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, 

, 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, 

,
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, 

, 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, 

, 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, 

, 416 (1971); see
also Motor Vehicle Manufacturers Ass’n v. State Farm Mutual Au-
tomobile Ins. Co., 

, 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, 

, 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, 

, 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., 

, 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, 

, 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.