and hydric soil are present but hydrology indicators may
    be lacking due to normal variations in rainfall or runoff, hu-
    man activities that destroy hydrology indicators, and other
    factors.’” Id. at 31–32, quoting Corps Supplement at 118 (em-
    phasis added here). One such option when hydric soils and
    hydrophytic vegetation are present, but “human activity” has
    apparently “destroyed hydrology,” is to “determine hydrol-
    ogy with a reference site.” Id., quoting Corps Supplement at
    118, 121. “That describes Un1 and Un2 exactly,” says the gov-
    ernment, and “That is what NRCS did here.” Id. at 31–32.
    This critical passage in the government’s argument has
    several important errors and omissions. First, the entire point
    of the cited passage of the Corps Supplement—as reflected in
    the passages quoted above—is to address situations where,
    despite the seeming absence of hydrology, both hydric soils
    and hydrophytic vegetation are present. That does not “describe
    Un1 and Un2 exactly.” The thrust of this portion of the man-
    ual’s guidance is to assist NRCS experts assess situations
    where hydrology is lacking, be it due to a “dry season,” “dry
    year,” “normal variations in rainfall or runoff,” or “human ac-
    tivities.” Corps Supplement at 118. Thus, this section outlines
    approaches NRCS experts can take when making site visits
    40                                                  No. 16-1654
    “during the dry season,” or during “Periods with below-nor-
    mal rainfall,” or “Drought years.” Id. at 119–21.
    Even if this guidance did not specifically describe “ap-
    proaches that can be used to determine whether wetland hy-
    drology is present on sites where indicators of hydrophytic
    vegetation and hydric soil are present but hydrology indica-
    tors may be lacking,” it is still instructive to review the guid-
    ance. Corps Supplement at 118. At Step 1, NRCS agents must
    “Verify that indicators of hydrophytic vegetation and hydric
    soil are present, or are absent due to disturbance or other
    problem situations,” and “proceed to step 2.” Id. at 119. The
    NRCS agents must then “Verify that the site is in a geo-
    morphic position that is likely to collect or concentrate water,”
    such as a “Concave surface (e.g., depression or swale).” Id.
    Again, the only scientific or technical evidence in this case’s
    Administrative Record as to site topography is that Un1 and
    Un2 are not concave or in a depression—unlike Field 7.
    Assuming for purpose of argument that “the landscape is
    appropriate,” i.e., is in a “depression,” we might “proceed to
    step 3.” Corps Supplement at 119. NRCS agents may now
    “use one or more” of several described “approaches to deter-
    mine whether wetland hydrology is present and the site is a
    wetland,” but must in “the data form or in the delineation re-
    port, explain the rationale for concluding that wetland hy-
    drology is present even though indicators of wetland hydrol-
    ogy described in [this manual] were not observed.” Id. Of
    course, the only rationale in the record for the NRCS experts
    believing hydrology sufficient for wetlands had once been
    present was their erroneous assumption, made by the agency
    from 2002 forward, that drainage tile had been installed in
    Un1 and Un2 after 1985—an untested and undocumented
    No. 16-1654                                                            41
    hypothesis that Mrs. Boucher’s evidence thoroughly refuted.
    But let’s proceed.
    As the USDA points out, the use of “Reference sites” for
    hydrology determinations is permitted, but with the follow-
    ing guidance: “If indicators of hydric soil and hydrophytic
    vegetation are present on a site that lacks wetland hydrology
    indicators, the site may be considered to be a wetland if the
    landscape setting, topography, soils, and vegetation are substan-
    tially the same as those on nearby wetland reference areas.” Corps
    Supplement at 121 (emphasis added). Again, this guidance
    presumes the comparison of extant hydrophytic vegetation
    and insists that the topography be “substantially the same.”
    Neither condition applies to this case. 12
    NRCS agents, however, are not left high and dry, so to
    speak, if they suspect hydrology has been altered and there is
    no adequate reference site. The guidance outlines additional
    “Hydrology Tools,” which are “a collection of methods that
    can be used to determine whether wetland hydrology is pre-
    sent on a potential wetland site that lacks [hydrology] indica-
    tors due to disturbance or other reasons, particularly on lands
    used for agriculture.” Corps Supplement at 122. These are
    back-up tools, to “be used only when an indicator-based wet-
    land hydrology determination is not possible or would give
    misleading results,” and they include: analyzing stream and
    lake gauge data; estimating runoff volumes, duration, and
    frequency of ponding in depressional areas based on
    12 At oral argument, the government read from this portion of the Corps
    Supplement but did not include Step 2’s direction to verify the disputed
    site’s geomorphic position and Step 3’s insistence that, at a minimum, any
    reference site should share that same topography—just as it omitted those
    points in its appellate brief.
    42                                                No. 16-1654
    precipitation; evaluating the frequency of wetness signatures
    on repeated aerial photography; and estimating the “scope
    and effect” of ditches or any subsurface drainage systems—
    such as the drainage system the NRCS agents presumed was
    present in this case. Id.
    It is disappointing that the NRCS agents failed to try an
    “indicator-based wetland hydrology” approach and made no
    effort to use any of these back-up tools. The guidance also spe-
    cifically advises: “A hydrologist may be needed to help select
    and carry out the proper analysis” in situations where poten-
    tial lack of hydrology is an issue. Corps Supplement at 122.
    Hydrology was at the heart of this dispute, but none of the
    experts NRCS sent to the Boucher farm was a hydrologist. The
    USDA’s selection of Field 7 as a reference site remains defi-
    cient, even under the most favorable reading of the govern-
    ment’s preferred regulation and guidance, making the result-
    ing decision arbitrary and capricious.
    B. Forgoing Hydrology Altogether
    The USDA’s second argument, taken to its logical conclu-
    sion, would eliminate the need to choose a comparison site at
    all, but would run directly counter to controlling statutes and
    regulations. The government relies on 
    7 C.F.R. § 12.32
    (a)(2) to
    assert that removal of woody hydrophytic vegetation from
    hydric soils is—by itself—sufficient to declare the area “con-
    verted wetland.” If one reads only the highlighted provision,
    and not the governing statute, other regulations, or any guid-
    ance manuals, this could be a plausible reading:
    Where woody hydrophytic vegetation has been
    removed from hydric soils, for the purpose of or
    permitting the production of an agricultural
    No. 16-1654                                                   43
    commodity, the area will be considered to be
    converted wetland.
    
    7 C.F.R. § 12.32
    (a)(2).
    This argument reaches too far. First, here is the prefatory
    language for this provision: “Converted wetland shall be
    identified by determining whether the wetland was altered so as
    to meet the definition of converted wetland,” and “in making
    this determination, the following factors are to be consid-
    ered”—including whether woody hydrophytic vegetation
    has been removed. 
    7 C.F.R. § 12.32
    (a) (emphasis added). This
    language indicates that the regulation—as a whole—does in
    fact comport with the statutory requirement that soils, vege-
    tation, and hydrology are to be considered by first determin-
    ing whether the land is wetland, before considering various
    factors (including the removal of woody hydrophytic vegeta-
    tion) to determine whether it is converted wetland. Gunn v.
    U.S. Dep’t of Agriculture, 
    118 F.3d 1233
    , 1236–37 (8th Cir. 1997)
    (reading § 12.32(a) as “list[ing] factors that are to be consid-
    ered in determining whether a wetland has been converted,
    such as whether woody hydrophytic vegetation has been re-
    moved”).
    The government’s reading of this provision—the removal
    of woody hydrophytic vegetation from hydric soil is sufficient
    by itself to deem the site a converted wetland, without refer-
    ence to hydrological factors—conflicts with the statutory def-
    inition’s focus on hydrology. As defined in 
    16 U.S.C. § 3801
    (a)(7)(A), a “converted wetland” is:
    wetland that has been drained, dredged, filled,
    leveled, or otherwise manipulated (including
    any activity that results in impairing or
    44                                                      No. 16-1654
    reducing the flow, circulation, or reach of water)
    for the purpose or to have the effect of making
    the production of an agricultural commodity
    possible.
    Agency guidance also maintains a focus on hydrology: “If in-
    dicators of either hydric soil or [one primary or two secondary
    indicators of] wetland hydrology are absent,” then NRCS
    should conclude that “the area is likely non-wetland.” Corps
    Supplement at 105. 13 More practically, if the agency experts in
    this case actually believed they were operating under this in-
    terpretation of 
    7 C.F.R. § 12.32
    (a)(2), why bother to select the
    comparison site of Field 7 and spend years defending that (ar-
    bitrary) decision?
    This is not to say that the wholesale removal or decimation
    of hydrophytic vegetation could not affect an area’s hydrol-
    ogy. In fact, the potential for such cause and effect is acknowl-
    edged in the agency’s guidance because the presence or ab-
    sence of vegetation can result in either more water or less wa-
    ter than would otherwise have been present. The Corps Man-
    ual specifically cautions NRCS agents that the “[t]ype and
    amount of plant cover affect both degree of inundation and
    duration of saturated soil conditions.” Corps Manual at 29.
    For example, in some areas of “abundant plant cover,” the
    “[e]xcess water drains more slowly … thereby increasing fre-
    quency and duration of inundation and/or soil saturation.” 
    Id.
    “On the other hand,” the Manual notes, it is possible for “tran-
    spiration rates [to be] higher in areas of abundant plant cover,
    13The Corps Supplement adds at page 105 the caveat “unless soil and/or
    hydrology are also disturbed or problematic,” but on appeal the USDA
    has abandoned any claim the fields were drained.
    No. 16-1654                                                  45
    which may reduce the duration of soil saturation.” 
    Id.
     In short,
    the agency sends experts out into the field because someone
    must assess whether and how a change in vegetation has af-
    fected the area’s hydrology.
    The agency’s assertion—that the removal of nine trees re-
    moved wetland hydrology from several acres of land—is in-
    compatible not just with common sense, but with the NRCS
    experts’ notation in their 2013 Atypical Situation Data Sheet
    for the Boucher farm. They noted: “Soils were minimally dis-
    turbed from land clearing and the site was not filled,” such
    that there was only a “Minimal impact” on the area. App. 57.
    There is no indication in the experts’ documentation that the
    removal of trees altered the disputed site’s hydrology, despite
    the agency’s position in subsequent litigation.
    Furthermore, even if the USDA had adequate support for
    the convenient but arbitrary shortcut offered by its proposed
    interpretation of 
    7 C.F.R. § 12.32
    (a)(2), that provision still
    would not support a finding of converted wetland based on
    the record evidence in this case. Despite the government’s as-
    sertion in oral argument that “there has never been a dispute
    as to whether or not the trees that were removed were hydro-
    phytic vegetation,” Mr. Boucher documented that the nine
    trees he removed from the disputed site were categorized as
    FACU—i.e., Facultative Upland Plants that are unlikely to oc-
    cur in wetland areas, a point made repeatedly by Mrs. Bou-
    cher during the agency appeals process. See supra at 28–29.
    An assertion by the agency’s appeal coordinator during the
    June 2013 hearing that some trees located on the comparison
    site appeared to be in the same “family” as the removed trees
    is not contrary evidence, and is not, at least without more,
    even relevant. The agency’s mandate is to make an
    46                                                        No. 16-1654
    assessment of sufficient reliability that benefits can be denied.
    This post hoc observation by an agency officer is not enough
    for us to avoid finding the agency’s decision arbitrary and ca-
    pricious.
    The agency has also claimed on appeal that the NRCS ex-
    perts could identify the species of removed trees via a series
    of aerial photography slides. There is no evidence in the rec-
    ord that any agency expert has attempted to identify the re-
    moved trees’ species from aerial photographs. We confirmed
    at oral argument which photographs might be relied upon by
    the agency. While it is possible those blurry photographs are
    of the Boucher farm, it is highly implausible that anyone—
    even USDA experts—could use these aerial photographs to
    identify species of trees. 14
    As Mrs. Boucher correctly pointed out during the agency
    appeal process, the NRCS experts did not attribute the altera-
    tion of hydrology to the removal of the nine trees, and the
    agency presented no evidence that the tree removal altered
    the wetland hydrology. The USDA hearing officer and appel-
    late officer failed to engage meaningfully with this point,
    thereby ignoring a crucial factor under the agency’s
    14 Post hoc agency arguments notwithstanding, it is unclear from the ad-
    ministrative record what slides the agency experts reviewed before sub-
    mitting their 2013 Final Technical Determination. During the June 2013
    agency hearing, Mrs. Boucher referenced aerial photographs to support
    her contention that the disputed parcels had never sustained a vegetation
    coverage like that on Field 7. The USDA appeals coordinator inquired how
    Mrs. Boucher had obtained the referenced slides because the agency’s
    technical team had “misplaced” their copies and had never requested any
    replacement slides from FSA. There are some notes from Karen Hauer’s
    initial 2002 assessment reviewing slides, but those notes do not mention
    any species of trees.
    No. 16-1654                                                 47
    interpretation of this regulation, rendering the decision arbi-
    trary and capricious.
    The district court’s judgment is REVERSED and the case
    is REMANDED with instructions to enter judgment for Mrs.
    Boucher, awarding all appropriate relief.